December 2012 Volume 22, Number 12 pp.

611–692

Special Issue – Synthetic Cell Biology

Editor
Editorial
Rebecca Alvania
Executive Editor, Cell Biology 611 Cell Biology 2.0 Wendell A. Lim, Rebecca Alvania, and
Deborah Sweet Wallace F. Marshall
Journal Manager
Jeanette Bakker
Forum: Science & Society
Journal Administrators 613 Science ⴙ dance = bodystorming Carl Flink and David J. Odde
Patrick Scheffmann
Ria Otten
Opinion
Advisory Editorial Board
Wolfgang Baumeister, Munich, Germany 617 Synthetic multicellularity Michel M. Maharbiz
Dominique Bergmann, Stanford, USA
Pascale Cossart, Paris, France
David Drubin, Berkeley, USA Reviews
William C. Earnshaw, Edinburgh, UK
Scott Emr, Ithaca, USA
624 Designer nucleic acids to probe and Yamuna Krishnan and Mark Bathe
Anne Ephrussi, Heidelberg, Germany program the cell
Susan Gasser, Geneva, Switzerland
David Goldfarb, New York, USA
634 Towards a bottom-up reconstitution Ariadna Martos, Mercedes Jiménez,
Ari Helenius, Zurich, Switzerland of bacterial cell division Germán Rivas, and Petra Schwille
Nobutaka Hirokawa, Tokyo, Japan
644 Engineered, harnessed, and hijacked: Brian S. Goodman, Nathan D. Derr, and
Alan Rick Horwitz, Charlottesville, USA
Tony Hunter, La Jolla, USA synthetic uses for cytoskeletal systems Samara L. Reck-Peterson
Chris Marshall, London, UK
Sascha Martens, Vienna, Austria 653 Principles for designing ordered protein Yen-Ting Lai, Neil P. King, and
Andrew Matus, Basel, Switzerland assemblies Todd O. Yeates
James Nelson, Stanford, USA
Hugh Pelham, Cambridge, UK 662 Designing biological compartmentalization Anna H. Chen and Pamela A. Silver
Hidde Ploegh, Cambridge, USA
671 Directed cytoskeleton self-organization Timothée Vignaud, Laurent Blanchoin,
James R. Woodgett, Toronto, Canada
and Manuel Théry
Editorial Enquiries 683 Directing the assembly of spatially Jennifer S. Liu and Zev J. Gartner
Trends in Cell Biology
Cell Press organized multicomponent tissues from
600 Technology Square the bottom up
Cambridge, MA 02139, USA
Tel: +1 617 386 2105
Fax: +1 617 397 2810
E-mail: tcb@cell.com

Cover: In recent years, the bottom-up approach of synthetic biologists has yielded new insight into fundamental aspects of cell biology.
In this special issue, co-guest edited by Wendell A. Lim and Wallace F. Marshall (editorial on pages 611–612), we highlight some of the
exciting work that has sprung from this intersection between synthetic and cell biology. On the cover, the construction of a single cell is
depicted via an instruction sheet similar to that which might be found in a child’s game. The cover is meant to represent the constructionist
approach to understanding the inner workings of the cell. Cover design by Yvonne Blanco.
 Editorial
Special Issue – Synthetic Cell Biology
Cell Biology 2.0
Wendell A. Lim1, Rebecca Alvania3, and Wallace F. Marshall2
1
Department of Cellular and Molecular Pharmacology, University of California, San Francisco, Genentech Hall – N412E,
600 16th Street, San Francisco, CA 94158, USA
2
Department of Biochemistry and Biophysics, University of California, San Francisco, 600 16th Street, San Francisco,
CA 94143-2200, USA
3
Editor, Trends in Cell Biology
‘Verum esse ipsum factum’, the true is in the made –
Giambattista Vico
Synthetic Cell Biology sounds intriguing, but the name
begs the question – why should we try to rebuild or repro-
gram the cell, especially when we barely understand how
cells work? This issue of TiCB explores this emerging area
in which scientists are taking apart, rebuilding, repro-
gramming, and repurposing parts of the cell. The reviews
cover a wide range of scales, from microscopic molecular
machines to macroscopic, multicellular tissues. These
reviews highlight the fact that, in addition to its role in
harnessing and unleashing the power of cells for new and
future applications, synthetic biology also has an impor-
tant role to play in facilitating the understanding of com-
plex cellular processes. In short, we can learn much about
cells through the process of trying to build them (or parts of
them). And if we understand cells, we can start to really
harness their power.
Cell biology has historically been primarily an observa-
tional science, inextricably linked with tools like the mi-
croscope, which first enabled humans to peer into the
remarkable world of the cell, with its diversity of forms
and complex and dynamic inner structure. Yet today, even
as we know the genomic parts list of the cell, we realize that
there is a huge gap in our mechanistic understanding of the
logic of how these systems work. On the one hand, we can
observe these beautiful systems that yield complex cellular
structures, movements, and regulatory decisions; on the
other, we have a list of molecular parts that somehow
underlie these behaviors and decisions. But the middle
ground linking how these parts fit together and work as a
system is often missing. We have a serious gap in the
mesoscopic mechanistic understanding of how microscopic
molecules work together to yield complex macroscopic
behavior.
Synthetic biology as applied to cell biology can help to
address this gap, because in many ways it is a philosophical
descendant of old-school reconstitution biochemistry. It
offers a unique and powerful way to construct simple,
stripped down, and manipulable molecular systems, which
can then be used gain insight into the fundamental rules of
how some complex biological process takes place. But now,
thanks to the power of genetic engineering, much of this
Corresponding authors: Lim, W.A. (lim@cmp.ucsf.edu);
Marshall, W.F. (Wallace.Marshall@ucsf.edu).
0962-8924/$ – see front matter ß 2012 Published by Elsevier Ltd. http://dx.doi.org/10.1016/j.tcb.2012.10.004 Trends in Cell Biology, December 2012, Vol. 22, No. 12
manipulation and perturbation of systems can take place in
the milieu of the living cell, with its basic metabolic and
transcriptional systems and complex spatial environments.
The cell, or even artificially constructed cells, can serve as
the new test tube for logical and systematic analysis of a
process. As Vico first suggested (and Richard Feynman later
reiterated), if we can build it, we can understand it.
Predictive, mechanistic understanding of how the parts
of a cell work together to generate cell behaviors is not only
interesting in its own right; it would provide a rational
basis to turn cell biology into an engineering discipline.
Synthetic cell biology has the promise of remarkable bio-
technology applications. Many advances have been made
in harnessing cells to create new and useful molecules, by
cobbling together new strings of metabolic enzymes. But in
many ways, this is only the tip of the iceberg of what cells
could do. Current efforts have largely ignored the cell, its
structure and regulation, as a tool to harness. Cells are
some of the most powerful and capable devices we know,
especially in producing complex products. Manipulation of
cellular organelles, transport systems, and secretion sys-
tems and their regulation could, if properly applied, have
dramatic effects on cell-based production of fuels, drugs,
nutrients, and chemicals. Ultimately, learning how to
reprogram the machinery of the cell extends far beyond
biofermentation – it may lead to dramatic advances in
designer cells (or cell-inspired devices) that can execute
precision therapeutic or regenerative functions. It may
even be possible to create new cells that have hybrid
functions and structures that are completely novel. The
next few decades offer tantalizing promise of how our
understanding of cell biology could be applied, and it is
time for cell biologists to start thinking creatively about
how cells as machines could be harnessed.
The reviews in this issue cover several fundamental
issues. Thery, Schwille, and Odde discuss different ways
that reconstitution or synthetic approaches can be used to
understand the fundamental problem of self-organization
– how molecules dynamically interact with one another to
yield higher-order structure systems such as the cytoskel-
eton and the cell division machinery. Odde explores how
cellular processes can be reconstituted or visualized using
dancers as a medium.
Maharbiz and Gartner explore the intriguing problem
of multicellular assembly, which is intimately linked to
developmental biology. Can we understand the rules of
611
Editorial
how molecular systems can lead to hierarchical assembly
of multicellular structures like tubes and spheres and
eventually into complex tissues? How might we manipu-
late this to gain a better understanding, and is it eventu-
ally possible that we can build novel and arbitrary
macroscopic structures from reengineered cells?
Bathe, Yeates, Reck-Petersen, and Silver explore differ-
ent ways in which we can harness and repurpose the
machinery of the cell. Bathe explores the diversity of
ways in which nucleic acids, with their simple Watson–
Crick base pairing can be used as a substrate to build
complex assemblies, carry out catalysis, and build
612
Trends in Cell Biology December 2012, Vol. 22, No. 12
information-processing systems. Yeates explores the way
in which known protein modules can be used to build novel
assemblies. Reck-Petersen discusses how we can harness
the power of cytoskeleton-based transport machines, and
Silver reviews progress towards harnessing cellular
compartmentalization for metabolic engineering.
These articles underscore the promise of synthetic cell
biology, both as an engineering discipline and as an alter-
native route to understanding the secret of how cells work.
But this issue is not so much a roadmap, as an invitation to
participate. Figuring out how to reimagine the cell biology
of the future is left as an exercise for the reader.
Forum: Science & Society
Special Issue – Synthetic Cell Biology
Science + dance = bodystorming
Carl Flink1 and David J. Odde2
1
Department of Theatre Arts & Dance, University of Minnesota, Minneapolis, MN 55455, USA
2
Department of Biomedical Engineering, University of Minnesota, Minneapolis, MN 55455, USA
In everyday life, gravity and inertial forces often domi-
nate our movements; in the cell, these forces pale in
comparison to thermal forces. The violent, collisional
world of the cell, where water moves faster than a jet
airliner, can be difficult to imagine. To develop our
intuitive understanding of cellular and molecular pro-
cesses, we are exploring the concept of ‘bodystorming’,
where human ‘movers’ act as molecules that diffuse,
undergo reactions, and generate/absorb forces.
Getting started
For the past few years, we have worked together to explore
how science can be informed through dance and vice
versa. As a starting point, we considered the concept of
‘catastrophe’, an important feature of microtubule self-as-
sembly [1], where an abrupt, stochastic switch from an
assembling state to a disassembling state allows dynamic
microtubules rapidly to explore the intracellular environ-
ment. Catastrophe is vital for proper formation of the mitotic
spindle, and some of the more widely used anticancer
drugs (e.g., paclitaxel) specifically target microtubule self-
assembly.
We came to focus on catastrophe after a chance meeting
sponsored by the Institute for Advanced Study (IAS) at the
University of Minnesota, through which we had separately
developed collaborative projects under the general theme
of ‘time’. One project focused on ‘Catastrophe’, a research
collaborative organized by D.O. and his brother Thomas C.
Odde (film studies), that explored disruption of time from
the multiple perspectives of art-making, biology, and engi-
neering. The other project was entitled ‘Wreck’, an explo-
ration of the depths of physical and psychological
endurance in the face of impending destruction, developed
by C.F. as artistic director of the Black Label Movement
(BLM) Dance Company. Under the direction of Ann Waltner,
IAS required us to meet with other collaborative leaders,
which prompted us to propose The Moving Cell Project,
where the concept of microtubule catastrophe would be
explored through dance (Box 1).
Basic moves
The cytoplasm is a violent place. The root-mean-square
speed of a molecule is:
rffiffiffiffiffiffiffiffiffiffiffiffi
3kB T
vrms ¼ [1]
m Which model?
Corresponding author: Odde, D.J. (oddex002@umn.edu).
where kB is Boltzmann’s constant, T is absolute tempera-
ture, and m is the mass of the molecule. For water, the most
abundant molecule in the cell, the root-mean-square speed
exceeds 600 m/s (1400 m.p.h.). As water molecules move at
these speeds, they collide with other molecules moving at
similar speeds, creating an extreme collisional environ-
ment. It is these collisions that give rise to diffusion, the
apparently random movements of molecules. For a dance
to convey a cellular process, such as microtubule assembly,
it is important to first appreciate the thermally driven
environment within which this process occurs. Fortunate-
ly, BLM embraces physicality and seeks safe ways in which
collision and contact can be incorporated into dance. To
convey this philosophy, BLM often refers to its dancers as
movers (Figure 1).
To simulate diffusion, we started with training exercises
where a coin flip determined the left–right and forward–
backward movements of the dancers. Although we have not
formally checked that the movers are executing truly ran-
dom walks, one gets the impression that they are (it is a
challenge to demonstrate true randomness; we routinely use
computers to simulate diffusion even though we know that
computers are deterministic machines). Initially, movers
were protected by body suits, but we found that these were
too cumbersome. Instead, we focused on developing trust
within the movers. An overarching principle is that movers
should neither seek nor avoid collisions. In fact, we routinely
adjust the rate of random stepping to adjust the contact level,
equivalent to increasing/decreasing the temperature.
Once we had movers diffusing, we added bond formation.
If the movers face each other when bonded, they can form
dimers. Because the Debye screening length (the distance
over which attractive forces are felt between macromole-
cules) is typically of the order of 1 nm or less, the bonds do not
form until the movers are in direct contact. When the
bonding is front to back, self-assembled linear filaments
can form. We simulate bond breaking in various ways,
typically by throwing balls at the bonded partners. We then
enclose all the movers in a circle of movers holding hands to
simulate enclosure of the diffusion and reactions within a
membrane bilayer. Working with a professional dance com-
pany means that the movers have a high degree of physical
intelligence and can build trust quickly. This allows us to
move rapidly with a robust range of physicality available to
express the scientific concepts.
Initially, our plan was to convey scientific concepts, such as
microtubule catastrophe, to a lay audience through dance.
613
Forum: Science & Society
Box 1. Useful links
Robert Hammel Vimeo Site
http://vimeo.com/30346802 – video documenting The Moving Cell
Project (20 minutes).
Twin Cities Public Television
http://www.mnoriginal.org/episode/hit-the-moving-cell-project/ – in-
terview with D.O. and C.F. and project clips (10 minutes)
http://www.mnoriginal.org/episode/black-label-movement-hit/
– edited extended excerpt of BLM performing ‘HIT’ (10 minutes)
Institute for Advanced Study
http://ias.umn.edu/2011/09/22/moving-cell-2/ – video podcast of a
joint class between BMEN 5351 Cell Engineering (D.O, instructor)
and DNCE 3010 Modern Dance Technique (C.F., instructor) held in
the Barker Dance Studio at the University of Minnesota (85 minutes)
Chicago Tribune review of ‘HIT’
http://articles.chicagotribune.com/2011-03-11/features/ct-live-0312-
same-planet-dance-review20110311_1_dancer-dance-center-
collisions
Twin Cities Star Tribune review of ‘HIT’
ht tp: //w ww. star tri bun e.c om/ ent e rta inm ent /st agea ndar ts/
139153774.html?refer=y
BLM website
http://www.blacklabelmovement.com/
David Odde’s laboratory
http://oddelab.umn.edu/
Although microtubules are central to eukaryotic life
and the target of many cancer treatments, they remain
virtually unknown to the general public. Using dance to
convey basic ideas is not entirely novel; perhaps the most
famous example is a dance performed by Stanford Univer-
sity students in the 1970s entitled ‘A Protein Primer’
(http://www.youtube.com/watch?v=u9dhO0iCLww). Since
the 1970s, advances in computer animation have led to
beautifully rendered depictions of cellular processes, such
as the ‘Inner Life of the Cell’ video from XVIVO Scientific
Animation and BioVisions Multimedia at Harvard Univer-
sity (http://multimedia.mcb.harvard.edu/anim_innerlife_
music.html). Nevertheless, the medium of dance remains
a vital means of communicating science, as evidenced by
the success of the ‘Dance Your Ph.D.’ contest, an annual
event organized by John Bohannon, a correspondent for
Science. John has joined our Moving Cell Project and
TRENDS in Cell Biology
Figure 1. Black Label Movement ‘movers’ Eddie Oroyan and Carl Flink collide in a
simulation of the thermally driven environment of the cell.
614
Trends in Cell Biology December 2012, Vol. 22, No. 12
together we have explored the extent to which dance can
be used as a communication medium in science, most
recently exemplified in a dance performance at the
2011 TedX-Brussels conference (http://www.ted.com/talks/
john_bohannon_dance_vs_powerpoint_a_modest_proposal.
html).
In either dance or computer animation, key decisions
must be made about what to include in the performance/
animation and what not to include. Naturally, every sci-
entist will have his or her own unique scientific views, just
as choreographers will have their own unique perspectives.
The question then becomes, which model for a cellular
process should be depicted? A model from a major textbook
or a recent review article? Or perhaps a new model that a
scientist is developing? In choreographing a dance, a sci-
entist is making key decisions about which model to depict.
In choosing to depict a cellular process in a way that differs
from how another scientist would depict it, we have the
beginning of a potentially constructive scientific discus-
sion. It occurred to us that dance could potentially be
informative not only to the general public, but also to
scientists at the cutting edges of scientific knowledge as
they articulate their hypotheses through the dances that
they choreograph.
Bodystorming: rapid prototyping of hypotheses
One might wonder whether one can ‘prove’ anything using
dance; why not just use mathematics or computer simula-
tions to make testable predictions from hypotheses? In
fact, this is a standard approach in our research and is
fast becoming standard in cell biology research generally,
as described in this special issue on synthetic biology.
Because of the complexity of cellular processes, we gener-
ally need to solve equations numerically or perform Monte
Carlo simulations, depending on whether the dynamics are
fundamentally deterministic or stochastic [2]. In practice,
doing theory via computer simulation requires a signifi-
cant investment in computers and time spent program-
ming and debugging code. It is often weeks or months
before useful simulation output is obtained; when the
simulations are done, we are left with the task of inter-
preting the output. Often we need to revise the output
statistics from the simulation and try different ways of
plotting the results or making animations of the output.
Even when the model makes testable predictions that we
test experimentally and find to be correct, we often do not
intuitively understand why the model behaves in the way
that it does. This last phase of modeling, model deconstruc-
tion, is now taking us at least as much time as the model
construction phase. A modeling project can continue for
months before we start to feel confident that the model is
physically sound, makes testable predictions that are cor-
rect, and is sufficiently deconstructed that we understand
it intuitively.
Dance provides us with a useful complement to comput-
er simulation, allowing us to rapidly prototype hypotheses
without having to code and debug each one on a computer.
Using the basic moves described above, we can rapidly
explore the consequences of our hypotheses using movers.
Because these dances are most useful at the earlier brain-
storming stage, we refer to the process of dancing out our
Forum: Science & Society
TRENDS in Cell Biology
Figure 2. Scientists and Black Label Movement ‘movers’ simulate filament self-
assembly, an example of bodystorming.
hypotheses as bodystorming (Figure 2). Often we see in the
first few moments of a bodystorming dance that the
experimentally observed behavior will not emerge under
the hypothesized rules. What is especially interesting is
when the reasons for the discrepancy are not evident to
the scientists as the dance unfolds. At this point, we
engage in a dialog with the movers: why is the expected
behavior not occurring? Unlike the computer, the movers
can engage in discussion with the scientists and with
each other. In addition, the scientists themselves can
become movers and inhabit the cell that they are imag-
ining, which can be a profound and visceral experience. If
there are N people engaged in the dance, there are of the
order of N2 possible discussions that can take place after
the dance. By contrast, in the traditional approach, the
computer is mute and the task is left to the scientist to
interpret the numerical output in tabular, graphical, or
animated output form. It is interesting to note that there
is a general trend toward agent-based simulations in cell
biology [2]. In the case of bodystorming, the agents are
human beings who can reflect on their experience in the
simulation.
To see whether BLM movers could facilitate scientific
dialog and debate, we invited biochemists Dyche Mullins
(University of California San Francisco) and Enrique De
La Cruz (Yale University) to join us in a rapid prototyp-
ing experiment. Along with John Bohannon, who sug-
gested the term bodystorming during these sessions, we
explored a wide range of scientific questions. Perhaps
most stimulating was the question of the effect of mac-
romolecular crowding on association rates in vivo. Most
of our kinetic and thermodynamic measurements have
been made in vitro, but we do not yet fully understand
how the macromolecules of the cytoplasm affect kinetics
and thermodynamics [3]. Therefore, understanding
crowding is vital to linking our understanding of in vitro
biochemistry to in vivo biochemistry and cell biology.
Although we did not settle any issues definitively, we
developed a better intuitive understanding of crowding,
and the issues that were raised prompted us to go back to
the laboratory to set up rigorous computer simulations,
work that is ongoing. In this case, bodystorming
allowed us to engage in rapid prototyping to identify
key issues for further exploration by more traditional
approaches.
Trends in Cell Biology December 2012, Vol. 22, No. 12
A new way of teaching?
One exciting aspect of the collaboration is how the BLM
movers have responded positively to the science. The BLM
company members are highly trained professional dancers
with scientific backgrounds of varying extent. We have
found that abstract concepts, such as diffusion and self-
assembly, are relatively easy for a dancer to understand
when put into concrete physical terms. When we teach
science, students are normally seated and the teaching is
largely verbal and visual (e.g., lecture with whiteboard
and/or computer slides). When we teach science to dancers,
both teacher and students are on their feet and in motion.
This allows the teacher and student to bring physicality
into the learning process, which we suspect is especially
effective for students whose thinking is intimately linked
to physicality (e.g., dancers and athletes). The integration
of bodystorming into our educational system seems rela-
tively unexplored. We envision possible collaborations be-
tween teachers in physical education, arts, and science,
with students encouraged to learn actively through partic-
ipation in a cellular (or physical–chemical) process. A
future science classroom might look more like a dance
studio or gymnasium than a room filled with desks aimed
at a whiteboard, or teaching can even move out of the
classroom and into the outdoors.
Mutations and engineering
One early challenge in the collaboration was that the
dancers would sometimes propose to move in ways that
were incongruent with our current understanding of the
scientific concept we were trying to convey. The motivation
for these proposals was often to develop dance movements
with aesthetic value. Although it was tempting for the
scientist to dismiss these suggestions as scientifically in-
valid, we found that it was best not to be too strict in the
artistic interpretations. First, the BLM movers are artists,
and for a two-way collaboration to work effectively there
needs to be room for artistic freedom, so that the movers do
not simply feel like automatons executing an algorithm.
Second, the scientists should keep an open mind. Perhaps
the dancers have injected a new idea into the process.
These proposals could be regarded as ‘mutations’ and allow
the sort of ‘what if?’ experimentation that helps us to build
our physical intuition and reasoning. Can we successfully
predict how this mutated dance will unfold, even though
we highly doubt that a cell could ever work that way? The
ability to understand how a mutant dance will unfold
connects very directly to engineering, where we attempt
to steer the system in a direction that has a health or
technological benefit. This could help us to address the
issue of which molecules to target (e.g., via drugs) to
achieve a desired therapeutic outcome. Making up new
and bizarre rules should not be discouraged.
Feeding back into dance
Another advantage of collaborating with a professional
dance company is that the movements developed through
the collaboration may inspire new works of performance
art. In our case, the basic moves described above (e.g.,
diffusion, reaction), led to a broader exploration of the use
of impact in creating dance works. In particular, the explo-
615
Forum: Science & Society
TRENDS in Cell Biology
Figure 3. Patrick Jeffrey and Eddie Oroyan create a curling microtubule
protofilament in the new Black Label Movement dance ‘HIT’.
ration of the technique of impact required to manifest the
violent interior of the cell suggested to us that there was a
choreographic application of this vocabulary useful for
examining human relationships. In a quartet entitled
‘HIT’, Carl removed the prohibition of violent physicality
in healthy human relationships to examine a dysfunctional
community of people. When microtubules disassemble,
their protofilaments curl outward releasing mechanical
616
Trends in Cell Biology December 2012, Vol. 22, No. 12
strain energy, a process that is now integrated into ‘HIT’
(Figure 3) to capture a moment of violent control between
two men. ‘HIT’ is an example of how this collaboration is
leading to serious and recognized art-making (see links to
critical reviews of ‘HIT’ in Box 1).
Future outlook
The Moving Cell Project has led us in many surprising
directions. We recently provided advice to the Citizens
League of Minnesota, a public policy group that is inter-
ested in exploring the application of bodystorming to com-
plex policy issues. In our own work, we are now planning
joint group meetings in a dance studio, which will allow
even greater integration of science and dance into our
work. Last summer we visited the Marine Biological Lab-
oratory (MBL) in Woods Hole, MA and provided a body-
storming capability to the community there through the
physiology course. Students and faculty participated in
bodystorming and rapidly prototyped hypotheses driving
their research projects during the course. In general, the
close integration of dance and science offers new opportu-
nities to learn, teach, and drive new discoveries across
disciplinary boundaries.
Acknowledgments
The Moving Cell Project is funded by the Institute for Advanced Study at
the University of Minnesota, directed by Ann Waltner. The authors thank
BLM, the Odde laboratory group, John Bohannon, Enrique De La Cruz,
and Dyche Mullins for their participation and valuable input into the
collaborative. We are grateful to have worked with the late Robert
Hammel and to Perimeter Productions for their video documentation of
our work. We also thank the MBL, especially Dyche Mullins, Diana
Kenney, Bill Reznikoff, and Gary Borisy, for supporting a BLM residency
in Woods Hole. Finally, we thank Ann Waltner for her commitment to
interdisciplinary research and for catalyzing our collaboration.
References
1 Mitchison, T.J. and Kirschner, M.W. (1984) Dynamic instability of
microtubule growth. Nature 312, 237–242
2 Mogilner, A. and Odde, D. (2011) Modeling cellular processes in 3D.
Trends Cell Biol. 21, 692–700
3 Zhou, H.X. et al. (2008) Macromolecular crowding and confinement:
biochemical, biophysical, and potential physiological consequences.
Annu. Rev. Biophys. 37, 375–397
0962-8924/$ – see front matter ß 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.tcb.2012.10.005 Trends in Cell Biology, December 2012,
Vol. 22, No. 12
 Opinion
Special Issue – Synthetic Cell Biology
Synthetic multicellularity
Michel M. Maharbiz
Department of Electrical Engineering and Computer Science, University of California, Berkeley, CA 94720, USA
The ability to synthesize biological constructs on the
scale of the organisms we observe unaided is probably
one of the more outlandish, yet recurring, dreams
humans have had since they began to modify genes.
This review brings together recent developments in
synthetic biology, cell and developmental biology, com-
putation, and technological development to provide
context and direction for the engineering of rudimenta-
ry, autonomous multicellular ensembles.
What is multicellularity?
Macroscopic organisms generally comprise numerous cells,
usually from a common genetic parent, differentiated
through environmentally sensitive genetic programs. That
is, they are multicellular. But what is multicellularity?
How could we approach the multifaceted challenge of
engineering autonomous multicellular ensembles? What
is missing today from the synthetic biological repertoire?
Are there approaches that veer away from the paths taken
by nature? Lastly, is there a good reason to do this at all?
A discussion of the definition of multicellularity – par-
ticularly when appended to the word organism or when
discussed in the context of complex bacterial communities
– is beyond the scope of this short paper and excellent
reviews already exist [1–4]. Needless to say, any student
contemplating these issues should take a developmental
biology course to lose themselves in the myriad routes to
multicellularity in the natural record; classic textbooks
include Wolpert [5] and Gilbert [6]. The seminal synthesis
by D.W. Thomson, On Growth and Form [7], and the many
highly readable monographs by J.T. Bonner [8–10], al-
though in many ways dated, issue forth inspiration from
almost any page. Lastly, the elegant little book Vehicles:
Experiments in Synthetic Psychology [11] provides a con-
ceptual roadmap toward engineered complexity that is
well worth the read.
I will avoid in this paper the notion of coopting existing
mammalian cell lines to form synthetic multicellular
ensembles; that is primarily the province of tissue engi-
neering. This is not to say that is not a useful endeavor, of
course; but these lines already contain within them com-
plex – and poorly understood – systems for producing
specific metazoan forms. They harbor millions of years of
evolutionary baggage, so to speak. To form a working
definition of multicellularity (to which we can later add
the label synthetic), we would like to start much closer to
the beginning.
Corresponding author: Maharbiz, M.M. (maharbiz@eecs.berkeley.edu).
Keywords: multicellular; abiotic; interfaces; development; emergent;
cell–cell adhesion.
0962-8924/$ – see front matter ß 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tcb.2012.09.002 Trends in Cell Biology, December 2012, Vol. 22, No. 12
Multicellularity and emergent behavior
At its most fundamental level, multicellularity arises when
cells come together and find means to couple their internal
states in such a way that the connections result in emer-
gent behavior –generally with improved fitness for a set of
problems – that arises from the collective of cells. This is a
somewhat weak definition of multicellularity, because it
allows for numerous collective ensembles (e.g., bacterial
biofilms or intestinal microflora) that are themselves not
considered organisms, but for the engineer it allows a
certain flexibility when approaching the problem. There
is now a well-developed literature in applied mathematics,
computer science, and the biology of social insects that
deals with the analysis and design of similar multi-agent
systems. It has long been known, for instance, that the
complex functions carried out by social insect communities
(foraging and detection of environmental signals, construc-
tion of complex structures [12], and even specification of
insect phenotype [13]) emerge from a beautiful interplay
between environmental factors, ‘simple’ behaviors encoded
in the individual insects, and the exchange of information
between individuals. This cooperation results in a type of
‘swarm intelligence’ that is robust to insult and environ-
mental change and can carry out optimization tasks
[14–16]. Numerous engineering efforts, notably in computer
science and applied mathematics, have taken inspiration
from these phenomena to develop software packages for
routing and optimization of many real-world problems
[17], to explore the theoretical aspects of emergent systems
as computational engines [18,19], and to build computation
into physical structures [20,21] (an active endeavor that is
fundamentally similar to the topic that concerns us here).
That many of these ‘social’ behaviors bear striking resem-
blance to functions and properties of multicellular organ-
isms motivates a closer inspection. The point here is that
before we delve into how multicellular ensembles might be
constructed, we should spend some time learning the rich
computational background that now exists for designing
and analyzing systems with emergent properties. In this
regard, the Dictyostelium discoideum community has per-
haps some of the best demonstrations of the remarkably
complex, yet computationally tractable, emergent behaviors
arising from cells executing relatively simple processes.
From Bonner [10,22] through Nakagaki’s maze-finding
slime molds [23] to the so-called Tokyo subway experiment
[24], Dictyostelium has become one of the models for study-
ing emergence and multicellularity [25]. The broader point
is that, as with almost all engineering efforts initially
inspired by nature, once we computationally understand
emergence, we are likely to find solutions not available from
617
Opinion
the observable natural catalog (e.g., consider the robot Rhex
[26] or synthetic nacre [27], both inspired by but very
different from natural systems). Armed with this conviction,
we can now ask what ‘parts’ are needed to synthesize
collective behavior in cells.
The transition to multicellularity
Where might we turn for a first look at what genetic and
epigenetic modules are needed to engineer multicellulari-
ty? The biomolecular complexity of the classic model
systems in developmental biology (e.g., Drosophila mela-
nogaster, Caenorhabditis elegans) can be daunting [5]; even
Hydra [28] and the simple flatworm [29] are already far
down the path into multicellularity. Luckily, the past
decade has seen an explosion of work on species bordering
the transition from the unicellular to multicellular lifestyle
[30–34]. It is here that, stripped of all but the barest
essentials, we may begin to identify the specific compo-
nents we must engineer. Sometime around 600 million
years ago, colonial and truly multicellular ensembles
began to emerge from unicellular flagellates [34]. Today,
some of the closest relatives to metazoans are the choano-
flagellates (Figure 1) – a collection of small (3–10 mm),
flagellated eukaryotes, some of which are unicellular, some
multicellular, and some that can adopt either phenotype
[35]. What is so intriguing about these creatures is not only
that certain species exhibit both unicellular and multicel-
lular phenotypes – which makes it possible to observe in
detail the genetic and structural changes that allow multi-
cellularity [36,37] – but that spectra of closely related
species exist that together capture the heritable transition
to multicellular behavior. Moreover, genomic and proteo-
mic analysis has shown that, remarkably, components of
many of the genetic systems once thought specific to
metazoans and bilaterians (cadherins, receptor tyrosine
kinases [RTKs], bilaterian miRNAs, Piwi-interacting
RNAs [piRNAs]) and thought to be crucial in the develop-
ment and maintenance of complex forms are present in
choanoflagellates [30,32,33,38]. Some of these pathways
10 um By DHZanee
N~230
TRENDS in Cell Biology
Figure 1. A spherical colony of Sphaeroeca choanoflagellates. Reproduced, with
permission, from http://commons.wikimedia.org/wiki/File:Sphaeroeca-colony.jpg.
618
Trends in Cell Biology December 2012, Vol. 22, No. 12
(e.g., cadherins) appear to have arisen before multicellu-
larity, were involved in environmental and prey–predator
detection, and were coopted during the transition[39].
Numerous studies [1,31,36,40–44] point to the handful
of basic ‘enabling technologies’ we must consider engineer-
ing: the excretion of an extracellular matrix (ECM), the
presence of cytoplasmic bridges, cell–cell adhesion, and
molecular differentiation (or inherited functional speciali-
zation). A fifth observation, likely to be emergent, is that
choanoflagellate colonies appear to form not due to aggre-
gation, but due to non-separation after division (with the
concomitant production of a matrix and cell junctions). Let
us consider these enabling technologies one at a time.
A recent review [45] provides an overview of ECM
domains as well as the known phylogeny in the catalog;
Özbek et al. provides an evolutionary viewpoint [46]. Choa-
noflagellates can secrete multiple types of ECM [36], al-
though apparently none with the complexity of true
metazoan ECM [45]. The type prevalent in colonial phe-
notypes of Salpingoeca rosetta appears ‘amorphous, loose
and space-fitting’, whereas unicellular phenotypes of the
same species are capable of producing denser, more pre-
cisely shaped ECM as well (the so-called theca ‘goblet’) [36].
Cells of the multicellular alga Volvox embed themselves in
a complex ECM and convert part of their cell walls to ECM;
the massive amount of ECM secreted by these cells allows
for very large colonies (in proportion to their cell number)
[30]. In more complex eukaryotes, the ECM constitutes not
only a mechanical support layer, but is a fundamental
communication channel between cells, allowing chemical
and mechanical signals to be exchanged by cells attached
to it [46,47].The fact that ECM is usually laminar with
respect to some axis of the ensemble provides a ready
reference to establish polarity [6,47]. Bacteria are known
to also secrete some sort of matrix, particularly when
forming biofilms, but they seem to lack the same richness
of communication modes as eukaryotes capable of more
complex multicellularity despite the fact that they can
form differentiated spatial structures with different struc-
tural ECM for different functional components [45].
This, of course, brings us to cell–cell adhesion mole-
cules, principally the cadherins. Choanoflagellates pos-
sess cadherins [39] and integrin domains (but no true
integrins) [45], putatively allowing for cell–cell and cell–
ECM adhesion. This enables the coupling of external
mechanical events and information to internal changes
in cell state. This is one of the functions most sorely lacking
in the bacterial synthetic biology repertoire. Although
ongoing work has demonstrated that the MscL family of
channels in bacteria are sensitive to membrane strain and
affect transmembrane potential [48,49], there are, at
present, no well-characterized modules for coupling strain
or adhesion to genetic expression. Cadherins are crucial
enablers not only of cell polarity, but also of emergent and
coordinated relative motion and reorganization within a
cell ensemble, a topic of recent interest [50–53]. Cyto-
plasmic bridges, which in choanoflagellates seem to arise
from incomplete cytokinesis [36], also seem to play a role,
both in maintaining physical connections among colonial
members and in the constrained exchange of signaling
molecules.
Opinion
Lastly, we come to the topic of differentiation (or heri-
table cell state). In eukaryotic systems, this is a vast field of
inquiry, well outside the scope of this review. Within
microbial synthetic biology, switches and feedback loops
that allow an inducible, switchable state and that can pass
this state to daughter cells have been demonstrated; I will
return to this below in the context of circuits and systems
already developed by synthetic biologists. It is worth not-
ing that, in flagellates, there is an inherent trade-off
between motility and division, because both processes
make use of microtubule-organizing centers (MTOCs).
King [34], Buss [54], and Michod [55] have pointed out
that differentiation between flagellated cells on the surface
of a multicell colony and non-flagellated cells in its center
arose as a solution to the MTOC trade-off, because colonies
containing flagellated and non-flagellated cells would be
under strong environmental pressure to expose as many
motile cells on the surface of the colony as possible. This, of
course, is what one sees in many species of colonial fla-
gellates. More importantly, it points to the idea that di-
rected evolution [56] in the presence of some of the basic
components described above might allow us to recapitulate
something as powerful as gastrulation [5,6] and spatial
organization [55]. In fact, almost 30 years ago, Edelman
suggested in his ‘regulator hypothesis’ that selection acting
on just cell–cell adhesion genes and differentiation genes
could account for stable and varied body plans ‘within
relatively short periods of evolutionary time’ [57].
Moving forward – synthetic engineering beyond the
genetic circuit
Efforts over the past decade have yielded a catalog of basic
cell state circuit motifs [58,59] that include bistable
switches and memory elements [60–63], oscillators [64–
66], simple logic gates [67,68], and amplifiers [69,70]. The
complexity of genetic circuitry will no doubt increase in the
coming years, enabling more sophisticated programming of
cell state and, thus, at least some degree of synthetic
differentiation in microbial models. What else then should
we focus on in the quest for multicellularity?
Local relationships lead to global organization
Numerous synthetic intercellular signaling systems have
been demonstrated in the past few years [59,71], with
quorum sensing-derived circuits gaining widespread use
[72]. Quorum sensing in bacteria relies on the well-known
acyl homoserine lactone (AHL) family of freely diffusible
and membrane-permeable compounds. The ability to both
induce the production of AHLs via the LuxI gene product
homologs and detect AHL concentration via AHL–LuxR
complexes acting on inducible promoters allows for the
introduction of diffusible intercellular signaling modules;
an early seminal result was the coupling of an AHL sig-
naling circuit to a cell ‘killer’ gene to synthetically regulate
bacterial culture growth curves [73]. The number of truly
orthogonal AHL communication channels is somewhat
limited, but that catalog may grow. One of the outstanding
challenges in the field is the demonstration of a robust
framework for coupling intercellular signaling and circuits
encoding cell state to generate programmable patterns in
cell state across an initially homogeneous multicellular
Trends in Cell Biology December 2012, Vol. 22, No. 12
population. Early seminal efforts demonstrated that
groups of ‘receiver’ cells sensitive to the presence of AHL
could be made to respond in a tunable fashion to the
presence of ‘sender’ cells that produced diffusible AHL
[74]. This work was subsequently extended to demonstrate
consensus cooperation in microbial communities [75], syn-
thetic predator–prey systems [76], and intercellular com-
munication enabling edge detection across light/dark areas
[77], among others. A series of papers from the Hasty
laboratory have since demonstrated the use of AHL-cou-
pled oscillators for entraining and synchronizing popula-
tions [66,78,79]. More recently, AHL-based signaling was
used to provide chemical ‘wires’ between cell colonies, each
carrying out distinct computations to enable more complex
functions than those possible within single cells [80]. Much
more remains to be done in this area. True symmetry-
breaking systems such as the classic Turing [81] and
Meinhardt–Gierer [82] diffusible signal pattern generators
have not yet been robustly demonstrated, despite recent
experimental [Ting, L. et al. (2008) Pattern formation in a
synthetic multicellular system. 2008 APS March Meeting.
53 (http://meetings.aps.org/link/BAPS.2008.MAR.J17.7)]
and theoretical results [83]. Despite the early work on
the subject, little has been done to create systems in
bacteria with multiple, simple, orthogonal cell–cell com-
munication motifs operating to produce complex emergent
behavior in communities of cells. As cited above, there is a
wide literature in computer science and applied mathe-
matics concerning cellular automata and emergent sys-
tems, many of which have been shown to be capable of
propulsion, reproduction and size control and to be Turing
complete [19,84–86]. As additional orthogonal communi-
cation channels become available (see below), this type of
spatial programming should become accessible. Moreover,
emergent behavior need not arise exclusively via chemical
dynamics. A recent example elegantly demonstrates how
motility and density can be coupled via a synthetic AHL-
mediated circuit to produce stripe formation in Escherichia
coli [87]. Decades of theoretical treatment of the mechani-
cal underpinnings of metazoan morphogenesis have pro-
vided several testable models of how mechanical
communication via cell–cell and cell–ECM coupling would
result in complex shape changes and modulation of gene
expression [51,88–91]. Emergent pattern formation from
simple ECM-mediated processes has been documented. An
old example [92] demonstrates how fibroblast traction on
collagen matrix itself aligns the underlying collagen fibers
into tracks toward the cells, driving further aggregation.
This simple mechanical instability was capable of driving
cell density-dependent Turing-like pattern formation. This
result reflects a still rather unexplored facet of cellular
pattern formation: the interplay between cell processes
and physicochemical forces in the environment, which
together drive complex emergent behavior [51].
Missing links
It is believed that several of the genetic and epigenetic
processes that regulate multicellularity in eukaryotes are
missing in bacteria; these include cadherins, tyrosine
kinases, and complex ECM production (see above). Given
that bacteria as a whole do exhibit some matrix production,
619
 Opinion
Matrix
secreon
Bacterium
Cell-cell adhesion
+ aggregaon
or division
Figure 2. What motifs are necessary for rudimentary synthetic multicellularity? Shown here are distinct ‘enabling technologies’ strung into a putative roadmap to colonial
ensembles. The various ‘missing links’ are discussed in the text.
cell–cell contact-mediated signal transduction [93–95],
strain sensing [49], rapid membrane potential depolariza-
tion events [96], and even direct electron transfer channels
[97–100], it seems likely that analogs of eukaryotic pro-
cesses can be synthetically engineered into or mined from
bacteria. Three specific missing components would be in-
valuable to the pursuit of multicellularity and are likely to
be obtainable in the near term (Figure 2): a stress/strain
sensor coupling either membrane stress or stress on an
extracellular process (i.e., pili, flagella, or cilia) to gene
regulation; a cell–cell contact-mediated channel for either
cytoplasmic contact or signal exchange; and a system for
inducing and maintaining cell polarity in bacteria relative
to a matrix plane. Coupled with synthetic cell–cell adhe-
sion, a stress/strain sensor might allow bacteria to ap-
proach the level of cytomechanical control multicellular
eukaryotes employ for cell sorting and multicell motion. On
the issue of cytoplasmic bridging, some reports appear to
indicate the presence of cytoplasmic exchange – including
plasmid transfer – across ‘nanotubes’ seen between Bacil-
lus subtilis, Staphylococcus aureus, and E. coli [101].
Again, choanoflagellates may be invaluable, because it is
thought that metazoan adhesion proteins may be derived
from proteins used by heterotrophic flagellates to recognize
and bind bacterial prey [39]. A fourth, and far more diffi-
cult, need is for modular RNA- or protein-based systems
capable of switching states much more rapidly than
allowed via transcriptional regulation (i.e., for fast
responses to the environment) [102,103]. Although compo-
nents do exist, they are not yet sufficiently well understood,
modular, or orthogonal to enable the flexible engineering
of, for example, sensor-to-flagellar control modules in bac-
teria (i.e., fast synthetic navigation systems).
Putting it all together
Paralleling the increasing complexity of the constructs in
Valentino Braitenberg’s seminal book Vehicles: Experi-
ments in Synthetic Psychology, it is tempting to entertain
the notion of a series of grand challenges. Who can be
the first to demonstrate a synthetic, motile colony? With
programmable sensor-to-motility response? With pro-
grammable form? With cell lineage differentiation? With
reproduction? With memory?
620
Trends in Cell Biology December 2012, Vol. 22, No. 12
Strain
Cell-cell contact sensor
communicaon Cell polarity
Le-right symmetry
In-out symmetry ‘Fast’ non-transcripon
networks drive molity
TRENDS in Cell Biology
Here arises the issue of assembly. The classical position
of the genetic engineer is to look for the gene or network of
genes, however mythical, that will generate a desired
phenotype when introduced into a cell. This is, in essence,
a self-assembly approach: we genetically engineer desired
behaviors in cells and then allow them to grow and form
multicellular systems. A crucial area of work in the context
of engineering multicellularity via self-assembly is the so-
called ‘decomposition problem’ [104–106]. That is, given a
final form and a palette of possible operations or gene
circuits, can I compute the sequence of such operations
that form the final form? Can I determine the minimal set
of ‘functions’ (that is, cellular behaviors such as adhesion,
force production, migration, and signal production) re-
quired to arrive at a certain multicellular form?
Alternatively, we can impose external constraints on
the cells that direct the formation of multicellularity. This
is an example of constrained assembly, common to classic,
industrial age manufacturing wherein form or function is
engineered via a sequence of top-down steps (as in the
assembly of a mechanical watch). Biology, in fact, makes
extensive use of constrained assembly. The Drosophila egg
chamber, for example, is in part an apparatus for symme-
try breaking and driving the constrained assembly of the
early fly embryo. Developmental biology courses usually
focus on the beautiful sequence of events that unfold in the
fertilized embryo starting with the Bicoid gradient (speci-
fying, for example, segments and assembling organs) and
ending with an adult fly. What is relevant to this discussion
is that the initial anterior–posterior and dorsal–ventral
axes of polarity are determined by sequestered mRNA laid
down in unfertilized eggs by maternal nurse cells in the egg
chamber; thus, symmetry breaking is accomplished in the
embryo by the mother [5,107]. Many examples of these
maternal effects can been documented phenomenological-
ly, but detailed mechanistic understanding appears to be
emerging [108].
When considering synthetic multicellular engineering,
it is likely that abiotic ‘maternal effect’ technologies – that
is, machines that break symmetry for the developing sys-
tem or otherwise impose boundary conditions on the mul-
ticellular ensemble to guide development – could play a
role. A vast repertoire of non-biological devices can be
Opinion
brought to bear. Whether or not natural colonial and
multicellular systems arise through self-assembly, human
engineers need not build them that way. Modern advances
have resulted in a wide variety of devices for specifying the
chemical, mechanical, optical, and electrical milieu both
intra- and extracellularly. An exhaustive review is outside
the scope of this piece, but some examples are provided
below. Given the medical utility of many of these methods
(including drug delivery, tissue engineering, and clinically
relevant sensing [109]), they continue to advance rapidly.
Simply put, having established an engineering goal (e.g., a
1-mm diameter, 1024-cell motile, flagellated colony that
swims up a specified concentration gradient and lyses at a
programmed threshold), we should explore technologies
beyond those available in natural systems to engineer it.
Hybrid abiotic/biotic systems are likely to be useful even
beyond assembly tasks. Consider again our flagellar colo-
ny. Why not assemble such a colony on abiotic parts
properly functionalized with adhesion proteins? A simple
computational engine fabricated in a modern semiconduc-
tor process can be as small as 100 mm on a side. Chemical
input/output [110–114], electrochemical sensors for DNA
hybridization [115], nanoscale and integrated light detec-
tors and LEDs [116], polymer strain gauges [117,118],
surface energy switching [119], and even nanowire trans-
membrane recordings [120] can be integrated into very
small systems. From my work in integrating abiotic com-
ponents with live insects for flight control [121], it has
become clear that hybrid systems can use the best of both
worlds. Biological systems did not evolve radios or 22-nm
transistors; man-made communication and computational
systems have reached a staggering degree of sophistication
in incredibly small, low-power packages (e.g., the off-the-
shelf ATtiny10 from Atmel is a 12-MHz, 8-bit microcon-
troller with a 1024-byte flash memory in a package 2 mm
square; unpackaged, it is about half this size). The power
requirements for the 100-mm computational engine could
be <10 mW on average and powered entirely from a tiny
solar cell on-chip. Conversely, man-made energy sources,
motility systems, and material properties still pale in
comparison with the natural arsenal. It is not feasible with
modern abiotic technology to make an autonomous 1-mm
swimming robot capable of locomoting indefinitely; we
cannot build good synthetic chemotrophs, our actuator
technologies are poorly suited to that scale, and communi-
cation via radio or optics is not easy in real underwater
environments. There are other, ancillary advantages of
hybrid approaches. Abiotic parts would not be subject to
genetic changes or selection pressure; hybrid systems
would not be able to reproduce, alleviating some safety
concerns, and have finite lifetimes.
Why do this?
Lastly is the issue of motivation: is this something worth
doing? Much is still to be worked out to achieve scalable
synthetic control circuits [71], not to mention fundamental
technology development [122,123]. It is not clear what we
do with the simple ‘colonial’ constructs I have sketched
above and it is a long way to anything resembling a
metazoan. The attraction of the problem, however, is
threefold.
Trends in Cell Biology December 2012, Vol. 22, No. 12
First, a grand challenge around true, synthetic multi-
cellularity would provide an immediate test bed for the
development of many other platform technologies crucial
to synthetic biology. With properly formulated systems
goals, we could outline specifications for the various mod-
ules needed and present multicellularity as a competitive
challenge wherein groups arrive at unique solutions. Con-
sider some medium-term applications. Multicellular, mo-
tile colonies with silicon payloads could be deployed in
rivers and lakes as environmental health monitoring
devices, capable of interacting with the aquatic biology
at various depths and communicating with man-made
devices. Organized films of swarmer cells could clean
surfaces: ingesting small particles, digesting larger ones
and carrying indigestible particulates (including excess
cell mass) to set locations. Cell collectives capable of taxiing
into and around each other in weaves (and then depositing
matrix) could assemble and repair simple textiles on de-
mand in remote places. Sophisticated biofilms could detect
strain non-linearities indicative of the cracking of an un-
derlying stratum and produce binding matrix.
Second, this endeavor may drive the development of
toolsets for use in non-canonical organisms like the choa-
noflagellates (among others), something of great value to
fundamental molecular and comparative biology.
Lastly, and most importantly, the demonstration of
synthetic multicellularity with a complexity eventually
rivaling that of metazoans or terrestrial plants must surely
be one of the long-term goals of synthetic biology. This
arises not from an argument of biomimicry – which, fol-
lowed blindly, leads us down fruitless technological paths –
but because we will never truly understand the limits of
developmental regulation, stability, and plasticity until we
have recapitulated developmental processes on our syn-
thetic platforms. There are likely to be stable areas of the
metazoan parameter space that evolution has either se-
lected out or not explored and that possess technological
value for our society. Just as the 20th century made it
obvious that humans could arrive at molecules and chem-
istries not found in nature starting from the same constit-
uent atoms, it is likely we will find multicellular solutions
with technological value not arrived at via natural selec-
tion. That this has profound ethical and societal conse-
quences cannot be overstated.
Concluding remarks
There is additionally – and perhaps controversially – an
ultimately ecological rationale for this vision. For the most
part, the technological base created by the industrial revo-
lution communicates poorly with the underlying organic
technology of the planet. The rapid explosion of man-made,
acellular, resource-consuming, and waste-producing con-
structs is in large part responsible for the ecological and
climactic mess we are in. Mindful of the vast ethical and
societal questions raised, it is worth considering a future
wherein our homes, our factories, and our consumer gad-
gets can ‘understand’ the language of the organic systems
around them and form part of a related or hybrid frame-
work of information and material exchange. This notion –
that our societal artifacts should be mindful of their natu-
ral surroundings – has long and deep roots in many
621
Opinion
cultures and has modern reflections in, for example, the
natural building movements. It is my contention that the
development of synthetic multicellular – and likely hybrid
– systems is a step down this transformative road.
References
1 Nedelcu, A.M. (2001) Evolution of Multicellularity, John Wiley & Sons
2 Webb, J.S. et al. (2003) Bacterial biofilms: prokaryotic adventures in
multicellularity. Curr. Opin. Microbiol. 6, 578–585
3 Monds, R.D. and O’Toole, G.A. (2009) The developmental model of
microbial biofilms: ten years of a paradigm up for review. Trends
Microbiol. 17, 73–87
4 McShea, D.W. (2002) A complexity drain on cells in the evolution of
multicellularity. Evolution 56, 441–452
5 Wolpert, L. and Tickle, C. (2011) Principles of Development, (4th edn),
Oxford University Press
6 Gilbert, S.F. (2010) Developmental Biology, (9th edn), Sinauer
Associates
7 Thompson, D.A.W. (1942) On Growth and Form, The University Press
8 Bonner, J.T. (2001) First Signals: The Evolution of Multicellular
Development, Princeton University Press
9 Bonner, J.T. (1955) Cells and Societies, Princeton University Press
10 Bonner, J.T. (1998) The origins of multicellularity. Integr. Biol. 1, 27–
36
11 Braitenberg, V. (1986) Vehicles: Experiments in Synthetic Psychology,
MIT Press
12 Hölldobler, B. and Wilson, E.O. (1994) Journey to the Ants: A Story of
Scientific Exploration, Belknap Press
13 Patalano, S. et al. (2012) Shifting behaviour: epigenetic
reprogramming in eusocial insects. Curr. Opin. Cell Biol. 24, 367–
373
14 Dussutour, A. et al. (2005) Amplification of context: the ants. Proc. R.
Soc. B: Biol. Sci. 272, 705–714
15 Reznikova, Z. (2008) Experimental paradigms for studying cognition
and communication in ants (Hymenoptera: Formicidae). Myrmecol.
News 11, 201–214
16 Panteleeva, S. et al. (2010) Using ideas of Kolmogorov complexity for
studying animal behavioural patterns. In Proceedings of Measuring
Behavior 2010 (Eindhoven, The Netherlands, August 24-27, 2010)
(Spink, A.J. and et, al., eds), pp. 174–176, Noldus Information
Technology
17 Michael, L. (2005) Ant-based computing. Artif. Life 15, 337–349
18 Schiff, J.L. (2011) Cellular Automata: A Discrete View of the World,
John Wiley & Sons
19 Wolfram, S. (2002) A New Kind of Science, Wolfram Media
20 Griffith, S. et al. (2005) Robotics: self-replication from random parts.
Nature 437, 636
21 Pfeifer, R. et al. (2007) Self-organization, embodiment, and
biologically inspired robotics. Science 318, 1088–1093
22 Bonner, J.T. (2010) Brainless behavior: a myxomycete chooses a
balanced diet. Proc. Natl. Acad. Sci. U.S.A. 107, 5267–5268
23 Nakagaki, T. et al. (2000) Intelligence: maze-solving by an amoeboid
organism. Nature 407, 470
24 Tero, A. et al. (2010) Rules for biologically inspired adaptive network
design. Science 327, 439–442
25 Devreotes, P. (1989) Dictyostelium discoideum: a model system for
cell-cell interactions in development. Science 245, 1054–1058
26 Saranli, U. et al. (2001) RHex: a simple and highly mobile hexapod
robot. Int. J. Robot Res. 20, 616–631
27 Tang, Z. et al. (2003) Nanostructured artificial nacre. Nat. Mater. 2,
413–418
28 Steele, R.E. (2002) Developmental signaling in Hydra: what does it
take to build a ‘‘simple’’ animal? Dev. Biol. 248, 199–219
29 Gentile, L. et al. (2010) The planarian flatworm: an in vivo model for
stem cell biology and nervous system regeneration. Dis. Model. Mech.
4, 12–19
30 Abedin, M. and King, N. (2010) Diverse evolutionary paths to cell
adhesion. Trends Cell Biol. 20, 734–742
31 Ruiz-Trillo, I. et al. (2007) The origins of multicellularity: a multi-
taxon genome initiative. Trends Genet. 23, 113–118
32 Young, S.L. et al. (2011) Premetazoan ancestry of the Myc – Max
network. Mol. Biol. Evol. 28, 2961–2971
622
Trends in Cell Biology December 2012, Vol. 22, No. 12
33 Nichols, S.A. et al. (2006) Early evolution of animal cell signaling and
adhesion genes. Proc. Natl. Acad. Sci. U.S.A. 103, 12451–12456
34 King, N. (2004) The unicellular ancestry of animal development. Dev.
Cell 7, 313–325
35 King, N. (2005) Choanoflagellates. Curr. Biol. 15, 113–114
36 Dayel, M.J. et al. (2011) Cell differentiation and morphogenesis in the
colony-forming choanoflagellate Salpingoeca rosetta. Dev. Biol. 357,
73–82
37 Fairclough, S.R. et al. (2010) Multicellular development in a
choanoflagellate. Curr. Biol. 20, 875–876
38 Grimson, A. et al. (2008) Early origins and evolution of microRNAs
and Piwi-interacting RNAs in animals. Nature 455, 1193–1198
39 Abedin, M. and King, N. (2008) The premetazoan ancestry of
cadherins. Science 319, 946–948
40 Kirk, D.L. (2005) A twelve-step program for evolving multicellularity
and a division of labor. Bioessays 27, 299–310
41 Kirk, D.L. (1999) Evolution of multicellularity in the volvocine algae.
Curr. Opin. Plant Biol. 2, 496–501
42 Sachs, J.L. (2008) Resolving the first steps to multicellularity. Trends
Ecol. Evol. 23, 245–248
43 Grosberg, R.K. and Strathmann, R.R. (2007) The evolution of
multicellularity: a minor major transition? Annu. Rev. Ecol. Evol.
Syst. 38, 621–654
44 Michod, R.E. (2007) Evolution of individuality during the transition
from unicellular to multicellular life. Proc. Natl. Acad. Sci. U.S.A. 104,
8613–8614
45 Hynes, R.O. (2012) The evolution of metazoan extracellular matrix. J.
Cell Biol. 196, 671–679
46 Özbek, S. et al. (2010) The evolution of extracellular matrix. Mol. Biol.
Cell 21, 4300–4305
47 Hynes, R.O. (2009) The extracellular matrix: not just pretty fibrils.
Science 326, 1216–1219
48 Booth, I.R. and Blount, P. (2012) The MscS and MscL families of
mechanosensitive channels act as microbial emergency release
valves. J. Bacteriol. 194, 4802–4809
49 Booth, I.R. et al. (2011) Sensing bilayer tension: bacterial
mechanosensitive channels and their gating mechanisms. Biochem.
Soc. Trans. 39, 733–740
50 Borghi, N. and James Nelson, W. (2009) Intercellular adhesion in
morphogenesis: molecular and biophysical considerations. Curr. Top.
Dev. Biol. 89, 1–32
51 Newman, S. et al. (2006) Before programs: the physical origination of
multicellular forms. Int. J. Dev. Biol. 50, 289–299
52 Brien, L.E.O. et al. (2002) Building epithelial architecture: insights
from three-dimensional culture models. Nat. Rev. Mol. Cell Biol. 3, 1–
7
53 Borghi, N. et al. (2010) Regulation of cell motile behavior by crosstalk
between cadherin- and integrin-mediated adhesions. Proc. Natl.
Acad. Sci. U.S.A. 107, 13324–13329
54 Buss, L.W. (1987) The Evolution of Individuality, Princeton
University Press
55 Solari, C.A. et al. (2006) A hydrodynamics approach to the evolution of
multicellularity: flagellar motility and germ-soma differentiation in
volvocalean green algae. Am. Nat. 167, 537–554
56 Cobb, R.E. et al. (2012) Directed evolution: an evolving and enabling
synthetic biology tool. Curr. Opin. Chem. Biol. 16, 285–291
57 Edelman, G.M. (1984) Cell adhesion and morphogenesis: the
regulator hypothesis. Proc. Natl. Acad. Sci. U.S.A. 81, 1460–1464
58 Wolf, D.M. and Arkin, A.P. (2003) Motifs, modules and games in
bacteria. Curr. Opin. Microbiol. 6, 125–134
59 Boyle, P.M. and Silver, P.A. (2009) Harnessing nature’s toolbox:
regulatory elements for synthetic biology. J. R. Soc. Interface 6,
S535–S546
60 Palani, S. and Sarkar, C.A. (2011) Synthetic conversion of a graded
receptor signal into a tunable, reversible switch. Mol. Syst. Biol. 7,
480
61 Atkinson, M.R. et al. (2003) Development of genetic circuitry
exhibiting toggle switch or oscillatory behavior in Escherichia coli.
Cell 113, 597–607
62 Chang, D-E. et al. (2010) Building biological memory by linking
positive feedback loops. Proc. Natl. Acad. Sci. U.S.A. 107, 175–180
63 Sallee, N.A. et al. (2007) Engineering modular protein interaction
switches by sequence overlap. J. Am. Chem. Soc. 129, 4606–4611
Opinion
64 Fung, E. et al. (2005) A synthetic gene-metabolic oscillator. Nature
435, 118–122
65 Stricker, J. et al. (2008) A fast, robust and tunable synthetic gene
oscillator. Nature 456, 516–519
66 Mondragon-Palomino, O. et al. (2011) Entrainment of a population of
synthetic genetic oscillators. Science 333, 1315–1319
67 Anderson, J.C. et al. (2007) Environmental signal integration by a
modular AND gate. Mol. Syst. Biol. 3, 133
68 Wang, B. et al. (2011) Engineering modular and orthogonal genetic
logic gates for robust digital-like synthetic biology. Nat. Commun. 2,
508
69 Nistala, G. et al. (2010) A modular positive feedback-based gene
amplifier. J. Biol. Eng. 4, 1–8
70 Nandagopal, N. and Elowitz, M.B. (2011) Synthetic biology:
integrated gene circuits. Science 333, 1244–1248
71 Purnick, P.E.M. and Weiss, R. (2009) The second wave of synthetic
biology: from modules to systems. Nat. Rev. Mol. Cell Biol. 10, 410–422
72 Jayaraman, A. and Wood, T.K. (2008) Bacterial quorum sensing:
signals, circuits, and implications for biofilms and disease. Annu.
Rev. Biomed. Eng. 10, 145–167
73 You, L. et al. (2004) Programmed population control by cell–cell
communication and regulated killing. Nature 428, 868–871
74 Basu, S. et al. (2005) A synthetic multicellular system for programmed
pattern formation. Nature 434, 1130–1134
75 Brenner, K. et al. (2007) Engineering bidirectional communication
mediates a consensus in a microbial biofilm consortium. Proc. Natl.
Acad. Sci. U.S.A. 104, 17300–17304
76 Balagadde, F.K. et al. (2008) A synthetic Escherichia coli predator –
prey ecosystem. Mol. Syst. Biol. 4, 1–8
77 Tabor, J.J. et al. (2009) A synthetic genetic edge detection program.
Cell 137, 1272–1281
78 Danino, T. et al. (2010) A synchronized quorum of genetic clocks.
Nature 463, 326–330
79 Prindle, A. et al. (2012) A sensing array of radically coupled genetic
‘‘biopixels’’. Nature 481, 39–44
80 Tamsir, A. et al. (2011) Robust multicellular computing using
genetically encoded NOR gates and chemical ‘wires’. Nature 469,
212–215
81 Turing, A.M. (1952) The chemical basis of morphogenesis. Philos.
Trans. R. Soc. Lond. Ser. B: Biol. Sci. 237, 37–72
82 Gierer, A. and Meinhardt, H. (1972) A theory of biological pattern
formation. Biol. Cybern. 12, 30–39
83 Hsia, J. et al. (2012) A feedback quenched oscillator produces Turing
patterning with one diffuser. PLoS Comput. Biol. 8, e1002331
84 Wolfram, S. (1998) Cellular automata as models of complexity, In
Nonlinear Physics for Beginners: Fractals, Chaos, Solitons, Pattern
Formation, Cellular Automata, Complex Systems, World Scientific
85 Mitchell, M. et al. (2005) Computation in cellular automata: a selected
review, In In Non-Standard Computation, Wiley-VCH Verlag, pp. 95–
140
86 Gutowitz, H. (1991) Cellular Automata: Theory and Experiment, MIT
Press
87 Liu, C. et al. (2012) Sequential establishment of stripe patterns in an
expanding cell population. Science 334, 238–241
88 Beloussov, L.V. and Grabovsky, V.I. (2006) Morphomechanics: goals,
basic experiments and models. Int. J. Dev. Biol. 50, 81–92
89 Patwari, P. and Lee, R.T. (2008) Mechanical control of tissue
morphogenesis. Circ. Res. 103, 234–243
90 Gordon, R. (2006) Mechanics in embryogenesis and embryonics: prime
mover or epiphenomenon? Int. J. Dev. Biol. 50, 245–253
91 Ingber, D.E. (2006) Mechanical control of tissue morphogenesis
during embryological development. Int. J. Dev. Biol. 50, 255–266
92 Harris, A.K. et al. (1984) Generation of spatially periodic patterns by a
mechanical instability: a mechanical alternative to the Turing model.
J. Embryol. Exp. Morphol. 80, 1–20
93 Blango, M.G. and Mulvey, M. (2009) Bacterial landlines: contact-
dependent signaling in bacterial populations. Curr. Opin.
Microbiol. 12, 177–181
94 Aoki, S.K. et al. (2009) Contact-dependent growth inhibition causes
reversible metabolic downregulation in Escherichia coli. J. Bacteriol.
191, 1777–1786
95 Thanassi, D.G. (2011) The long and the short of bacterial adhesion
regulation. J. Bacteriol. 193, 327–328
Trends in Cell Biology December 2012, Vol. 22, No. 12
96 Kralj, J.M. et al. (2011) Electrical spiking in Escherichia coli probed
with a fluorescent voltage-indicating protein. Science 333, 345–348
97 Jensen, H.M. et al. (2010) Engineering of a synthetic electron conduit
in living cells. Proc. Natl. Acad. Sci. U.S.A. 107, 19213–19218
98 Ross, D.E. et al. (2011) Towards electrosynthesis in Shewanella:
energetics of reversing the Mtr pathway for reductive metabolism.
PLoS ONE 6, e16649
99 Shi, L. et al. (2007) Respiration of metal (hydr)oxides by Shewanella
and Geobacter: a key role for multihaem c-type cytochromes. Mol.
Microbiol. 65, 12–20
100 Baron, D. et al. (2009) Electrochemical measurement of electron
transfer kinetics by Shewanella oneidensis MR-1. J. Biol. Chem.
284, 28865–28873
101 Dubey, G.P. and Ben-yehuda, S. (2011) Intercellular nanotubes
mediate bacterial communication. Cell 144, 590–600
102 Lucks, J.B. et al. (2011) Versatile RNA-sensing transcriptional
regulators for engineering genetic networks. Proc. Natl. Acad. Sci.
U.S.A. 108, 8617–8622
103 Qi, L. et al. (2012) Engineering naturally occurring trans-acting non-
coding RNAs to sense molecular signals. Nucleic Acids Res. 40, 5775–
5786
104 Nagpal, R. (2001) Programmable Self-assembly: Constructing Global
Shape Using Biologically-Inspired Local Interactions and Origami
Mathematics. PhD Thesis, MIT Artificial Intelligence Laboratory
Technical Memo 2001-008, MIT
105 Galle, J. et al. (2009) From single cells to tissue architecture—a
bottom-up approach to modelling the spatio-temporal organisation
of complex multi-cellular systems. J. Math. Biol. 58, 261–283
106 Kondacs, A. (2003) Biologically-inspired self-assembly of two-
dimensional shapes using global-to-local compilation. In
Proceedings of the 18th International Joint Conference on Artificial
Intelligence (Gottlob, G. and Walsh, T., eds), pp. 633–638, Morgan
Kaufmann
107 St Johnston, D. and Nüsslein-Volhard, C. (1991) The origin of pattern
and polarity in the Drosophila embryo. Cell 68, 201–219
108 Wolf, J.B. and Wade, M.J. (2009) What are maternal effects (and what
are they not)? Philos. Trans. R. Soc. B: Biol. Sci. 364, 1107–1115
109 Khademhosseini, A. et al. (2006) Microscale technologies for tissue
engineering and biology. Proc. Natl. Acad. Sci. U.S.A. 103, 2480–
2487
110 Bansal, T. et al. (2009) Patterned delivery and expression of gene
constructs into zebrafish embryos using microfabricated interfaces.
Biomed. Microdevices 11, 633–641
111 Park, J. et al. (2006) A microsystem for sensing and patterning
oxidative microgradients during cell culture. Lab Chip 6, 611–622
112 Frevert, C.W. et al. (2006) Measurement of cell migration in response
to an evolving radial chemokine gradient triggered by a microvalve.
Lab Chip 6, 849–856
113 Ellis-Davies, G.C.R. (2007) Caged compounds: photorelease
technology for control of cellular chemistry and physiology. Nat.
Methods 4, 619–628
114 Huschka, R. et al. (2010) Visualizing light-triggered release of
molecules inside living cells. Nano Lett. 10, 4117–4122
115 Gooding, J.J. (2002) Electrochemical DNA hybridization biosensors.
Electroanalysis 14, 1149–1156
116 Jiang, X. et al. (2011) Rational growth of branched nanowire
heterostructures with synthetically encoded properties and
function. Proc. Natl. Acad. Sci. U.S.A. 108, 12212–12216
117 Cohen, D.J. et al. (2012) A highly elastic, capacitive strain gauge
based on percolating nanotube networks. Nano Lett. 12, 1821–1825
118 Laukhina, E. et al. (2010) Ultrasensitive piezoresistive all-organic
flexible thin films. Adv. Mater. 22, 977–981
119 Kuroki, H. et al. (2012) Responsive surfaces for life science
applications. Annu. Rev. Mater. Res. 42, 343–372
120 Duan, X. et al. (2012) Intracellular recordings of action potentials by
an extracellular nanoscale field-effect transistor. Nat. Nanotechnol. 7,
174–179
121 Sato, H. et al. (2009) Remote radio control of insect flight. Front.
Integr. Neurosci. 3, 24
122 Ellis, T. et al. (2011) DNA assembly for synthetic biology: from parts to
pathways and beyond. Integr. Biol. 3, 109–118
123 Ma, S. et al. (2012) DNA synthesis, assembly and applications in
synthetic biology. Curr. Opin. Chem. Biol. 16, 260–267
623
Review
Special Issue – Synthetic Cell Biology
Designer nucleic acids to probe and
program the cell
Yamuna Krishnan1 and Mark Bathe2
1
National Centre for Biological Sciences, TIFR, GKVK, Bellary Road, Bangalore 560 065, India
2
Department of Biological Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
Recent advances in nucleic acid sequencing, structural,
and computational technologies have resulted in dramat-
ic progress in our understanding of nucleic acid structure
and function in the cell. This knowledge, together with the
predictable base-pairing of nucleic acids and powerful
synthesis and expression capabilities now offers the
unique ability to program nucleic acids to form precise
3D architectures with diverse applications in synthetic
and cell biology. The unique modularity of structural
motifs that include aptamers, DNAzymes, and ribozymes,
together with their well-defined construction rules,
enables the synthesis of functional higher-order nucleic
acid complexes from these subcomponents. As we illus-
trate here, these highly programmable, smart complexes
are increasingly enabling researchers to probe and pro-
gram the cell in a sophisticated manner that moves well
beyond the use of nucleic acids for conventional genetic
manipulation alone.
Introduction to functional nucleic acid nanostructures
Over 50 years ago, it was discovered that RNA had cata-
lytic activity in addition to its role as an information-
carrying polymer, giving rise to the hypothesis that life
may have evolved from a precursor ‘RNA World’ without
proteins or DNA [1,2]. Today, the diverse structures and
activities adopted by RNAs are revealing a panoply of
cellular functions that were barely imaginable at the time
of this postulation by Alexander Rich and colleagues [1,2].
Unlike polypeptides, polynucleotides are structurally
highly programmable polymers due to the specific, pair-
wise Watson–Crick interactions of nucleic acid bases that
result in well-defined secondary structures of predictable
form. This, together with a simple set of construction rules
established for DNA in a landmark paper [3], the rapidly
decreasing cost of nucleic acid synthesis, the programma-
bility of RNA expression in living cells, and our increasing
knowledge of higher-order RNA structure and chemical
activity, is resulting in an explosion in the number and
diversity of nucleic acid structures that are being designed
for diverse applications in science and technology, with a
recently emerging focus on designer nucleic acids to probe
and program the cell.
Corresponding authors: Krishnan, Y. (yamuna@ncbs.res.in);
Bathe, M. (mark.bathe@mit.edu).
Keywords: synthetic biology; RNA nanotechnology; DNA nanotechnology.
624 0962-8924/$ – see front matter ß 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tcb.2012.10.001 Trends in Cell Biology, December 2012, Vol. 22, No. 12
As with most emerging technologies, with novelty comes
expectation and anticipation. Notwithstanding, a number
of recent works show real promise for transforming our
abilities to measure and manipulate cell function in highly
programmable ways that cannot be achieved with existing
technologies. We review a number of these works and
provide our perspective on what we expect to be promising
avenues of research to pursue in this rapidly evolving,
largely uncharted territory.
Programming nucleic acid nanostructure shape
via self-assembly
A hallmark of DNA nanotechnology is the ability to pro-
gram structurally well-defined, thermodynamically stable
3D polyhedra using Watson–Crick base-pairing and con-
struction rules [3,4], where the edges of the polyhedra
comprise individual DNA duplexes. Now, conventional
3D shapes that may be produced with high yield and
sample homogeneity include tetrahedra, icosahedra,
cubes, dodecahedra, and, more recently, larger-scale, con-
siderably more complex wireframe structures [5] using the
DNA origami approach [6], in which approximately 200
short strands are used to ‘staple’ together a longer single-
stranded DNA scaffold strand, reviewed in [7–9]
(Figure 1a–c). Advanced computational tools enable the
rational design of these structures accounting for thermo-
dynamic and mechanical stability key to function [9–12].
The principal advantages of using DNA to design
nucleic acid architectures include its long chemical half-
life and highly specific Watson–Crick base-pairing that
offers the ability to design stable nanostructures of precise
3D form [9,12]. By contrast, non-canonical base-pairing
acquired by RNA primary structures results in consider-
ably more diverse, dynamic secondary structures [13,14]
that are additionally subject to hydrolytic instability. This
Glossary
3WJ: a DNA or RNA motif comprising three duplexes radiating from a single
node; for example, the replication fork is a naturally occurring DNA 3WJ.
Fluorescence in situ hybridization (FISH): a technique used to visualize
endogenous DNA or RNA molecules in cells and tissues using fluorescently
labeled antisense oligonucleotides.
Hybridization chain reaction (HCR): a process by which metastable hairpins
assemble by a cascade of hybridization events to form a long nicked duplex.
Systematic evolution of ligands by exponential enrichment (SELEX): uses in
vitro evolution to confer specific chemical properties to DNA and RNA
molecules, typically high-affinity binding to a target small molecule.
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a) (b) (c)

2.2–3 nm

+ 21 bp
+ +

al
ne
An

Ligate
D

D
B

C
B

E E Ligate
F F
A A

(d) (e) (f)

8 nm
FP
siRNAs
scaffold
Funconalizaon

Dicing

siRNA NPs

A B C D

TRENDS in Cell Biology

Figure 1. (a) Left: schematic of the assembly of four DNA strands to form a DNA tetrahedron via annealing and ligation. Reproduced, with permission, from [18]. Right:
atomic model and cryo-electron microscopy reconstruction of the tetrahedron. Reproduced, with permission, from [102]. (b) Left: overview of the scaffolded DNA origami
concept, showing crossovers formed by staple strands every 21 base pairs to form a 2D rectangle [9]. Right: atomic model of a 52 nm diameter wireframe structure [5,11]
compared with 2D transmission electron microscopy. Reproduced, with permission, from [5]. (c) Top: atomic model of a tetrahedron functionalized with DNA oligomers at
its vertices and cryo-electron microscopy reconstruction. Reproduced, with permission, from [24]. Bottom: schematic of physical encapsulation of fluorescent cargo (yellow
oval) within a DNA icosahedron that binds anionic ligand-binding receptors (ALBRs) to map the pH of endosomes as they mature. Reproduced, with permission, from [22].
(d) Functionalizable RNA nanoring scaffolds (grey) assembled through kissing-loop interactions between RNA modules. Nanoring scaffolds are functionalized with RNAs
(red) that can penetrate the siRNA pathway. Reproduced, with permission, from [28]. (e) RNA cubes constructed from six prefolded RNA building blocks. Reproduced, with
permission, from [27]. (f) Atomic model and 3D cryo-electron microscopy reconstruction of a non-uniform square antiprism constructed from self associating tRNA-based
motifs. Reproduced, with permission, from [31].

relative lack of structural fidelity with respect to DNA,
however, is largely offset by its rich conformational and
catalytic space that may be used to program associations of
high affinity and specificity using aptamers [15], enzymatic
activity using ribozymes, and allostery using riboswitches
[16,17] (Figure 2). Further, the genetic encodability of RNA
offers the exciting opportunity to program cells to express
precise nanoarchitectures for sensing, scaffolding, and
signaling, among other functions, positioning RNA nano-
technology as a powerful medium to probe and program
cell function [18].
Programming DNA architectures for targeted delivery
The tetrahedron has arisen as a popular 3D template due
to its minimal DNA content and high synthetic yield [18]
(Figure 1a,c). An important early advance was the demon-
stration that a single covalently linked protein molecule
(cytochrome C) could be positioned within the tetrahedral
cavity, opening up the possibility to use the interior space
as a protected cargo hold for delivery [19]. To display such
protected cargo to external receptors, an innovative barrel-
shaped structure was designed [20] that sequesters T cell-
stimulating Fab fragments (the cargo) within its interior,
hidden from the external milieu when the barrel is closed
(Figure 3a). The hexagonal barrel remains closed with the
help of hasps comprising an aptamer and its complement
that dissociate when specific combinations of ligands bind
the aptameric strand. These act as AND gates for the detec-
tion of NKL cells from a mixture of NKL cells and leukocytes,
with the subsequently opened barrel exposing key Fab frag-
ments that trigger specific cellular processes such as T cell
activation or growth arrest. This work demonstrated that
allosteric DNA-based containers may be programmed to
expose cargo to cell-surface receptors based on selective,
specific, logic-based multivalent binding in cell culture.
In cases where the covalent linkage of molecular cargo
to the DNA scaffold is infeasible either due to lack of
reactive groups or because molecular activity would be
compromised, alternative approaches are still needed.
One means of eliminating chemical conjugation is to use
physical entrapment, whereby the cargo is physically con-
strained within DNA polyhedra. This strategy was
employed [21] to retain fluorescent dextrans or gold nano-
particles in DNA nanocages, and it was subsequently
shown that these loaded polyhedra could be targeted spe-
cifically to single Caenorhabditis elegans cell types that
expressed receptors for the DNA cage without compromis-
ing cargo functionality [22] (Figure 1c). Programmed
625
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

Aptamer

Aptazyme

Ribozyme/ Mn+
DNAzyme

Allosteric
+
aptamer

Riboswitches

Transcriponal Translaonal Metabolite-binding

TRENDS in Cell Biology

Figure 2. Schematic showing different functional nucleic acid motifs. Red stars and black circles represent reporter fluorophores. Aptamer: a DNA or RNA molecule that
specifically binds a small molecule or biomolecule (green oval). Aptazyme: a DNA or RNA molecule that is comprised of an aptamer domain fused to a catalytic domain,
where the latter becomes functional upon ligand-binding to the aptameric domain. Ribozyme/DNAzyme: an RNA/DNA molecule with catalytic properties, typically
consisting of a self-cleaving reaction in the presence of metal ions (denoted Mn+). Similar to aptazymes, the cleaving reaction is typically followed by strand dissociation.
Allosteric aptamer: aptameric modules sensitive to distinct ligands, typically an analyte and a reporter molecule, fused to one another so that the reporter molecule (red
star) may bind only when the analyte (green oval) is present. Riboswitches: regulatory RNA elements that act in cis by binding a small molecule, switching gene expression
on or off. Switches control either (left, Transcriptional) the formation of transcription terminator hairpins (red), (middle, Translational) accessibility to the Shine-Dalgarno
sequence (blue), or (right, Metabolite-binding) degradation of the transcript. Adapted from [17] and [79].

release of nanoparticles loaded within DNA scaffolds has
also been achieved by altering the pore size of the scaffold
using a complementary DNA strand [23]. This suggests
that it may be possible to both load and release encapsu-
lated molecular cargo from containers in a programmed
manner. Interfacing cell targeting with programmed re-
lease of non-tethered cargo, however, remains an open
challenge.
In addition to their ability to encapsulate or sequester
cargo for programmed release, DNA nanostructures have
the capacity to act as pegboards using small-molecule
conjugation to position functional biomolecules precisely
in space due to their well-defined structure and address-
able scaffold. One such demonstration was achieved using
DNA tetrahedra, in which short hairpins were grafted onto
the vertices of the scaffold without perturbing the tetrahe-
dral geometry [24]. Such morphology-rich nucleic acid
motifs displayed on DNA tetrahedra were shown to retain
functionality [25] when it was demonstrated that CpG
626
motifs displayed in a tetrahedral fashion could effect im-
munostimulation of macrophages by interacting with Toll-
like receptor 9.
Importantly, such multivalent CpG motifs showed en-
hanced immunostimulatory effects, presumably due to
increased ligand density. This increased ligand density
afforded by the immobilization of functional nucleic acid
motifs on a minimal DNA scaffold display was borne out in
vivo [26], when a DNA tetrahedron was designed that
displayed small interfering RNAs (siRNAs) as well as a
targeting ligand on each edge of the tetrahedron projecting
out from the tetrahedral scaffold (Figure 3b). The authors
showed impressive control over dosage by creating stoi-
chiometrically identical particles in bulk with fixed ratios
of the number of siRNA duplexes per tetrahedron. Fur-
ther, spatial orientation of the targeting ligands affected
the efficiency of RNAi, presumably by affecting the uptake
that is in turn dependent on the surface density of the
ligands. This suggests that aspects of endocytosis, such as
 Review
(a)
Figure 3. (a) Functional design of a hexagonal barrel bearing Fab fragments (magenta) internally that opens in response to the binding of a signaling molecule (red circle) to
the aptameric strand (blue) of the hasps keeping the barrel closed. Reproduced, with permission, from [20]. (b) Schematic of tetrahedron assembly and structure obtained
from six DNA strands to which siRNA (orange) is hybridized. The tetrahedron bears one to six strands of siRNA in fixed stoichiometries. siRNA-bearing DNA tetrahedra are
shown to be specifically targeted to a tumor in a mouse model. Reproduced, with permission, from [26]. (c) Schematic of RNA aptamers used to localize protein synthesis
machinery to an RNA scaffold to create a larger-scale synthesis pathway in higher-order 1D and 2D structures. Reproduced, with permission, from [43].
ligand–receptor spacing and distinct uptake mechanisms
that facilitate endosomal escape or lysosomal delivery
may be interrogated using 3D designer DNA scaffolds [22].
Programming RNA architectures for targeted delivery
Advanced 3D designer RNA scaffolds have also been real-
ized, including nanocubes [27] and nanorings [28], using
the assembly principle ‘RNA tectonics,’ which exploits the
modular nature of RNA molecules to form new architec-
tures [29] (Figure 1d–f). RNA-based particles may be
functionalized with siRNA motifs of controllable stoichi-
ometry and incorporate chemically modified RNA and
DNA analogs due to their synthetic nature, and may
therefore prove valuable for delivery in nanomedicine
[18,30]. Larger-scale RNA nanostructures approaching
the size of DNA origami objects of 5–25 nm have also
recently been achieved based on tectonics of tRNAs [31],
yielding fully addressable polyhedra functionalized with
proteins inside and outside the nanostructure. The bacte-
riophage phi29 DNA packaging motor pRNA has been used
as an alternative construction module [32], comprising a
117-nucleotide RNA that folds into a three-way junction
and further associates into dimers, trimers, or hexamers
[33,34]. This enabled the construction of functional hetero-
dimeric RNA nanoparticles bearing a targeting moiety
such as folate, as well as an siRNA cargo that brings about
apoptosis in nasopharyngeal carcinomas [35].
One important limitation of RNA nanoparticles assem-
bled through intermolecular tertiary interactions for cell-
based applications, however, is the need for metal salts to
stabilize intermolecular junctions. Recently, this limitation
was surmounted by a new thermodynamically stable 3WJ
junction based on the pRNA junction that circumvents this
requirement and retains its structure at physiological con-
centrations of Mg2+. Further, this junction was programmed
using three distinct cargoes: a targeting moiety, an siRNA
against an antiapoptotic marker, and an aptamer or ribo-
zyme, all of which proved to be functional in vivo [36].
Additionally, the realization of RNase-resistant yet biologi-
cally active pRNA [37] suggests that assembly of preformed
RNA components from native biological constructs will
Trends in Cell Biology December 2012, Vol. 22, No. 12
(b) (c)
H F H
Cargo
F
Protein H
D0
aptamer F +
RNA aptamer
Scaffold
D0
Epifluorescence
1.8
Unscaffolded A B
1.6 (unorganized)
1.4
1.2
Scaffolded A B
1.0 (organized)
0.8 One- Two-
Discrete
dimensional dimensional
TRENDS in Cell Biology
remain an attractive alternative to de novo assembly from
synthesized nucleic acids until complex tertiary structures
can reliably be predicted computationally [38,39] or robust
principles of DNA-based assembly can be adopted in the
RNA domain. An alternative approach is to use protein
fragments to guide the assembly of nucleic acid building
blocks into well-defined structures. For example, the L7Ae
protein is known to bind the k-turn motif and force it to adopt
a bent conformation [40]. This has been used to form closed
polygonal shapes, with the protein occupying the vertices of
the associating RNA building blocks [41].
Programming biomolecular spatial arrays for synthesis
and signaling
Organizing biomolecules in 3D to control, for example,
multi-enzyme or signaling cascades is another attractive
aim for nanostructure design. For example, increasing
local concentrations of reactants by spatially localizing
enzymes in close proximity resulted in increased down-
stream mevalonate production by approximately 70-fold
[42] using synthetic protein scaffolds in E. coli. Limitations
of protein-based scaffolds, however, include propensity for
nonspecific aggregation, lack of scaffold addressability, and
lack of scaffold structural fidelity due to a limited ability to
program precisely higher-order protein nanostructures. As
an alternative, extended 1D and 2D RNA-based scaffolds
were designed [43] to organize catalytic enzymes in bacte-
ria to achieve enhanced rates of hydrogen synthesis of up to
approximately 50-fold, demonstrating in vivo structural
engineering of the intracellular environment (Figure 3c).
Similarly, increased metabolic flux in Escherichia coli was
achieved using a plasmid-based DNA scaffold and zinc-
finger binding motifs [44]. Synthetic microcompartments
for sequestration of reactants similar to natural organelles,
such as carboxysomes, and scaffolding motifs for signaling
modules with allosteric functionality [45,46] represent
other enticing possibilities. In vitro scaffolded enzyme
cascades [47,48] studied in conjunction with computational
models will be useful in defining the optimal design criteria
for maximizing chemical synthesis yield and efficiency in
the cell.
627
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a) (c)

Detecon stage Amplificaon stage

HBI (in GFP)
mRNA target H1 Metastable
Probe binding sites H2 hairpins 13-2 RNA DMHBI + DMHBI +
DMHBI
Spinach Spinach-DFHBI
H1 H2 13-2 RNA control RNA
fluorescent complex
Probe set Tethered
DMHBI
Hybridize probes
and wash Fluorescent H1 H2 fluorescent
in situ HCR amplificaon
and wash polymers
Iniators H2
H1

0 min 15 min 30 min

mRNA target

Minimal Target binding Fluorescent complex

90 min 120 min 150 min
fluorescence reinforces structure formaon

(b) (d)
Tg(flk1:egfp) tpm3 elavl3 ntla sox10

Donor Acceptor D/A

H+

5 min
O1 O2
O3 OH-

17 min
z = 20 µm z = 28 µm

60 min
H+

z = 44 µm z = 52 µm 7

TRENDS in Cell Biology

Figure 4. (a) Hybridization chain reaction (HCR) principle whereby metastable fluorescent RNA hairpins H1 and H2 self-assemble into amplification polymers in the
presence of a specific target RNA initiator. An amplification cascade using fluorescent hairpin pairs results in localized amplification of fluorescence. Reproduced, with
permission, from [55]. (b) HCR applied to the multiplexed detection of five distinct mRNAs in whole-mount zebrafish embryos. Reproduced, with permission, from [55]. (c)
Top: chemical structures of 4-hydroxybenzylidene imidazolinone (HBI) (green) and 3,5-dimethoxy-4-hydroxybenzylidene imidazolinone (DMHBI) (black) are shown in
addition to the RNA aptamer Spinach that binds DFHBI but requires the formation of stem 2. Complexation of DMHBI with its aptamer (13-2) enhances its fluorescence.
Reproduced, with permission, from [58]. Bottom: S-adenosyl methionine (SAM) (magenta hexagon) sensor using RNA aptamer region Spinach (black), a transducer
domain (orange), and a SAM aptamer (blue). SAM binding of the sensor promotes the binding of DFHBI (green circle) to Spinach, leading to fluorescence. Bottom right:
distinct patterns of SAM accumulation after adding methionine to Escherichia coli expressing the sensor. From [60]. Reprinted with permission from AAAS. (d) Top left:
Schematic of an i-motif powered, Förster resonance energy transfer (FRET)-based DNA pH sensor, the I-switch, that undergoes a pH triggered conformational change.
Bottom left: the I-switch is internalized by specific receptors present on coelomocytes (yellow cells) in Caenorhabditis elegans and is able to map spatiotemporal pH
changes associated with maturing endosomes on the anionic ligand-binding receptor (ALBR) pathway. Right: pH heat maps of early endosomes, late endosomes and
lysosomes along the ALBR pathway. Reproduced, with permission, from [63].

Probing the cell with fluorescence-based sensors that forms the basis of localized amplification of the fluo-
Biological sensors aim to recognize molecular inputs with rescent signal. This approach enabled multiplexed RNA–
high specificity and sensitivity to transduce them into FISH in whole-mount embryos of impressive clarity, albeit
readouts with minimal perturbation to the cell. Molecular still in fixed samples (Figure 4b).
beacons are a simple early form of nucleic acid-based Directly visualizing RNA molecules in live cells to probe
biosensor comprising a single hairpin with a complemen- RNA trafficking, localization, and transcriptional bursting
tary probe sequence and quenched Förster resonance en- remains an important challenge [52,57] that currently
ergy transfer (FRET) pair that fluoresces on hybridization limits our understanding of the burgeoning RNA layer of
to the target mRNA/DNA [49,50]. Genomic DNA loci and gene regulation. A promising recent development towards
single RNA transcripts are also now routinely measured in realizing live-cell mRNA imaging exploits the photophysi-
fixed cells using fluorescence in situ hybridization (FISH) cal properties of the green fluorescent protein (GFP) fluor-
[51–54], which tiles numerous fluorescent probes along the ophore using RNA aptamers to yield multispectral RNA
target sequence to enhance reporter signal. Although use- mimics of GFP [58]. 4-Hydroxybenzylidene imidazolinone
ful for fixed cells and tissue sections that require only (HBI), the fluorescent moiety in GFP, fluoresces when
limited imaging depth, FISH still suffers from a limited caged in the beta-barrel core of GFP due to proper fluores-
signal-to-noise ratio due to the finite number of fluoro- cence emission from the excited fluorophore state, entering
phores that can be tiled to the target transcript without the dark state when exposed to bulk solution. SELEX was
amplification and limited probe specificity for short used [58] to obtain an RNA aptamer that binds DMHBI, a
sequences. To overcome these limitations and enable sin- chemical analog of HBI, resulting in fluorescence emission
gle-step, multiplexed RNA imaging, a triggered hybridiza- in the bound state (Figure 4c). Further, the authors syn-
tion chain reaction was used [55,56] to dramatically thesized the small molecule DFHBI that is exclusively in
increase the signal-to-noise ratio (Figure 4a). Here, RNA the phenolate form at neutral pH and also obtained an
probes folded into metastable hairpins harboring fluores- aptamer, termed Spinach, that binds this target molecule.
cent reporters restructure in the presence of a desired DFHBI exhibits fluorescence when bound to Spinach that
mRNA sequence into a nicked double-stranded polymer is approximately half as bright as enhanced GFP (eGFP),
628
Review
yet brighter than many other fluorescent proteins, consid-
erably more photo-stable than GFP, and does not require
the maturation time of GFP. Impressive live-cell imaging
of 5S RNA translocation in mammalian cells was achieved
when incorporated into the 30 UTR, which does not affect
localization.
To achieve conditional functionality, first described [59]
using the cytotoxic malachite green (MG), a metabolite-
binding aptamer was fused [60] to Spinach using a commu-
nication module that confers proper folding of Spinach and
subsequent fluorescence in the presence of DFHBI and the
metabolite in a manner that is dependent on concentration of
the latter (Figure 4c). This enabled visualization of S-ade-
nosyl methionine distribution in living bacteria, which, to-
gether with the ability to evolve aptamers to bind nearly any
small molecule, provides an exciting opportunity to image
bioactive small molecules and RNAs in living systems.
In addition to RNA-based switches, DNA-based switches
may also exhibit large-scale conformational rearrange-
ments in response to molecular cues including pH that offer
the unique ability to map the latter in live cells. Spatiotem-
poral maps of pH in endocytic compartments within living
cells have provided important insights into the stages of
endosomal maturation, viral entry mechanisms, and their
crosstalk with secretory pathways [61]. Imaging endosomal
maturation within a multicellular, live organism was re-
cently achieved using a FRET-based, pH-sensitive DNA
switch called the I-switch [62] that undergoes a conforma-
tional change under acidic conditions (Figure 4d). This
conformational change was mediated by i-motif formation
of C-rich sequences on the I-switch and the device was first
validated in cell culture [63]. Importantly, the functionality
of this pH-triggered device was quantitatively recapitulated
from the in vitro to the in vivo context, providing such
bistable DNA devices with a new role inside living systems
as probes of the intracellular chemical environments of
specific proteins.
Programming the cell to perform desired functions
Synthetic biology seeks to program novel functions into
living systems, for example, convert or synthesize biomole-
cules including biofuels, sense and destroy biological mate-
rial including pathogens, harvest, convert, and store solar
energy, and test working hypotheses on the function of
complex living systems [64,65]. Engineered biological sys-
tems must therefore have the capacity to receive molecular
cues, perform context-dependent computation, and produce
outputs that lead to desired changes in cell state or function,
and achieve all of this under the noisy, heterogeneous con-
ditions in which cells operate. Unlike solid-state machines
that principally seek to store and compute large amounts of
information, typically containing only a single mode of input
and output, biological computation exploits an opposing side
of the computational spectrum, often comprising only a few
‘states’ with large numbers of inputs comprising small
molecules, metabolites, RNAs, and protein, and outputs
at the levels of transcription, translation, ligation, cleavage,
and small-molecule signaling [17,64–67]. Here, we focus on
selected recent findings of relevance to a breadth of cell
biological problems and refer the reader to the preceding in-
depth reviews for additional background.
Trends in Cell Biology December 2012, Vol. 22, No. 12
Digital and analog controls of protein expression in
living systems are essential tools for probing the role of
protein function in determining cell state and behavior.
Controlling translation using small-molecule inducers is
of major interest for this purpose, both for the reversible
control intrinsic to such methods per se, and specifically for
the study, for example, of processes and pathogens not
amenable to conventional genetic tools, such as the ma-
larial parasite Plasmodium falciparum. The adenine
riboswitch is one such ‘translational ON’ switch that
allows translation to occur in the presence of adenine
by restructuring the riboswitch to unmask the Shine-
Dalgarno sequence. This switch was reengineered [68]
to produce two different riboswitches that are responsive
either to ammeline or to 2-aminopurine. In subsequent
work [69], the authors went on to independently control in
the same bacterium the coexpression of eGFP and DsRed
induced by ammeline and 2-aminopurine, respectively, in
a dose-dependent manner (Figure 5a). A major advantage
of these approaches over analogous protein-based systems
is the elimination of crosstalk and/or the ability to control
protein expression profiles more finely. Further, protein
stoichiometries may possibly be sculpted in a multiplexed
fashion using this approach with a cocktail of inducers at
different concentrations. Such systems thereby facilitate
quantitative approaches to probing and understanding
protein function in live cells. In this respect, a recent
ion-sensitive fluoride riboswitch represents a valuable
addition for temporal control of gene expression using
orthogonal triggers, because fluoride does not result in
metabolites that could potentially crosstalk with the trig-
ger, is cell-permeable, and may be washed out post-induc-
tion [70].
The RNA motifs and small molecules described above
may be used not only to modulate the expression of any
unrelated gene, but also to turn on a pathway that meta-
bolizes the chemical trigger. It is therefore feasible to
engineer synthetic functions into living systems, such as
sensing and responding to a non-natural molecule. As an
example, a synthetic riboswitch was used to control motili-
ty in E. coli by introducing an atrazine aptamer in the 50
UTR of the cheZ gene, a key player in the chemotactic E.
coli motility machinery, in the background of an atrazine-
metabolizing gene [71]. This resulted in ‘seek-and-destroy’
bacteria that were programmed to seek out and degrade
the toxic herbicide atrazine. This approach to controlling
chemotaxis is potentially generalizable to myriad chemi-
cals where E. coli may be engineered to clear unwanted
chemicals in the environment or other systems [72].
Apart from ON/OFF riboswitches that respond to a
unique molecular cue, there are rare natural examples of
complex logical control over gene expression through tan-
dem riboswitches on a single transcript [73]. Such control
was realized synthetically using tandem riboswitches cou-
pled to the hammerhead ribozyme (HHR) on the 30 UTR of a
given mRNA transcript that mediated logical control over
mRNA cleavage and degradation (Figure 5b). Thus, yeast-
enhanced GFP (yeGFP) was expressed as a function of AND,
OR, NAND, and NOR logic by positionally coupling multiple
aptamers via crucial connecting modules on the mRNA
transcript such that the addition of two small-molecule
629
Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a) (b)
AND gate
A Input A Input B
AB Output
B
OFF-state ON-state
3′ [SI 1.2]
Start A B Output
RBS theo tc GFP
0 0 0
3′ 0 1 0
5′ 5′ 1 0 0
AAAAA
1 1 1

Output protein (high when AB)

250

125
5:1 (+)
4:1
NOR gate
Ammeline (μM)

64 Input A Input B
3:1 A
A+B Output
32 2:1 B
[SI 1.3]
1:1 A B Output
16
1:2 theo tc GFP
8 1:3 0 0 1
1:4
0 1 0
4 1:5 (+)
1 0 0
AAAAA
1 1 0
0
0 4 8 16 32 64 125 250
2-aminopurine (μM)

Output protein (high when A+B)

TRENDS in Cell Biology

Figure 5. (a) Top: translational mechanism of gene regulation by the engineered adenine riboswitch responsive to either ammeline or 2-aminopurine (2AP) (blue hexagon).
Binding to the inducer releases the normally sequestered ribosome-binding site (RBS) and start codon, allowing mRNA translation. Bottom: contour plots showing relative
coexpression ratios of enhanced green fluorescent protein (eGFP): DsRed for the full range of ammeline and 2-aminopurine. Reproduced, with permission, from [69]. (b)
Top: an RNA device incorporated into an mRNA that expresses yeast-enhanced GFP (yeGFP) according to AND logic and the associated truth table. In the absence of either
A or B, at least one functional hammerhead ribozyme (HHR) motif cleaves the transcript. In the presence of both A and B, both HHR motifs are inactivated and the yeGFP is
produced. Reproduced, with permission, from [74]. Bottom: an RNA device incorporating two analogous inactivated HHR motifs that become activated on addition of A or B
serves as a NOR gate. Reproduced, with permission, from [74].

inputs resulted in activation or deactivation of mRNA high-throughput approaches are likely to play an impor-
cleavage by HHR [74]. tant future role in identifying suitable candidates for RNA-
Examples of implementation of logical control of cell based function.
function in mammalian systems are limited [75,76,77],
representing a major challenge of interest going forward. Concluding remarks
In a recent success, specific signatures of microRNAs Fifty years after the postulate that RNA stands at the
characteristic of cancer cells were read into a logic-gated crossroads of the biocatalytic and information-bearing
computer and used to induce apoptosis using network logic molecules: proteins and DNA, our rapidly advancing un-
[76]. The logic framework implemented is completely gen- derstanding of nucleic acid form and function, together
eral for use in detecting specific gene expression profiles in with our ability to program these polymers for custom
mammalian cells [78], utilizing a combination of micro- purposes offer immense room for their application to basic
RNA inputs together with different logic gates such as and applied cell biology.
AND and NOT to produce a specific programmed cellular We envision an eventual palette of made-to-order RNA-
output. Various means of using riboswitches, aptamers, based devices to probe cellular machinery including levels
and allosteric aptamers to switch on or off gene expression and localization of metabolites, small molecules, ions, and
are comprehensively discussed in [79]. diverse coding and non-coding RNAs in a multiplexed
The aforementioned studies have largely used manual manner with sensitivities that are inaccessible to existing
approaches to achieve RNA-based programs that modulate technologies that are largely protein-based (Figure 6). Ex-
gene expression. In an important development, a high- ploration of intracellular spaces in an organelle-specific
throughput approach was delineated [80] that screens manner to probe their proteomic composition in individual,
endogenous RNA interactions to identify orthogonal living cells, as well as intercellular spaces such as the
mRNA-regulatory RNA pairs. Using an integrated model- neuronal or immunological synapse [81,82] with such de-
ing–experimental framework based on the free energies of signer devices represent intriguing uncharted territories.
RNA hybridization, the authors realized a library of engi- Utilizing the ability to precisely engineer higher-order
neered RNA regulators for predictable interactors. Such structural scaffolds and nucleic acid architectures from
630
Review Trends in Cell Biology December 2012, Vol. 22, No. 12

Neu

lar
rotra

struc molecu
Scaffolds Proteomics

rs
nsm

ture

to
ro

mo
ier

Mac

lar
s

cu
Fo

ole
ule

rce
M
lec

M
et o
ab
oli -m s
te gle sic
s Sin ophy
bi

s
ure
l
-cel

DN
DNA Cell acons

uct
Azy
e r
int

str
me

no
s

Na
Super-resoluon
DNA Signaling
microscopy cascades
Sensors Ribozymes Aptamers Delivery
RNA

Ions he
s
Diagn
oscs

mi
itc

RN
osw

A
Rib

The
A rap
RN e u
cs
Programs

Lig
ns
ei

ht
ot

-h
Pr

ar
Ene versio
n

ve
con
ao

s
rgy
Biosynth
diaon

gn
gul
ore

n
Bioreme
Rib

esis

TRENDS in Cell Biology

Figure 6. Overview showing the distinct layers of tools based on DNA and RNA and their diverse applications to cellular and synthetic biology.

DNA and RNA templates at the scale of macromolecular of force sensation and feedback on binding interactions
assemblies such as the nuclear pore complex, chaperones, mediated by focal adhesions [96] and cell–cell junctions
and the ribosome further offers the unique ability to probe [97], will be key to mechanistically understanding morpho-
the effect of the specific higher-order organization of di- genesis, development, and regeneration.
verse macromolecular structures, including relative posi- As evidenced by the macromolecular architectures and
tioning and orientation, on their function. programs presented in this review, nucleic acids offer the
Active scaffolds integrated with precisely positioned pro- unique capability to design scaffolds with high precision for
teins and protein domains may additionally be designed diverse applications in cell biology. Among the many chal-
with allosteric response to probe signaling cascades [45], lenges we face in bringing these structures to bear broadly
interrogate cell circuits, probe the RNA transcriptome [83], on the study of cell biology are an understanding of how to
modulate enzyme cascades and metabolic pathways [42– control and circumvent crosstalk between thousands of
44], perform single-molecule manipulations, and provide molecules and pathways in the cell to leverage sophisticat-
platforms for structural biology [84,85] as well as super- ed nucleic acid-based in vitro programs [66,98–100], as well
resolution imaging [86,87]. Extrinsic control of nucleic acid as the immense heterogeneity and variability in molecular
duplex stability and conformation with light, as demonstrat- levels that feed back to determine cell state and response,
ed extensively in vitro [88–90], will offer novel means to particularly in complex mammalian systems [17,64,66,75–
spatially and temporally regulate the activity of scaffolds, 78,101]. Although the challenge is great, the ever-increas-
sensors, and inhibitors in living systems from individual ing arsenal of new tools and technologies at our disposal
cells to developing embryos [91]. Finally, construction of ensures that meaningful progress will be achieved in the
extracellular origami using DNA to position cells in 3D near term. Along the way, developing and applying nucleic
[92,93] to probe their interaction with their environment acid-based tools to address problems that are currently
in artificial tissue scaffolds and on 2D substrates [94,95] will intractable using conventional technologies alone will un-
provide insight into their function in organizing higher- doubtedly aid the advance of synthetic biology in its aim to
order tissues. Such studies, together with an understanding engineer biological systems for the benefit of society.
631
Review Trends in Cell Biology December 2012, Vol. 22, No. 12

Acknowledgments 27 Afonin, K.A. et al. (2010) In vitro assembly of cubic RNA-based

M.B. acknowledges the Office of Naval Research (ONR N000141210621), scaffolds designed in silico. Nat. Nanotechnol. 5, 676–682
the Army Research Office (MURI W911NF-12-1-0420), and MIT Faculty 28 Grabow, W.W. et al. (2011) Self-assembling RNA nanorings based on
Start-up Funds for financial support. Y.K. acknowledges the Innovative RNAI/II inverse kissing complexes. Nano Lett. 11, 878–887
Young Biotechnologist Award and the Wellcome Trust-DBT India 29 Jaeger, L. and Chworos, A. (2006) The architectonics of programmable
Alliance Senior Research Fellowship. The authors are grateful to RNA and DNA nanostructures. Curr. Opin. Struct. Biol. 16, 531–543
Camille Delebecque, Hendrik Dietz, Shawn Douglas, Luc Jaeger, Samie 30 Afonin, K.A. et al. (2011) Design and self-assembly of siRNA-
Jaffrey, Takayuki Kato, Hyukjin Lee, Chengde Mao, Jason Micklefield, functionalized RNA nanoparticles for use in automated
Niles Pierce, Christina Smolke, and Andrew Turberfield, for providing nanomedicine. Nat. Protoc. 6, 2022–2034
high-resolution versions of published figures that are reproduced here, as 31 Severcan, I. et al. (2010) A polyhedron made of tRNAs. Nat. Chem. 2,
well as Matthew Adendorff for help with rendering figures. The authors 772–779
appreciate discussions and feedback from numerous colleagues in the 32 Guo, P.X. et al. (1998) Inter-RNA interaction of phage phi 29 pRNA to
preparation of this work, and sincerely apologize to those authors whose form a hexameric complex for viral DNA transportation. Mol. Cell 2,
work was not included due to space limitations. 149–155
33 Chen, C.P. et al. (2000) A dimer as a building block in assembling RNA
- a hexamer that gears bacterial virus phi29 DNA-translocating
References
machinery. J. Biol. Chem. 275, 17510–17516
1 Rich, A. (1962) On the problems of evolution and biochemical
34 Shu, D. et al. (2004) Bottom-up assembly of RNA arrays and
information transfer. In Horizons in Biochemistry (Kasha, M. and
superstructures as potential parts in nanotechnology. Nano Lett. 4,
Pullmann, B., eds), pp. 103–126, Plenum Press
1717–1723
2 Gilbert, W. (1986) Origin of life - the RNA world. Nature 319, 618
35 Guo, S. et al. (2006) Construction of folate-conjugated pRNA of
3 Seeman, N.C. (1982) Nucleic acid junctions and lattices. J. Theor. Biol.
bacteriophage phi29 DNA packaging motor for delivery of chimeric
99, 237–247
siRNA to nasopharyngeal carcinoma cells. Gene Ther. 13, 814–820
4 Seeman, N.C. (2010) Nanomaterials based on DNA. Annu. Rev.
36 Shu, D. et al. (2011) Thermodynamically stable RNA three-way
Biochem. 79, 65–87
junction for constructing multifunctional nanoparticles for delivery
5 Dietz, H. et al. (2009) Folding DNA into twisted and curved nanoscale
of therapeutics. Nat. Nanotechnol. 6, 658–667
shapes. Science 325, 725–730
37 Liu, J. et al. (2011) Fabrication of stable and RNase-resistant RNA
6 Rothemund, P.W.K. (2006) Folding DNA to create nanoscale shapes
nanoparticles active in gearing the nanomotors for viral DNA
and patterns. Nature 440, 297–302
packaging. ACS Nano 5, 237–246
7 Nangreave, J. et al. (2010) DNA origami: a history and current
38 Cruz, J.A. et al. (2012) RNA-puzzles: a CASP-like evaluation of RNA
perspective. Curr. Opin. Chem. Biol. 14, 608–615
three-dimensional structure prediction. RNA 18, 610–625
8 Shih, W.M. and Lin, C.X. (2010) Knitting complex weaves with DNA
39 Westhof, E. (2012) RNA modeling, naturally. Proc. Natl. Acad. Sci.
origami. Curr. Opin. Struct. Biol. 20, 276–282
U.S.A. 109, 2691–2692
9 Castro, C.E. et al. (2011) A primer to scaffolded DNA origami. Nat.
40 Turner, B. et al. (2005) Induced fit of RNA on binding the L7Ae protein
Methods 8, 221–229
to the kink-turn motif. RNA 11, 1192–1200
10 Douglas, S.M. et al. (2009) Rapid prototyping of 3D DNA-origami
41 Ohno, H. et al. (2011) Synthetic RNA-protein complex shaped like an
shapes with caDNAno. Nucleic Acids Res. 37, 5001–5006
equilateral triangle. Nat. Nanotechnol. 6, 115–119
11 Kim, D-N. et al. (2012) Quantitative prediction of 3D solution shape
42 Dueber, J.E. et al. (2009) Synthetic protein scaffolds provide modular
and flexibility of nucleic acid nanostructures. Nucleic Acids Res. 40,
control over metabolic flux. Nat. Biotechnol. 27, 753–759
2862–2868
43 Delebecque, C.J. et al. (2011) Organization of intracellular reactions
12 Dirks, R.M. and Pierce, N.A. (2003) A partition function algorithm for
with rationally designed RNA assemblies. Science 333, 470–474
nucleic acid secondary structure including pseudoknots. J. Comput.
44 Conrado, R.J. et al. (2012) DNA-guided assembly of biosynthetic
Chem. 24, 1664–1677
pathways promotes improved catalytic efficiency. Nucleic Acids
13 Isambert, H. (2009) The jerky and knotty dynamics of RNA. Methods
Res. 40, 1879–1889
49, 189–196
45 Good, M.C. et al. (2011) Scaffold proteins: hubs for controlling the flow
14 Shapiro, B.A. et al. (2008) Protocols for the in silico design of RNA
of cellular information. Science 332, 680–686
nanostructures. Methods Mol. Biol. 474, 93–115
46 Bashor, C.J. et al. (2010) Rewiring cells: synthetic biology as a tool to
15 Ellington, A.D. and Szostak, J.W. (1990) In vitro selection of RNA
interrogate the organizational principles of living systems. Annu. Rev.
molecules that bind specific ligands. Nature 346, 818–822
Biophys. 39, 515–537
16 Guo, P.X. (2010) The emerging field of RNA nanotechnology. Nat.
47 Fu, J.L. et al. (2012) Interenzyme substrate diffusion for an enzyme
Nanotechnol. 5, 833–842
cascade organized on spatially addressable DNA nanostructures. J.
17 Isaacs, F.J. et al. (2006) RNA synthetic biology. Nat. Biotechnol. 24,
Am. Chem. Soc. 134, 5516–5519
545–554
48 Wilner, O.I. et al. (2009) Enzyme cascades activated on topologically
18 Goodman, R.P. et al. (2005) Rapid chiral assembly of rigid DNA
programmed DNA scaffolds. Nat. Nanotechnol. 4, 249–254
building blocks for molecular nanofabrication. Science 310, 1661–
49 Tyagi, S. and Kramer, F.R. (1996) Molecular beacons: probes that
1665
fluoresce upon hybridization. Nat. Biotechnol. 14, 303–308
19 Erben, C.M. et al. (2006) Single-molecule protein encapsulation in a
50 Bratu, D.P. et al. (2003) Visualizing the distribution and transport
rigid DNA cage. Angew. Chem. Int. Ed. 45, 7414–7417
of mRNAs in living cells. Proc. Natl. Acad. Sci. U.S.A. 100, 13308–
20 Douglas, S.M. et al. (2012) A logic-gated nanorobot for targeted
13313
transport of molecular payloads. Science 335, 831–834
51 Femino, A. et al. (1998) Visualization of single RNA transcripts in situ.
21 Bhatia, D. et al. (2009) Icosahedral DNA nanocapsules by modular
Science 280, 585–590
assembly. Angew. Chem. Int. Ed. 48, 4134–4137
52 Tyagi, S. (2009) Imaging intracellular RNA distribution and dynamics
22 Bhatia, D. et al. (2011) A synthetic icosahedral DNA-based host-cargo
in living cells. Nat. Methods 6, 331–338
complex for functional in vivo imaging. Nat. Commun. 2, 339
53 Capodieci, P. et al. (2005) Gene expression profiling in single cells
23 Lo, P.K. et al. (2010) Loading and selective release of cargo in DNA
within tissue. Nat. Methods 2, 663–665
nanotubes with longitudinal variation. Nat. Chem. 2, 319–328
54 Raj, A. et al. (2008) Imaging individual mRNA molecules using
24 Zhang, C.A. et al. (2010) Exterior modification of a DNA tetrahedron.
multiple singly labeled probes. Nat. Methods 5, 877–879
Chem. Commun. 46, 6792–6794
55 Choi, H.M.T. et al. (2010) Programmable in situ amplification for
25 Li, J. et al. (2011) Self-assembled multivalent DNA nanostructures for
multiplexed imaging of mRNA expression. Nat. Biotechnol. 28, 1208–
noninvasive intracellular delivery of immunostimulatory CpG
1212
oligonucleotides. ACS Nano 5, 8783–8789
56 Dirks, R.M. and Pierce, N.A. (2004) Triggered amplification by
26 Lee, H. et al. (2012) Molecularly self-assembled nucleic acid
hybridization chain reaction. Proc. Natl. Acad. Sci. U.S.A. 101,
nanoparticles for targeted in vivo siRNA delivery. Nat.
15275–15278
Nanotechnol. 7, 389–393

632
Review
57 Armitage, B.A. (2011) Imaging of RNA in live cells. Curr. Opin. Chem.
Biol. 15, 806–812
58 Paige, J.S. et al. (2011) RNA mimics of green fluorescent protein.
Science 333, 642–646
59 Stojanovic, M.N. and Kolpashchikov, D.M. (2004) Modular aptameric
sensors. J. Am. Chem. Soc. 126, 9266–9270
60 Paige, J.S. et al. (2012) Fluorescence imaging of cellular metabolites
with RNA. Science 335, 1194
61 Weisz, O.A. (2003) Acidification and protein traffic. Int. Rev. Cytol.
226, 259–319
62 Surana, S. et al. (2011) An autonomous DNA nanomachine maps
spatiotemporal pH changes in a multicellular living organism. Nat.
Commun. 2, 340
63 Modi, S. et al. (2009) A DNA nanomachine that maps spatial and
temporal pH changes inside living cells. Nat. Nanotechnol. 4, 325–330
64 Khalil, A.S. and Collins, J.J. (2010) Synthetic biology: applications
come of age. Nat. Rev. Genet. 11, 367–379
65 Andrianantoandro, E. et al. (2006) Synthetic biology: new engineering
rules for an emerging discipline. Mol. Syst. Biol. 2, 2006.0028
66 Benenson, Y. (2012) Biomolecular computing systems: principles,
progress and potential. Nat. Rev. Genet. 13, 455–468
67 Chen, X. and Ellington, A.D. (2010) Shaping up nucleic acid
computation. Curr. Opin. Biotechnol. 21, 392–400
68 Dixon, N. et al. (2010) Reengineering orthogonally selective
riboswitches. Proc. Natl. Acad. Sci. U.S.A. 107, 2830–2835
69 Dixon, N. et al. (2012) Orthogonal riboswitches for tuneable
coexpression in bacteria. Angew. Chem. Int. Ed. 51, 3620–3624
70 Baker, J.L. et al. (2012) Widespread genetic switches and toxicity
resistance proteins for fluoride. Science 335, 233–235
71 Mandelbaum, R.T. et al. (1995) Isolation and characterization of a
pseudomonas sp that mineralizes the s-triazine herbicide atrazine.
Appl. Environ. Microbiol. 61, 1451–1457
72 Sinha, J. et al. (2010) Reprogramming bacteria to seek and destroy an
herbicide. Nat. Chem. Biol. 6, 464–470
73 Sudarsan, N. et al. (2006) Tandem riboswitch architectures exhibit
complex gene control functions. Science 314, 300–304
74 Win, M.N. and Smolke, C.D. (2008) Higher-order cellular information
processing with synthetic RNA devices. Science 322, 456–460
75 Leisner, M. et al. (2010) Rationally designed logic integration of
regulatory signals in mammalian cells. Nat. Nanotechnol. 5, 666–670
76 Xie, Z. et al. (2011) Multi-input RNAi-based logic circuit for
identification of specific cancer cells. Science 333, 1307–1311
77 Ausländer, S. et al. (2010) A ligand-dependent hammerhead ribozyme
switch for controlling mammalian gene expression. Molecular
BioSystems 6, 807–814
78 Rinaudo, K. et al. (2007) A universal RNAi-based logic evaluator that
operates in mammalian cells. Nat. Biotechnol. 25, 795–801
79 Vinkenborg, J.L. et al. (2011) Aptamers for allosteric regulation. Nat.
Chem. Biol. 7, 519–527
80 Mutalik, V.K. et al. (2012) Rationally designed families of orthogonal
RNA regulators of translation. Nat. Chem. Biol. 8, 447–454
81 Schratt, G. (2009) MicroRNAs at the synapse. Nat. Rev. Neurosci. 10,
842–849
Trends in Cell Biology December 2012, Vol. 22, No. 12
82 Manz, B.N. and Groves, J.T. (2010) Spatial organization and signal
transduction at intercellular junctions. Nat. Rev. Mol. Cell Biol. 11,
342–352
83 Wan, Y. et al. (2011) Understanding the transcriptome through RNA
structure. Nat. Rev. Genet. 12, 641–655
84 Douglas, S.M. et al. (2007) DNA-nanotube-induced alignment of
membrane proteins for NMR structure determination. Proc. Natl.
Acad. Sci. U.S.A. 104, 6644–6648
85 Berardi, M.J. et al. (2011) Mitochondrial uncoupling protein 2
structure determined by NMR molecular fragment searching.
Nature 476, 109–113
86 Opazo, F. et al. (2012) Aptamers as potential tools for super-resolution
microscopy. Nat. Methods 9, 938–939
87 Lin, C. et al. (2012) Submicrometre geometrically encoded fluorescent
barcodes self-assembled from DNA. Nat. Chem. 4, 832–839
88 Nishioka, H. et al. (2012) A photon-fueled DNA nanodevice that
contains two different photoswitches. Angew. Chem. Int. Ed. 51,
1165–1168
89 Wang, C.M. et al. (2011) Caged molecular beacons: controlling nucleic
acid hybridization with light. Chem. Commun. 47, 5708–5710
90 Asanuma, H. et al. (2007) Synthesis of azobenzene-tethered DNA for
reversible photo-regulation of DNA functions: hybridization and
transcription. Nat. Protoc. 2, 203–212
91 Toettcher, J.E. et al. (2011) The promise of optogenetics in cell biology:
interrogating molecular circuits in space and time. Nat. Methods 8,
35–38
92 Gartner, Z.J. and Bertozzi, C.R. (2009) Programmed assembly of 3-
dimensional microtissues with defined cellular connectivity. Proc.
Natl. Acad. Sci. U.S.A. 106, 4606–4610
93 Selden, N.S. et al. (2012) Chemically programmed cell adhesion with
membrane-anchored oligonucleotides. J. Am. Chem. Soc. 134, 765–
768
94 Hung, A.M. et al. (2010) Large-area spatially ordered arrays of gold
nanoparticles directed by lithographically confined DNA origami.
Nat. Nanotechnol. 5, 121–126
95 Kershner, R.J. et al. (2009) Placement and orientation of individual
DNA shapes on lithographically patterned surfaces. Nat.
Nanotechnol. 4, 557–561
96 Geiger, B. et al. (2009) Environmental sensing through focal
adhesions. Nat. Rev. Mol. Cell Biol. 10, 21–33
97 Levayer, R. and Lecuit, T. (2012) Biomechanical regulation of
contractility: spatial control and dynamics. Trends Cell Biol. 22,
61–81
98 Winfree, E. et al. (1998) Design and self-assembly of two-dimensional
DNA crystals. Nature 394, 539–544
99 Seelig, G. et al. (2006) Enzyme-free nucleic acid logic circuits. Science
314, 1585–1588
100 Zhang, D.Y. et al. (2007) Engineering entropy-driven reactions and
networks catalyzed by DNA. Science 318, 1121–1125
101 Xie, Z. et al. (2010) Logic integration of mRNA signals by an RNAi-
based molecular computer. Nucleic Acids Res. 38, 2692–2701
102 Kato, T. et al. (2009) High-resolution structural analysis of a DNA
nanostructure by cryoEM. Nano Lett. 9, 2747–2750
633
Review
Special Issue – Synthetic Cell Biology
Towards a bottom-up reconstitution of
bacterial cell division
Ariadna Martos1, Mercedes Jiménez2, Germán Rivas2, and Petra Schwille1
1
Max Planck Institute of Biochemistry, Am Klopferspitz 18, D-82152 Martinsried, Germany
2
Centro de Investigaciones Biológicas, CSIC, Ramiro de Maeztu 9, 28040-Madrid, Spain
The components of the bacterial division machinery
assemble to form a dynamic ring at mid-cell that drives
cytokinesis. The nature of most division proteins and
their assembly pathway is known. Our knowledge about
the biochemical activities and protein interactions of
some key division elements, including those responsible
for correct ring positioning, has progressed considerably
during the past decade. These developments, together
with new imaging and membrane reconstitution tech-
nologies, have triggered the ‘bottom-up’ synthetic ap-
proach aiming at reconstructing bacterial division in the
test tube, which is required to support conclusions
derived from cellular and molecular analysis. Here, we
describe recent advances in reconstituting Escherichia
coli minimal systems able to reproduce essential func-
tions, such as the initial steps of division (proto-ring
assembly) and one of the main positioning mechanisms
(Min oscillating system), and discuss future perspectives
and experimental challenges.
The strategy of constructive (bottom-up) synthetic
biology of bacterial cell division
The aim of synthetic biology is to dissect biological entities
into functional modules (macromolecular assemblies, cells,
tissues, organisms, and their networks), to modify them
rationally, or to engineer entirely new ones with the goal of
implementing novel biomimetic functionalities [1,2]. The
bottom-up synthetic biology approach has strong physico-
chemical foundations; it is interdisciplinary and has a
rigorous quantitative character. It aims at constructing
biological parts, devices, and systems with increasing
levels of complexity toward a minimal cell-like scaffold
[3–5]. The ‘top-down’ approach relies rather on biotechnol-
ogy and systems biology, seeking primarily the program-
ming of novel gene circuits within an electronic and
computational engineering framework [6,7].
In the context of bottom-up approaches, the in vitro
reconstitution of minimal bacterial cell division machinery
is an obvious and attractive goal. Considering the highly
dynamic and membrane-associated nature of this essential
machine, and the difficulty of assaying its functional reac-
tions, it is a challenging prospect. However, in the past few
years, new technologies for imaging and functional protein
Corresponding authors: Rivas, G. (grivas@cib.csic.es); Schwille, P.
(schwille@biochem.mpg.de).
Keywords: constructive synthetic biology; cytomimetic biochemistry; macromolecular
interactions; FtsZ; Min system; Escherichia coli.
634 0962-8924/$ – see front matter ß 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tcb.2012.09.003 Trends in Cell Biology, December 2012, Vol. 22, No. 12
Glossary
Amphitropic proteins: proteins that can exist as either soluble cytoplasmic or
membrane-attached forms depending on physiological conditions of activa-
tion. Their reversible interaction with membranes seems to be mediated by
specific lipids. This concept was proposed by Paul Burn (1988).
Atomic force microscopy (AFM): an imaging and measurement tool that
provides a 3D map of a sample’s surface. It allows the collection of topography
images with resolution in the nanometer scale and, at the same time, maps
other interesting properties of the samples. With the capacity to observe and
manipulate biological surfaces under physiological conditions, AFM can be
used to explore biological structures at the single-molecule level.
Division ring: a multiprotein complex whose presence at mid-cell is derived
from the coincident location of several proteins (e.g., FtsZ, FtsA, ZipA, FtsN),
because they are independently visualized as rings in dividing cells [8].
Divisome: the complete macromolecular machinery able to effect division in
the living cell. All the proteins that form the division ring at mid-cell are part of
the divisome. Other components such as periplasmic and outer membrane
proteins and the cell envelope may also be integral parts of the divisome. The
term divisome was proposed by Nanne Nanninga (1998).
Fluorescence recovery after photobleaching (FRAP): the observation and
measurement of FRAP allows scientists to investigate the diffusion and motion
of fluorescently labelled biological macromolecules. The fluorescently tagged
biomolecule of interest is visualized at low light intensity followed by
photobleaching of a user-defined region of interest with high-intensity light,
causing the fluorophores to bleach in the selected region. The measure of the
rate of recovery of fluorescence allows one to infer how fast macromolecules
move.
Fluorescence resonance energy transfer (FRET): a technique for measuring
interactions between two molecules. It relies on the ability to capture
fluorescent signals from the interactions of labeled molecules in single living
or fixed cells. If FRET occurs, the donor channel signal will be quenched and
the acceptor channel signal will be sensitized or increased. This allows
investigation of the colocalization of the donor- and acceptor-labeled probes
and the molecular associations at close distances.
Minimal divisome: the theoretical minimal set of elements capable of effecting
division. It can be defined genetically, when using mutants defective in some
of the divisome elements, or biochemically when using artificial cell envelope
assemblies in the test tube.
Membrane-targeting sequence (mts): among other amphitropic proteins, FtsA
and MinD present a highly conserved C-terminal sequence motif that it is
thought to form an amphipathic helix that mediates the direct interaction of
these proteins with lipid bilayers. The mts from MinD was incorporated into
chimeric FtsZ forms to bypass the need for membrane-anchoring proteins.
Proto-ring: the natural initial subset of the division ring able to be anchored to
the cytoplasmic membrane. In E. coli, it is formed by FtsZ, FtsA, and ZipA and it
is assumed to be the initial step of divisome assembly [10].
Total internal reflection fluorescence microscopy (TIRFM): a powerful fluores-
cence microscopy technique that allows imaging but also the measurement of
dynamics and single-molecule events taking place in a thin section (100 nm)
in the contact surface between the specimen and the glass slide. An
evanescent field propagates into the specimen on total reflection of the
incident beam at the liquid/glass interface. In this way, only the fluorophores
close to the surface are excited, thus reducing background fluorescence due to
out-of-focus molecules.
Z-ring: a structure present at mid-cell of dividing cells and visualized by
microscopy techniques using FtsZ-specific reagents, such as fusions to
fluorescent proteins or antibodies against FtsZ. It is postulated to be a ring
of variable diameter formed by the accumulation of FtsZ-containing complexes
[8]. The Z- ring has been assumed to provide a scaffold on which the remaining
division elements assemble due to their ability to interact directly or indirectly
with FtsZ. The term Z-ring was proposed by Joe Lutkenhaus (1991).
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

reconstitution in membranes have put this goal within
reach. Today, we know the main elements, their biological
function, and a large part of their biochemistry; therefore,
it seems not unreasonable to expect that the application of
this bottom-up synthetic strategy to construct minimal
divisomes (see Glossary) in the test tube is feasible and
will yield functional assemblies. This constructive synthet-
ic approach will help us to support conclusions already
derived from cellular and molecular analysis and, there-
fore, to complete our understanding of how the bacterial
division machinery works.
The bacterial cell division machinery
Bacterial cell division is mediated by a multiprotein ma-
chine whose elements interact dynamically at mid-cell to
form the division ring, a structure that drives cytokinesis,
forming part of the divisome [8,9]. The position of the ring
is highly regulated by two negative control systems that
inhibit its incorrect formation (see below). In the case of E.
coli, the division ring is formed by at least ten essential
proteins, most of them integrated at the membrane. Their
assembly pathway has been mostly derived from the be-
havior of mutants in which divisome elements have been

A
(a) B
Z K Q W I N
L
Zip

Proto-ring Periplasmic Pepdoglycan

connector factory

Cytoplasmic ring Periplasmic ring

Early assembly Late assembly

FtsA-independent assembly

(b) FtsQ
FtsL FtsB FtsL FtsN
FtsK
Out FtsW

Membrane

In
FtsA ZipA

FtsZ
GDP GTP

Nanodiscs Microbeads Bilayers Vesicles

TRENDS in Cell Biology

Figure 1. Assembly of the essential components of the Escherichia coli division ring. (a) First row: FtsZ interacts with FtsA and ZipA at the cytoplasmic membrane forming
the initial assembly of the division machinery, the proto-ring. For the sake of simplicity other division proteins (such as FtsE/X or Zap proteins) are not shown in this scheme.
FtsK is then added sequentially, followed by a group formed by FtsQ, B and L. FtsW and I (PBP3), two proteins involved in septal peptidoglycan synthesis, are then added
together to the ring, which is completed by addition of FtsN. The rest of the rows describe the functionality and localization of the ring components. Republished with
permission from [10]. (b) Proto-ring in cells and reconstruction of proto-rings in membrane systems. Diagram illustrating the position of essential divisome elements in the
living cell and reconstruction of proto-rings in biomimetic membranes: nanodiscs, microbeads, supported lipid bilayers, and vesicles. The proto-ring components are
colored; other elements of the divisome are shown in dark grey. The light grey circles in the upper part of the diagram represent macromolecules that contribute to the
crowding of the cytoplasm. Vesicles show the two alternative orientations of the reconstructed proto-ring relative to the vesicle lumen.

635
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

genetically altered (Figure 1a) [10]. The first macromolec-
ular assembly of the divisome is the proto-ring, comprising
three proteins – FtsZ, FtsA, and ZipA – that associate at
the cytoplasmic membrane forming the scaffold structure
into which the rest of the essential division proteins are
incorporated in a sequential and concerted manner
(Figure 1a).
Central to assembly of the division machinery is FtsZ
(40 kDa), the best-known prokaryotic division protein,
which is widely conserved in bacteria and the ancestor
of eukaryotic tubulin [11,12]. Although the basic unit of
FtsZ is a monomer and the tubulin protomer is a hetero-
dimer, both proteins bind and hydrolyze GTP and form
polymers in a GTP-dependent manner, making FtsZ poly-
mers very plastic and polymorphic. The GTP-dependent
FtsZ assembly/disassembly cycle it is thought to be essen-
tial for Z-ring function [12–14]. Current research on FtsZ is
aimed at describing the polymerization and force-genera-
tion mechanisms and evaluating the roles of nucleotide
exchange and hydrolysis [11,15]. FtsZ is cytosolic and
attaches to the membrane though interactions with FtsA
or ZipA. FtsA (45 kDa) belongs to the actin family and
contains a short amphipathic helix that it is thought to
mediate its association with the membrane [16]. ZipA (35
kDa) contains an amino-terminal domain that is integrat-
ed into the membrane and connected to a cytoplasmic FtsZ-
interacting domain via a flexible linker [17,18]. The
attachment of FtsZ to the membrane requires either ZipA
or FtsA, but in the simultaneous absence of both, no
localization of FtsZ occurs. Gain-of-function mutants of
FtsA, such as FtsA*, can bypass the need for ZipA for
recruitment of downstream proteins to form a division ring
[13,19]. Addition of FtsA* to FtsZ should be sufficient, in
principle, to form an active, membrane-anchored proto-
ring without ZipA. However, there is no evidence suggest-
ing that the behaviors of FtsA* and native FtsA are the
same.
Spatial modulators of divisome assembly – the
oscillating Min CDE system
In E. coli, the positioning of the division ring at mid-cell at
the end of the cell cycle is regulated by two independent
mechanisms, the Min system and nucleoid occlusion, that
cooperate to target the formation of the ring at the right
place [9,14,20,21]. The control exerted by the bacterial
nucleoid (bulk chromosomal DNA), largely due to crowd-
ing/electrostatic effects and the action of the protein SlmA
[22], is nevertheless not able to block the formation of polar
Z-rings, suggesting that the task of restricting septation to
mid-cell might significantly rely on the Min system.
The Min system comprises MinD, E, and C proteins,
which oscillate repeatedly from one pole of the cell to the
other (Figures 2 and 3) [23,24]. A description of the bio-
chemical and molecular details that account for the inter-
change between membrane-bound and cytoplasmic
locations of the Min proteins are reviewed elsewhere
[25,26]. Briefly, MinC inhibits FtsZ assembly [24,27,28].
MinD, a membrane-bound ATPase member of the Walker

(a) (b) MinE-Cy5 MinD-Cy3

70 µm

(c) (d)

+ATP /-ADP 1
Rear Middle Front
intensity (AU)
Normalized

2
3
4b
5 –P
4a

MinD
MinE
Key: MinD-ADP MinC MinC
MinD-ATP MinE

TRENDS in Cell Biology

Figure 2. Min oscillating system. (a) MinE fluorescence (red) is observed as it pushes the green fluorescent signal of MinD (yellow). Scheme from [78]. (b) Confocal
fluorescence micrographs showing spiral waves of MinD and E in the presence of ATP on a supported lipid bilayer in vitro. (c) Simplified scheme of the biochemical
reactions underlying the oscillating behaviour. ATP-bound MinD binds to the membrane as dimers. Both MinC and MinE can bind to membrane-bound MinD, but MinE is
able to displace MinC from MinD. MinE stimulates the ATPase activity of MinD and, on hydrolysis of ATP, MinD and MinE are released from the membrane [26]. (d)
Normalized fluorescence intensity profiles of Min D, E, and C corresponding to waves equivalent to the ones shown in (b). Intensity profiles reveal the relative position of
the Min proteins within the wave. Black arrows indicate the traveling direction of the protein waves.

636
 Review
(a)
(b)
10 min
173 µm
0 µm
20 µm
Figure 3. MinD/E waves propagating on the outer surface of giant unilamellar vesicles (GUVs). (a) Sequential confocal images of the equatorial plane of a GUV in which
MinD/E oscillations are observed on its outer surface are shown (images starting on the upper left toward the right side of the panel; taken every 5 s). The waves appear as
an alternate increase and decrease in the fluorescence intensity of Cy5-labeled MinE. (b) Kymograph of the GUV shown in (a). The rim of the vesicle is analyzed as a straight
line and its fluorescence intensity represented over time. This reconstruction closely resembles the kymographs of the waves previously observed in supported lipid
bilayers.
A cytoskeletal ATPase (WACA) family, powers the oscilla-
tion by ATP consumption [25,29,30]. It activates MinC’s
inhibitory activity and recruits it to the membrane, while
MinE activates MinD’s ATPase to detach MinD from the
membrane and drive the oscillation [31]. Their combined
action results in FtsZ assembly into a ring at mid-cell,
where the local time-averaged concentration of MinC in-
hibitor is the lowest. Fusions of the Min proteins with
fluorescent proteins allow following of these pole-to-pole
oscillations in vivo. In this way, another prominent feature
of the Min system can be observed, the so called ‘E-ring’,
comprising a sharply defined band of MinE that travels
behind MinD and MinC [23].
The use of biomimetic membranes to reconstruct
minimal divisome assemblies
One major advantage of in vitro reconstitution experi-
ments in biomimetic membranes is that both biochemical
and biophysical parameters can be controlled precisely,
providing an opportunity to investigate the complex spa-
tial and temporal dynamics of membrane-based processes
under defined conditions (Figure 1b and Box 1) [32–34].
Supported membrane structural organization (e.g., the
presence of membrane microdomains) can be studied in
detail by fluorescence microscopy, but are also perfectly
suited to be studied by surface probe techniques such as
atomic force microscopy (AFM), to yield nanoscopic resolu-
tion. Employing total internal reflection fluorescence
(TIRF) as an excitation scheme, the dynamics of single
lipid and protein molecules in membranes can be visual-
ized and dynamic interactions of soluble proteins with
membranes can be resolved [35]. Another model mem-
brane system useful in reconstitution experiments is giant
unilamellar vesicles (GUVs), which are closed lipid
bilayers with diameters well above the optical resolution
limit, from several to hundreds of micrometers. Due to
their large size, they are perfectly suited to investigations
Trends in Cell Biology December 2012, Vol. 22, No. 12
MinE-Cy5
173 µm
TRENDS in Cell Biology
by optical microscopy, and the behavior of macromolecules
(divisome proteins) trapped in the GUV interior can be
studied by microspectroscopy techniques [36]. Finally,
coated microbeads, comprising lipid bilayers deposited
on solid, spherical surfaces, are a useful novel membrane
model to reconstitute macromolecular assemblies. They
offer a robust procedure to generate lipid surfaces with
uniform curvatures and cell-like sizes, in which both lipid
composition and the material of the bead can be modified to
obtain the desired properties of the system [3,32].
Nanodiscs, formed by a membrane scaffold protein
encircling a phospholipid mixture [37], are a promising
novel tool to study assembly on a lipidic surface while the
system under study remains soluble. When a target mem-
brane protein is included in the mixture, each nanodisc
self-assembles and incorporates a single target molecule,
preserving its natural structural and functional properties.
Targets may be single proteins (e.g., ZipA) or even a protein
complex (e.g., ZipA or FtsA plus FtsZ). Due to their small
size, nanodiscs remain soluble, allowing the use of solution
biochemical and biophysical tools [38,39] to characterize
the embedded protein and to monitor and measure its
interactions with other proteins.
Reconstruction of a minimal divisome in the test tube,
step-by-step
For successful reconstitution of bacterial cell division, two
essential phenomena observed in intact cells need to be
achieved by a minimal protein machinery: (i) the assembly
of the divisome components at discrete regions along the
cell length (this occurs usually at mid-cell in normal bac-
terial division, but might vary in mutants such as those
defective in the Min system); and (ii) the function of the
force-generating elements that initiate constriction of the
membrane. We will first discuss approaches toward recon-
stituting the Min positioning system and then touch on the
perspectives of realizing a contractile proto-ring.
637
Review
Box 1. The membrane as an element of the divisome
In the bacterial cell, it is likely that membrane component
redistributions during cell division may create lipid domains [79]
and regions highly concentrated in specific components that can
trigger membrane deformations (favouring curved structures)
[80,81] that might contribute to the accurate selection of the division
site and to the force generation that drives constriction. In addition,
membrane fluidity might also modulate bacterial division, as
suggested by recent observations that increasing fluidity favours
cytokinesis of L forms [82] and FtsZ polymers modify the
viscoelastic properties of E. coli lipid monolayers to become more
fluid and plastic [83,84]. The composition of a bilayer defines its
mechanical properties [85]. Three major lipids constitute the inner
membrane of E. coli: phosphatidylethanolamine (74%), phosphati-
dylglycerol (19%), and cardiolipin (3%). This mixture confers a net
negative charge to the membrane at physiological pH of 7.5. The
relevance of the lipid composition takes place at two levels,
modulating the mechanical properties of the membrane and
offering specificity as recruiting factors. First, enriched domains on
cardiolipin were detected at both the cell poles and the septal region
of dividing E. coli cells [86]. This conical phospholipid tends to form
high-curvature rather than bilayer structures, indicating that its
accumulation might provoke mechanical destabilization of the
bilayer thus facilitating morphological transformations of the
membrane. Second, its anionic character, together with its marked
distribution in key areas of the membrane during cell division, make
cardiolipin a suitable target for amphitropic proteins such as FtsA
[16]. These differences in the distribution of cardiolipin in the
membrane along the cell suggest an active role of this lipid in cell
division. In addition, MinD has been shown to bind preferentially to
domains of anionic phospholipids on the inner membrane [87].
Related to this, it has been reported that membrane potential might
be relevant for the binding of MinD and FtsA to the membrane [88].
The high protein content present on the inner membrane of E.
coli, where the lipid/protein ratio is 0.4, is notable [89]. This is likely
to have a major effect on the physical properties of the membrane,
because imposition of physical constraints on its plasticity would
make it less prone to change. In summary, the restoration of the
membrane can also be considered a strategy towards the recon-
struction of bacterial division. Accordingly, the manipulation of
membrane composition in membrane models would contribute to
revealing its relationship with the constriction process.
Reconstituting Min oscillations in vitro
The Min system has been described as a reaction–diffusion
system in which, starting from a homogeneous distribu-
tion, the oscillatory patterns emerge by self-organization
on dynamic instability powered by the hydrolysis of ATP.
In fact, these dynamics largely depend on collective mem-
brane binding and unbinding. Thus, a crucial step toward a
minimal system for Min oscillations was achieved by the
reconstruction of in vivo cycling behaviour on artificial
bilayers mimicking the inner membrane of E. coli
[26,40]. In the presence of ATP, protein waves emerged
on E. coli lipid bilayers on incubation of purified and
fluorescently labelled Min proteins at physiological ratios
[41,42] (Figure 2). In the absence of spatial boundaries,
these waves propagated in arbitrary directions across the
bilayer. Using confocal and single molecule TIRF micros-
copy, all three proteins were simultaneously imaged and
the order of events during wave propagation could be
determined (Figure 2).
Furthermore, this in vitro system reproduces some
relevant features of the in vivo oscillations [40], such as
a similar speed, about 0.5 mm/s, which depends on the
concentrations of MinD and MinE [26]. In addition, the
obtained distribution profiles are compatible with the
638
Trends in Cell Biology December 2012, Vol. 22, No. 12
dynamic behaviour displayed in vivo, such that MinE
accumulates following MinD and MinC toward the rear
of the wave, resembling the E-ring in cells. It was con-
firmed that MinC is not required for the cycling, but
assumes the role of a passenger. Interestingly, the wave-
length of the in vitro pattern is 10 times larger than that in
an E. coli cell, which has been proposed to be explained by
crowding effects or higher reaction rates [26], although
neither has been confirmed experimentally yet.
According to single-molecule imaging of Min dynamics
[26], the patterns emerge from an isotropically distributed
carpet of MinD on the membrane [40]. A difference in the
membrane residence times between MinD and MinE, with
MinE being attached longer than MinD, originates small
fluctuations in the MinD/MinE relative ratio, such that
accumulation of MinE initiates a dominant wave of protein
detachment, leading to faster detachment of MinD toward
the end of the wave. Fluorescence resonance energy trans-
fer (FRET) experiments on these supported bilayers con-
firmed that MinE binds to MinD as a dimer [40].
Regarding a plausible explanation for the formation of
the E-ring in vivo, evidence from several experiments
points in favor of persistent binding of MinE [43–45] rather
than formation of higher-order complexes [46]. The differ-
ence in residence times between MinD and MinE, together
with a decrease in the diffusion coefficient of MinE by a
factor of two in the back of the wave, may explain the
accumulation of MinE in the rear and therefore the E-ring
in vivo [40]. Fluorescence recovery after photobleaching
(FRAP) experiments demonstrated that incorporation of
MinE starts from the front of the wave, further corrobora-
ting accumulation – in contrast to cooperative binding – of
MinE [26]. Experiments using MinE C1, a mutant unable
to bind to the membrane, revealed that transient mem-
brane binding of MinE could be partially involved in the
reported persistent binding [43,47,48]. Unlike wild-type
MinE, the fluorescence intensity profiles of the much less
regular waves based on MinE C1 do not show its accumu-
lation at the rear, nor displacement of MinC from mem-
brane-bound MinD [40].
The mechanistic picture that emerges from the data
obtained by reconstruction of a minimal Min system is
summarized in Figure 2. Recently, it has been verified that
the Min patterns can be equally well established on vesicle
membranes (Figure 3). The obvious next step is now to
encapsulate the Min protein dynamics into a closed com-
partment and convert the propagating waves – which do
not allow directing Z-ring assembly to a specific site – to
true oscillations with a clear minimum in Z-ring assembly
inhibition.
Minimal bacterial proto-rings in vesicles
The activity, interactions, and assembly properties of the
proto-ring elements can be quantitatively characterized in
minimal membrane systems structured as nanodiscs, coat-
ed-microbeads, lipid bilayers, and, finally, vesicles
(Figure 1b). The first example of this synthetic approach
[49] demonstrated the assembly of a purified variant of
FtsZ ring inside unstable vesicles that spontaneously
changed in shape over time, suggesting that FtsZ is able
to at least partly constrict the region of the liposome in
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a) (I) (II)

C-terminal MTS N-terminal MTS FtsZ / ZipA

Unilamellar liposomes

3 9 15 21
Mullamellar tubular
vesicles

(b)

TRENDS in Cell Biology

Figure 4. Reconstruction of the contractile FtsZ element. (a.I) Reconstruction of chimeric FtsZ (C-terminal or N-terminal mutants) in unilamellar liposomes and tubular
vesicles. FtsZ is able to narrow the region of the liposome in which it is located. Composed image of liposome contour (from phase contrast microscopy) and fluorescence
of chimeric FtsZ colored in yellow [49]. In unilamellar liposomes, chimeric FtsZ deforms the bilayer according to the location of the mutant construct, suggesting that FtsZ
presents an intrinsic angle of polymerization. In tubular vesicles, FtsZ is able to condense into several rings, narrowing the region in which it is located. Adapted from [49].
(a.II) Membrane tubulation of giant vesicles containing ZipA driven by native FtsZ polymers as demonstrated by confocal microscopy using Alexa 488-labeled FtsZ (A.
Martos et al., unpublished). Scale bars = 10 mm. (b) (Left panel) Fluorescence image of copolymers of YFP-mts-FtsZ and FtsZ assembled on glass substrates having curved
grooves. The profile of the surface is shown on the right. (Right panel) Atomic force microscopy image of filaments assembled on a similar substrate. Red arrows show
filaments on planar surfaces and white arrows show filaments on the curved grooves. Scale bars = 1 mm [57].

which it is located (Figure 4a.I). The authors used an
artificially membrane-targeted FtsZ variant (membrane-
targeting sequence [mts]-FtsZ) to tether it to the liposome
in the absence of physiological membrane-anchoring pro-
teins. The cylindrical geometry of the tubular vesicles
seems to be important, because no Z-rings were found
inside spherical liposomes. The incorporation of mts-FtsZ
polymers into the outside of liposomes produced visible
deformations (liposome tubulation) [50], an observation
recently confirmed using a vesicle reconstruction contain-
ing natural FtsZ and ZipA (Figure 4a.II) (A. Martos et al.,
unpublished). FtsZ polymer bending (associated with the
natural curvature of FtsZ) is thought to be a relevant
feature in generating the force that promotes membrane
deformation, because the geometry of the distortion
depends on the FtsZ terminal region at which the mem-
brane-targeting sequence is attached. These experiments
proved that FtsZ polymers alone are capable of generating
force to promote lipid vesicle deformation, although the
biological implications of this force within the (much more
challenging) task of full membrane constriction and cell
division remain unclear [49–51].
The next step towards proto-ring reconstruction has
been achieved by incorporating the native FtsZ form (in-
stead of the mts-FtsZ variant) together with FtsA inside
giant vesicles obtained from bacterial inner membranes
(Figure 5). Being more complex in composition than artifi-
cial lipid vesicles, they maintain a distribution of proteins
and lipids resembling more closely the native membrane,
and contain the essential membrane-anchoring factor ZipA
naturally inserted. FtsZ polymerization alters the spatial
distribution of the proto-ring complexes inside the vesicle,
resulting in dislodging of FtsA from the membrane to
associate with FtsZ fibers in the vesicle lumen, showing
that, under assembly-promoting conditions, the interac-
tion between FtsZ monomers is stronger than the binding
639
 Review
(b)
FtsZ GDP (a) FtsA GDP
FtsZ GTP (c) FtsA GTP (d)
Contrast (e) Superimposed
(f)
TRENDS in Cell Biology
Figure 5. Reconstruction of proto-ring elements in giant unilamellar vesicles
(GUVs). The figure shows the spatial distribution of FtsZ and FtsA inside GUVs
made from inner bacterial membrane. Confocal images of Alexa 488-labeled FtsZ
(green; a and c) and Alexa 647-labeled FtsA (red; b and d), formed in the absence (a
and b) and in the presence (c and d) of GTP. The superposition of (c) and (d) can be
observed in (f). Differential interference contrast image of the giant unilamellar
inner membrane vesicle (e). Scale bars = 10 mm. Republished with permission
from [52].
strength of FtsA to the membrane [52]. These findings
suggest that in addition to the known role in the associa-
tion of FtsZ with the membrane during division, FtsA may
have other roles related to its amphitropic character, which
may involve signalling of proto-ring assembly completion
or the activation of late divisome functions. Interestingly,
FtsA interacts with FtsN [53], a late-assembling protein
essential in septation, because its absence results in de-
stabilization of the already assembled divisome, even af-
fecting the proto-ring elements [54]. These results show
that the interaction between elements of the divisome is far
more complex than what could be deduced from the visual
inspection of the assembly of rings in a living cell. These
approaches can help to compare the in vivo and in vitro
systems to obtain new knowledge on the process of division.
Structural features of minimal proto-rings on supported
and free-standing lipid bilayers
The structural and dynamic organization of FtsZ polymers
has been studied by AFM on mica and on lipid bilayers
containing a soluble variant of ZipA, suggesting that they
are very plastic [55,56]. FtsZ polymers anchored to lipid
640
Trends in Cell Biology December 2012, Vol. 22, No. 12
bilayers through ZipA form dynamic 2D networks that
retain their capacity to dynamically restructure, to frag-
ment, to anneal, and to condense laterally [56]. In a recent
approach [57], microstructured substrates to support
membranes were employed to investigate the curvature-
dependent membrane attachment, and thus curvature
preference, of FtsZ rings. It could be shown that Z-rings
comprise pre-curved FtsZ filaments, which preferentially
orient themselves along curvatures resembling the native
inner diameter of E. coli cells. FtsZ filaments seem to avoid
smaller diameters (resembling constriction stages during
cell division), which can be evidenced by the slanted orien-
tation of filaments in membrane-coated grooves of greater
curvature (Figure 4b). This speaks for the requirement of
additional factors to facilitate constriction during cell divi-
sion and against FtsZ filaments or rings being the sole
force-generation system. However, conclusive direct mea-
surements of forces exerted by FtsZ filaments with or
without additional divisome proteins, similar to motor
protein assays, are still lacking.
Biochemical features of proto-ring proteins in coated
beads and nanodiscs
Lipid- and membrane-coated beads allow measurement of
the binding affinities of soluble proteins to membrane
structures. A recent example is the incorporation of FtsA
in these structures [58], showing that the apparent affinity
of FtsA to the inner membrane is tenfold higher than to the
E. coli lipids (without integral proteins). The reconstitution
of the three proto-ring elements in coated beads is feasible,
because these membrane systems appear to retain their
capability to interact with FtsZ (Monterroso et al., unpub-
lished). Beads encoded with gold/silver sensor particles
before membrane coating are being exploited to design
assays to monitor interactions between proto-ring ele-
ments [59].
The nanodisc technology has recently been exploited to
elucidate biochemical features of elements of the bacterial
division machinery and their interactions under well-de-
fined lipid environments [38]. A single copy of the full-
length ZipA protein from E. coli has been incorporated in
phospholipid bilayer nanodiscs. ZipA in nanodiscs binds
FtsZ oligomers (GDP forms) and polymers (GTP forms)
equally with moderate affinity, which is also similar to the
binding strength of a soluble form of ZipA to FtsZ, suggest-
ing that the transmembrane segment of ZipA has no role in
complex formation. These findings have provided new
insights into the role of ZipA as a passive anchoring device,
supporting the notion that FtsZ tethering to the membrane
has enough plasticity to allow the remodelling of the
architecture of the division ring during constriction.
Outlook: experimental challenges
The bottom-up synthetic approaches described here, to test
the properties of the elements of the division machinery
and their interactions in well-defined and controllable
environments, have helped us to gain new mechanistic
insights into their functions. However, regarding possible
reconstitution of full bacterial cell division in minimal
systems, many more experimental challenges await us.
First, the true origin of a constriction force will have to be
Review
elucidated. A model of action of FtsZ (and the potential
influence of companion division proteins) needs to be com-
pleted to reconcile the FtsZ force-generating properties in
vitro [60] and the FtsZ dynamics in vivo [12], with their
implication on division. Moreover, the mechanical role of
the cytoplasmic membrane (Box 1) and cell wall synthesis
during septum formation are far from being understood.
Second, the correct positioning of the division site by the
Min protein system will have to be verified, by introducing
the Min system as well as the required proto-ring elements
into deformable closed containers, preferably vesicles.
These goals will require the successful introduction of
functional proteins into the vesicles at physiological salt
conditions without disrupting them (which has proved to
be rather complicated at large scales so far) and the
identification of physical preconditions to convert traveling
Min waves into steady oscillations. Progress in current
research toward the optimization of procedures to encap-
sulate protein complexes with high yield inside vesicles in
a controlled manner using permeable membranes and
microfluidic technologies will be crucial [61–64]. Moreover,
lipid chemistry and microsystems technology will probably
be required to establish vesicle size and shape as tunable
parameters in reconstitution experiments. In a recent
study, [65] it was demonstrated that Min waves and pat-
terns already respond quite strongly to the structure of
membranes. In a 2D assay with patterned membranes, it
was found that the waves could be directed by sensing the
dimensions of membrane patches, without any confine-
ment in 3D being required. This raises the interesting
question of whether the inner surface, rather than the
volume of bacterial cells, provides the spatial cue for divi-
sion-related processes.
The nucleoid, a highly charged structure that occupies a
large volume of the intracellular space, may significantly
influence protein interaction networks in the cytoplasm
and at the membrane; for example, it has been recently
shown that large DNA molecules act as an exclusion zone
for actin fibers within a microsphere [66]. Therefore, future
reconstitution approaches may need to use environments
that faithfully reproduce natural crowding and its associ-
ated electrostatic/excluded-volume effects when attempt-
ing to assemble functional divisomes (Box 2). Research
toward a better understanding of the physicochemical
and biochemical properties of highly concentrated hetero-
geneous mixtures of proteins and/or other macromolecules
[67,68] and the production of artificial organelles [69] will
be essential. Along the same lines, the reconstitution of
fortifying shells around membranes, mimicking the pepti-
doglycan layer, appears to be a highly attractive, albeit
complicated, mosaic stone in a full mechanical understand-
ing. However, cell division studies on so-called L forms –
bacterial cells deprived of their cell walls – argue against a
decisive role of this feature in cell division [70]. Their
division, which does not require FtsZ, is far from being
understood.
In addition to these fundamental questions, nearly all of
the subsystems discussed above, such as the Min and Fts
proteins, deserve a more quantitative study of their bio-
chemical properties to better understand and modulate
their functional reactivity. Of particular interest are their
Trends in Cell Biology December 2012, Vol. 22, No. 12
Box 2. Numbers in E. coli cell division
Proto-ring elements
It is estimated that each E. coli cell contains 4000–15 000 FtsZ [90,91],
600–3000 FtsA [90], and 1000–1500 ZipA [90,92] molecules. In the
case of FtsZ, this would correspond to an average concentration of
approximately 3–10 mM (0.1–0.4 mg/ml). Non-essential division
proteins (e.g., Z-ring-stabilizing elements) are estimated to be less
than 5% of FtsZ numbers [93]. Because FtsZ assembly follows a
cooperative mechanism, it would be distributed into soluble and
polymeric forms whose relative abundance would be modulated in
the cell by several parameters, such as protein concentration, other
companion proteins, physiological ligands such as magnesium,
potassium or the ratio of GTP to GDP [51,94], and nonspecific
electrostatic and crowding effects (see below). Taking all these
factors together, the amount of FtsZ that is part of the Z-ring is
estimated not to be over two-thirds of the total cellular FtsZ [12,94],
which is enough to encircle the cell.
Min proteins
The abundance of MinE molecules per cell has been estimated to be
1100–1400; MinD is present at 2000 [42] and MinC at about 400 [41].
The crowded bacterial interior
It is estimated that 20–30% of the bacterial cytoplasm is occupied by
macromolecules, among them 200–300 mg/ml protein, 75–120 mg/
ml RNA, and 11–15 mg/ml DNA [95,96]. These macromolecules
inside the cell seem to be (transiently) clustered in replicative,
structural, and metabolic networks rather than homogeneously
distributed [97]. The abundance of macromolecules in bacteria
causes volume exclusion effects, both in solution and in mem-
branes. These effects substantially increase the effective concentra-
tion and reactivity of the proteins and may modify the extent and
rate of protein–protein and protein–membrane interactions [68,98].
Because crowding leads to phase separation, it may also modify the
spatial arrangement of macromolecular networks involved in
essential processes [99].
specific ways of interacting with the membrane as one key
template for all division-related processes. New and im-
proved biophysical methods need to be introduced to bac-
terial cell biology, to allow for a better understanding of
membrane binding and dissociation, such as super-resolu-
tion and single-molecule imaging, dynamic techniques like
fluorescence correlation spectroscopy, and plasmonic
methods invoking the potential of nano-optics [71,72].
Concluding remarks
To conclude, the strategy discussed in this review together
with complementary bottom-up synthetic approaches
using either cell-free expression systems inside giant vesi-
cles [73–75] or bacteria with simplified genetic content will
provide the best chance for making progress in the syn-
thetic approach toward studying bacterial division. It is
fair to say that, with optical technology entering the sub-
microscopic regime [76], there are many fundamental
discoveries to be made on bacterial systems that take us
a step closer to the phenomenon of life in its minimal
representation. However, the most important steps toward
its fundamental understanding will probably still have to
come through bottom-up reconstitution [77].
Acknowledgments
We thank Professor Miguel Vicente (CNB-CSIC) for useful comments and
especially for sharing with us unpublished material and contributing the
cell cycle definitions in the Glossary. We also thank the anonymous
reviewers for useful remarks. Work in the authors’ laboratories is
supported by Human Frontiers Science Program through grant
641
Review
RGP0050/2010-C102 (to G.R. and P.S.), by the Spanish Government
through grant BIO2011-C03-03 (to G.R.), by the European Commission
though contract HEALTH-F3-2009-223432 (to G.R.), and by the DFG
Leibniz Prize (to P.S.). We apologize to the authors whose work we were
not able to include in this review owing to space limitations.
References
1 Endy, D. (2005) Foundations for engineering biology. Nature 438, 449–
453
2 de Lorenzo, V. and Danchin, A. (2008) Synthetic biology: discovering
new worlds and new words. EMBO Rep. 9, 822–827
3 Liu, A.P. and Fletcher, D.A. (2009) Biology under construction: in vitro
reconstitution of cellular function. Nat. Rev. Mol. Cell Biol. 10, 644–
650
4 Chiarabelli, C. et al. (2012) Approaches to chemical synthetic biology.
FEBS Lett. 586, 2138–2145
5 Schwille, P. and Diez, S. (2009) Synthetic biology of minimal systems.
Crit. Rev. Biochem. Mol. Biol. 44, 223–242
6 Purnick, P.E. and Weiss, R. (2009) The second wave of synthetic
biology: from modules to systems. Nat. Rev. Mol. Cell Biol. 10, 410–422
7 Nandagopal, N. and Elowitz, M.B. (2011) Synthetic biology: integrated
gene circuits. Science 333, 1244–1248
8 Vicente, M. et al. (2006) Septum enlightenment: assembly of bacterial
division proteins. J. Bacteriol. 188, 19–27
9 Wu, L.J. and Errington, J. (2011) Nucleoid occlusion and bacterial cell
division. Nat. Rev. Microbiol. 10, 8–12
10 Vicente, M. and Rico, A.I. (2006) The order of the ring: assembly of
Escherichia coli cell division components. Mol. Microbiol. 61, 5–8
11 Mingorance, J. et al. (2010) Strong FtsZ is with the force: mechanisms
to constrict bacteria. Trends Microbiol. 18, 348–356
12 Erickson, H.P. et al. (2010) FtsZ in bacterial cytokinesis: cytoskeleton
and force generator all in one. Microbiol. Mol. Biol. Rev. 74, 504–
528
13 Pichoff, S. et al. (2012) FtsA mutants impaired for self-interaction
bypass ZipA suggesting a model in which FtsA’s self-interaction
competes with its ability to recruit downstream division proteins.
Mol. Microbiol. 83, 151–167
14 Adams, D.W. and Errington, J. (2009) Bacterial cell division: assembly,
maintenance and disassembly of the Z ring. Nat. Rev. Microbiol. 7, 642–
653
15 Salvarelli, E. et al. (2011) Independence between GTPase active sites in
the Escherichia coli cell division protein FtsZ. FEBS Lett. 585, 3880–
3883
16 Pichoff, S. and Lutkenhaus, J. (2005) Tethering the Z ring to the
membrane through a conserved membrane targeting sequence in
FtsA. Mol. Microbiol. 55, 1722–1734
17 Hale, C.A. and de Boer, P.A. (1997) Direct binding of FtsZ to ZipA, an
essential component of the septal ring structure that mediates cell
division in E. coli. Cell 88, 175–185
18 Ohashi, T. et al. (2002) Structural evidence that the P/Q domain of ZipA
is an unstructured, flexible tether between the membrane and the C-
terminal FtsZ-binding domain. J. Bacteriol. 184, 4313–4315
19 Geissler, B. et al. (2007) The ftsA* gain-of-function allele of Escherichia
coli and its effects on the stability and dynamics of the Z ring.
Microbiology 153, 814–825
20 Vats, P. et al. (2009) The dynamic nature of the bacterial cytoskeleton.
Cell. Mol. Life Sci. 66, 3353–3362
21 Lutkenhaus, J. (2007) Assembly dynamics of the bacterial MinCDE
system and spatial regulation of the Z ring. Annu. Rev. Biochem. 76,
539–562
22 Bernhardt, T.G. and de Boer, P.A. (2005) SlmA, a nucleoid-associated,
FtsZ binding protein required for blocking septal ring assembly over
chromosomes in E. coli. Mol. Cell 18, 555–564
23 Raskin, D.M. and de Boer, P.A. (1999) Rapid pole-to-pole oscillation of a
protein required for directing division to the middle of Escherichia coli.
Proc. Natl. Acad. Sci. U.S.A. 96, 4971–4976
24 Raskin, D.M. and de Boer, P.A. (1999) MinDE-dependent pole-to-pole
oscillation of division inhibitor MinC in Escherichia coli. J. Bacteriol.
181, 6419–6424
25 Lutkenhaus, J. (2008) Min oscillation in bacteria. Adv. Exp. Med. Biol.
641, 49–61
26 Loose, M. et al. (2011) Protein self-organization: lessons from the Min
system. Annu. Rev. Biophys. 40, 315–336
642
Trends in Cell Biology December 2012, Vol. 22, No. 12
27 Hu, Z. et al. (1999) The MinC component of the division site selection
system in Escherichia coli interacts with FtsZ to prevent
polymerization. Proc. Natl. Acad. Sci. U.S.A. 96, 14819–14824
28 Dajkovic, A. et al. (2008) MinC spatially controls bacterial cytokinesis
by antagonizing the scaffolding function of FtsZ. Curr. Biol. 18, 235–
244
29 Raskin, D.M. and de Boer, P.A. (1997) The MinE ring: an FtsZ-
independent cell structure required for selection of the correct
division site in E. coli. Cell 91, 685–694
30 Zhou, H. and Lutkenhaus, J. (2003) Membrane binding by MinD
involves insertion of hydrophobic residues within the C-terminal
amphipathic helix into the bilayer. J. Bacteriol. 185, 4326–4335
31 Hu, Z. and Lutkenhaus, J. (2001) Topological regulation of cell division
in E. coli spatiotemporal oscillation of MinD requires stimulation of its
ATPase by MinE and phospholipid. Mol. Cell 7, 1337–1343
32 Arumugam, S. et al. (2011) Protein-membrane interactions: the virtue
of minimal systems in systems biology. Wiley Interdiscip. Rev. Syst.
Biol. Med. 3, 269–280
33 Loose, M. and Schwille, P. (2009) Biomimetic membrane systems to
study cellular organization. J. Struct. Biol. 168, 143–151
34 Vogel, S.K. and Schwille, P. (2012) Minimal systems to study
membrane-cytoskeleton interactions. Curr. Opin. Biotechnol. 23,
758–765
35 Garcia-Saez, A.J. and Schwille, P. (2010) Surface analysis of membrane
dynamics. Biochim. Biophys. Acta 1798, 766–776
36 Garcia-Saez, A.J. et al. (2010) Fluorescence correlation spectroscopy for
the study of membrane dynamics and organization in giant
unilamellar vesicles. Methods Mol. Biol. 606, 493–508
37 Ritchie, T.K. et al. (2009) Chapter 11 - Reconstitution of membrane
proteins in phospholipid bilayer nanodiscs. Methods Enzymol. 464,
211–231
38 Hernandez-Rocamora, V.M. et al. (2012) Dynamic interaction of the
Escherichia coli cell division ZipA and FtsZ proteins evidenced in
nanodiscs. J. Biol. Chem. 287, 30097–30104
39 Nath, A. et al. (2010) Single-molecule fluorescence spectroscopy
using phospholipid bilayer nanodiscs. Methods Enzymol. 472, 89–
117
40 Loose, M. et al. (2008) Spatial regulators for bacterial cell division self-
organize into surface waves in vitro. Science 320, 789–792
41 Szeto, T.H. et al. (2001) The dimerization function of MinC resides in a
structurally autonomous C-terminal domain. J. Bacteriol. 183, 6684–
6687
42 Shih, Y.L. et al. (2002) Division site placement in E.coli: mutations that
prevent formation of the MinE ring lead to loss of the normal midcell
arrest of growth of polar MinD membrane domains. EMBO J. 21, 3347–
3357
43 Park, K.T. et al. (2011) The Min oscillator uses MinD-dependent
conformational changes in MinE to spatially regulate cytokinesis.
Cell 146, 396–407
44 Derr, J. et al. (2009) Self-organization of the MinE protein ring in
subcellular Min oscillations. Phys. Rev. E: Stat. Nonlin. Soft Matter
Phys. 80, 011922
45 Meacci, G. and Kruse, K. (2005) Min-oscillations in Escherichia coli
induced by interactions of membrane-bound proteins. Phys. Biol. 2, 89–
97
46 Kang, G.B. et al. (2010) Crystal structure of Helicobacter pylori MinE, a
cell division topological specificity factor. Mol. Microbiol. 76, 1222–
1231
47 Arjunan, S.N. and Tomita, M. (2010) A new multicompartmental
reaction-diffusion modeling method links transient membrane
attachment of E. coli MinE to E-ring formation. Syst. Synth. Biol. 4,
35–53
48 Hsieh, C.W. et al. (2010) Direct MinE-membrane interaction
contributes to the proper localization of MinDE in E. coli. Mol.
Microbiol. 75, 499–512
49 Osawa, M. et al. (2008) Reconstitution of contractile FtsZ rings in
liposomes. Science 320, 792–794
50 Osawa, M. et al. (2009) Curved FtsZ protofilaments generate bending
forces on liposome membranes. EMBO J. 28, 3476–3484
51 Osawa, M. and Erickson, H.P. (2011) Inside-out Z rings–constriction
with and without GTP hydrolysis. Mol. Microbiol. 81, 571–579
52 Jimenez, M. et al. (2011) Reconstitution and organization of
Escherichia coli proto-ring elements (FtsZ and FtsA) inside giant
Review
unilamellar vesicles obtained from bacterial inner membranes. J. Biol.
Chem. 286, 11236–11241
53 Busiek, K.K. et al. (2012) The early divisome protein FtsA interacts
directly through its 1c subdomain with the cytoplasmic domain of the
late divisome protein FtsN. J. Bacteriol. 194, 1989–2000
54 Rico, A.I. et al. (2010) Role of Escherichia coli FtsN protein in the
assembly and stability of the cell division ring. Mol. Microbiol. 76, 760–
771
55 Mateos-Gil, P. et al. (2012) Depolymerization dynamics of individual
filaments of bacterial cytoskeletal protein FtsZ. Proc. Natl. Acad. Sci.
U.S.A. 109, 8133–8138
56 Mateos-Gil, P. et al. (2012) FtsZ polymers bound to lipid bilayers
through ZipA form dynamic two dimensional networks. Biochim.
Biophys. Acta 1818, 806–813
57 Arumugam, S. and Schwille, P. (2012) Surface topology engineering of
membranes for mechanical investigation of tubulin homologue FtsZ.
Angew. Chem. Int. Ed. Engl. 51, 1–6
58 Martos, A. et al. (2012) Isolation, characterization and lipid-binding
properties of the recalcitrant FtsA division protein from Escherichia
coli. PLoS ONE 7, e39829
59 Ahijado-Guzman, R. et al. (2012) Surface-enhanced raman scattering-
based detection of the interactions between the essential cell division
FtsZ protein and bacterial membrane elements. ACS Nano 6, 7514–
7520
60 Erickson, H.P. (2009) Modeling the physics of FtsZ assembly and force
generation. Proc. Natl. Acad. Sci. U.S.A. 106, 9238–9243
61 Abkarian, M. et al. (2011) Continuous droplet interface crossing
encapsulation (cDICE) for high throughput monodisperse vesicle
design. Soft Matter 7, 4610–4614
62 Richmond, D.L. et al. (2011) Forming giant vesicles with controlled
membrane composition, asymmetry, and contents. Proc. Natl. Acad.
Sci. U.S.A. 108, 9431–9436
63 Stachowiak, J.C. et al. (2009) Inkjet formation of unilamellar lipid
vesicles for cell-like encapsulation. Lab Chip 9, 2003–2009
64 Takiguchi, K. et al. (2011) Transformation of actoHMM assembly
confined in cell-sized liposome. Langmuir 27, 11528–11535
65 Schweizer, J. and Schwille, P. (2012) Geometry sensing by self-
organized protein patterns. Proc. Natl. Acad. Sci. U.S.A. 109,
15283–15288
66 Negishi, M. et al. (2010) Cooperation between giant DNA molecules
and actin filaments in a microsphere. Phys. Rev. E Stat. Nonlin. Soft
Matter Phys. 81, 051921
67 Long, M.S. et al. (2005) Dynamic microcompartmentation in synthetic
cells. Proc. Natl. Acad. Sci. U.S.A. 102, 5920–5925
68 Zhou, H.X. et al. (2008) Macromolecular crowding and confinement:
biochemical, biophysical, and potential physiological consequences.
Annu. Rev. Biophys. 37, 375–397
69 Marguet, M. et al. (2012) Polymersomes in ‘‘gelly’’ polymersomes:
toward structural cell mimicry. Langmuir 28, 2035–2043
70 Leaver, M. et al. (2009) Life without a wall or division machine in
Bacillus subtilis. Nature 457, 849–853
71 Baciu, C.L. et al. (2008) Protein-membrane interaction probed by single
plasmonic nanoparticles. Nano Lett. 8, 1724–1728
72 Ament, I. et al. (2012) Single unlabeled protein detection on individual
plasmonic nanoparticles. Nano Lett. 12, 1092–1095
73 Pereira de Souza, T. et al. (2009) The minimal size of liposome-based
model cells brings about a remarkably enhanced entrapment and
protein synthesis. Chembiochem 10, 1056–1063
74 Maeda, Y. et al. (2012) Assembly of MreB filaments on liposome
membranes: a synthetic biology approach. ACS Synth. Biol. 1, 53–59
75 Noireaux, V. et al. (2011) Development of an artificial cell, from self-
organization to computation and self-reproduction. Proc. Natl. Acad.
Sci. U.S.A. 108, 3473–3480
Trends in Cell Biology December 2012, Vol. 22, No. 12
76 Li, Z. et al. (2007) The structure of FtsZ filaments in vivo suggests a
force-generating role in cell division. EMBO J. 26, 4694–4708
77 Alberts, B. (2011) A grand challenge in biology. Science 333, 1200
78 Meinhardt, H. and de Boer, P.A. (2001) Pattern formation in
Escherichia coli: a model for the pole-to-pole oscillations of Min
proteins and the localization of the division site. Proc. Natl. Acad.
Sci. U.S.A. 98, 14202–14207
79 Mileykovskaya, E. and Dowhan, W. (2005) Role of membrane lipids in
bacterial division-site selection. Curr. Opin. Microbiol. 8, 135–142
80 Huang, K.C. et al. (2006) A curvature-mediated mechanism for
localization of lipids to bacterial poles. PLoS Comput. Biol. 2, e151
81 Norris, V. et al. (2004) A hypothesis to explain division site selection in
Escherichia coli by combining nucleoid occlusion and Min. FEBS Lett.
561, 3–10
82 Mercier, R. et al. (2012) Crucial role for membrane fluidity in
proliferation of primitive cells. Cell Rep. 1, 417–423
83 Lopez-Montero, I. et al. (2010) Lipid domains and mechanical plasticity
of Escherichia coli lipid monolayers. Chem. Phys. Lipids 163, 56–63
84 Lopez-Montero, I. et al. (2012) Active membrane viscoelasticity by the
bacterial FtsZ-division protein. Langmuir 28, 4744–4753
85 Janmey, P.A. and Kinnunen, P.K. (2006) Biophysical properties of
lipids and dynamic membranes. Trends Cell Biol. 16, 538–546
86 Mileykovskaya, E. and Dowhan, W. (2000) Visualization of
phospholipid domains in Escherichia coli by using the cardiolipin-
specific fluorescent dye 10-N-nonyl acridine orange. J. Bacteriol.
182, 1172–1175
87 Mazor, S. et al. (2008) Mutual effects of MinD-membrane interaction: I.
Changes in the membrane properties induced by MinD binding.
Biochim. Biophys. Acta 1778, 2496–2504
88 Strahl, H. and Hamoen, L.W. (2010) Membrane potential is important for
bacterial cell division. Proc. Natl. Acad. Sci. U.S.A. 107, 12281–12286
89 Devaux, P.F. and Seigneuret, M. (1985) Specificity of lipid-protein
interactions as determined by spectroscopic techniques. Biochim.
Biophys. Acta 822, 63–125
90 Rueda, S. et al. (2003) Concentration and assembly of the division ring
proteins FtsZ, FtsA, and ZipA during the Escherichia coli cell cycle. J.
Bacteriol. 185, 3344–3351
91 Lu, C. and Erickson, H.P. (1999) The straight and curved conformation
of FtsZ protofilaments-evidence for rapid exchange of GTP into the
curved protofilament. Cell Struct. Funct. 24, 285–290
92 Hale, C.A. and de Boer, P.A. (2002) ZipA is required for recruitment of
FtsK, FtsQ, FtsL, and FtsN to the septal ring in Escherichia coli. J.
Bacteriol. 184, 2552–2556
93 Gueiros-Filho, F.J. and Losick, R. (2002) A widely conserved bacterial
cell division protein that promotes assembly of the tubulin-like protein
FtsZ. Genes Dev. 16, 2544–2556
94 Gonzalez, J.M. et al. (2003) Essential cell division protein FtsZ
assembles into one monomer-thick ribbons under conditions
resembling the crowded intracellular environment. J. Biol. Chem.
278, 37664–37671
95 Zimmerman, S.B. and Trach, S.O. (1991) Estimation of macromolecule
concentrations and excluded volume effects for the cytoplasm of
Escherichia coli. J. Mol. Biol. 222, 599–620
96 Record, M.T., Jr et al. (1998) Biophysical compensation mechanisms
buffering E. coli protein-nucleic acid interactions against changing
environments. Trends Biochem. Sci. 23, 190–194
97 Vendeville, A. et al. (2011) An inventory of the bacterial
macromolecular components and their spatial organization. FEMS
Microbiol. Rev. 35, 395–414
98 Rivas, G. et al. (2004) Life in a crowded world. EMBO Rep. 5, 23–27
99 Walter, H. and Brooks, D.E. (1995) Phase separation, due to
macromolecular crowding, is the basis for microcompartmentation.
FEBS Lett. 361, 135–139
643
Review
Special Issue – Synthetic Cell Biology
Engineered, harnessed, and hijacked:
synthetic uses for cytoskeletal systems
Brian S. Goodman, Nathan D. Derr, and Samara L. Reck-Peterson
Department of Cell Biology, Harvard Medical School, Boston, MA 02115, USA
Synthetic biology re-imagines existing biological systems
by designing and constructing new biological parts,
devices, and systems. In the arena of cytoskeleton-based
transport, synthetic approaches are currently used in two
broad ways. First, molecular motors are harnessed for
non-physiological functions in cells. Second, transport
systems are engineered in vitro to determine the biophys-
ical rules that govern motility. These rules are then ap-
plied to synthetic nanotechnological systems. We review
recent advances in both of these areas and conclude by
discussing future directions in engineering the cytoskele-
ton and its motors for transport.
Harnessing the cytoskeleton
The eukaryotic cytoskeleton is composed of fibers that
exert and respond to force and serve as highways for
motors. In cells this cytoskeleton performs an array of
functions that include organizing subcellular compart-
ments, transporting diverse cargos [1], and producing
unique functional structures such as the mitotic spindle
[2,3] and cilia [4,5]. Although synthetic approaches to
harness and engineer the cytoskeleton are in the early
stages, recent work utilizes the diversity of mechanochem-
ical attributes of both the cytoskeletal motors and their
tracks. In this review we first introduce the key compo-
nents of the cytoskeleton. We then discuss recent endea-
vors to employ these components for novel synthetic
purposes. Next, we highlight how engineering molecular
motors is providing a modular toolbox for synthetic in vitro
transport [6–8]. Finally, we review synthetic transport
systems composed of either cytoskeletal motors that occur
in nature (‘natural motors’), modified natural motors, or
synthetic motors. We conclude by describing promising
future directions in the synthetic biology of transport.
The parts list: tracks, motors, and cargos
Long-distance transport within eukaryotic cells occurs on
two cytoskeletal track systems: actin filaments and micro-
tubules (Box 1, Figure I). Both tracks are polarized, with fast
growing ends (termed ‘plus’ ends) and slow growing ends
(termed ‘minus’ ends). Generally, dynamic actin filaments
are found near the cell periphery, often with their plus ends
polymerizing near the plasma membrane. Microtubules are
usually nucleated from a perinuclear-organizing center and
emanate from this point; thus, microtubule plus ends tend to
Corresponding author: Reck-Peterson, S.L. (reck-peterson@hms.harvard.edu).
Keywords: dynein; kinesin; myosin; actin; microtubule; motor.
644 0962-8924/$ – see front matter ß 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tcb.2012.09.005 Trends in Cell Biology, December 2012, Vol. 22, No. 12
be located in the cell periphery and minus ends near the
nucleus (Box 1, Figure Ia,b). The dynamic growth and
shrinkage of both actin filaments and microtubules is capa-
ble of exerting pushing and pulling forces [9]. Large, highly
polarized cells, such as neurons or epithelial cells, are
especially reliant on the cytoskeleton and cytoskeletal trans-
port for maintaining spatial and temporal localization of
intracellular components [10,11], as evidenced by the grow-
ing list of human diseases resulting from defects in cyto-
skeleton-mediated transport [12].
Most naturally occurring cytoskeletal motors move uni-
directionally along either actin filaments or microtubules.
Dynein and kinesin motors move on microtubules, whereas
myosin motors move on actin filaments (Box 1, Figure Ic,d).
For the purposes of synthetic biology, the most useful motors
may be those that are capable of moving cargo processively
over long distances because these are capable of individually
producing prolonged movements of cargos and/or filaments,
in contrast to non-processive motors which require larger
arrays to produce similar results. Of the approximately 100
genes coding for cytoskeletal motors, at least 20 of their
protein products are capable of moving multiple types of
intracellular cargo over long distances [1]. For microtubules,
these cargo-transporting motors include the minus-end-di-
rected cytoplasmic dyneins-1 and -2 [13,14], and the plus-
end-directed kinesins-1, -2, and -3 [15]. For actin filaments,
the plus-end-directed class V myosins are the best-charac-
terized cargo-transporting motors [16]. Many organisms
have expanded on the types of kinesins and myosins through
gene duplication and divergent evolution. Although some of
these motors do not function as long-distance cargo trans-
porters endogenously, their range of biophysical properties
could be useful for synthetic purposes. For example, there
are classes of kinesins and myosins that move in reverse
compared to most of their other family members: kinesin-
14s are minus-end-directed kinesins [15,17] and class VI
myosins are minus-end-directed myosins [18].
Cargos of motor proteins vary widely in size, shape, and
function, and they include membranous organelles, protein-
aceous signaling molecules, and ribonucleoproteins. For
synthetic biology uses it is instructive to understand how
motors attach to cargo such that those attachment mecha-
nisms can be hijacked or mimicked. Although the molecular
connections linking natural motors to their physiological
cargos are only beginning to be defined, they include both
protein and lipid receptors [19,20]. The filaments them-
selves can also be cargos because motors can slide filaments
Review Trends in Cell Biology December 2012, Vol. 22, No. 12

Box 1. Cellular tracks and motors

(a) In many mammalian cells, actin filaments form a meshwork
throughout the cell and near the cell membrane, whereas micro-
tubules are organized radially and emit outward from a perinuclear
microtubule-organizing center. (b) In mammalian neurons microtu-
bules are similarly polarized in axons, but have mixed polarity in
dendrites. (c) Microtubules are polymers of a- and b-tubulin
heterodimers. They are composed of 13–15 protofilaments (linear
arrays of tubulin dimers) and can be many microns in length. In cells,
they predominately polymerize and depolymerize from their dynamic
plus ends near the cell cortex. The major cargo-transporting
microtubule-based motors are kinesin-1, -2, and -3 (plus-end-directed)
and cytoplasmic dynein (minus-end-directed). Most kinesins and
dyneins are homodimers of motor-containing subunits. Cytoplasmic
dynein contains a number of additional subunits that play a role in
cargo binding. Kinesin-14 is a minus-end-directed kinesin. (d)
Filamentous actin (F-actin) is a polymer of globular actin (G-actin),
and rapid polymerization in cells occurs primarily at the plus end and
depolymerization at the minus end. Myosin-V moves along actin
filaments toward the plus end, whereas myosin-VI moves toward the
minus end.

(a) (b)
Plus Dendrite
Minus end
end

Acn Minus end

Microtubules

Plus end

Axon
(c)
Microtubule-based transport
Kinesin-1 Cytoplasmic dynein

Kinesin-14
Plus
end
Minus
end

Microtubule

(d)
Acn-based transport

Myosin-V Myosin-VI
Minus Plus
end end

Filamentous acn
TRENDS in Cell Biology

Figure I. Cellular tracks and motors.

with respect to one another or relative to a fixed position in
the cell [21]. Thus, motors and the dynamic tracks they
propel are versatile building blocks for engineering synthet-
ic systems.
Pathogens: model hijackers of cytoskeletal systems
Pathogenic organisms have already evolved multiple
mechanisms for harnessing the power of the cytoskeleton.
Here we discuss examples wherein viruses or bacteria
645
Review
co-opt the microtubule or actin cytoskeleton of infected
cells to propagate their pathogenic life cycle. Understand-
ing these mechanisms can provide synthetic biologists with
inspiration for re-engineering the cytoskeleton.
Many viruses hijack the microtubule cytoskeleton at
two major stages of their life cycle. Invading viruses use
dynein to reach the nucleus, where they replicate. Later in
the infection kinesin is used to reach the cell membrane,
where viral budding and exit occurs [22]. Considerable
progress has been made in identifying the specific virus–
motor interactions required for these transport events,
creating a useful resource for synthetic biologists. Inter-
estingly, viruses have evolved a variety of mechanisms for
recruiting dynein or kinesin [23]. For example, dynein is
recruited by adenovirus via interactions between the dy-
nein light intermediate and intermediate chains and a
viral capsid subunit termed hexon [24]. This interaction
is pH-sensitive, providing an elegant mechanism for en-
suring that hexon-coated adenoviruses only associate with
dynein after entering the low-pH environment of the lyso-
somal compartment. By contrast, herpes simplex virus
(HSV) interacts with dynein via an interaction between
the dynein light chains RP3/TcTex1 and its capsid protein,
VP26 [25]. In the case of kinesin, vaccinia virus binds to
kinesin-1 via an interaction with the kinesin light chain
[26], whereas HSV binds directly to kinesin via a region of
the kinesin heavy chain [27]. These diverse viral binding
and recruiting mechanisms highlight methods by which
motors could be co-opted for other non-physiological func-
tions.
In addition to using the cytoskeletal motors, some
pathogens hijack the polymerization power of the cytoskel-
etal filaments [28]. The best-characterized example of this
is the bacterium Listeria monocytogenes, which recruits the
actin polymerization process to propel itself rapidly
through the cytoplasm at speeds of up to 30 nm/s. When
the pathogen hits the cell cortex, actin polymerization
generates enough force for the host cell to drive itself into
an adjacent cell in a finger-like projection, which is then
phagocytosed by that neighboring cell. This process mimics
non-pathogenic lamellipodia formation in which actin net-
works produce membrane protrusion and ruffling at the
leading edge of a motile cell [29]. Other pathogens such as
Clostridium difficile use toxins to alter microtubule dy-
namics, causing large microtubule-filled projections to em-
anate out from the cell. These projections entangle the
extracellular pathogen, and are thus hypothesized to help
it adhere to the cell [28,30]. These distinct mechanisms
demonstrate the malleability of the cytoskeletal compo-
nents to be subverted for many non-physiological purposes
and can provide insight for designing novel synthetic
systems.
These examples highlight how understanding the mech-
anisms that pathogens use to co-opt the cytoskeleton may
provide inspiration for synthetic designs. Although purely
synthetic attachment methods using standard connection
technologies have been employed recently [31,32], endoge-
nous and virus-based attachments have the added benefits
of being sensitive to endogenous regulation and cellular
conditions. If these mechanisms are better understood,
future synthetic attachment schemes could more closely
646
Trends in Cell Biology December 2012, Vol. 22, No. 12
mimic the dynamic and versatile connections that nature
has provided. Such systems would allow the more complete
integration of artificial transport with endogenous process-
es. In addition, synthetic in vivo cargo systems could be
utilized to function in response to internal cues, a feature
that current attachment systems cannot yet provide.
Harnessing transport systems to perform non-
physiological functions
A major goal of synthetic biology as it relates to transport is
to harness existing cellular transport systems for novel
purposes, such as delivering recombinant DNA to the
nucleus, targeting intracellular organelles to new loca-
tions, and mapping neuronal connectivity.
In a process similar to an invading virus navigating the
cytoplasm to reach the nucleus, effective synthetic delivery
of recombinant DNA to the cell nucleus is a crucial hurdle
to overcome for many applications of gene manipulation. A
goal for synthetic biologists is to design a modular system
capable of bypassing all barriers from cell entry through to
transcription of the recombinant DNA. The first experi-
mental attempts to achieve this linked plasmid DNA to the
microtubule cytoskeleton [33]. Although engineered viral
pathogens are commonly used as vectors, they pose signif-
icant safety risks, making non-viral methods largely pref-
erable [34]. Taking cues from viruses, recent work has
sought to deliver DNA to the nucleus by coupling DNA
to a dynein-recruiting element [35,36]. For example, re-
combinant dynein light chain (LC8) linked to a DNA-
binding domain can interact with plasmid DNA; transfec-
tions in the presence of this reagent are more efficient
compared to common transfection reagents or naked DNA
[35]. Here, a likely barrier to achieving higher transfection
efficiency is the ability of the plasmid DNA–dynein light
chain complex to escape membrane-bound endosomes [37].
A challenge for the future will be to integrate motor
recruitment with other steps required to increase the
efficiency of nuclear delivery, such as the ability to escape
endosomes and import across nuclear pores.
Another promising synthetic application uses actin- or
microtubule-based motors to target existing intracellular
cargo to new locations. This causes spatial sequestering of
cargo, resulting in either blocked or modified functions. Two
recent proofs of concept showed that synthetic recruitment
of motors to peroxisomal organelles could drive organelle
redistribution [31,32]. In the first study [32], opposite polar-
ity microtubule- (cytoplasmic dynein and kinesin-1 or kine-
sin-3) or actin- (myosin-V and myosin-VI) based motors were
recruited to peroxisomes in a drug-dependent manner using
the FKBP–FRB system [38] (Figure 1a). The recruitment of
each type of motor yielded a different peroxisomal localiza-
tion pattern in fibroblasts. In the second study, similar
results were obtained in hippocampal neurons [31]. Here,
axon-specific cargo was redistributed to dendrites by
recruiting dynein via the FKBP–FRB system [31]. This
approach is both rapid and reversible; organelle redistribu-
tion occurs on a timescale of minutes and can be reversed by
drug removal. Finally, the modularity of the FKBP–FRB
system can also be used to drive dynein activation in cells
through motor dimerization [31], which is necessary for
motor processivity [39] (Figure 1b). In future experiments,
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a) RapalogA Cargo

localizaon
FKBP* domain
element
RapalogB

FRB domain Cargo

protein
FKBP domain

(b)

RapalogA

(c)

R
RapalogB

(d)

RapalogA RapalogB

TRENDS in Cell Biology

Figure 1. FKBP/FRB motor recruitment system. (a) The FKBP–FRB system is a drug-induced, protein dimerization system used by several different groups to control motor
properties and attachments to cargo [32,39,47]. A mutant form of FKBP, referred to as FKBP*, homodimerizes in the presence of the small rapamycin analogue (AP20187)
referred to as RapalogA [31]. In addition, heterodimers can be produced using a different mutant FKBP and the protein domain FRB in the presence of the small molecule,
RapalogB (AP21967) [96]. (b) Fusing FKBP* to two dynein monomers can produce an inducible dynein dimer. This construction technique can be used to activate dynein-
based transport by enabling processive stepping in the presence of RapalogA. (c) Fusing FKBP to a cargo of interest through a cargo localization element (e.g. a PEX domain
for peroxisome localization), allows recruitment of a recombinant FRB–motor to the cargo in an inducible fashion upon the addition of RapalogB. (d) The orthogonal nature
of this system provides the ability to combine their activities for novel complex uses. For example, a protein of interest (tan circle) fused to several copies of FKBP* and a
single copy of FKBP could be selectively aggregated and sequestered by the addition of RapalogA, and subsequently delivered to a site of interest by recruiting a
recombinant FRB–motor by the addition of RapalogB. These strategies (panels a–d) were successfully used to study trafficking in neurons; aspects of this figure were
adapted from [31].

the FKBP–FRB system could also allow inducible spatial
control of protein, RNA, or organellar cargo through the
recruitment of endogenous cytoskeletal motors
(Figure 1c,d). Constitutive methods for co-opting motors
have also been developed. By generating a myosin–kinesin
chimera containing the motor domain of a kinesin and the
cargo-binding domain of a myosin, fission yeast cells were
rewired to use microtubules in place of actin filaments [40].
These studies highlight the versatility of motors to move
non-physiological cargo and provide promise that additional
innovative uses for synthetic technology will be discovered
in the future.
Recruitment of motor proteins can also be used as a tool
to harness intracellular sorting mechanisms, allowing
mapping of neural circuits [41]. Many viruses that infect
neurons, such as rabies virus, spread unidirectionally
across synapses from the peripheral nervous system to
the central nervous system. This type of cell-to-cell infec-
tion requires navigation of the microtubule network within
neurons. Recently, it was demonstrated that glycoproteins
from different viruses were sufficient to provide directional
viral particle movement [42]. This was achieved by engi-
neering vesicular stomatitis virus (VSV) with glycopro-
teins from viruses known to move either towards or
647
Review Trends in Cell Biology December 2012, Vol. 22, No. 12

away from the neuron cell body [42]. A current missing link
is to determine how these viral glycoproteins differentially
recruit dynein and kinesin motors, and which specific iso-
forms they use. These experiments demonstrate the feasi-
bility of using synthetic approaches to recruit intracellular
motors, allowing the mapping of complex neural circuits
and perhaps more complex tissue organization in the
future.
Interchangeable parts for molecular transport machines
Although engineering the cytoskeleton in cells for synthet-
ic purposes is still in its infancy, significant work has been
done in minimal systems in vitro. Such experiments have
focused on engineering systems to gain insight into
motor mechanism. These approaches have paved the
way for a toolbox of interchangeable motor parts, allowing
the design of self-assembling, synthetic motile devices

Box 2. Motility assays

In vitro motility assays provide powerful methods for studying or
mimicking the behavior of cellular transport. (a) In a cargo transport
assay crude or purified cellular components are observed to move on
immobilized filaments by observing the motion of the cargo. Motors
are either absorbed to beads nonspecifically such that their orienta-
tion and relative positions on the bead or cargo are not controlled, or
specifically through adaptors such as antibodies or chemical linkers.
(b) In gliding assays, filaments are moved by the action of motors
attached to a coverslip surface. Motors are typically attached to
coverslips through antibody linkages. (c) In a single-molecule motility
assay the filament is immobilized to a coverslip and fluorescently
labeled motors are observed to move along the filament.

(a) In vitro cargo transport assay

Direcon of
dynein movement

- +

(b) Filament gliding assay

Direcon of
microtubule gliding

+ -

Direcon of
dynein stepping

Y Y Y Y Y Y

(c) Single-molecule molity assay

Direcon of
dynein movement

- +

TRENDS in Cell Biology

Figure I. Motility assays.

648
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(termed ‘molecular shuttles’) destined for performing pro-
grammed transport tasks for biomedical and nanomanu-
facturing applications.
Motor proteins have been studied in vitro in both native
and recombinant forms (Box 2, Figure I). Initial single-
molecule experiments with kinesin, myosin, and dynein
relied on genetically engineering these motors by truncat-
ing them to their smallest essential components, and
adding sites for tags to aid in protein expression, purifica-
tion, and labeling (see e.g., [39,43,44]). These tags were
originally intended for attaching fluorophores for visuali-
zation purposes, but recent advances allow them to serve
as attachment sites for linking elements that provide
connectivity [45–49]. This connectivity enables disparate
motor and cargo components to become modular and link-
able to myriad other objects. Such linking can be accom-
plished not only with protein, but also with DNA, arguably
the most versatile connection medium [50].
DNA connections provide not only a guaranteed linking
specificity through unique sequences, but are inexpensive,
easily modifiable, allow tunable affinity, and come with
many enzymatic tools already provided by nature, such as
restriction enzymes, polymerases, and ligases. The engi-
neering properties of DNA have recently been employed to
understand the intramolecular interactions between
domains in kinesin and dynein, both of which are normally
homodimers. For kinesin, DNA was used to connect two
motor-containing monomer subunits at non-native loca-
tions [48]. With dynein, DNA replaced the native dimer-
ization domains such that two motor-containing monomer
subunits were linked to form a functional heterodimer [46].
This approach also enabled orthogonal fluorophore label-
ing of each motor subunit. Similar work using the FRB–
FKBP system [38] made use of the concept of interchange-
able parts to study the mechanism of wild type and mutant
dynein protomers [39,47]. Although these components
have yet to be categorized into truly ‘standard’ parts (such
as has been done for nucleic acid elements [51,52]), they are
interchangeable, with the potential for standardization
(Figure 2). Thus, a future goal for the field is to create a
toolbox of mechanical components that can be applied to
controlling cargo transport.
Initial work demonstrating the utility, promise, and
challenges associated with employing natural motors for
engineering purposes has led to several devices, including
molecular transporters and sorters, and has been reviewed
recently [6–8,53]. Most of these efforts to produce molecu-
lar transport devices have relied on the filament-gliding

Interchangable parts Ideal component properes

Membrane-bound
Funconal elements

organelles
Nucleic
acids
· Flexible cargo type/funcon
· Task specific tools
Protein · Mulple funconal units
complexes

· Diverse aachment chemistries

Plaorm

Adaptor/scaffold · Modular construcon

· Orthogonal aachment sites
· Constuve or inducible

· Programmable motor type

Motors

· Biological or synthec
· Tunable motor funcons
· Recombinant or nave

TRENDS in Cell Biology

Figure 2. Synthetic cargo systems. An ideal synthetic motile system would have three major sets of components: task-specific functional elements (top), a scaffolding
platform (middle), and motor machinery (bottom). The scaffolding platform provides connectivity for all active components of the motile system and should be modular,
allowing multiple orthogonal attachments sites to bind motors and functional elements. Depending on in vitro or in vivo applications, it can be purely synthetic in origin, or
composed of biologically produced material. In addition, the scaffolding platform can be activated or induced as needed. The platform could be driven by motors of
synthetic or biological origin. The motors are recruited to the platform through motor-specific linking chemistries, such as DNA or Rapalogs. Motors are chosen from an
array of possibilities depending on need, where the potential task-specific motor attributes include directionally, velocity, track selection, and exogenous control. Finally,
specific functional elements can be bound to the platform. These might include various cargo payloads or specific tools for particular applications.

649
Review
assay architecture (Box 2, Figure I) with actin or micro-
tubules serving as the cargo scaffold (see e.g., [54,55]).
However, a modular cargo framework based on the cargo
transport assay architecture (Box 2, Figure I) with multi-
ple attachment possibilities allows the construction of
more complicated assemblies with programmable func-
tions. Once a motor or motor component is connected to
a chemical linker, attachments to larger scaffolding struc-
tures are possible. This has recently been performed with
myosin [56,57] and kinesin [58], using either quantum dots
[56] or DNA [57,58] as scaffolds. This work with DNA-
based scaffolds made use of a single DNA double helix;
however, DNA origami [59,60] was recently used to create
a more complicated scaffolding structure containing up to 7
dynein and kinesin motors [61]. Other structure-building
techniques [45] in vitro or in vivo [62] are also possible.
Such systems could lead to the creation of more sophisti-
cated, varied, and modular molecular shuttles.
Molecular shuttles and other sophisticated synthetic
applications of cytoskeletal motors require the ability to
control the basic motile properties of the motors that drive
them. Methods for exogenous control over motors and
motor systems will be required to achieve such applica-
tions. Although progress has been made in understanding
how the motile properties of biologically derived motors are
modulated by the cell [14,15,63], there is a need for greater
mechanistic understanding of this process. Crucial motor
parameters that require innovative methods of control
include regulation and control of cargo attachment, load-
ing, and unloading; directionality; processivity; bidirec-
tional versus unidirectional movement; and reuse or
recycling of motors. Recent work to address these issues
using synthetic approaches include protein engineering to
affect directionality [64,65], processivity [66–68], and di-
merization [66]; chemical control over directionality and
processivity [69–71]; motor copy-number effects on direc-
tionality [72]; and photo-based control over motor function
[73]. Of particular note is a study that engineered a
switchable, bidirectional myosin motor [64].
Designing synthetic motors de novo
As a complement to the biologically evolved cytoskeletal
motors, recent work has focused on the creation of synthet-
ic motors capable of motion along a track. Although notable
progress has been made in this area [74], artificial motors
have yet to reach the capabilities of naturally occurring
motors, particularly with regard to velocity and processiv-
ity. Nevertheless, there are many advantages to the de
novo design and fabrication of synthetic motors – attempts
to create a functional motor could allow highly specific
tasks to be performed and lead to a better understanding of
the mechanistic details of naturally occurring motors.
Many fundamental considerations regarding how a syn-
thetic motor might operate correlate directly with our
attempts to understand better the existing natural cytoskel-
etal motors. For example, how is directionality programmed
into the structures and functions of the motors and their
tracks? What enables the motors to operate processively?
What is the source of fuel and how is it used? Given this fuel,
what are the energy and entropic considerations that allow
work to be done? Can multiple motors work together to
650
Trends in Cell Biology December 2012, Vol. 22, No. 12
achieve higher functions? These questions should help drive
the design principles when creating synthetic motors de
novo; however, purely synthetic systems also offer design
possibilities beyond what is available from the biologically
derived tools. For example, the structural organization of
the motor could depart from cytoskeletal models and be
multimeric, with many discrete track and cargo-binding
domains. Track and motor design offers the potential to
create many track–motor combinations or dynamically pro-
grammable routes through a network of tracks. In addition,
motor tasks could be algorithmically determined depending
on the conditions and instruction sets contained within the
system (see e.g., [75,76]).
Given the versatility of DNA as a nanotechnological
platform [50,77–79], it has become a primary construction
material for synthetic molecular motors. These synthetic
motors tend to be significantly smaller than cytoskeletal
motors and have focused on design issues of autonomy,
processivity, velocity, directionality, track selection, and
visualization. Some early synthetic DNA motors are termed
‘clocked’ because they required the sequential addition of
fuel strands to propagate motility (see e.g., [80,81]). Howev-
er, other designs allow for continuous autonomous motion
[82–88] and in some cases with coordinated stepping
[84,88,89]. Some of these motors consist solely of DNA
[84,89]; however, others make use of complementary com-
ponents such as DNAzymes with an RNA substrate [82,84]
or ligating [83] and nicking [83,85] enzymes. In these sys-
tems, directionality can be controlled by loading the motor
onto the track at a specific location [82,85] or by the use of
polar tracks [83,84]. In addition, instructions contained
within the motor’s fuel itself [88–90] can imbue directionali-
ty, providing a directionality mechanism distinct from those
of cytoskeletal motors. In contrast to cytoskeletal motor
systems, in some cases the track is modified or destroyed
as the synthetic motor moves along it [82,84–87].
Recently, route-based selection motifs included in the
track design have allowed motors to navigate particular
paths by following instructions in solution or intrinsic
parts of the motor [90,91]. Specific construction tasks have
also been demonstrated; an autonomous motor created
sequential peptide bonds between building blocks analo-
gously to a ribosome [92], and a clocked motor performed a
nanomanufacturing task by assembling clusters of gold
nanoparticles along an assembly line [93]. Furthermore, a
DNA motor visualized in real time was found to take
continuous, discrete steps [85], as was recently observed
for myosin-V by atomic force microscopy [94]. The visuali-
zation of single synthetic motor steps has also been
achieved using single-molecule motility assays (Box 2,
Figure I) [82,95]. Future designs are likely to incorporate
hybrid approaches that make use of both de novo and
natural motor elements. By building modular components,
new designs and functions could be rapidly created and
tested. One promising possibility is to use biological parts
as the core machinery, with synthetic adapters that are
able to sense and interact with their environment.
Concluding remarks
Nature has provided excellent nanoscale transporters, and
our understanding of how cells control them is rapidly
Review
Box 3. Outstanding questions
The cytoskeleton has proved to be an amenable system for
engineering novel synthetic functions. Some outstanding chal-
lenges in the field include the following questions:
 Can the cytoskeleton and motor proteins be used to create
additional novel functionality? This is of particular interest in
large, polarized cells where cargo sorting heavily relies on
transport mechanisms.
 Can the cytoskeleton be reorganized for novel transport require-
ments? For example, can actin or microtubules be nucleated to
grow from new sites?
 Can motors be co-opted to enhance the efficiency of synthetic in
vivo processes already developed?
 Can a synthetic toolbox be built to recruit motors to any cargo of
interest with standardized parts?
 Can motors be exogenously controlled to produce molecular
shuttles that start, stop, and change direction based on outside
input?
 Can the cytoskeleton be harnessed for efficient directed transport
of foreign entities to the nucleus?
 Can motors be designed de novo to have the efficiency, velocity,
and processivity of biological motors? Can these synthetic motors
be imbued with greater functionality than biological motors?
increasing. The regulatory control issues cells face with
molecular motors are similar to those that synthetic biol-
ogists encounter in harnessing and reengineering these
motors for novel tasks. Similarly, these engineering pro-
blems overlap with areas of active research in cell biology.
The complementary approaches of working on solutions to
these design problems, and understanding how the cell has
solved these same issues, will enrich the utility and un-
derstanding of molecular motors.
Although in the early stages of research, the ease of
building in vitro transport systems shows that the core
motor components are amenable to modification and in-
clusion within modular engineered systems. These new
constructs will continue to increase in both standardiza-
tion and complexity, and in the future will be used to probe
cellular functions and provide novel synthetic applications.
Purely synthetic motors will continue to inform how en-
dogenous motors function. The programmability of artifi-
cial motors will expand their role in synthetic
nanomanufacturing, possibly outpacing the use of natural
motors.
Despite this exciting progress, there are a number of
outstanding questions in the field and challenges for the
future (Box 3). The constant feedback between synthetic
and cellular biologists promises to allow this nascent field
to uncover the mechanisms of cellular motility and engi-
neer novel transport systems.
Acknowledgments
We thank J. Tytell, X. Su, and S. Wickham for comments on the
manuscript.
References
1 Vale, R.D. (2003) The molecular motor toolbox for intracellular
transport. Cell 112, 467–480
2 Gadde, S. and Heald, R. (2004) Mechanisms and molecules of the
mitotic spindle. Curr. Biol. 14, R797–R805
3 Wittmann, T. et al. (2001) The spindle: a dynamic assembly of
microtubules and motors. Nat. Cell Biol. 3, E28–E34
4 Gibbons, I.R. (1996) The role of dynein in microtubule-based motility.
Cell Struct. Funct. 21, 331–342
Trends in Cell Biology December 2012, Vol. 22, No. 12
5 Fisch, C. and Dupuis-Williams, P. (2011) Ultrastructure of cilia and
flagella – back to the future! Biol. Cell 103, 249–270
6 Goel, A. and Vogel, V. (2008) Harnessing biological motors to engineer
systems for nanoscale transport and assembly. Nat. Nanotechnol. 3,
465–475
7 van den Heuvel, M.G.L. and Dekker, C. (2007) Motor proteins at work
for nanotechnology. Science 317, 333–336
8 Hess, H. (2011) Engineering applications of biomolecular motors.
Annu. Rev. Biomed. Eng. 13, 429–450
9 McIntosh, J.R. et al. (2010) Tubulin depolymerization may be an
ancient biological motor. J. Cell Sci. 123, 3425–3434
10 Conde, C. and Cáceres, A. (2009) Microtubule assembly, organization
and dynamics in axons and dendrites. Nat. Rev. Neurosci. 10, 319–332
11 Müsch, A. (2004) Microtubule organization and function in epithelial
cells. Traffic 5, 1–9
12 Hirokawa, N. et al. (2010) Molecular motors in neurons: transport
mechanisms and roles in brain function, development, and disease.
Neuron 68, 610–638
13 Scholey, J.M. (2008) Intraflagellar transport motors in cilia: moving
along the cell’s antenna. J. Cell Biol. 180, 23–29
14 Kardon, J.R. and Vale, R.D. (2009) Regulators of the cytoplasmic
dynein motor. Nat. Rev. Mol. Cell Biol. 10, 854–865
15 Verhey, K.J. and Hammond, J.W. (2009) Traffic control: regulation of
kinesin motors. Nat. Rev. Mol. Cell Biol. 10, 765–777
16 Hammer, J.A. and Sellers, J.R. (2012) Walking to work: roles for class V
myosins as cargo transporters. Nat. Rev. Mol. Cell Biol. 13, 13–26
17 Walker, R.A. et al. (1990) The Drosophila claret segregation protein is a
minus-end directed motor molecule. Nature 347, 780–782
18 Wells, A.L. et al. (1999) Myosin VI is an actin-based motor that moves
backwards. Nature 401, 505–508
19 Akhmanova, A. and Hammer, J.A. (2010) Linking molecular motors to
membrane cargo. Curr. Opin. Cell Biol. 22, 479–487
20 Kamal, A. and Goldstein, L.S.B. (2002) Principles of cargo attachment
to cytoplasmic motor proteins. Curr. Opin. Cell Biol. 14, 63–68
21 Kulic, I.M. et al. (2008) The role of microtubule movement in
bidirectional organelle transport. Proc. Natl. Acad. Sci. U.S.A. 105,
10011–10016
22 Döhner, K. et al. (2005) Viral stop-and-go along microtubules: taking a
ride with dynein and kinesins. Trends Microbiol. 13, 320–327
23 Dodding, M.P. and Way, M. (2011) Coupling viruses to dynein and
kinesin-1. EMBO J. 30, 3527–3539
24 Bremner, K.H. et al. (2009) Adenovirus transport via direct interaction
of cytoplasmic dynein with the viral capsid hexon subunit. Cell Host
Microbe 6, 523–535
25 Douglas, M.W. et al. (2004) Herpes simplex virus type 1 capsid
protein VP26 interacts with dynein light chains RP3 and Tctex1 and
plays a role in retrograde cellular transport. J. Biol. Chem. 279,
28522–28530
26 Ward, B.M. and Moss, B. (2004) Vaccinia virus A36R membrane
protein provides a direct link between intracellular enveloped
virions and the microtubule motor kinesin. J. Virol. 78, 2486–2493
27 Diefenbach, R.J. et al. (2002) Herpes simplex virus tegument protein
US11 interacts with conventional kinesin heavy chain. J. Virol. 76,
3282–3291
28 Haglund, C.M. and Welch, M.D. (2011) Pathogens and polymers:
microbe-host interactions illuminate the cytoskeleton. J. Cell Biol.
195, 7–17
29 Lambrechts, A. et al. (2008) Listeria comet tails: the actin-based
motility machinery at work. Trends Cell Biol. 18, 220–227
30 Schwan, C. et al. (2009) Clostridium difficile toxin CDT induces
formation of microtubule-based protrusions and increases adherence
of bacteria. PLoS Pathog. 5, e1000626
31 Kapitein, L.C. et al. (2010) Mixed microtubules steer dynein-driven
cargo transport into dendrites. Curr. Biol. 20, 290–299
32 Kapitein, L.C. et al. (2010) Probing intracellular motor protein activity
using an inducible cargo trafficking assay. Biophys. J. 99, 2143–2152
33 Lechardeur, D. et al. (2005) Intracellular routing of plasmid DNA
during non-viral gene transfer. Adv. Drug Deliv. Rev. 57, 755–767
34 Edelstein, M.L. et al. (2007) Gene therapy clinical trials worldwide to
2007 – an update. J. Gene Med. 9, 833–842
35 Toledo, M.A.S. et al. (2012) Development of a recombinant fusion
protein based on the dynein light chain LC8 for non-viral gene
delivery. J. Control. Release 159, 222–231
651
Review
36 Bergen, J.M. and Pun, S.H. (2007) Evaluation of an LC8-binding
peptide for the attachment of artificial cargo to dynein. Mol. Pharm.
4, 119–128
37 Miller, A.M. and Dean, D.A. (2009) Tissue-specific and transcription
factor-mediated nuclear entry of DNA. Adv. Drug Deliv. Rev. 61, 603–
613
38 Banaszynski, L. et al. (2005) Characterization of the FKBP.
rapamycin.FRB ternary complex. J. Am. Chem. Soc. 127, 4715–4721
39 Reck-Peterson, S.L. et al. (2006) Single-molecule analysis of dynein
processivity and stepping behavior. Cell 126, 335–348
40 Lo Presti, L. and Martin, S.G. (2011) Shaping fission yeast cells by
rerouting actin-based transport on microtubules. Curr. Biol. 21, 2064–
2069
41 Enquist, L.W. and Card, J.P. (2003) Recent advances in the use of
neurotropic viruses for circuit analysis. Curr. Opin. Neurobiol. 13, 603–
606
42 Beier, K.T. et al. (2011) Anterograde or retrograde transsynaptic
labeling of CNS neurons with vesicular stomatitis virus vectors.
Proc. Natl. Acad. Sci. U.S.A. 108, 15414–15419
43 Warshaw, D.M. et al. (2005) Differential labeling of myosin V heads
with quantum dots allows direct visualization of hand-over-hand
processivity. Biophys. J. 88, L30–L32
44 Rice, S. et al. (1999) A structural change in the kinesin motor protein
that drives motility. Nature 402, 778–784
45 Diehl, M.R. et al. (2006) Engineering cooperativity in biomotor–protein
assemblies. Science 311, 1468–1471
46 Qiu, W. et al. (2012) Dynein achieves processive motion using both
stochastic and coordinated stepping. Nat. Struct. Mol. Biol. 19, 193–200
47 DeWitt, M.A. et al. (2012) Cytoplasmic dynein moves through
uncoordinated stepping of the AAA+ ring domains. Science 335,
221–225
48 Miyazono, Y. et al. (2010) Strain through the neck linker ensures
processive runs: a DNA–kinesin hybrid nanomachine study. EMBO
J. 29, 93–106
49 Guydosh, N.R. and Block, S.M. (2009) Direct observation of the binding
state of the kinesin head to the microtubule. Nature 461, 125–128
50 Michelotti, N. et al. (2012) Beyond DNA origami: the unfolding
prospects of nucleic acid nanotechnology. Wiley Interdiscip. Rev.
Nanomed. Nanobiotechnol. 4, 139–152
51 Shetty, R.P. et al. (2008) Engineering BioBrick vectors from BioBrick
parts. J. Biol. Eng. 2, 5
52 Canton, B. et al. (2008) Refinement and standardization of synthetic
biological parts and devices. Nat. Biotechnol. 26, 787–793
53 Korten, T. et al. (2010) Towards the application of cytoskeletal motor
proteins in molecular detection and diagnostic devices. Curr. Opin.
Biotechnol. 21, 477–488
54 Schmidt, C. et al. (2012) Tuning the ‘roadblock’ effect in kinesin-based
transport. Nano Lett. 12, 3466–3471
55 Malcos, J.L. and Hancock, W.O. (2011) Engineering tubulin:
microtubule functionalization approaches for nanoscale device
applications. Appl. Microbiol. Biotechnol. 90, 1–10
56 Ali, M.Y. et al. (2011) Myosin Va and myosin VI coordinate their steps
while engaged in an in vitro tug of war during cargo transport. Proc.
Natl. Acad. Sci. U.S.A. 108, E535–E541
57 Lu, H. et al. (2012) Collective dynamics of elastically-coupled myosinV
motors. J. Biol. Chem. 287, 27753–27761
58 Rogers, A.R. et al. (2009) Negative interference dominates collective
transport of kinesin motors in the absence of load. Phys. Chem. Chem.
Phys. 11, 4882–4889
59 Shih, W.M. and Lin, C. (2010) Knitting complex weaves with DNA
origami. Curr. Opin. Struct. Biol. 20, 276–282
60 Krishnan, Y. and Bathe, M. (2012) Designer nucleic acids to probe and
program the cell. Trends Cell. Biol. 22, 624–633
61 Derr, N.D. et al. (2012) Tug of war in motor protein ensembles revealed
with a programmable DNA origami scaffold. Science 338, 662–665
62 Delebecque, C.J. et al. (2011) Organization of intracellular reactions
with rationally designed RNA assemblies. Science 333, 470–474
63 Hartman, M.A. et al. (2011) Principles of unconventional myosin
function and targeting. Annu. Rev. Cell Dev. Biol. 27, 133–155
64 Chen, L. et al. (2012) Engineering controllable bidirectional molecular
motors based on myosin. Nat. Nanotechnol. 7, 252–256
652
Trends in Cell Biology December 2012, Vol. 22, No. 12
65 Carter, A.P. et al. (2008) Structure and functional role of dynein’s
microtubule-binding domain. Science 322, 1691–1695
66 Shishido, H. and Maruta, S. (2012) Engineering of a novel Ca2+-
regulated kinesin molecular motor using a calmodulin dimer linker.
Biochem. Biophys. Res. Commun. 423, 386–391
67 Clancy, B.E. et al. (2011) A universal pathway for kinesin stepping.
Nat. Struct. Mol. Biol. 18, 1020–1027
68 Hodges, A.R. et al. (2007) Engineering the processive run length of
Myosin V. J. Biol. Chem. 282, 27192–27197
69 Gerson-Gurwitz, A. et al. (2011) Directionality of individual kinesin-5
Cin8 motors is modulated by loop 8, ionic strength and microtubule
geometry. EMBO J. 30, 4942–4954
70 Cochran, J.C. et al. (2012) A metal switch for controlling the activity of
molecular motor proteins. Nat. Struct. Mol. Biol. 19, 122–127
71 Walter, W.J. et al. (2012) Two independent switches regulate
cytoplasmic dynein’s processivity and directionality. Proc. Natl.
Acad. Sci. U.S.A. 109, 5289–5293
72 Roostalu, J. et al. (2011) Directional switching of the kinesin Cin8
through motor coupling. Science 332, 94–99
73 Yamada, M.D. et al. (2007) Photocontrol of kinesin ATPase activity
using an azobenzene derivative. J. Biochem. 142, 691–698
74 von Delius, M. and Leigh, D.A. (2011) Walking molecules. Chem. Soc.
Rev. 40, 3656–3676
75 Qian, L. et al. (2011) Neural network computation with DNA strand
displacement cascades. Nature 475, 368–372
76 Soloveichik, D. et al. (2010) DNA as a universal substrate for chemical
kinetics. Proc. Natl. Acad. Sci. U.S.A. 107, 5393–5398
77 Seeman, N.C. (2010) Nanomaterials based on DNA. Annu. Rev.
Biochem. 79, 65–87
78 Bath, J. and Turberfield, A.J. (2007) DNA nanomachines. Nat.
Nanotechnol. 2, 275–284
79 Aldaye, F.A. et al. (2008) Assembling materials with DNA as the guide.
Science 321, 1795–1799
80 Sherman, W.B. and Seeman, N.C. (2004) A precisely controlled DNA
biped walking device. Nano Lett. 4, 1203–1207
81 Shin, J-S. and Pierce, N.A. (2004) A synthetic DNA walker for
molecular transport. J. Am. Chem. Soc. 126, 10834–10835
82 Lund, K. et al. (2010) Molecular robots guided by prescriptive
landscapes. Nature 465, 206–210
83 Peng, Y. et al. (2004) A unidirectional DNA walker that moves
autonomously along a track. Angew. Chem. Int. Ed. Engl. 43, 4906–
4911
84 Omabegho, T. et al. (2009) A bipedal DNA Brownian motor with
coordinated legs. Science 324, 67–71
85 Wickham, S.F.J. et al. (2011) Direct observation of stepwise movement
of a synthetic molecular transporter. Nat. Nanotechnol. 6, 166–169
86 Pei, R. et al. (2006) Behavior of polycatalytic assemblies in a substrate-
displaying matrix. J. Am. Chem. Soc. 128, 12693–12699
87 Bath, J. et al. (2005) A free-running DNA motor powered by a nicking
enzyme. Angew. Chem. Int. Ed. Engl. 44, 4358–4361
88 Bath, J. et al. (2009) Mechanism for a directional, processive, and
reversible DNA motor. Small 5, 1513–1516
89 Green, S.J. et al. (2008) Coordinated chemomechanical cycles: a
mechanism for autonomous molecular motion. Phys. Rev. Lett. 101,
238101
90 Wickham, S.F.J. et al. (2012) A DNA-based molecular motor that can
navigate a network of tracks. Nat. Nanotechnol. 7, 169–173
91 Muscat, R.A. et al. (2011) A programmable molecular robot. Nano Lett.
11, 982–987
92 He, Y. and Liu, D.R. (2010) Autonomous multistep organic synthesis in
a single isothermal solution mediated by a DNA walker. Nat.
Nanotechnol. 5, 778–782
93 Gu, H. et al. (2010) A proximity-based programmable DNA nanoscale
assembly line. Nature 465, 202–205
94 Kodera, N. et al. (2010) Video imaging of walking myosin V by high-
speed atomic force microscopy. Nature 468, 72–76
95 Masoud, R. et al. (2012) Studying the structural dynamics of bipedal
DNA motors with single-molecule fluorescence spectroscopy. ACS
Nano 6, 6272–6283
96 Clackson, T. et al. (1998) Redesigning an FKBP–ligand interface to
generate chemical dimerizers with novel specificity. Proc. Natl. Acad.
Sci. U.S.A. 95, 10437–10442
 Review
Special Issue – Synthetic Cell Biology
Principles for designing ordered
protein assemblies
Yen-Ting Lai1, Neil P. King2, and Todd O. Yeates3,4
1
Biomedical Engineering Interdepartmental Degree Program, University of California, Los Angeles, CA, USA
2
Department of Biochemistry, University of Washington, Seattle, WA, USA
3
Department of Chemistry and Biochemistry, University of California, Los Angeles, CA, USA
4
University of California Los Angeles (UCLA)–Department of Energy Institute for Genomics and Proteomics, University of
California, Los Angeles, CA, USA
In nature, many proteins have evolved to have self-
complementary shapes. This drives them to assemble
into supramolecular structures, sometimes of great
complexity, and often carrying out sophisticated cellular
functions. Designing novel proteins that can self-assem-
ble into similarly complex structures is a longstanding
goal in bioengineering. New ideas, combined with con-
tinually improving computer algorithms, are making it
possible to advance on that goal, bringing wide-ranging
applications in synthetic biology within reach. Prospec-
tive applications range from vaccine design to molecular
delivery to bioactive materials. Recent strategies and
examples of successfully designed protein cages, layers,
and crystals are reviewed.
Protein assemblies as a synthetic biology goal
The emerging research area of synthetic biology seeks to
recreate various complex phenomena exhibited by biologi-
cal systems, especially at the molecular level [1]. The
phenomena of interest are often characterized by a high
degree of order – either in time or space. The emergent
behavior or ‘output’ of synthetic systems can be consider-
ably more complex than the behavior of the individual
components [2]. Often, this arises from the introduction
of non-trivial or self-referring interactions between compo-
nents. For example, if rather than simply A affecting B,
instead B also affects C, and C affects A, then an output
behavior may arise where molecular concentrations oscil-
late in time [3]. Likewise, if rather than simply A binding to
B, instead A binds to itself in multiple ways, surprisingly
large and complex molecular assemblies can arise, leading
to spatial organization of various types, such as compart-
mentalization or long-range propagation of forces by rigid
structures. We concern ourselves here with strategies for
designing protein molecules that self-associate to produce
large, complex assemblies with potential synthetic biology
applications.
Diverse efforts in the area of designing protein-based
assemblies and materials can be divided into two groups:
stochastic and deterministic [4,5]. The stochastic group
encompasses several design strategies where the self-
assembling protein molecule is highly flexible [4,6]. When
Corresponding author: Yeates, T.O. (yeates@mbi.ucla.edu).
Keywords: protein design; symmetry; cages; biomaterials; nanomaterials; assembly.
0962-8924/$ – see front matter ß 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tcb.2012.08.004 Trends in Cell Biology, December 2012, Vol. 22, No. 12
a highly flexible molecule self-associates to form a higher-
order assembly, the result is typically an extended and
geometrically irregular material. Such network or mesh-
like materials have interesting bulk properties, which can
sometimes be modulated in useful ways [7]. The second
group, where the assembly behavior is intended to be
deterministic, encompasses those strategies aimed at pro-
ducing specific 3D structures [5]. These structures, which
may be finite in size (e.g., clusters or cages) or indefinite in
extent (e.g., arrays or crystals), can be built with atomic
level features in mind.
Early work on designing geometrically specific protein
assemblies focused on filamentous structures as design
targets. That early focus reflects the relatively simple
design requirements for filamentous assemblies: a single
self-associating interface can produce end-to-end polymer-
ization. Specific well-studied self-associating protein
motifs have been a rich source of building blocks for
designing filamentous structures. Helical coiled-coils have
been especially useful [8,9]. Cyclic polypeptides composed
of b-strand-preferring amino acids have provided another
self-associating motif, in this case leading to rigid tubular
assemblies [10]. Variations on filamentous designs have
sought to produce more complex patterns, such as branch-
ing [11], but the end-to-end polymerization strategies cen-
tral to filament design do not extend easily to the problem
of creating highly specific 3D architectures.
In this review we focus on strategies for designing
proteins that self-assemble to give defined structures with
complex architectures, including cages and extended 2D
and 3D crystalline arrays.
Underlying principles
In nature, wherever supramolecular structures are built
up by the assembly of multiple copies of the same subunit
(or similar subunits), the subunits are nearly always as-
sembled in a symmetrical fashion [12,13]. The reason for
this was anticipated as early as 1956 by Crick and Watson
[14]: symmetric assemblies require fewer distinct kinds of
specific interaction interfaces compared to asymmetric
assemblies. It is natural then that efforts to design ordered
protein assemblies should rely on principles of symmetry.
We articulate three connected ideas that we believe are
important to permit full exploitation of symmetry-based
653
Review
approaches for designing self-assembling molecular sys-
tems:
(i) A symmetric molecular assembly, whether it is a finite
structure or an extended array, is characterized by its
symmetry group. A symmetry group is an exact
mathematical idea that expresses the complete set of
spatial operations that interrelate a set of individual
components, in this case a set of structurally identical
protein subunits in a 3D assembly. For example, a 24-
subunit assembly built with the symmetry of a cube
(referred to as octahedral symmetry) is described by a
symmetry group with 24 spatial operations. Every pair
of subunits is related by some operation in the group.
For chiral biological molecules, such as proteins, these
operations must be rotations, potentially combined
with translations, but excluding reflections.
(ii) Any interface formed between two subunits in a
symmetric assembly corresponds to one of the
operations of the symmetry group (vide supra).
(iii) For a hypothetical, symmetric constellation of sub-
units (whether finite or indefinite) to constitute a
plausible, physically connected assembly – in other
words, to not be two or more disjoint sub-complexes –
the following condition is both necessary and suffi-
cient. Taking any individual subunit as a reference,
the distinct interface types it uses to contact its
neighbors comprise a subset of the operations of the
complete symmetry group, and this subset of opera-
tions must be capable of generating the full symmetry
group. That is, repeated combinations of this typically
small subset of operations must be able to reform the
whole symmetry group. In physical terms, this is
equivalent to stating that it must be possible to trace
a path from any one molecule to any other molecule in
the assembly through the contacts between molecules
in the assembly (Box 1). Otherwise, the collection of
molecules would be disjoint.
The first two ideas are relatively well-known to those
familiar with crystallography or molecular symmetry, but
the third is less obvious. It was first articulated in the
context of crystal symmetries [15], and then in the context
of designed protein assemblies [16]. Particularly surprising,
and important from a design perspective, was the realiza-
tion that large, complex symmetries (including some 3D
crystal symmetries) could be generated using only two
distinct symmetry elements [16]. This translates to the idea
that if one can design two distinct, geometrically specific,
self-associating interfaces into a single molecule, then a
wide range of assembly architectures can be realized. Al-
though more than the minimum number of distinct contact
types can be introduced in a design – and large natural
assemblies such as viral capsids almost always exhibit more
than the minimum number of distinct contact types [17] –
the minimum contact number (which is just two for many
cases) establishes an important design principle.
Design strategies and successes
Owing largely to the complexity of protein molecules, and
our incomplete understanding of the rules by which they
fold and recognize each other, designing complex protein
654
Trends in Cell Biology December 2012, Vol. 22, No. 12
assemblies has been a difficult challenge. However, efforts
along multiple lines are beginning to bear fruit. Recalling
the discussions above on the design requirement of intro-
ducing two (or more) modes of self-association into a single
protein molecule, varied approaches to the problem of
designing self-assembling proteins can be grouped accord-
ing to the strategy used to satisfy this central requirement.
Different strategies rely to different degrees on natural (or
native-like) protein–protein interfaces versus novel inter-
faces created by amino acid sequence design (Figure 1).
Fusion of natural oligomers
An early idea for introducing two self-associating interfaces
into a single protein molecule emerged at a time when the
prospects for designing novel protein–protein interfaces by
computational methods were limited. In 2001, we proposed
that genetically fusing two different, naturally oligomeric
domains into a single protein chain could satisfy the design
requirement of combining two self-associating motifs [16].
The problem was how to dictate, or at least predict in
advance, what the relative orientation would be between
two genetically fused domains; free backbone rotations
occur at a point of fusion, and this would make the final
geometry unpredictable. The solution for how to predict the
relative orientation in advance was to use oligomeric
domains that began or ended in an a-helix, such that the
protein backbone (and any a-helix-preferring amino acids
introduced as a linker) might adopt an unbroken a-helix
running from within one oligomeric domain into the next. In
this way, pairs of oligomeric domains of known structure
would be combined in hypothetical, predicted arrange-
ments, and a search could be made for pairs that would
satisfy specific geometric rules for constructing different
architectures such as cages or crystals. The same paper laid
out geometric rules for how the symmetry axes of the
component domains would have to be oriented relative to
each other to obtain various architectures ([16], Box 1). That
initial set of geometric rules only covered combinations of
dimers and trimers because the protein structure database
at that time contained relatively few proteins with both
higher oligomeric symmetry and subunits ending in helices
[16]. A tremendous range of assembly architectures can be
achieved using higher symmetry building blocks; those
possibilities have not been completely enumerated yet.
The oligomeric fusion method was first used in the design
of protein filaments – a relatively easy design target – and a
12-subunit molecular cage with a tetrahedral shape, repre-
senting the first protein construction of its kind. However,
that initially designed protein sequence formed cage-like
assemblies whose sizes were too heterogeneous to charac-
terize in atomic detail; crystals could not be obtained [16].
This barrier was surmounted in recent work. When the
original design was revisited, and two amino acid changes
were made based on a visual identification of potential steric
conflicts, a homogenous 12-subunit cage was obtained and
crystallized [18]. The designed cage is roughly 16 nm in
diameter and contains a central opening about 5 nm in
diameter (Figure 2a,b). Interestingly, despite an overall
match to the design, the observed assembly exhibited sig-
nificant deviations (about 8 Å root-mean-square deviation,
RMSD) from perfect symmetry [18].
Review
Box 1. Principles for designing ordered protein assemblies
A general requirement for engineering complex self-assembling
structures is to create a protein molecule bearing multiple distinct
self-associating interfaces (Figure I, top). This can be achieved
through the use of natural oligomeric proteins or computational
sequence design, separately or in combination. The multiple
interfaces must be combined according to specific geometric rules
[16] if defined structures are to be created (Figure I, bottom). The
geometric design requirements are given for some example
Generang mulple self-associang interfaces by fusion or sequence design
or
Natural + natural (fusion)
Examples of designed architectures based on geometrically controlled
combinaons of symmetric interfaces
Cages or shells
4 & 2, 45 4 & 3, 54.7
24 subunits 24 subunits
symmetry O symmetry O
Other possibilies (not shown):
3 & 2, 54.7 12 subunits, symmetry T [18,34]
3 & 2, 35.3 24 subunits, symmetry O [34]
3 & 2, 20.9 60 subunits, symmetry I
5 & 2, 31.7 60 subunits, symmetry I
5 & 3, 37.4 60 subunits, symmetry I
Figure I. Strategies and geometric rules for combining multiple protein interfaces.
Variations on the fusion approach have been introduced
by others. In one study [19], natural oligomers were used to
supply the two self-associating interfaces required for an
extended 2D array of protein molecules, but in this case the
components were connected by binding to biotin instead of
by genetic fusion; one of the oligomers was streptavidin, a
D2 tetramer, and the other was aldolase (RhuA), a C4
tetramer. Arrays of only limited extent were obtained,
probably owing to flexibility of the biotin linker and imper-
fect control over the relative orientation of the component
oligomers. In another study, two types of building blocks
were generated by fusing PDZ domains or PDZ-binding
peptides to the termini of a tetrameric (D2 symmetry)
Trends in Cell Biology December 2012, Vol. 22, No. 12
assemblies. In each case, the two types of symmetry elements are
noted in bold, together with the angle they must form. The
symmetries of the resulting assemblies are given (T: tetrahedral;
O: octahedral; I: icosahedral). For 2D layers and 3D crystals (not
shown), rules for constructing the full range of possible symmetries
have not been articulated yet [16], though specific instances of 2D
layers and 3D crystals have been successfully engineered
[22,35,36].
or
Natural + designed Designed + designed
Layers
4 & 2 , parallel
layer symmetry p422
Other possibilies (not shown):
- numerous [16,22]
TRENDS in Cell Biology
superoxide reductase [20]. These building blocks formed
linear filaments upon mixing. Smaller protein motifs have
also been used in oligomeric fusion approaches. One study
connected a trimeric coiled-coil to a pentameric coiled-coil
with the goal of generating 60-subunit icosahedral parti-
cles [21]. A disulfide bond was introduced to force a bend
between the two helical components; an angle of precisely
37.48 is required to satisfy icosahedral symmetry. The
particles formed by this design appeared polymorphic
but were roughly the intended size of 16 nm.
A notable extension to the original oligomer fusion
strategy was recently introduced [22] (Figure 2c,d). These
authors noted that, if higher-order oligomers are used as
655
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a) (e)

(b) (c) (d)

TRENDS in Cell Biology

Figure 1. Strategies for introducing a new oligomerization interface. (a) Oligomeric fusion strategy. Two different oligomeric proteins can be fused together to generate two
oligomerization interfaces within a single protein subunit, thereby driving the assembly of complex structures [16]. A single protein chain within the complex is shown in
the black rectangle. In the example shown, the green domain derives from a natural trimer and the magenta domain derives from a natural dimer. Linking residues are
yellow. (b) New interfaces generated through metal-binding. Multiple histidine residues incorporated into the side of an exposed a-helix can constitute a new dimeric
interface [30]. Two different chains are shown in green and yellow, histidine residues are shown in black lines, and metal ions are shown as orange spheres. (c) a-Helix-
based oligomerization. a-Helices can be designed to form different oligomerization states based on well-studied coiled-coil motifs [32]. Different chains are shown in
different colors. (d) b-Sheet-based oligomerization. The open edge of a b-sheet can be used as a site for designing a dimeric interaction [31]. Two different chains are
colored in green and cyan. (e) De novo design of a new interface. A new interface can be introduced into a natural oligomer of relatively low initial symmetry to generate
higher-order assemblies [34]. In the case illustrated, multiple copies of a natural trimer are shown in different colors. One trimer, shown in green, is lifted from its assembled
position. The de novo designed interaction patches are shown in red. The designed interactions between the green subunit and other subunits are indicated by thin lines.

the components, the two oligomers to be genetically fused
can be aligned at an axis of symmetry that they both share.
In such a scenario, there will be two (or more) polypeptide
linkers instead of one running between adjacent oligo-
mers. The requirement for a rigid a-helical linker could
then be removed, with the necessary orientational control
resulting instead from the multiple chain connections.
This symmetry-matching fusion protein strategy success-
fully generated a linear filament and 2D arrays. A notable
success rate of 40% was achieved when constructing 2D
arrays, and one of the arrays exhibited exceptionally good
long-range order. The design of 3D crystals based on this
strategy led to large solid aggregates having crystal-like
morphology, but the long-range order required for crystal-
lographic analysis has not been reported yet.
In theory, the shared symmetry axis method can be used
to create 2D and 3D arrays obeying a variety of different
symmetries, the full range of which has not been articu-
lated yet. However, an unavoidable constraint of the meth-
od is that only extended materials (e.g., 2D arrays and 3D
crystals) can be generated, but not finite structures such as
molecular cages.
Interface design
Computer programs for designing protein–protein inter-
actions are becoming increasingly powerful [23–25], mak-
ing it possible to design sequence mutations that drive
specific modes of symmetric self-association. This has
opened up more direct strategies for designing large pro-
tein assemblies and extended materials. The most conser-
vative approaches for designing such large assemblies,
including most of the successful experiments reported so
far, rely on a natural (or native-like) oligomerization
656
motif to provide one of the modes of self-association,
and computer algorithms are then used to introduce an
additional interface (or interfaces) (Box 1). This approach
minimizes the number of novel interfaces that must be
successfully designed computationally. Strategies that
combine a natural oligomeric interface with the computa-
tional design of an additional interaction were presaged in
experiments [26] in which assemblies such as double-ring
structures were generated by introducing relatively sim-
ple interfacial features such as a hydrophobic patch into
simpler, single-ring, natural protein assemblies. The suc-
cess rate of that strategy was higher when the starting
structures contained more subunits (for example, C4 com-
pared to C2), presumably owing to the higher multiplicity
of the newly introduced interaction sites. In addition, the
observed assemblies typically showed substantial devia-
tions from the intended structures, presumably reflecting
the limited geometric precision provided by the sequence
design strategy.
Numerous sequence design strategies for promoting
self-association have been demonstrated, with strategies
that deliver the highest geometric specificity placing the
greatest demand on the sequence design process
(Figure 1b–e). At one end of the spectrum, a simple se-
quence element such as two histidines at positions i and
i+4 on the exposed surface of an a-helix has proved to be a
straightforward approach for driving protein self-associa-
tion in the presence of metals [27–29]. However, the spe-
cific geometries of those associations have been hard to
predict [30]. Higher levels of geometric specificity have
been demonstrated in the design of self-associating inter-
actions involving an exposed b-strand in one case [31], and
a-helical bundles in several other studies [32,33]. At the
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a) (b)

2
3

16 nm

(c) 4 2 (d)

c N 2 xn
28 nm
2

2 2
2
2

TRENDS in Cell Biology

Figure 2. Designed assemblies based on two types of oligomer-fusion strategies. (a) Fusion through a semi-rigid helix linker. Two natural oligomeric domains (green and
purple; left) are held in a predetermined orientation by genetic fusion of the two domains and an a-helical linker (yellow; middle) [16]. Multiple copies of the fusion protein
can then self-assemble into a well-defined shape (right). (b) Crystal structure of a successful example by using the helix-linker fusion strategy [18]. Twelve different chains in
the assembly are colored differently; a central sphere of about 5 nm is shown in black. (c) Fusion by common symmetry axis. Short and relatively flexible linkers (green; left)
are used to fuse natural oligomeric domains (pink and yellow) such that the shared symmetry axes tend to align with each other (blue arrow; left) [22]. Multiple subunits can
then self-assemble into extended arrays (right). (d) Molecular model of a layer assembled by the common-symmetry-axis fusion strategy, which was consistent with
electron micrographs of the resulting material. Panels (a) and (b) adapted, with permission, from [18]. Panel (c) adapted, with permission, from [55]. Panel (d) adapted, with
permission, from [22].

extreme end, the sequence design of entirely novel,
geometrically specific self-associating interfaces (i.e., not
modeled on well-characterized natural motifs) has been
demonstrated in a newly published study [34]. Together,
these varied modes of interface design have been exploited
in a recent flurry of experiments demonstrating the crea-
tion of complex assemblies and extended materials.
Metal site design, combined with additional interface
design, was used to create a small protein that self-assem-
bles into a series of different architectures [35]. A cyto-
chrome protein that had been engineered by the
introduction of coordinating histidine residues to form a
metal-induced dimer was stabilized in that specific geomet-
rical configuration by further mutations using computer-
aided interface design. In the resulting configuration, the
two symmetric metal-binding motifs each had one open
coordination site to promote a further twofold association
between dimers. Typically, a tail-to-tail association between
head-to-head dimers tends to produce a filament or strip of
molecules. In this study [35] it appears that different exper-
imental conditions promoted different arrangements and
associations of such strips, such as side-by-side associations
to give sheets, or helical wrappings to give tubular struc-
tures (Figure 3). Assemblies produced by the metal-mediat-
ed strategy have the potential to be fine-tuned through the
adjustment of the protein:metal ratio or pH. At the present
stage, the assembly outcomes in metal-mediated designs
must be established experimentally, and the generation of
predictable structures constitutes an important future chal-
lenge.

(a) (b) (c)

Bend & twist

Zn1 Zn3
i-face i-face

Zn2
i-face
Key: Zn1 Zn2 Zn3

TRENDS in Cell Biology

Figure 3. Layers and nanotubes by metal-mediated assembly. (a) Zinc was used to trigger the assembly of protein subunits with designed di-histidine motifs in two
perpendicular directions into a layer, as shown in (b) [35]. The equivalent part of (a) is shown in the green box. Three differently located zinc ions are colored in cyan,
orange, and red. (c) Bending and twisting around the binding site for zinc #3 caused the layer to curl into a nanotube. Adapted, with permission, from [35].

657
 Review
(a)
Figure 4. A designed 3D protein crystal using coiled-coil building blocks. (a) An assembly obeying P6 crystal symmetry was designed using a parallel 3-helix coiled-coil as
the starting motif [36]. Computational sequence design was used to promote lateral and vertical contacts between trimers. Distinct chains in the trimer are colored in green,
cyan, and magenta. Minor breakage of the intended P6 symmetry is visible. (b) The vertical stacking of layers is illustrated in a perpendicular view from the position of the
black triangle in (a). The ends of the helices in one layer form hydrogen bonds with helices in neighboring layers, forming semi-continuous helices along the longitudinal
crystal dimension. The layer shown in (a) is located between the two solid lines.
The successful design of a self-assembling 3D protein
crystal was recently demonstrated [36]. First, a protein
backbone conformation based on a well-studied protein
motif, a three-helix bundle, was placed in a chosen crystal
lattice. In this case, the template trimer was placed on the
threefold symmetry axis of a P6 crystal unit cell (Figure 4).
Favorable unit cell spacings and trimer orientations were
then sampled, after which the sequence of the protein was
optimized based on its interactions with symmetry-based
neighboring protein molecules. Following production of the
designed protein, two different protein crystal forms were
obtained, one of which agreed very closely with the designed
crystal model, with deviations in the range of about 1 Å.
Two different test cases recently demonstrated the ability
to design a large, novel self-associating interface into a
naturally oligomeric protein subunit to produce large molec-
ular cages and clusters [34] (Figure 5). Novel homodimeric
interfaces were designed into two different naturally trimer-
ic proteins in geometries intended to yield a 24-subunit cage
with octahedral symmetry in one case and a 12-subunit
assembly with tetrahedral symmetry in the other. This
design strategy relied on a computational docking procedure
that restrained the rigid body and conformational sampling
to configurations obeying the target symmetry [37,38]. Mul-
tiple crystal structures of the designed proteins revealed
that the resulting assemblies matched the design models at
atomic-level resolution, exhibiting backbone RMSDs of
about 1 Å or less. It should be noted that the initially
designed sequence for the tetrahedral case yielded an as-
sembly that was slightly less accurate (backbone RMSD of
2.7 Å); introducing three additional mutations yielded the
more accurate assembly. This study demonstrated that
designing interfaces based on many weak interactions over
large interface areas – similar to the interfaces found in
natural protein assemblies – can result in highly ordered
and accurately assembled structures. However, the two
successes were accompanied by 39 other designed proteins
658
Trends in Cell Biology December 2012, Vol. 22, No. 12
(b)
1.7 nm
4 nm
TRENDS in Cell Biology
that either expressed insolubly or failed to assemble,
emphasizing the need for improved methods for designing
protein–protein interactions if this approach is to become
widely used.
Strategic variations
The basic design ideas discussed above admit numerous
variations, some of which have been demonstrated already.
For extended protein assemblies, one immediate obstacle
is the production of recombinant proteins in soluble form
during bacterial protein expression. The building blocks for
such assemblies tend to associate as soon as they are
produced inside the host cell, leading to large aggregates
and inclusion bodies that seriously complicate the problem
of protein purification. Several groups have overcome this
problem by splitting the assemblies into two or more
distinct components that can be produced in soluble forms
in separate bacterial strains [19,20,22]. Purified, soluble
components can then be mixed to form the intended as-
semblies in situ. Other methods to trigger protein assem-
blies rely on the addition of metals or other ligands into the
otherwise soluble protein building blocks [35,39,40]. These
approaches offer the additional advantage that the metals
or ligands can be chelated or scavenged to revert the
protein assemblies to their soluble building blocks.
The surfaces of protein building blocks can also be
decorated with various functional groups to generate de-
sirable properties. In one case, a rhodamine chemical
group was conjugated to building blocks to promote asso-
ciation between protein layers [35]. In another example,
protein sidechains in a repeating pattern along a helical
peptide were chosen to adhere to the periodic pattern of
carbon atoms in single-walled carbon nanotubes [41]. The
helical peptides wrapping around the carbon nanotubes
further served as a periodic registry for gold nanoparticle
attachment, highlighting the potential for generating hy-
brid bio-nanomaterials.
 Review
(a) Twofold
35.2
Computaonal docking
and interface design
Oligomeric protein Self-assembling
building blocks protein nanomaterials
(b) 54.7
Computaonal docking
and interface design
Figure 5. Accurate design of large symmetric protein assemblies based on novel interface design. Among the wide variety of symmetric architectures that can be obtained
by interface design, two large cubic assemblies have been demonstrated [34]. Crystal structures of designed (a) 24-subunit octahedral and (b) 12-subunit tetrahedral
assemblies are shown, viewed along their twofold and threefold symmetry axes. Within each assembly, individual trimers are depicted in different colors. The designed
interfaces [regions boxed in (a) and (b)] in the crystal structures (green/magenta) of the (c) octahedral and (d) tetrahedral assemblies closely agree with the designed models
(white). Backbone RMSDs between the observed structures and the designed models were 1.1 Å and 0.6 Å over all subunits for the octahedral and tetrahedral assemblies,
respectively.
Further variations for designing protein assemblies have
been conceptualized. In one example, the building blocks
comprise smaller fragments of natural proteins, which
would then be stitched together to produce molecules having
specific self-association properties [42]. Although this ap-
proach has not been realized yet in practice, it highlights the
variety of creative ideas being brought to bear on the prob-
lem of designing ordered assemblies.
Different design strategies offer their own advantages
and disadvantages. Recent studies emphasize key distinc-
tions between the two main strategies discussed here –
oligomeric fusion and interface design. The former ap-
proach avoids the introduction of mutations within natu-
rally occurring protein domains, and hence minimizes the
chance that the natively folded structure will be disrupted.
On the other hand, de novo interface design should ulti-
mately enable the use of arbitrary protein domains as
building blocks, without being limited to those natural
oligomers offered by nature.
Concluding remarks
If engineered protein assemblies are to become a reliable
technology for real-life applications, the success rate of the
strategies discussed here will need to be improved. For the
oligomeric fusion strategies, the major challenge is to
control better the relative orientation of the fusion part-
ners. This is especially true for the helix-fusion strategy
given that the recent crystal structure of the designed cage
revealed considerable bending and twisting of the helical
linkers [18]. For the symmetry-matching fusion strategy,
although the presence of multiple linkers between joined
oligomers promotes alignment of the shared symmetric
axes, twisting about this axis may be harder to control
[22]. Such rotational flexibility may need to be dealt with if
a high success rate is to be achieved, particularly in the
Trends in Cell Biology December 2012, Vol. 22, No. 12
Threefold (c) (d)
13 nm
11 nm
TRENDS in Cell Biology
challenging area of extended 3D materials. For strategies
that rely on creating new interfaces by computational
design, the key challenge is in achieving geometrically
precise interfaces. The successful design of a protein crys-
tal based on an a-helical polypeptide [36] suggests that
this may be relatively straightforward for protein building
blocks consisting of simple motifs. By comparison, design-
ing precise interfaces into larger proteins of arbitrary
shape appears to present more complex problems. Early
experiments showed that single rings of protein oligomers
could be reliably converted to double rings by incorporat-
ing a few nonpolar residues on the surface, but the assem-
blies obtained generally deviated from the designs in
detail [26]. King et al. [34] twice succeeded in designing
extensive, geometrically-accurate interfaces into globular
protein subunits, which consequently gave rise to large,
symmetric cubic assemblies in very close agreement with
the designed models. However, the successful designs
were only a small fraction of the total designs attempted,
and in one of the two successful designs a second round of
sequence mutations was necessary to fine-tune the initial
design. The successes (and accompanying failures) of var-
ious recent design efforts suggest that further improve-
ments in interface design algorithms could have an
important impact.
An ultimate challenge will be to create a large, complex
assembly without relying on existing oligomerization
interfaces or well-known self-associating motifs. This will
require designing at least two distinct interfaces into a
naturally monomeric protein. In addition to the high accu-
racy required, the need to introduce multiple interfaces
compounds the success-rate problem, and also increases
the likelihood that the sequence changes will be so many as
to compromise protein folding and stability. The latter
issue may call for more careful consideration of which
659
Review
protein folds might be the most suitable as starting tem-
plates for design.
The success-rate problem might be mitigated if efficient
screening methods can be developed. King et al. imple-
mented a multistep screening method to identify which of
their many computationally designed protein sequences
were soluble and compatible with self-assembly [34]. Al-
though their screening protocol, which involved gel elec-
trophoresis or liquid chromatography, was suitable for the
number of designs examined (50), design strategies that
exploit much larger sequence pools could benefit from the
development of higher-throughput screening strategies
[43].
The next stage in developing designed protein assemblies
will need to focus on applications [44,45]. The possibilities
are myriad, arising from different combinations of function-
al elements (such as catalytic sites or recognition motifs)
with various types of spatial organization, such as encapsu-
lation or ordered display [46,47]. Finite structures such as
cages and shells invite the idea of encapsulation, as well as
targeted delivery based on externally displayed motifs
[48,49]. Motifs displayed on the surface of large assemblies
could be useful in their own right, for example in the design
of synthetic vaccines [50]. Extended materials offer addi-
tional modes of spatial organization [51]. Numerous areas of
synthetic biology could be advanced by materials presenting
ordered arrangements of enzymes or receptors [52,53].
Across all the possible architectures that might be explored,
the ability to trigger assembly or disassembly should enable
a broad range of responsive materials [54].
Acknowledgments
The authors thank Jennifer Padilla for helpful discussions and critical
reading of the manuscript.
References
1 Schwille, P. (2011) Bottom-up synthetic biology: engineering in a
tinkerer’s world. Science 333, 1252–1254
2 Tan, C. et al. (2009) Emergent bistability by a growth-modulating
positive feedback circuit. Nat. Chem. Biol. 5, 842–848
3 Elowitz, M.B. and Leibler, S. (2000) A synthetic oscillatory network of
transcriptional regulators. Nature 403, 335–338
4 Zhang, S.G. et al. (2002) Design of nanostructured biological materials
through self-assembly of peptides and proteins. Curr. Opin. Chem. Biol.
6, 865–871
5 Yeates, T.O. and Padilla, J.E. (2002) Designing supramolecular protein
assemblies. Curr. Opin. Struct. Biol. 12, 464–470
6 Zhang, S.G. (2003) Fabrication of novel biomaterials through molecular
self-assembly. Nat. Biotechnol. 21, 1171–1178
7 Petka, W.A. et al. (1998) Reversible hydrogels from self-assembling
artificial proteins. Science 281, 389–392
8 Ogihara, N.L. et al. (2001) Design of three-dimensional domain-swapped
dimers and fibrous oligomers. Proc. Natl. Acad. Sci. U. S. A. 98,
1404–1409
9 Woolfson, D.N. (2010) Building fibrous biomaterials from alpha-helical
and collagen-like coiled-coil peptides. Biopolymers 94, 118–127
10 Ghadiri, M.R. et al. (1993) Self-assembling organic nanotubes based on
a cyclic peptide architecture. Nature 366, 324–327
11 Ryadnov, M.G. and Woolfson, D.N. (2003) Introducing branches into a
self-assembling peptide fiber. Angew. Chem. Int. Ed. Engl. 42, 3021–
3023
12 Levy, E.D. et al. (2008) Assembly reflects evolution of protein
complexes. Nature 453, U1262–U1266
13 Goodsell, D.S. and Olson, A.J. (2000) Structural symmetry and protein
function. Annu. Rev. Biophys. Biomol. Struct. 29, 105–153
660
Trends in Cell Biology December 2012, Vol. 22, No. 12
14 Crick, F.H.C. and Watson, J.D. (1956) Structure of small viruses.
Nature 177, 473–475
15 Wukovitz, S.W. and Yeates, T.O. (1995) why protein crystals favor
some space-groups over others. Nat. Struct. Biol. 2, 1062–1067
16 Padilla, J.E. et al. (2001) Nanohedra: using symmetry to design self
assembling protein cages, layers, crystals, and filaments. Proc. Natl.
Acad. Sci. U. S. A. 98, 2217–2221
17 Plevka, P. et al. (2012) Crystal structure of human enterovirus 71.
Science 336, 1274
18 Lai, Y.T. et al. (2012) Structure of a 16-nm cage designed by using
protein oligomers. Science 336, 1129
19 Ringler, P. and Schulz, G.E. (2003) Self-assembly of proteins into
designed networks. Science 302, 106–109
20 Usui, K. et al. (2009) Nanoscale elongating control of the self-assembled
protein filament with the cysteine-introduced building blocks. Protein
Sci. 18, 960–969
21 Raman, S. et al. (2006) Structure-based design of peptides that self-
assemble into regular polyhedral nanoparticles. Nanomedicine 2,
95–102
22 Sinclair, J.C. et al. (2011) Generation of protein lattices by fusing
proteins with matching rotational symmetry. Nat. Nanotechnol. 6,
558–562
23 Leaver-Fay, A. et al. (2011) Rosetta3: an object-oriented software suite
for the simulation and design of macromolecules. Methods Enzymol.
487, 545–574
24 Fleishman, S.J. et al. (2011) Computational design of proteins
targeting the conserved stem region of influenza hemagglutinin.
Science 332, 816–821
25 Huang, P.S. et al. (2007) A de novo designed protein-protein interface.
Protein Sci. 16, 2770–2774
26 Grueninger, D. et al. (2008) Designed protein–protein association.
Science 319, 206–209
27 Radford, R.J. and Tezcan, F.A. (2009) A superprotein triangle driven by
nickel(II) coordination: exploiting non-natural metal ligands in protein
self-assembly. J. Am. Chem. Soc. 131, 9136–9137
28 Salgado, E.N. et al. (2010) Metal templated design of protein interfaces.
Proc. Natl. Acad. Sci. U. S. A. 107, 1827–1832
29 Der, B.S. et al. (2012) Metal-mediated affinity and orientation
specificity in a computationally designed protein homodimer. J. Am.
Chem. Soc. 134, 375–385
30 Salgado, E.N. et al. (2010) Metal-directed protein self-assembly. Acc.
Chem. Res. 43, 661–672
31 Stranges, P.B. et al. (2011) Computational design of a symmetric
homodimer using beta-strand assembly. Proc. Natl. Acad. Sci. U. S. A.
108, 20562–20567
32 Zaccai, N.R. et al. (2011) A de novo peptide hexamer with a mutable
channel. Nat. Chem. Biol. 7, 935–941
33 Korendovych, I.V. et al. (2010) De novo design and molecular assembly
of a transmembrane diporphyrin-binding protein complex. J. Am.
Chem. Soc. 132, 15516–15518
34 King, N.P. et al. (2012) Computational design of self-assembling
protein nanomaterials with atomic level accuracy. Science 336,
1171–1174
35 Brodin, J.D. et al. (2012) Metal-directed, chemically tunable assembly
of one-, two- and three-dimensional crystalline protein arrays. Nat.
Chem. 4, 375–382
36 Lanci, C.J. et al. (2012) Computational design of a protein crystal. Proc.
Natl. Acad. Sci. U. S. A. 109, 7304–7309
37 Andre, I. et al. (2007) Prediction of the structure of symmetrical protein
assemblies. Proc. Natl. Acad. Sci. U. S. A. 104, 17656–17661
38 Wang, C. et al. (2007) Protein-protein docking with backbone flexibility.
J. Mol. Biol. 373, 503–519
39 Carlson, J.C.T. et al. (2006) Chemically controlled self-assembly of
protein nanorings. J. Am. Chem. Soc. 128, 7630–7638
40 Fegan, A. et al. (2010) Chemically controlled protein assembly:
techniques and applications. Chem. Rev. 110, 3315–3336
41 Grigoryan, G. et al. (2011) Computational design of virus-like protein
assemblies on carbon nanotube surfaces. Science 332, 1071–1076
42 Tsai, C.J. et al. (2007) Principles of nanostructure design with protein
building blocks. Proteins 68, 1–12
43 Worsdorfer, B. et al. (2011) Directed evolution of a protein container.
Science 331, 589–592
Review
44 Woolfson, D.N. and Mahmoud, Z.N. (2010) More than just bare scaffolds:
towards multi-component and decorated fibrous biomaterials. Chem.
Soc. Rev. 39, 3464–3479
45 Howorka, S. (2011) Rationally engineering natural protein assemblies
in nanobiotechnology. Curr. Opin. Biotechnol. 22, 485–491
46 Tschiggerl, H. et al. (2008) Exploitation of the S-layer self-assembly system
for site directed immobilization of enzymes demonstrated for an
extremophilic laminarinase from Pyrococcus furiosus. J. Biotechnol.
133, 403–411
47 Minten, I.J. et al. (2009) Controlled encapsulation of multiple proteins
in virus capsids. J. Am. Chem. Soc. 131, 17771–17773
48 Tong, G.J. et al. (2009) Viral capsid dna aptamer conjugates as
multivalent cell-targeting vehicles. J. Am. Chem. Soc. 131, 11174–11178
49 Han, M. et al. (2011) Targeted vault nanoparticles engineered with an
endosomolytic peptide deliver biomolecules to the cytoplasm. ACS
Nano 5, 6128–6137
Trends in Cell Biology December 2012, Vol. 22, No. 12
50 Smith, M.L. et al. (2006) Modified tobacco mosaic virus particles as
scaffolds for display of protein antigens for vaccine applications.
Virology 348, 475–488
51 Ilk, N. et al. (2011) S-layer fusion proteins – construction principles and
applications. Curr. Opin. Biotechnol. 22, 824–831
52 Delebecque, C.J. et al. (2011) Organization of intracellular
reactions with rationally designed RNA assemblies. Science 333,
470–474
53 Dueber, J.E. et al. (2009) Synthetic protein scaffolds provide modular
control over metabolic flux. Nat. Biotechnol. 27, 753–759
54 de la Rica, R. and Matsui, H. (2010) Applications of peptide and
protein-based materials in bionanotechnology. Chem. Soc. Rev. 39,
3499–3509
55 Yeates, T.O. (2011) Nanobiotechnology: protein arrays made to order.
Nat. Nanotechnol. 6, 541–542
661
Review
Special Issue – Synthetic Cell Biology
Designing biological
compartmentalization
Anna H. Chen1 and Pamela A. Silver1,2
1
Department of Systems Biology, Harvard Medical School, Boston, MA 02115, USA
2
Wyss Institute for Biologically Inspired Engineering, Harvard University, Boston, MA 02115, USA
Intracellular organization is a key factor in cell metabo-
lism. Cells have evolved various organizational systems
to solve the challenges of toxic pathway intermediates,
competing metabolic reactions, and slow turnover rates.
Inspired by nature, synthetic biologists have utilized pro-
teins, nucleic acids, and lipids to construct synthetic
organizational systems that mimic natural systems. Many
of these systems have been applied to metabolic path-
ways and shown to significantly increase the production
of industrially and commercially important chemicals.
Further engineering and characterization of synthetic
organizational systems will allow us to better understand
native cellular strategies of spatial organization. Here, we
discuss recent advances and ongoing efforts in designing
and characterizing synthetic compartmentalization sys-
tems to mimic natural strategies and increase metabolic
yields of engineered pathways.
Compartmentalization benefits natural and engineered
systems
Biological complexity requires varying degrees of organi-
zation. Cells require spatial organization to perform the
various enzymatic reactions and processes necessary to
sustain life [1]. This is achieved through compartmentali-
zation, the physical separation of biological reactions.
Examples of compartmentalization include membrane-
bound organelles, bacterial microcompartments [2,3], mul-
tienzyme complexes, and others [4,5].
Inspired by nature, synthetic biologists have recently
devised strategies to mimic cellular organizational sys-
tems. These synthetic systems have been predominantly
designed toward metabolic engineering of pathways, har-
nessing the capability of cells to produce industrially [6,7]
or pharmaceutically useful [8,9] compounds.
In this review, we describe the various difficulties faced
by the cell when performing metabolic reactions and natu-
ral compartmentalization systems that solve these pro-
blems. We review recent advances in designing synthetic
compartments that provide modular solutions to overcome
these same challenges. These systems capture the benefits
of spatial organization and apply them to engineered path-
ways. This has also recently been reviewed in [10–12]. Our
review will discuss the latest progress and challenges in
Corresponding author: Silver, P.A. (pamela_silver@hms.harvard.edu).
Keywords: compartmentalization; synthetic scaffolds; bacterial microcompartments;
biological organization.
662 0962-8924/$ – see front matter ß 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tcb.2012.07.002 Trends in Cell Biology, December 2012, Vol. 22, No. 12
designing compartmentalization, especially in building
bacterial microcompartments and RNA scaffolds. We also
analyze the degree to which these mimic natural systems
and discuss how they aid in our understanding of the
biological organization of the cell.
The need for intracellular organization
Cells face many challenges that benefit from compartmen-
talization (Figure 1a). First, some enzymes, such as ribu-
lose 1,5-bisphosphate carboxylase oxygenase (RuBisCO)
[13], suffer from slow turnover, which results in flux imbal-
ances or bottlenecks in pathways. Reliance on such
enzymes may require establishing local concentration gra-
dients of substrates. This would increase reaction rates to
support adequate pathway flux [14]. Second, diffusion of
volatile intermediates through the cell membrane results
in their loss from the cell [15]. Third, biosynthetic path-
ways can generate toxic intermediates that inhibit growth,
such as hydrogen sulfide accumulated during bacterial
sulfur metabolism [16]. Finally, metabolites can partici-
pate in multiple competing reactions, reducing their avail-
ability for any single pathway. An example of this is
malonyl-CoA, an intermediate that is consumed in fatty
acid and phospholipid production but is also used in the
biosynthesis of polyketides and flavonoids [17].
Nature’s solutions
To deal with these challenges, nature has evolved compart-
mentalization strategies (Figure 1b), such as large enzyme
complexes [10,18,19] and organelles [2,20], to spatially
organize metabolism. In eukaryotes, compartmentaliza-
tion in the form of membrane-bound organelles is common.
The peroxisome, for example, encapsulates reactions that
generate or consume hydrogen peroxide, a toxic interme-
diate from the breakdown of organic substrates in oxida-
tive reactions [21].
Until recently, prokaryotes were generally thought to
lack internal organization [22]. However, researchers have
recently discovered different types of bacterial microcom-
partments that partition the internal space of the bacterial
cell for specialized functions [2,23]. In cyanobacteria and
other autotrophic prokaryotes, carboxysomes encapsulate
RuBisCO and carbonic anhydrase, enzymes involved in the
rate-limiting step of the Calvin cycle [24,25] (Figure 2a).
These proteinaceous microcompartments are the primary
‘carbon-concentrating mechanism’ in these bacteria. They
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

Key: = enzymes
(a)
(2) Escape through membrane = metabolites

(1) Flux imbalance

Enzyme 1 Enzyme 2 Enzyme 3

Intermediate A Intermediate B Product

Side Product
Enzyme 4

(3) Toxicity
(4) Compeng reacons

Cell

(b)

(1) Increase local Compartmentalizaon

(2) Sequester concentraon System
intermediates
Enzyme 2
Enzyme 1
Enzyme 3

Intermediate A Intermediate B
(3) Reduce toxicity Product

(4) Reduce
compeon
Enzyme 4
Cell
TRENDS in Cell Biology

Figure 1. Compartmentalization: nature’s solution to various challenges. (a) Nature faces many challenges when conducting the chemical reactions of the cell. (1) Differing
enzyme kinetics may result in flux imbalances. (2) Intermediates may be lost through the cell membrane. (3) Toxic intermediates can result in growth inhibition. (4)
Competing reactions can divert flux through undesired pathways. (b) Compartmentalization systems specifically solve challenges 1–4, respectively, by: (1) creating areas of
local concentrations to favor reaction kinetics; (2) sequestering intermediates; (3) reducing toxicity; and (4) reducing competition. (Adapted from [11].)

are proposed to help overcome the slow turnover rate of
RuBisCO by providing a high local concentration of carbon
dioxide to the enzyme [26,27].
Two other bacterial proteinaceous microcompartments
protect the cell from toxic aldehyde intermediates. The
ethanolamine utilization (Eut) microcompartment
sequesters acetaldehyde, a volatile and toxic intermediate
of the ethanolamine utilization pathway [28]. Likewise,
the 1,2-propanediol utilization (Pdu) microcompartment
encapsulates propionaldehyde, minimizing its toxicity
[29]. These and numerous other bacterial microcompart-
ments have been found in approximately 400 microbial
genomes [2].
Another method of compartmentalization found in na-
ture is multienzyme complexes, which directly link
enzymes involved in a given pathway. Ideally, this results
in substrate channeling, the process by which intermedi-
ates are directly transferred between the active sites of two
enzymes that catalyze sequential reactions in the pathway
[30]. Substrate channeling prevents the loss of intermedi-
ates and minimizes competing cross-reactions. A classic
example is tryptophan synthase, a multienzyme complex
that catalyzes the last two reactions in the biosynthesis of
L-tryptophan [18,31]. The intermediate, indole, is chan-
neled from one active site to the next without being
released into the surrounding environment. This is advan-
tageous for the cell not only because indole is reactive and
easily lost through the cell membrane, but also because, in
the absence of indole, tryptophan synthase catalyzes
dehydration of serine to pyruvate at 5% the rate of trypto-
phan formation [32]. Other multi-enzyme complexes found
in nature include polyketide synthase [33], carbamoyl
phosphate synthetase [34], and cellulosomes [35], which
may function similarly by increasing reaction kinetics and
reducing the loss of intermediates.
Synthetic compartmentalization
The goal of metabolic engineering is to optimize a given
biosynthetic pathway to increase production of a particular
substance [7,36]. Many of these pathways present the same
challenges of toxic intermediates, competing reactions,
and flux imbalances found in nature [10,14]. Therefore,
663
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a) Shell RuBisCO

10 nm
CA 2 µm
Pentameric verces
Hexamers with
pores regulang transport

PDZ
(b) (c) HMGR

PDZ
HMGR

}
HMGR
HMGS

PDZ
GBD
AtoB

HMGR SH3 SH3

Enzymes

PDZ
HMGR

AtoB
HMGS
Pepde
}

AtoB
GBD SH3 SH3

HMGS
Ligands PDZ

( )( ) ( ) }

HMGS
HMGS

HM
PDZ
GBD Scaffold

AtoB
SH3 PDZ

HMGS
SH3

HMGS
GBD

AtoB

HMGRGR
x y z SH3

HMGR

HMGR

HMGS
HMGS
GBD SH3 SH3 PDZ

PD
Z
GR PDZ
HM HM
GR SH3

PDZ
SH3 GBD

TRENDS in Cell Biology

Figure 2. Protein scaffolds: utilizing nature’s building material for functional complexes. Nature compartmentalizes reactions via enzyme complexes. Proteins often serve
as scaffolds, bringing together pathway enzymes and increasing flux. (a) Bacterial microcompartments are protein structures that encapsulate reactions [2]. Carboxysomes
contain ribulose 1,5-bisphosphate carboxylase oxygenase (RuBisCO) and carbonic anhydrase. The shell comprises pentamers and hexamers, some of which contain pores
that are thought to regulate substrate transport (left). Carboxysomes spatially align in the cell [51] (right). Heterologous expression of microcompartments and further
engineering of targeting and transport has the potential for making microfactories that may be capable of encapsulating various reactions. (Left: reprinted from [46] with
permission from the authors. Right: image courtesy of David F. Savage, Berkeley, CA, USA). (b) Inspired by natural functional complexes, synthetic protein scaffolds were
created [40]. Eukaryotic protein–protein interaction domains (GBD, SH3, PDZ) were expressed as a single polypeptide, creating a synthetic scaffold. Modular protein
domain–ligand interactions were used to colocalize three enzymes of the mevalonate pathway: AtoB, HMGS, and HMGR. The stoichiometries of the enzymes could be
controlled by changing the numbers (x, y, z) of each of the scaffold domains. Scaffolds balance metabolic flux, sequester a toxic intermediate, HMG-CoA, and create a high
local concentration of intermediates, effectively mimicking nature’s strategies of compartmentalization. (c) Scaffolds are hypothesized to oligomerize into large complexes
that resemble metabolite microdomains, trapping intermediates in the interior and quickly converting them to product before they escape [11].

synthetic biologists have drawn inspiration from nature to
design synthetic compartmentalization systems.
Compartmentalization by tethering using protein
scaffolds
Nature uses proteins and protein–protein interactions to
build functional multienzyme complexes. In some cases,
non-catalytic scaffold proteins are used to assemble these
complexes. One example of such a scaffold protein is Ste5,
which selectively brings together MAPKK Ste7 and its
substrate MAPK Fus3, promoting phosphorylation in this
signaling cascade and disfavoring competing substrates
[37]. Another example is scaffoldins, which organize pro-
tein subunits into the cellulosome and assist in the break-
down of cellulose [35]. Bacterial microcompartments, large
assemblies comprising thousands of protein subunits, also
function as organizational devices in the cell. Attempts to
mimic natural protein scaffolding mechanisms have in-
cluded crosslinking enzymes [38], enzyme immobilization
664
on solid substrates in vitro [39], and direct enzyme fusion
[10].
Inspired by naturally existing enzyme complexes, a syn-
thetic protein scaffold, capable of colocalizing enzymes and
increasing product titers, was produced [40] (Figure 2b).
Eukaryotic protein–protein interaction domains were fused
to form scaffolds. Each of the domains selectively recruited
enzymes fused to cognate peptide ligands, colocalizing them
to the scaffold.
Protein scaffolds have been tested on various pathways
and increases in titers were shown. The mevalonate path-
way, for example, produces a toxic intermediate and suf-
fers from flux imbalance due to differing enzyme turnover
rates. Scaffolding increased product yields by 77-fold [40].
The glucaric acid pathway was also scaffolded, and 5-fold
improvement was seen over a high baseline titer of 0.6 g/L
[41]. This pathway requires high levels of intermediate
myoinositol to drive myoinositol oxygenase kinetics. A final
example is biological hydrogen production, which suffers
Review
from competing reactions and requires close contact of the
two enzymes involved. This pathway benefited by about
fivefold from protein scaffolding [42]. In principle, these
synthetic protein scaffolds can replicate natural compart-
mentalization strategies found in multienzyme complexes,
thereby alleviating the problems sometimes found in engi-
neered metabolic reactions.
The extent of the increase in yields due to scaffolding
depends on scaffold stoichiometry. Nature uses gene du-
plication events and evolution of protein–protein interac-
tions, events that operate on a long timescale, to change
the ratio of each of the enzymes in multienzyme complexes.
Synthetic scaffolds similarly enable us to build and test
many different enzyme stoichiometries, but in a short
period of time, to determine the optimal ratio for produc-
tion of a compound of interest. Varying the number of each
interaction domain in the scaffold was shown to dramati-
cally affect product titers [40]. In this way, designing
stoichiometries and geometries of synthetic protein scaf-
folds may result in more effective compartmentalization
mechanisms.
Although the exact mechanism by which protein scaf-
folds increase yields remains unknown, it has been hypoth-
esized that they may form higher order complexes due to
enzyme oligomerization [11] (Figure 2c), similar to natu-
rally occurring metabolite microdomains, such as Ca2+ or
cAMP microdomains [43,44]. If true, intermediates pro-
duced within the complex would be quickly consumed
before they could escape into the cellular milieu, thereby
reducing their toxicity. This would allow protein scaffolds
to more closely approximate bacterial microcompartments.
Compartmentalization using protein encapsulation
Bacterial microcompartments are relatively closed to the
surrounding environment with protein-based pores that
likely regulate transport of substrate into and out of the
proteinaceous shell (Figure 2a) [45]. This strategy effec-
tively isolates reaction intermediates from the rest of the
cellular milieu and could create high local concentrations of
substrates, thereby increasing product yields.
Microcompartments have been heterologously
expressed and could be used to encapsulate foreign path-
ways. As such, these microcompartments have the poten-
tial to increase pathway flux, making them useful tools for
metabolic engineering. For example, carboxysomes from
cyanobacteria (Figure 2a) were heterologously expressed
in Escherichia coli and the encapsulated RuBisCO was
functional in in vitro assays [46]. The Eut and Pdu micro-
compartments from Salmonella have also been expressed
in E. coli [47,48].
Many challenges remain before heterologously
expressed microcompartments can be utilized in metabolic
engineering. Targeting of foreign enzymes to microcom-
partments is an area of ongoing research and localization
sequences have been found for the Pdu and Eut micro-
compartments [47,49]. However, although associations be-
tween carboxysome shell proteins and other protein
components have been shown [50], targeting novel
enzymes to the carboxysome remains not well understood
[46]. Another challenge is elucidating the mechanism of
substrate transport across the microcompartment shell. In
Trends in Cell Biology December 2012, Vol. 22, No. 12
carboxysomes, diffusion of small molecules across the shell
is likely, but pores that suggest active regulation of sub-
strate crossing have also been revealed in crystal struc-
tures [45]. Heterologously expressed microcompartments,
isolated from their native cellular environment, may help
us better understand these mechanisms.
Spatial organization of microcompartments can be reg-
ulated within a cell, providing the cell with another mech-
anism to optimize metabolic processes. For instance,
carboxysomes are aligned along the major axis of the cell
[51] (Figure 2a, right) and the bacterial cytoskeletal com-
ponent ParA is required for proper alignment to occur. The
exact interaction between the cytoskeleton and the bacte-
rial microcompartment remains unknown and is an active
area of research. This knowledge would enable the design
and construction of a novel synthetic cytoskeletal scaffold-
ing device. By inducing polymerization and depolymeriza-
tion of cytoskeletal proteins in response to substrate
availability or other environmental signals, we may be
able to control the spatial organization and number of
compartments tethered to this synthetic cytoskeleton. This
may allow dynamic regulation of metabolic reactions in
synthetic microcompartments.
Another bacterial microcompartment-like encapsula-
tion device was recently engineered. A lumazine synthase
capsid from Aquifex aeolicus was heterologously expressed
in E. coli [52]. Lumazine synthase, an enzyme catalyzing
riboflavin synthesis, forms icosahedral assemblies that do
not naturally encapsulate other enzymes. However, elec-
trostatic interactions were engineered, enabling it to en-
capsulate a toxic enzyme, HIV protease. Directed evolution
was used to select for a capsid with higher loading capacity.
In addition, the assembly state of lumazine synthase was
controlled through engineered point mutations, which
changed the quaternary structure of the capsid, thereby
changing the size and structure of the capsid, without
disrupting the secondary or tertiary structure [53]. The
ability to insulate toxicity from the cytoplasm using a
designable compartmentalization system is important
for metabolic engineering. It allows the expression of het-
erologous pathways that may otherwise be toxic, or even
lethal, to the cell.
Overall, protein assemblies have been used effectively
to organize metabolic reactions and increase product
yields. Thus far, however, such increases have been low.
Therefore, the present challenge is to further increase
yields to levels high enough for industrial applications.
To do so, the mechanism of scaffold action must be further
elucidated. In addition, scaffolds have been applied to only
a few select pathways as proof of principle. Generality and
scalability of scaffolds will require a better understanding
of localization to synthetic scaffolds.
Compartmentalization using nucleic acids
DNA and RNA nanotechnology is a field of research that
has many promising applications in medicine and industry
[54–56]. Short strands of DNA or RNA can fold into various
structures or assemble into dimer or multimer building
blocks (Figure 3c). These building blocks can then poly-
merize into various 3D structures in vitro, including
simple structures such as sheets, as well as more complex
665
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a) (b) Hydrogenase Ferredoxin } Enzyme

Protein
} adapter

} RNA aptamer

} Scaffold

{
(c)

Dimerize
Polymerize
{

Dimerizaon
{

domain
Polymerizaon
domain
TRENDS in Cell Biology

Figure 3. Nucleic acid scaffolds: mimicking the diverse geometries found in nature. (a) DNA and RNA nanotechnology can produce many different structures in vitro,
including tubes and sheets [86]. Although nucleic acids are not frequently used in nature for scaffolding, the ease with which synthetic biologists can design and build these
structures make them a versatile tool for building synthetic scaffolds that can mimic various natural compartmentalization strategies of different geometries. (Adapted, with
permission, from [86].) (b) Synthetic RNA scaffolds were designed in silico and expressed in vivo [75]. Binding between RNA aptamers and protein adapters facilitated
recruitment of enzymes to the scaffold. Hydrogenase and ferredoxin, components of a hydrogen production pathway, were scaffolded and pathway output was increased.
(c) Single strands of RNA with aptamers dimerize and these building blocks polymerize into 2D sheets and nanotubes. Polymerization domains are protected with hairpins
before dimerization to prevent self-binding and assure order of assembly. The versatile architectures of RNA nanostructures in vivo allow the mimicking of various natural
compartmentalization systems.

structures such as tubes and capsules [57,58] (Figure 3a).
In this way, DNA and RNA nanotechnology can be used to
build complexes that structurally mimic natural organiza-
tional systems, such as enzyme channels and bacterial
microcompartments.
Nucleic acid structures can also functionally mimic nat-
ural organizational systems. That is, proteins can be tar-
geted directly to the scaffolds. Several multienzyme systems
have been successfully organized using nucleic acids in vitro.
NADH-flavin mononucleotide (FMN) oxidoreductase and
luciferase, enzymes that catalyze two consecutive reaction
steps, were assembled onto DNA scaffolds using streptavi-
din–biotin linkages and coupled enzymatic activity was
increased relative to unscaffolded enzymes [59]. Glucose
oxidase (GOX) and horseradish peroxidase (HRP) were
immobilized through covalent conjugation to DNA oligonu-
cleotides and an increase in reactivity was seen in both
scaffolded systems compared with unscaffolded enzymes
alone [60]. Other enzyme cascades have been organized
using large supramolecular DNA scaffolds [61] and DNA
origami [62] and reactivity was found to depend on relative
enzyme positioning on the scaffold.
Although traditional in vitro assembly of DNA or RNA
nanostructures relies on physiologically unsustainable
temperatures and slow cool down to denature and anneal
designed strands correctly [63], isothermal strategies for
both DNA and RNA were recently developed [64–66]. This
paved the way for in vivo applications of RNA and DNA
scaffolds.
666
A DNA scaffold, in the form of a plasmid, was con-
structed. It recruited enzymes via zinc finger domains that
bound to specific motifs on the scaffold [67]. Several
enzymes were tested using this system, including path-
ways for the production of resveratrol, 1,2-propanediol,
and mevalonate. Yields were found to increase as a func-
tion of scaffold architecture, similar to protein scaffolds.
Scaffolds with more complex geometries were con-
structed using designed non-coding (nc)RNA (Figure 3b,c).
As a building material, RNA has several advantages.
Because the base-pairing interactions in RNA (and DNA)
are well studied [68,69], secondary and tertiary structures of
nucleic acids can be predicted in silico [70–72] with nano-
meter precision [57] and are relatively easy to design [55].
RNA can also be expressed in large amounts by engineered
cells and can stably exist outside the nucleus in eukaryotic
cells, making it portable to a wide range of hosts. Addition-
ally, RNA can recruit proteins fused to adapters via RNA
aptamers, which can be evolved to bind with high specificity
to various ligands [73,74].
A single molecule of RNA (approximately 100 bases) can
fold into a linear discrete scaffold that mimics multien-
zyme complexes in nature [75] (Figure 3b). The RNA
molecule comprises multiple different aptamer motifs,
which can recruit proteins, as well as complementation
regions, which form the scaffold (Figure 3b). This scaffold is
relatively easy to build and characterize and is suitable for
expression of specific stoichiometries of enzymes in a chain
[75]. Scale up, which may allow production at industrially
Review
feasible titers, is also possible; numerous RNA aptamer
domains can be connected in a single molecule of RNA. In
this way, almost 100 proteins were recruited in a chain for
visualization experiments [76], although this has not yet
been adapted for metabolic engineering purposes.
An RNA nanotube was built using short single-stranded
building blocks that polymerized in vivo [75] (Figure 3c).
Here, the directionality of enzyme loading was not speci-
fied; that is, enzymes were tethered to both the interior and
the exterior of the nanotube. However, loading enzymes
selectively to the interior has been shown in vitro [77]. If
such selective loading could be accomplished in vivo, a
modified version of this 1D scaffold could act like a pipe,
preventing intermediates from escaping and increasing
product yields. Another scaffold architecture, a 2D sheet,
was constructed using a similar strategy and product yield
was highest when using this scaffold (Figure 3c). Although
there is no directly analogous system in nature, this
2D-scaffold is likely to function by increasing the number
of neighbors any enzyme has in close proximity, allowing
exploration of the quantitative effect of colocalizing addi-
tional proteins in two dimensions; for example, by recruit-
ing them to a membrane.
As a demonstration of feasibility, [FeFe]-hydrogenase
and ferredoxin, enzymes involved in the biological produc-
tion of hydrogen [78], were recruited to RNA scaffolds. This
pathway is a useful test bed because it suffers from com-
peting reactions and requires the enzymes to be in close
proximity [42]. Hydrogen output was found to increase as a
function of scaffold architecture.
DNA and RNA scaffolds allow metabolic engineers to
mimic many different strategies implemented by nature
for substrate channeling and reducing competition to
increase pathway flux. In vitro work provides a wide array
of tested architectures and functionalities that mimic
various natural compartmentalization systems in differ-
ent ways. This ability to design precise scaffold geometry
is a major advantage of nucleic acid scaffolds over protein
and lipid compartmentalization systems, where structure
prediction and design remain difficult. However, the in
vitro to in vivo transition is a major challenge for nucleic
acid scaffolds and is not always possible due to many
constraints, such as isothermal assembly. Specific control
over scaffold architectures and the ability to design cus-
tomized shapes with RNA and DNA isothermally in vivo
are crucial next steps in utilizing this technology to in-
crease yields. If successful, RNA and DNA may even be
used to build scaffolds in architectures not found in nature.
Although the effects of such structures on the cell’s meta-
bolic load and the crosstalk with native systems may be
problematic, it raises interesting possibilities for metabol-
ic engineers to exceed nature’s capabilities to produce
useful compounds.
RNA scaffolds currently have an added advantage over
DNA and other types of scaffold in that the amount of
scaffold expression can be controlled. However, the effect of
expressing high amounts of nucleic acid on cell growth and
viability is not well understood and titration of scaffold
expression has yet to be performed.
The generality of nucleic acid scaffolds remains to be
demonstrated, because they have been used in only a
Trends in Cell Biology December 2012, Vol. 22, No. 12
few select pathways. Further application of nucleic acid
scaffolds will require generation of additional orthogonal
RNA aptamers or zinc finger domains. The application of
RNA scaffolds, with its diverse architectures, to various
pathways will not only allow further improvement of the
technology, but will also allow testing of the mechanism of
scaffold action. For example, the effect of scaffolding is
hypothesized to be greater for non-diffusible intermediates
compared with diffusible intermediates if local high con-
centration of intermediates is an important contributing
factor. Overall, RNA scaffolds are capable of becoming
versatile, designable scaffolds that can mimic natural
systems of differing geometries, compartmentalize various
metabolic pathways, and increase yields, possibly beyond
nature’s capabilities.
Compartmentalization using lipids
Lipids, often in the form of membranes, are widely used to
encapsulate reactions in nature. Lipid vesicles and oil
emulsions have been used to perform a wide variety of
reactions in vitro, such as gene expression [79], sequencing,
and evolution of new enzymes [80]. Even self-replicating
lipid systems have been created: protocells that are capa-
ble of catalyzing RNA reactions and processing ‘food’
micelles [81]. Although these systems are effective at
encapsulation of reactions, targeting specific enzymes to
synthetic lipid vesicles remains difficult because localiza-
tion signals or other specific targeting mechanisms to lipid
vesicles are unknown.
Another compartmentalization strategy utilizes natu-
rally existing membrane-bound organelles. Unlike lipid
vesicles, many membrane-bound organelles have well
studied localization mechanisms [82]. One study targeted
methyl halide transferase to yeast vacuoles to increase
methyl iodide production [83]. The increases seen were
likely to have been due to added access to a cofactor and
sequestration of halogenated products. Also, targeting of
terpenoid production pathway components to the mito-
chondria significantly increased yields [84]. The advantage
of this system is that it allows engineering of eukaryotic
cells such as commercially important yeast.
Lipid compartmentalization systems remain a relative-
ly unexplored area of synthetic biology, because they have
many weaknesses compared with nucleic acid or protein
scaffolds. For example, controlling the architecture of
compartments and stoichiometries of encapsulated
enzymes is challenging. Other major challenges for in vivo
use of lipid compartmentalization devices include secretion
of final products and, in the case of native organelle utili-
zation, potential interactions with the host metabolism. A
better understanding of these factors and studies in natu-
ral lipid compartmentalization systems may yield results
that can inform the development of synthetic lipid com-
partments.
Using synthetic scaffolds to understand the limitations
of natural systems
Studying engineered scaffolds may give us new insight into
the function and design principles of natural compartmen-
talization strategies. For example, scaffold architecture
and enzyme stoichiometries were found to be important
667
Review
to pathway output [40,75]. Through scaffolding pathways
with known stoichiometries and differing architectures, it
may be possible to probe the limitations of native systems
and even improve on them.
Knocking out a native compartment, replacing it with a
synthetic one, and studying the resulting effects on the cell
may allow us to decouple biological functions. This will
help elucidate the functions and limitations of the native
system. Conversely, studying heterologous expression of
natural compartments and targeting will allow us to
determine the components that are necessary and suffi-
cient for proper function. Mechanisms of localization,
assembly, and substrate gating may be elucidated from
such studies.
In addition, the well-studied examples of natural
compartmentalization systems have comprised primarily
proteins and lipid membranes. Synthetic systems have
shown, for example, that nucleic acids can function as
scaffolds. However, RNA and DNA are not commonly used
as materials for scaffolding in nature. Yeast telomerase is
one of a few known examples of RNA serving as a flexible
scaffold for protein subunits [85]. Given findings of the
capabilities of a synthetic RNA scaffold, it may be that an
as yet to be discovered function of ncRNA, which consti-
tutes 80% of total RNA, is scaffolding. Synthetic scaffolds,
in this case, may help us elucidate the organizational
function of the material from which the scaffold was made.
Concluding remarks
Cells have evolved strategies for regulating and maintain-
ing proper metabolic flux. Synthetic biologists have taken
inspiration from nature and developed synthetic systems
that solve some of the same problems. From direct mimicry
using proteins to more novel solutions using nucleic acids,
synthetic biologists have devised promising strategies for
increasing metabolic production of industrially useful che-
micals. However, some major challenges remain (Box 1).
For example, scaffolds must be applicable to various path-
ways outside the test pathways chosen in the proof-of-
principle studies conducted so far. Further increases in
yields must be obtained before scale up to industrially
relevant quantities can be achieved. Utilizing multiple
types of scaffold in the same system, if feasible, may help
accomplish this, because improvements may be made or-
thogonally by each scaffold. In addition, the effect of such
high levels of production on cell growth must be minimized.
To overcome these challenges, a better understanding of
scaffold mechanism of action and enzyme localization must
be achieved. If these efforts prove successful, scaffolds can
be widely applied as an orthogonal novel method of in-
creasing yields in metabolic engineering efforts. In addi-
tion, implementing and studying synthetic systems will
help us understand the native biological organization of
the cell.
Acknowledgments
The authors thank D. Ducat, C.J. Delebecque, J.E. Dueber, G. Sachdeva,
D.F. Savage, J. Torella, and G.C. Wu for discussions and helpful
suggestions. This work is supported by the National Institutes of
Health (Grant #GM080177) and a Graduate Research Fellowship from
the National Science Foundation to A.H.C. and by the Wyss Institute for
Biologically Inspired Engineering and Department of Defense to P.A.S.
668
Trends in Cell Biology December 2012, Vol. 22, No. 12
Box 1. Outstanding questions
 What is the mechanism by which scaffolds operate to increase
yield? Do they form higher order structures?
 How can we predictably engineer scaffold architecture and control
assembly?
 For what pathways are different scaffolds most useful? Is there a
pathway size limitation? Does the diffusability of intermediate
substrates influence the effectiveness of scaffold action?
 How do substrates enter and exit synthetic compartments? How
are enzymes localized?
 How can we further improve yields using synthetic scaffold
technologies? How do we scale up the existing technologies to
industrial scale production?
 Are there other undiscovered natural scaffolding systems that
utilize proteins, nucleic acids, and lipids? How does biological
organization function natively?
References
1 Hurtley, S. (2009) Spatial cell biology. Location, location, location.
Introduction. Science 326, 1205
2 Kerfeld, C.A. et al. (2010) Bacterial microcompartments. Annu. Rev.
Microbiol. 64, 391–408
3 Kerfeld, C.A. et al. (2005) Protein structures forming the shell of
primitive bacterial organelles. Science 309, 936–938
4 Berlatzky, I.A. et al. (2008) Spatial organization of a replicating
bacterial chromosome. Proc. Natl. Acad. Sci. U.S.A. 105, 14136–14140
5 Komeili, A. (2012) Molecular mechanisms of compartmentalization and
biomineralization in magnetotactic bacteria. FEMS Microbiol. Rev. 36,
232–255
6 Savage, D.F. et al. (2008) Defossiling fuel: how synthetic biology can
transform biofuel production. ACS Chem. Biol. 3, 13–16
7 Lee, S.K. et al. (2008) Metabolic engineering of microorganisms for
biofuels production: from bugs to synthetic biology to fuels. Curr. Opin.
Biotechnol. 19, 556–563
8 Chang, M.C.Y. and Keasling, J.D. (2006) Production of isoprenoid
pharmaceuticals by engineered microbes. Nat. Chem. Biol. 2, 674–681
9 Khalil, A.S. and Collins, J.J. (2010) Synthetic biology: applications
come of age. Nat. Rev. Genet. 11, 367–379
10 Conrado, R.J. et al. (2008) Engineering the spatial organization of
metabolic enzymes: mimicking nature’s synergy. Curr. Opin.
Biotechnol. 19, 492–499
11 Lee, H. et al. (2011) Spatial organization of enzymes for metabolic
engineering. Metab. Eng. 14, 242–251
12 Agapakis, C.M. et al. (2012) Natural strategies for the spatial
optimization of metabolism in synthetic biology. Nat. Chem. Biol. 8,
527–535
13 Eichelmann, H. et al. (2009) Rubisco in planta kcat is regulated in
balance with photosynthetic electron transport. J. Exp. Bot. 60, 4077–
4088
14 Jørgensen, K. et al. (2005) Metabolon formation and metabolic
channeling in the biosynthesis of plant natural products. Curr.
Opin. Plant Biol. 8, 280–291
15 Manjasetty, B.A. et al. (2003) Crystal structure of a bifunctional
aldolase-dehydrogenase: sequestering a reactive and volatile
intermediate. Proc. Natl. Acad. Sci. U.S.A. 100, 6992–6997
16 Liu, Y. et al. (2012) Methanogens: a window into ancient sulfur
metabolism. Trends Microbiol. 20, 251–258
17 Xu, P. et al. (2011) Genome-scale metabolic network modeling results
in minimal interventions that cooperatively force carbon flux towards
malonyl-CoA. Metab. Eng. 13, 578–587
18 Hyde, C.C. et al. (1988) Three-dimensional structure of the tryptophan
synthase alpha 2 beta 2 multienzyme complex from Salmonella
typhimurium. J. Biol. Chem. 263, 17857–17871
19 Pfeifer, B.A. and Khosla, C. (2001) Biosynthesis of polyketides in
heterologous hosts. Microbiol. Mol. Biol. Rev. 65, 106–118
20 Martin, W. (2010) Evolutionary origins of metabolic
compartmentalization in eukaryotes. Philos. Trans. R. Soc. Lond. B:
Biol. Sci. 365, 847–855
21 Gehrmann, W. and Elsner, M. (2011) A specific fluorescence probe for
hydrogen peroxide detection in peroxisomes. Free Radic. Res. 45, 501–
506
Review
22 Losick, R. and Shapiro, L. (1999) Changing views on the nature of the
bacterial cell: from biochemistry to cytology. J. Bacteriol. 181, 4143–
4145
23 Yeates, T.O. et al. (2008) Protein-based organelles in bacteria:
carboxysomes and related microcompartments. Nat. Rev. Microbiol.
6, 681–691
24 Tanaka, S. et al. (2008) Atomic-level models of the bacterial
carboxysome shell. Science 319, 1083–1086
25 Cannon, G.C. and Shively, J.M. (1983) Characterization of a
homogenous preparation of carboxysomes from Thiobacillus
neapolitanus. Arch. Microbiol. 134, 52–59
26 Tanaka, S. et al. (2009) Insights from multiple structures of the
shell proteins from the beta-carboxysome. Protein Sci. 18, 108–
120
27 Badger, M.R. (2003) CO2 concentrating mechanisms in cyanobacteria:
molecular components, their diversity and evolution. J. Exp. Bot. 54,
609–622
28 Penrod, J.T. and Roth, J.R. (2006) Conserving a volatile metabolite: a
role for carboxysome-like organelles in Salmonella enterica. J.
Bacteriol. 188, 2865–2874
29 Sampson, E.M. and Bobik, T.A. (2008) Microcompartments for B12-
dependent 1,2-propanediol degradation provide protection from DNA
and cellular damage by a reactive metabolic intermediate. J. Bacteriol.
190, 2966–2971
30 Zhang, Y-HP. (2011) Substrate channeling and enzyme complexes for
biotechnological applications. Biotechnol. Adv. 29, 715–725
31 Dunn, M.F. (2012) Allosteric regulation of substrate channeling and
catalysis in the tryptophan synthase bienzyme complex. Arch.
Biochem. Biophys. 519, 154–166
32 Xu, Y. and Abeles, R.H. (1993) Inhibition of tryptophan synthase by (1-
fluorovinyl)glycine. Biochemistry 32, 806–811
33 Tran, L. et al. (2010) Insights into protein-protein and enzyme-
substrate interactions in modular polyketide synthases. Chem. Biol.
17, 705–716
34 Thoden, J.B. et al. (1997) Structure of carbamoyl phosphate
synthetase: a journey of 96 A from substrate to product.
Biochemistry 36, 6305–6316
35 Fontes, C.M.G.A. and Gilbert, H.J. (2010) Cellulosomes: highly
efficient nanomachines designed to deconstruct plant cell wall
complex carbohydrates. Annu. Rev. Biochem. 79, 655–681
36 Keasling, J.D. (2010) Manufacturing molecules through metabolic
engineering. Science 330, 1355–1358
37 Zeke, A. et al. (2009) Scaffolds: interaction platforms for cellular
signalling circuits. Trends Cell Biol. 19, 364–374
38 Sheldon, R. (2007) Enzyme immobilization: the quest for optimum
performance. Adv. Synth. Catal. 349, 1289–1307
39 Mosbach, K. and Mattiasson, B. (1970) Matrix-bound enzymes. II.
Studies on a matrix-bound two-enzyme-system. Acta Chem. Scand.
24, 2093–2100
40 Dueber, J.E. et al. (2009) Synthetic protein scaffolds provide modular
control over metabolic flux. Nat. Biotechnol. 27, 753–759
41 Moon, T.S. et al. (2010) Use of modular, synthetic scaffolds for improved
production of glucaric acid in engineered E. coli. Metab. Eng. 12, 298–
305
42 Agapakis, C.M. et al. (2010) Insulation of a synthetic hydrogen
metabolism circuit in bacteria. J. Biol. Eng. 4, 3
43 Laude, A.J. and Simpson, A.W.M. (2009) Compartmentalized
signalling: Ca2+ compartments, microdomains and the many facets
of Ca2+ signalling. FEBS J. 276, 1800–1816
44 Baillie, G.S. (2009) Compartmentalized signalling: spatial regulation
of cAMP by the action of compartmentalized phosphodiesterases.
FEBS J. 276, 1790–1799
45 Klein, M.G. et al. (2009) Identification and structural analysis of a
novel carboxysome shell protein with implications for metabolite
transport. J. Mol. Biol. 392, 319–333
46 Bonacci, W. et al. (2012) Modularity of a carbon-fixing protein
organelle. Proc. Natl. Acad. Sci. U.S.A. 109, 478–483
47 Choudhary, S. et al. (2012) Engineered protein nano-compartments for
targeted enzyme localization. PLoS ONE 7, e33342
48 Parsons, J.B. et al. (2010) Synthesis of empty bacterial
microcompartments, directed organelle protein incorporation, and
evidence of filament-associated organelle movement. Mol. Cell 38,
305–315
Trends in Cell Biology December 2012, Vol. 22, No. 12
49 Fan, C. et al. (2010) Short N-terminal sequences package proteins into
bacterial microcompartments. Proc. Natl. Acad. Sci. U.S.A. 107, 7509–
7514
50 Kinney, J.N. et al. (2012) Elucidating the essential role of the conserved
carboxysomal protein CcmN reveals a common feature of bacterial
microcompartment assembly. J. Biol. Chem. 287, 17729–17736
51 Savage, D.F. et al. (2010) Spatially ordered dynamics of the bacterial
carbon fixation machinery. Science 327, 1258–1261
52 Wörsdörfer, B. et al. (2011) Directed evolution of a protein container.
Science 331, 589–592
53 Chen, H-N. and Woycechowsky, K.J. (2012) Conversion of a
dodecahedral protein capsid into pentamers via minimal point
mutations. Biochemistry 51, 4704–4712
54 Pinheiro, A.V. et al. (2011) Challenges and opportunities for structural
DNA nanotechnology. Nat. Nanotechnol. 6, 763–772
55 Guo, P. (2010) The emerging field of RNA nanotechnology. Nat.
Nanotechnol. 5, 833–842
56 Famulok, M. and Ackermann, D. (2010) RNA nanotechnology: inspired
by DNA. Nat. Nanotechnol. 5, 634–635
57 Aldaye, F.A. et al. (2008) Assembling materials with DNA as the guide.
Science 321, 1795–1799
58 Wei, B. et al. (2012) Complex shapes self-assembled from single-
stranded DNA tiles. Nature 485, 623–626
59 Niemeyer, C.M. et al. (2002) DNA-directed assembly of bienzymic
complexes from in vivo biotinylated NAD(P)H:FMN oxidoreductase
and luciferase. Chembiochem 3, 242–245
60 Müller, J. and Niemeyer, C.M. (2008) DNA-directed assembly of
artificial multienzyme complexes. Biochem. Biophys. Res. Commun.
377, 62–67
61 Wilner, O.I. et al. (2009) Enzyme cascades activated on topologically
programmed DNA scaffolds. Nat. Nanotechnol. 4, 249–254
62 Numajiri, K. et al. (2010) Discrete and active enzyme nanoarrays on
DNA origami scaffolds purified by affinity tag separation. J. Am. Chem.
Soc. 132, 9937–9939
63 Castro, C.E. et al. (2011) A primer to scaffolded DNA origami. Nat.
Methods 8, 221–229
64 Afonin, K.A. et al. (2010) In vitro assembly of cubic RNA-based scaffolds
designed in silico. Nat. Nanotechnol. 5, 676–682
65 Jungmann, R. et al. (2008) Isothermal assembly of DNA origami
structures using denaturing agents. J. Am. Chem. Soc. 130, 10062–
10063
66 Yin, P. et al. (2008) Programming biomolecular self-assembly
pathways. Nature 451, 318–322
67 Conrado, R.J. et al. (2012) DNA-guided assembly of biosynthetic
pathways promotes improved catalytic efficiency. Nucleic Acids Res.
40, 1879–1889
68 Jaeger, L. and Chworos, A. (2006) The architectonics of programmable
RNA and DNA nanostructures. Curr. Opin. Struct. Biol. 16, 531–
543
69 Rothemund, P.W.K. (2006) Folding DNA to create nanoscale shapes
and patterns. Nature 440, 297–302
70 Zadeh, J.N. et al. (2011) NUPACK: analysis and design of nucleic acid
systems. J. Comput. Chem. 32, 170–173
71 Zuker, M. (2003) Mfold web server for nucleic acid folding and
hybridization prediction. Nucleic Acids Res. 31, 3406–3415
72 Andronescu, M. (2003) RNAsoft: a suite of RNA secondary structure
prediction and design software tools. Nucleic Acids Res. 31, 3416–
3422
73 Ellington, A.D. and Szostak, J.W. (1990) In vitro selection of RNA
molecules that bind specific ligands. Nature 346, 818–822
74 Tuerk, C. and Gold, L. (1990) Systematic evolution of ligands by
exponential enrichment: RNA ligands to bacteriophage T4 DNA
polymerase. Science 249, 505–510
75 Delebecque, C.J. et al. (2011) Organization of intracellular reactions
with rationally designed RNA assemblies. Science 333, 470–474
76 Golding, I. and Cox, E.C. (2004) RNA dynamics in live Escherichia coli
cells. Proc. Natl. Acad. Sci. U.S.A. 101, 11310–11315
77 Douglas, S.M. et al. (2007) DNA-nanotube-induced alignment of
membrane proteins for NMR structure determination. Proc. Natl.
Acad. Sci. U.S.A. 104, 6644–6648
78 Ducat, D.C. et al. (2011) Rewiring hydrogenase-dependent redox
circuits in cyanobacteria. Proc. Natl. Acad. Sci. U.S.A. 108, 3941–
3946
669
Review
79 Nomura, S.M. et al. (2003) Gene expression within cell-sized lipid
vesicles. Chembiochem 4, 1172–1175
80 Channon, K. et al. (2008) Synthetic biology through biomolecular
design and engineering. Curr. Opin. Struct. Biol. 18, 491–498
81 Zhu, T.F. and Szostak, J.W. (2009) Coupled growth and division of
model protocell membranes. J. Am. Chem. Soc. 131, 5705–5713
82 Siddiqui, M.S. et al. (2012) Advancing secondary metabolite
biosynthesis in yeast with synthetic biology tools. FEMS Yeast Res.
12, 144–170
670
Trends in Cell Biology December 2012, Vol. 22, No. 12
83 Bayer, T.S. et al. (2009) Synthesis of methyl halides from biomass using
engineered microbes. J. Am. Chem. Soc. 131, 6508–6515
84 Farhi, M. et al. (2011) Harnessing yeast subcellular compartments for
the production of plant terpenoids. Metab. Eng. 13, 474–481
85 Zappulla, D.C. and Cech, T.R. (2004) Yeast telomerase RNA: a flexible
scaffold for protein subunits. Proc. Natl. Acad. Sci. U.S.A. 101, 10024–
10029
86 Douglas, S.M. et al. (2009) Self-assembly of DNA into nanoscale three-
dimensional shapes. Nature 459, 414–418
 Review
Special Issue – Synthetic Cell Biology
Directed cytoskeleton self-organization
Timothée Vignaud, Laurent Blanchoin, and Manuel Théry
Laboratoire de Physiologie Cellulaire et Végétale, Institut de Recherche en Technologies et Sciences pour le Vivant, CNRS/UJF/
INRA/CEA, 17 Rue des Martyrs, 38054, Grenoble, France
The cytoskeleton architecture supports many cellular
functions. Cytoskeleton networks form complex intra-
cellular structures that vary during the cell cycle and
between different cell types according to their physio-
logical role. These structures do not emerge spontane-
ously. They result from the interplay between intrinsic
self-organization properties and the conditions imposed
by spatial boundaries. Along these boundaries, cytoskel-
eton filaments are anchored, repulsed, aligned, or reor-
iented. Such local effects can propagate alterations
throughout the network and guide cytoskeleton assem-
bly over relatively large distances. The experimental
manipulation of spatial boundaries using microfabrica-
tion methods has revealed the underlying physical pro-
cesses directing cytoskeleton self-organization. Here we
review, step-by-step, from molecules to tissues, how the
rules that govern assembly have been identified. We
describe how complementary approaches, all based
on controlling geometric conditions, from in vitro recon-
struction to in vivo observation, shed new light on these
fundamental organizing principles.
Setting boundaries
The reproducible shape and spatial organization of organs
imply the existence of deterministic rules directing the
assembly of complex biological structures. Organ shape
depends on cell architecture, which is supported by cyto-
skeleton networks. The formation of defined and geomet-
rically controlled intracellular structures relies on the self-
organization properties of the cytoskeleton. The contribu-
tion of self-organization in cell biology is vast and now well
documented [1]. Cytoskeleton self-organization is a process
in which the consumption (physicists would say dissipa-
tion) of energy brings the cytoskeleton away from its
thermodynamic equilibrium (i.e., a disordered mixture of
poorly dynamic filaments) toward defined and reproducible
steady states. This differs from the process of self-assem-
bly, in which components assemble spontaneously – with-
out an external energy source – to form a structure
corresponding to the thermodynamic equilibrium. Depend-
ing on the rules regulating the interaction of cytoskeleton
components, complex structures may self-organize in a
robust manner. The purpose of much of the research
described in this review has been to identify and formulate
these rules to understand how physical principles direct
biological morphogenesis.
Corresponding author: Théry, M. (manuel.thery@cea.fr).
Keywords: actin; microtubule; architecture; polarity; microfabrication;
micropatterning.
0962-8924/$ – see front matter ß 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tcb.2012.08.012 Trends in Cell Biology, December 2012, Vol. 22, No. 12
Cytoskeleton self-organization is partially regulated by
the action of proteins modulating the biochemical rules of
filament growth and interactions. The combination of sim-
ple biochemical rules can lead to the formation of complex
structures [2]. Robust patterns can emerge from oriented
displacements of cytoskeleton filaments by molecular
motors in the absence of any external guidance [3–5].
However, these autonomous self-organization processes
are extremely sensitive to the presence of spatial boundary
conditions (SBCs). An SBC is an external geometrical cue,
within or at the periphery of the network, that can locally
affect the self-organization of the network. For tissues, an
SBC can be a frontier with an external fluid or a contact
with bone, muscle, or other organ. For a cell, an SBC can be
a neighboring cell or extracellular matrix (ECM). For
intracellular cytoskeleton networks, an SBC can be a cell
adhesion for the actin network, a centrosome for the mi-
crotubule (MT) network, or a frontier such as the plasma
membrane or an intracellular organelle.
How an SBC can direct an autonomous self-organization
process is the subject of this review. We describe recent
advances in the understanding of the role of SBCs in the
self-organization of actin networks and MT arrays, how
these processes are integrated in the internal organization
of a cell, and how this in turn affects tissue architecture. In
the formation of cytoskeleton networks, an SBC can bias
monomer diffusion and thereby the assembly process ([6]
and references therein). Here, we focus on the role of
geometrical constraints on the growth, orientation, anchor-
age, and production of mechanical forces during cytoskele-
ton assembly.
Actin network self-organization
Actin is an asymmetric protein that can self-assemble to
form polarized actin filaments [7]. This spontaneous pro-
cess can be accelerated and temporally regulated by the
energy liberated from the release of a phosphate group
from the nucleotide triphosphate bound to actin [8]. Actin
filaments can interact to form actin networks. Actin net-
works can self-organize into several types of structures in
cells: bundles comprising aligned long filaments and mesh-
works comprising branched and intermingled short fila-
ments. Bundles and meshworks form such complex
intricate networks in cells [9] that it is difficult to identify
the principles of their self-organization.
Biochemists have developed alternative methods to
analyze self-organization in controlled conditions in vitro
by mixing, in defined proportions, the individual compo-
nents (either purified from tissues or from recombinant
671
Review
bacteria or yeasts). The kinetic parameters of actin poly-
merization measured in vitro and how these parameters
vary in response to the presence of actin-associated pro-
teins has provided key information about the regulation of
actin assembly dynamics [10]. However, the rules guiding
the spatial organization of the network can be identified
only by using controlled geometric boundary conditions.
Symmetry break
Mechanical constraints in an actin network can induce a
symmetry break (i.e., the sudden occurrence of a singular
axis in isotropic conditions in which all directions were
previously equivalent). This propensity for symmetry
breaking in actin networks was elegantly revealed using
a spherical glass bead coated with actin nucleation factors
as a simple SBC [11,12]. Actin nucleation is induced from
the bead, and the presence of capping proteins, which block
filament elongation from their fast growing end, ensures
that the actin filaments are short and form a dense
branched meshwork. As the actin filaments grow at the
bead surface, material accumulates and the stress
increases in the network up to a critical value inducing
its rupture [13]. The rupture creates an asymmetry in the
pressure applied on the bead such that the bead is dis-
placed (Figure 1a). Repetition of this sequence of events
induces saltatory propulsion of the bead [14,15].
In this experimental system, the SBC can easily be
manipulated by changing its dimensional parameters.
For example, the larger the bead, the shallower the curva-
ture of the bead surface, leading to an increase in the
critical value of network thickness before rupture [16,17]
and the periodicity of the saltatory propulsion (Figure 1b).
An asymmetric SBC can be created using ellipsoidal beads.
The difference in surface curvature of the bead biases the
location of network rupture, which occurs preferentially in
line with or orthogonal to the long axis of the bead [18]
(Figure 1c). Higher aspect ratios, obtained by actin nucle-
ation on small glass rods, further increase the spatial bias
and branched network growth is restricted to being orthog-
onal to the long axis of the rod [15]. As the rod length
increases, several independent networks can form, reveal-
ing the existence of a critical length for subnetwork inter-
connections (Figure 1d). Interestingly, symmetry break
and asymmetric force production are not restricted to
branched meshworks of actin, but can also be induced
by the bundling and alignment of individual filaments
polymerizing against the bead surface [19].
Filament alignment
Several self-organization processes can induce actin fila-
ment alignment in response to an SBC. Filaments can
become aligned by steric interactions. When two long
filaments come close to each other, they prevent the inser-
tion of a short filament between them. Long filaments will
be further forced to align by the steric interactions of short
filaments around them. Steric interactions between long
filament bundles will then promote their orientation in line
with the long axis of the volume in which they are confined
[20] (Figure 1e). Steric interaction of filaments freely mov-
ing on a layer of molecular motors can also result in their
alignment along each other [4] and along the SBC [21].
672
Trends in Cell Biology December 2012, Vol. 22, No. 12
Filaments can become aligned by membrane tension. Two
filaments pushing orthogonally to a deformable membrane
will coalesce and align to reduce the elastic energy of the
membrane [22]. Preassembled filaments can become
aligned by defining the anchorage positions with regular
arrays of beads or micropillars and adding filamin to
crosslink filaments [23,24].
Filaments can become aligned in parallel or antiparallel
configurations by controlling the orientation of their
growth. Surface micropatterning can be used to manipu-
late precisely the geometrical boundary conditions of fila-
ment growth and orientation [25]. Selective adsorption of
actin nucleation-promoting factors on micropatterned
regions induces localized formation of a branched mesh-
work. Only non-branched filaments grow out of the micro-
pattern, with their barbed ends reproducibly oriented
outward. Steric interactions force growing filaments to
align parallel to each other, orthogonal to the nucleation
region (Figure 1g). Distant from the nucleation region, two
filaments growing toward each other in nearly opposite
directions tend to form antiparallel bundles; whereas two
filaments growing toward each other but at an oblique
angle tend to form parallel bundles (Figure 1g). However,
these tendencies can be biased because adjacent filaments
sterically affect each other. The reorientation of filaments
during bundle formation guides adjacent filaments also to
align with the bundle (Figure 1g). Bundle formation is thus
a combination of local probabilistic events, governed by
filament flexibility, and the propagation of the alignment
configuration to adjacent filaments by steric interactions
[25].
In egg extracts, biochemical conditions are less well
defined but closer to intracellular conditions. Encapsula-
tion of egg extracts in membrane vesicles revealed that
filaments nucleated at the periphery move inward and
align to form a central ring. Interestingly, the ring can
form only when nucleation is restricted to the vesicle
periphery and not distributed evenly throughout the entire
volume. A scaling law appears to regulate the ring size in
proportion to the vesicle diameter [26] (Figure 1f).
Network contraction
Myosins are oriented motors moving toward a defined ex-
tremity of actin filaments. Thus, they have specific actions
depending on actin network architecture. They walk along
parallel filaments, whereas they slide along antiparallel
filaments in opposite directions relative to each other and
thus contract the network [27] (Figure 1i). Myosins can also
induce the contraction of branched meshworks, because
these networks also contain antiparallel filaments. Howev-
er, the rate of contraction is reduced due to the resistance
associated with branches and network anchoring to nucle-
ation regions [27]. It has been shown, based on the use of
actomyosin bundles connecting beads, that the contraction
rate is proportional to bundle length [28]. In more complex
structures comprising various types of network, the contrac-
tion rate is determined by the local proportion of parallel and
antiparallel bundles and branched meshwork [27]. Varia-
tions of these proportions in a given architecture will induce
anisotropic contraction, although myosins are present
throughout the network (Figure 1i). Therefore, an SBC
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a) Symmetry break

Time

(b) Nucleaon size (c) Nucleaon shape (e) Container shape

(d) (f)

(g)

(h) (i)

Time Time

Key:
Acn bundles: Acn meshwork:
Acn nucleaon
Intermingled and Pressure
micropaern An-parallel filaments
branched filaments Tension
Confinement Parallel filaments
volume

TRENDS in Cell Biology

Figure 1. Actin network self-organization. (a) Actin meshwork polymerization around beads leads to symmetry breaking, meshwork rupture, and bead propulsion. (b) Bead
size regulates the period and size of meshwork rupture. (c) Bead asymmetry orients meshwork growth. (d) Bar length affects network coherence. (e) Long filaments self-
align to form bundles, which become oriented along the long axis of the container. (f) Inward flow of filaments nucleated at the vesicle periphery leads to the formation of a
ring, the size of which is in proportion with the vesicle diameter. (g) Filament nucleation and growth of micropatterned branched meshworks. The filament interaction angle
modulates the probability of association in either parallel (blue filaments) or antiparallel (red filaments) configurations. (h) Myosins induce the specific contraction and
disassembly of antiparallel bundles and branched meshworks while leaving parallel bundles unaffected. (i) Asymmetric distribution of the ratio between branched and
antiparallel networks leads to asymmetric contraction.

can define the type of network architecture, which in turn
can define its pattern of contraction.
MT network self-organization
Similar to the formation of actin filaments from the self-
assembly of actin monomers, tubulin forms asymmetric
dimers that can self-assemble into MTs. However,
the release of tubulin-bound nucleotide triphosphate is
required to accelerate the process [29]. Compared with
actin filaments, MTs are much more rigid and almost
straight in the dimensions of a single cell. MTs can sustain
higher compression forces than actin filaments. They can
form bundles, but they cannot form branched networks.
MTs are not as numerous as actin filaments in the cell
cytoskeleton. MT growth is characterized by long growth
phases alternated with short periods of rapid shortening.
673
Review
The ‘plus-end’ of the MT is much more dynamic than the
‘minus-end’, which can be attached to a MT-organizing
center (MTOC). In most animal cells, the MT network
forms as an aster in which MTs radiate from the MTOC.
As cells divide, the MTOC is duplicated and the network
forms a bipolar spindle.
Centering
The most straightforward way to investigate MT aster
positioning in response to an SBC has been to purify
MTOCs from cells and place them in microfabricated
chambers of defined dimensions [30]. Hence, the bound-
aries of the chamber can serve as an external SBC. As MT
plus-ends grow and push against the edges of a square
chamber, MTs are subjected to compression forces that
push the aster toward the geometrical center of the cham-
ber [30] (Figure 2a). When fluctuations cause the MTOC to
become off-center in a given direction, MT curvature and
pressure increases in that direction and pushes the MTOC
back toward the center. Thus, an isotropic array of MTs
pushing on peripheral barriers is sufficient to maintain the
aster at the center of the volume in which it is confined.
However, MTs sliding along the periphery could affect the
stability of this centering mechanism by reorienting MTs.
In such conditions, both pushing and pulling forces, by
minus-end-directed motors attached to the periphery, are
necessary to ensure efficient stabilization of the MT aster
at the geometrical center of the SBC [31] (Figure 2a).
By contrast, asters with opposite polarities (i.e., with
MT plus-ends at the center of the aster) cannot adopt the
same steady state. As long as MTs contacting the periphery
are short enough to release their elastic energy by straight-
ening, they gently push the aster toward the center. As
they get longer, the compression forces in bent MTs in-
crease. The clustering of dynamic plus-ends by kinesins at
the aster center is not strong enough to resist these forces
and so the aster fragments. The MT network then switches
to highly robust vortex-like structures [32] (Figure 2b).
Symmetry break
When an aster is trapped in a water droplet encapsulated in
oil, MTs cannot attach to the periphery. The spherical
water–oil interface has minimal tangential resistance and
is an effective SBC along which MT can slide easily. In these
conditions, symmetry breaks in the aster configuration can
occur [33] (Figure 2c). In a relatively large spherical volume,
few MTs reach the boundaries and the aster is stabilized
close to the geometrical center. As the size of the spherical
volume is reduced, MTs tend to be longer than the container
radius. To minimize their curvature and relax their elastic
energy, MTs slip along the edges and align with the SBC
[33,34]. This produces an asymmetric redistribution of MTs
that pushes the MTOC to the periphery of the droplet [33]
(Figure 2c). Interestingly, when the rigidity of the SBC is
reduced, clustered MTs push and deform it to the extent that
a tubular protrusion can be formed [33,35,36].
Alignment and spindle formation
The formation of bipolar mitotic spindles also depends on
geometrical boundary conditions defined by DNA and cell
shape. Two mechanisms contribute to mitotic spindle
674
Trends in Cell Biology December 2012, Vol. 22, No. 12
assembly around DNA: the focusing of the minus-ends of
MTs that are associated with large DNA clusters to form
spindle poles, and the antiparallel alignment of the plus-
ends of MTs that are anchored at the two MTOCs such that
the aster MTs overlap [37].
Multimeric minus-end-directed motors, such as
dyneins, induce the formation of spindle poles. DNA pro-
vides guidance cues for initial MT alignment and thus
biases bipolar spindle formation [38,39]. MTs tend to align
parallel to the surface of a DNA-coated bead. The intrinsic
molecular machinery supporting spindle pole focusing and
mitotic spindle spatial organization is robust and initially
appeared insensitive to configuration of the DNA complex
[40]. However, extensive manipulations of the amount of
DNA and its spatial distribution using microcontact print-
ing revealed the DNA directing role in spindle assembly
[41]. The increase in size of DNA aggregates induces
spindle lengthening (Figure 2d). Above a critical size, large
DNA aggregates can induce the formation of multiple poles
[41] (Figure 2d). Moderately asymmetric distribution of
DNA is sufficient to orient spindle formation [40,41]
(Figure 2e). Long bars coated with DNA result in the
formation of multiple repeats of spindles along the length
of the bar and thus revealed the existence of an intrinsic
spindle width (Figure 2f). This intrinsic spindle width
seems to be defined by the balance between motors forcing
the focusing MT ends and the elastic reaction force due to
MT bending. In a certain range of parameters defined by
the ratio between the DNA aggregate width and MT
length, the symmetry is broken and all MTs collapse on
one side of the DNA, resulting in an asymmetric configu-
ration of spindles with respect to the long axis of the bar
[41] (Figure 2f). Below this critical range, antiparallel MTs
from opposite poles (with the bar in between) interact to
stabilize the formation of symmetric bipolar spindles;
above this range, the two spindle configurations on oppos-
ing sides of the DNA bar are independent and both form
independent monopolar spindles.
Cellular self-organization
In cells, the organizing principles described above appear
applicable but more difficult to reveal and investigate.
Both actin and MT networks are regulated by hundreds
of different types of binding protein. In addition, the as-
sembly of actin filaments and MTs are affected by each
other through physical and biochemical interactions. Cy-
toskeleton network assembly is regulated at the scale of a
cell and is no longer solely dependent on local biochemical
and geometrical conditions. The implication of biochemical
signals forces the system to break its symmetry and define
an axis of polarity. Although actin or MT network assembly
is more complex in the cellular context than in vitro, some
self-organizing principles have been identified.
In simple conditions as near to cellular conditions as can
be achieved in experiments in vitro, similar self-organized
structures can be observed. Cytoskeleton networks in cells
from lymphatic lines or in other cells or cell fragments on
non-adhesive substrates are subjected to no other geomet-
rical constraints than the flexible plasma membrane. In
the absence of MT networks, the actin network contracts
and breaks symmetry after a local rupture occurs in the
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a)

(b)

(c)

(d) (e)

(f)

Key:
Plus-end directed motor Microtubule DNA Pressure

Minus-end directed motor (free)

Minus end anchor Tension
Minus-end directed motor (anchored)

TRENDS in Cell Biology

Figure 2. Microtubule (MT) network self-organization. (a) Aster off-centering with short MTs in a large container (left). Aster centering by MT sliding and pushing on the
container corners (middle). Highly efficient aster centering by pushing and pulling forces (right). (b) MT length-dependent aster formation and centering. Short MT ‘plus-
end’ coalescence by motors (left). Aster centering by a few MT ‘minus-ends’ reaching and pushing on container edges (middle). Aster fragmentation and vortex formation
by pushing forces exerted by long MTs on container edges (right). (c) MT length-dependent aster off-centering. Aster centering by few MT plus-ends reaching and pushing
on container edges (left). Symmetry break and aster off-centering by a few, sliding MTs pushing on container edges (middle). Cortical alignment of MTs and peripheral
localization of MT-organizing center (MTOC) due to numerous MTs sliding and pushing on container edges (right). (d) DNA cluster size regulates spindle size and pole
formation. (e) DNA cluster asymmetry regulates spindle orientation. (f) DNA cluster width regulates spindle symmetry.

network. With the symmetry breaking, an over-contracted
region propagates in the network [42–44]. The process of
rupture is similar to what happens in branched meshworks
around beads [12], except the occurrence of the rupture
results from the contractile force generated by myosins
rather than by the pushing force associated with actin
polymerization. In the absence of actin networks, MTs
pushing on a deformable membrane coalesce, align, and
675
Review
break symmetry by forming a long tubular protrusion
[42,43], reminiscent of their behavior in vesicles
[33,35,36]. However, in cellular conditions, actin and MT
networks interact and the SBCs are more complex that a
freely fluctuating plasma membrane. The ECM and cell
neighbors can represent adhesive SBCs. Hence, the precise
control and manipulation of cell adhesions, which are
cellular structures that interact with the cell’s structural
microenvironment, reveal how these SBCs could direct
intrinsic cytoskeleton self-organizing properties.
Directed shape
Cells spreading on a defined regular array of adhesion
spots revealed that the size and spacing between spots
was a critical regulator of cell shape. Cells need a minimum
spot size to assemble focal adhesions and cannot extend
over a maximal distance between these spots [45–48]. Cell
shape appears to result from the competition between the
force from adhesion-induced spreading and a reaction force
from the cell’s elasticity and other internal contraction
forces [49]. However, some cells, such as fibroblasts, have
an intrinsic mechanism to regulate the length of their long
axis regardless of their width, which seems to implicate
tight crosstalk between actin and the MT network [50,51].
Although cell shape elongation, cytoskeletal alignment,
and internal cell polarity orientation are usually correlat-
ed, cell shape does not determine actin and MT organiza-
tion. Modifying the actin network by fluid flow while
maintaining constant shape reorients the MT network
[52]. Similarly, modifying the MT network independently
of cell shape reorients the actin network [53]. Rather, there
is an intricate coupling between actin and MT networks
that affects their respective spatial organizations and the
axis of cell polarity.
Directed actin network architecture
The cellular actin network is organized by a balance be-
tween the assembly of a contractile network of aligned
filaments and the polymerization of a non-contractile
branched meshwork. This balance appears to be finely
regulated by the degree of cell adhesion [54].
The branched meshwork assembles at the cell periph-
ery. It is preferentially developed along convex rather than
concave cell edges [55]; thus, it promotes the formation of
larger membrane deformations at a cell apex [56]
(Figure 3a), the size of which increases as the angle of
the apex is reduced [57].
Contractile bundles of antiparallel filaments are pres-
ent throughout the cytoplasm. Peripheral bundles and
more interior bundles have distinct dynamics and contrac-
tion properties. Components of peripheral bundles move
toward the bundle center, whereas components of interior
bundles remain static with respect to the bundle organi-
zation [58] (Figure 3b). This probably reveals key differ-
ences in the polarity of filaments and thus specific
contraction properties of these two types of bundle. As cell
spreading or the cell aspect ratio increases, cell contraction
increases [59–61]. The cell aspect ratio increase induces
the alignment of contractile bundles, which form struc-
tures such as stress fibers or myofibrils (Figure 3c). Aligned
stress fibers and the associated anisotropic contraction
676
Trends in Cell Biology December 2012, Vol. 22, No. 12
along the cell’s basal surface are coupled to the assembly
of similar structures and force distribution along the cell’s
apical surface [62,63]. Aligned myofibrils tend to organize
their banding patterns in register [64].
Asymmetric SBCs, defined in cell culture by micropat-
terned adhesion sites, can lead to the development of
asymmetric actin networks. Bundles accumulate preferen-
tially along concave rather than convex cell edges [65]. As
the cell spreads over an adhesive region, conspicuous
contractile bundles are formed that connect this region
to other adhesive regions separated by non-adhesive
regions [47,66,61], revealing the development of larger
traction forces [67] (Figure 3d). A relatively larger distance
between adhesion sites leads to a reduced edge curvature
and thicker bundles and so probably reflects a larger force
between these sites [66,68] (Figure 3e).
Directed MT network
The MT network adapts its dynamics to the various con-
figurations of the actin network. MTs bend and grow along
actin contractile bundles, but stop growing when they
reach a branched actin meshwork [69]. Interestingly, al-
though an asymmetric actin network will lead to asym-
metric organization of MTs, the MTOC remains at its
central location (Figure 3f). Centrosome positioning
appears to depend on generation of forces by dyneins on
MTs [70,71], but also on the forces generated by the less
characterized connections with the actomyosin network
[71,72]. Centrosome central positioning is even more re-
markable given that a large part of the cytoplasm is
occupied by the nucleus, on which MTs can also push
and pull. The robust mechanism by which the centrosome
becomes positioned at the geometrical center of the contour
that describes the cell shape, where the actin network is
asymmetric and the nucleus occupies a large part of the
cytoplasmic volume, remains to be elucidated.
The positioning of the nucleus in a cell in culture is off-
center and distal from the cell’s adhesion to the ECM and
the actin branched meshwork, but is proximal to the
contractile bundles [69]. Therefore, the internal polarity,
as revealed by the nucleus–centrosome vector, is oriented
with respect to ECM and actin network asymmetries
[55,69]. Biochemical disruption of the actomyosin network,
the MT network or the nucleus–cytoskeleton connections
can perturb polarity orientation with respect to cell–ECM
SBCs [71,73,75].
The centrosome and nucleus are often described as
being in the cell spreading plane in culture, but they can
be positioned on an axis orthogonal to this. Moreover, their
relative positions can be switched on this axis, depending
on the degree of confinement imposed by the available
spreading area. Indeed, the centrosome is positioned to-
ward the apical curved surface, above the nucleus, in
confined cells and below the nucleus in cells that have
spread extensively [76] (Figure 3g).
Directed migration
The asymmetric cytoskeleton organization in response to
an asymmetric SBC can affect the direction of a motile cell.
The relationship between external asymmetry and orient-
ed motility is not straightforward, because it seems to
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a) (b) (c) (d) (e)

Angle Posioning Aspect rao Adhesion Distance

(f) Centrosome XY centering (g) Centrosome Z off-centering

Side view

(h) Directed moon (i) Polarized moon

Key: Acn filaments:

An-parallel Nucleus

Parallel Centrosome
ECM micropaern Branched Microtubule

TRENDS in Cell Biology

Figure 3. Cellular self-organization. (a) Branched meshwork polymerization in acute-angled regions of the cell periphery. (b) Inward treadmilling (arrows) in peripheral actin
bundles and absence of treadmilling of internal bundles may reveal differences in filament polarities. (c) Alignment of myofibrils in response to cell shape elongation. (d)
Formation of conspicuous actin bundles along non-adhesive edges and thin actin bundles along adhesive edges. (e) Longer peripheral bundles are also thicker. (f)
Microtubules (MTs) adapt their growth to local actin structures. The centrosome maintains its central position in symmetric (left) and asymmetric environments (right). (g)
Centrosome positioned above the nucleus, close to branched actin meshwork, in spatially confined cells (top). Centrosome positioned below the nucleus, close to actin
bundles, in spread cells (bottom). (h) Cells move toward confined spaces above a certain threshold (top) and toward open spaces below that threshold (bottom). (i) Spread
cells move with the centrosome toward the front (top), whereas confined cells move with the centrosome toward the back (bottom).

depend on cell type and the degree of asymmetry. A linear
track of repeated micropatterns in the shape of isosceles
triangles can lead cells to move toward the acute angle
apices of these triangles [77,78]. By contrast, an elongated
isosceles triangle with a very acute angle (several degrees
only) can lead cells away from the acute angle apex [79]
(Figure 3h).
Without an external bias such as those created by
asymmetric SBC, the direction of motility is defined by
intrinsic cell polarization mechanisms and can be observed
in motile cells on adhesive micropatterns in the shape of
bars. However, bar width affects actin network organiza-
tion. Variations in actin network assembly in response to
bar width are cell type-specific because keratocytes need
large transversal spreading to move relatively fast [80],
whereas fibroblast speed is greater when the bar width is
narrower [81]. Interestingly, the orientation of internal cell
polarity, revealed by the position of the centrosome with
respect to the nucleus, also depends on the width of the
adhesive micropattern. A cell migrates on a wide bar with
the centrosome nearer the leading edge, whereas on a
narrow bar, the centrosome is nearer the trailing edge
[82] (Figure 3i). How this centrosome positioning is related
to the different types of actin organization remains to be
investigated.
The speed of migrating cells and their persistence in
moving in a given direction are both affected in cells whose
internal polarity orientation process is defective [75]. The
systematic connection between the actin network machin-
ery powering cell migration and the degree of stability of
677
Review
spatial organization of the internal cell polarity was fur-
ther supported by the observation in around a hundred
different cell types of a correlation between cell speed and
the persistence of the cell in maintaining their direction of
migration [83].
Directed cell division
The adaptation of MT network architecture in relation to
the actin network architecture and to cell shape is manifest
during cell division. The tensions in astral MTs, radiating
from the spindle pole toward the cell cortex, exert a torque
on the spindle and direct its orientation. The tension in
these astral MTs is regulated by the presence of cortical
cues associated with the actin network, which orient the
cell division axis with respect to cell adhesion cues and the
architecture of the actin network [84–87]. Tension can also
be exerted throughout the cytoplasm and therefore be
proportional to astral MT length; differences in astral
MT length can differ with respect to cell shape elongation
and these variations can direct the orientation of the
division axis accordingly [88,89].
Tissue self-organization
At the level of tissue organization, the complexity of the
system increases with greater numbers of components.
Nevertheless, precise manipulation of the geometries of
SBCs has proven useful in identifying consistent self-or-
ganization rules.
Directed cell positioning
Self-organization of cells in a given space depends on the
balance of mechanical forces between the cells and the
surrounding matrix. Two cells in contact constitute a mini-
mal multicellular structure where cells can form cell–matrix
adhesions (CMAs) and cell–cell adhesions (CCAs). When
confined on a homogeneous micropattern (i.e., when the cell
basal surface is in contact with a continuous layer of ECM),
endothelial cells forming CCAs move regularly around each
other in the plane of the culture dish, whereas fibroblasts,
which cannot form CCAs, do not [90]. Therefore, the forma-
tion of CCAs appears to modulate the capacity of the two
cells to reach a mechanical balance. The two adhesive
systems – CCA and CMA – within a cell can mutually affect
their respective localizations [91]. Two cells of a given
epithelial cell type confined on micropatterned ECM within
a defined area can move or adopt a stationary position in
response to subtle changes in ECM geometry [92]
(Figure 4a). Indeed, the production of tensional forces on
the CCA depends on the spatial organization of the ECM.
Intercellular force is higher when the CCA is close to the
ECM. This directs the CCA away from the ECM and stabi-
lizes the cell position in this configuration, which corre-
sponds to global minimization of the overall contractile
energy [92] (Figure 4a). Conversely, the formation of a
CCA prevents the formation of proximal CMAs [74,93].
The mutual exclusions of the two adhesion systems lead
to their spatial segregation [91] and directs cell positioning.
Directed collective motion
The collective motion of a large multicellular group
depends on the production of intercellular forces, the spatial
678
Trends in Cell Biology December 2012, Vol. 22, No. 12
distribution of which directs the migration of cells with
respect to their neighbors. The sudden removal of SBCs,
allowing a previously confined group to migrate, revealed
that intercellular forces propagate from the migrating front
to the group’s rear [94] (Figure 4b). Intercellular forces did
not appear to pull cells forward but rather to orient the
traction force field they develop on the ECM to migrate.
An intriguing recent work revealed that global coher-
ence can emerge in the spatial organization and collective
motion of large cell groups [95]. Cells plated as multicellu-
lar groups on micropatterned discs do not display any
coherent global motion nor specific cell orientation. How-
ever, on a torus-shaped micropattern, there is a clear
asymmetry in cell orientations such that the long cell axes
tilt at similar angles with respect to the torus center
(Figure 4c). This appears to be reflected in the direction
cells adopt when motile on the torus. The angular direction
of cell motility at the peripheral edge of the torus (with
positive curvature) tends to be opposite to that at the
interior edge of the torus (with negative curvature)
(Figure 4c). Therefore, it appears that the symmetry break
imposed by the torus arises from this directional property
of cell motility at the edges of the torus that is propagated
throughout the entire group of cells. Surprisingly, the
angular bias of endothelial cell orientation is clockwise,
whereas with myoblasts it is counterclockwise. Thus, var-
iations in intracellular parameters presumably can be
manifested as specific asymmetries for different cell types.
However, no explanation has yet been proposed for this
geometrically simple organization resulting from a proba-
bly quite complex mechanism. One area where a mecha-
nism may be identified is in the regulation of cell polarity
and its relationship to oriented cell motility.
Directed cell polarity
The relationship between the locations of CCAs and CMAs
affects nucleus–centrosome axis orientation. The centro-
some, with respect to the nucleus, tends to adopt a more
distal position from CCAs and a more proximal position to
CMAs [73,74,96] (Figure 4d). Thus, the asymmetric loca-
tions of both CCAs and CMAs are sufficient to bias the
nucleus–centrosome axis [73,74]. CCAs seem to regulate
centrosome positioning [73,96], whereas CMAs seem to
regulate nucleus off-centering [69,73]. Both actin filaments
[96] and MTs [73] have been shown to be involved in the
regulation of centrosome positioning away from CCAs.
Therefore, the mechanisms by which the cytoskeleton
affects centrosome and nucleus positioning remain un-
clear. In addition, the orientation of cell polarity not only
depends on the position of CCAs, but also on the orienta-
tion of intercellular force fields [97].
Given that the self-organization of actin filament and
MT networks is highly sensitive to SBCs and to the distri-
bution of mechanical constraints, and that both types of
network have intrinsic capacities to break symmetry, per-
haps biased collective directional motility [95] results from
symmetry break in the intracellular actin networks and
the consequent asymmetric orientation of internal cell
organization [98] (Figure 4e). How these polarized signals
propagate to adjacent cells and result in collective oriented
motility remains to be elucidated. Particularly, the role of
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a) Directed cell posioning (c) Symmetry break in collecve migraon

(b) Directed collecve migraon (d) Directed cell polarity

Removable Removable
barrier barrier

(e) Directed morphogenesis

Side view

Key:
Intercellular
Acn filament bundle Centrosome
ECM micropaern tensional forces
ECM
Nucleus Branched acn network Microtubule tracon forces

TRENDS in Cell Biology

Figure 4. Tissue self-organization. (a) Two cells move regularly around each other (black arrows) when extracellular matrix (ECM) is present all along the periphery (left),
whereas they stop moving when the extremities of their ECM contact plane reach a region without ECM (right). The presence and absence of ECM regulate intra- and
intercellular forces in opposite ways. (b) Intercellular forces propagate from the front to the center of a migrating cell group. (c) A large multicellular group on a disk of ECM
displays no geometrical bias (left), whereas on a torus, symmetry is broken and cells bias their orientation and move (black arrows) in a coherent fashion (right). (d) Two
adherent cells orient their internal polarity away from their contact plane. (e) Speculation on coherent tissue polarity establishment. Symmetry break first occurs in the actin
network, followed by microtubule (MT) rerouting and internal polarity reorientation. The asymmetric distribution of internal forces associated with these changes is
counterbalanced by asymmetric intercellular forces, which further affect polarity in adjacent cells and propagate asymmetric orientation cues.

internal mechanics and intercellular force transmission
could be the key elements supporting intracellular inte-
gration of spatial signals and the establishment of coherent
cell polarities in dynamic multicellular structures.
Concluding remarks
SBCs play a major role in directing intrinsic cytoskeleton
self-organization properties, from the architecture of mac-
romolecular structures to the distribution of cells in tis-
sues. Investigations at each scale – on isolated cytoskeleton
components, more complex cell extracts, or entire cells –
provide complementary information. All contribute to the
establishment of a working framework, which should ulti-
mately allow us to formulate the exact rules of cytoskeleton
self-organization during morphogenesis. However, our un-
derstanding of the self-organization of minimal molecular
systems in vitro is not sufficient to account for genuine
cellular architectures and dynamics. Additional efforts
need to be initiated to connect in vitro and in vivo self-
organized cytoskeleton networks and fully to benefit from
the former in understanding the latter.
There is currently a gap between the few self-organized
structures that have been characterized in vitro and the
myriad different structures observed in cells. Efforts
should be made to reconstitute all of these structures
in vitro. This will become possible by: (i) using more
complex protein mixtures in vitro to recapitulate their
effects on cytoskeleton networks observed in cells; (ii)
identifying ways to engineer controlled SBCs mimicking
actual biological membrane; and (iii) modulating bio-
chemical signaling.
The regulation of network disassembly is as important
as the regulation of assembly in network dynamics. There
is a critical need to further understand how this network
679
Review
disassembly is modulated by SBCs. Progress in this direc-
tion should allow the reconstitution of dynamic steady
states in which manipulation of SBCs and network assem-
bly–disassembly could lead to conditions in which the
network persistently self-renews, with its overall structure
remaining unaffected. Technological developments are al-
so required to modulate SBCs in real time [47], especially
for analyzing dynamic systems and cytoskeleton adapta-
tion to external changes.
However, the considerable efforts made to understand
the regulation of the self-organization properties of actin
filament or MT networks will not be sufficient to under-
stand their self-organization in a cellular context, because
the two networks are not independent of each other. In-
stead, the two networks are physically and biochemically
coupled. It is necessary to design new, controlled in vitro
biochemical assays in which the two networks can interact
and regulate each other. Such assays should offer the
possibility to manipulate the geometry of network inter-
actions as well as the spatial distribution of crosslinking
proteins and regulating enzymes such as Rho-GTPases.
Physical SBCs need to be completed by biochemical SBCs
comprising surface-grafted, but also soluble and diffusible,
cues.
Notably, understanding of the basic laws governing
cytoskeleton assembly can not only provide insights into
cell and tissue morphogenesis, but may also have techno-
logical applications in the development of microdevices
requiring complex and dynamic architectures. A structure
whose precise architecture is regulated by deterministic
assembly rules, that can grow and self-repair because it
self-renews, has advantages over a fixed structure that
would have to be repaired or replaced by a prefabricated
static component. This new sort of manufacturing would be
a useful way to prepare novel biomaterials and should find
promising applications in microelectronics and robotics.
Acknowledgments
We apologize to authors whose work on cytoskeleton self-organization
was instructive and influential but not cited here because the purpose
was to focus on the specific role of SBCs. We thank all members of the
Physics of the Cytoskeleton and Morphogenesis Laboratory for their
experimental work and discussions. This work was supported by grants
from the Human Frontier Science Programs (RGP0004/2011 to L.B. and
RGY0088/2012 to M.T.) and Institut National du Cancer (PLBIO 2011-
141 to M.T.).
References
1 Karsenti, E. (2008) Self-organization in cell biology: a brief history. Nat.
Rev. Mol. Cell Biol. 9, 255–262
2 Huber, F. and Käs, J. (2011) Self-regulative organization of the
cytoskeleton. Cytoskeleton (Hoboken) 68, 259–265
3 Surrey, T. et al. (2001) Physical properties determining self-
organization of motors and microtubules. Science 292, 1167–1171
4 Schaller, V. et al. (2010) Polar patterns of driven filaments. Nature 467,
73–77
5 Sumino, Y. et al. (2012) Large-scale vortex lattice emerging from
collectively moving microtubules. Nature 483, 448–452
6 Cortès, S. et al. (2006) Microtubule self-organisation by reaction-
diffusion processes in miniature cell-sized containers and
phospholipid vesicles. Biophys. Chem. 120, 168–177
7 Pollard, T.D. and Cooper, J.A. (1986) Actin and actin-binding proteins.
A critical evaluation of mechanisms and functions. Annu. Rev.
Biochem. 55, 987–1035
680
Trends in Cell Biology December 2012, Vol. 22, No. 12
8 De La Cruz, E.M. et al. (2000) Polymerization and structure of
nucleotide-free actin filaments. J. Mol. Biol. 295, 517–526
9 Xu, K. et al. (2012) Dual-objective STORM reveals three-dimensional
filament organization in the actin cytoskeleton. Nat. Methods 9, 185–
188
10 Pollard, T.D. (2007) Regulation of actin filament assembly by Arp2/3
complex and formins. Annu. Rev. Biophys. Biomol. Struct. 36, 451–
477
11 Yarar, D. et al. (1999) The Wiskott-Aldrich syndrome protein directs
actin-based motility by stimulating actin nucleation with the Arp2/3
complex. Curr. Biol. 9, 555–558
12 Oudenaarden, A.V. and Theriot, J.A. (1999) Cooperative symmetry-
breaking by actin polymerization in a model for cell motility. Nat. Cell
Biol. 1, 493–499
13 van der Gucht, J. et al. (2005) Stress release drives symmetry breaking
for actin-based movement. Proc. Natl. Acad. Sci. U.S.A. 102, 7847–
7852
14 Bernheim-Groswasser, A. et al. (2005) Mechanism of actin-based
motility: a dynamic state diagram. Biophys. J. 89, 1411–1419
15 Achard, V. et al. (2010) A ‘‘primer’’-based mechanism underlies
branched actin filament network formation and motility. Curr. Biol.
20, 423–428
16 Noireaux, V. et al. (2000) Growing an actin gel on spherical surfaces.
Biophys. J. 78, 1643–1654
17 Bernheim-Groswasser, A. et al. (2002) The dynamics of actin-
based motility depend on surface parameters. Nature 417, 308–
311
18 Lacayo, C.I. et al. (2012) Choosing orientation: influence of cargo
geometry and ActA polarization on actin comet tails. Mol. Biol. Cell
23, 614–629
19 Michelot, A. et al. (2007) Actin-filament stochastic dynamics mediated
by ADF/cofilin. Curr. Biol. 17, 825–833
20 Soares e Silva, M. et al. (2011) Self-organized patterns of actin
filaments in cell-sized confinement. Soft Matter 7, 10631
21 Månsson, A. et al. (2012) Self-organization of motor-propelled
cytoskeletal filaments at topographically defined borders. J. Biomed.
Biotechnol. 2012, 647265
22 Liu, A.P. et al. (2008) Membrane-induced bundling of actin filaments.
Nat. Phys. 4, 789–793
23 Roos, W.H. et al. (2003) Freely suspended actin cortex models on arrays
of microfabricated pillars. Chemphyschem 4, 872–877
24 Uhrig, K. et al. (2009) Optical force sensor array in a microfluidic device
based on holographic optical tweezers. Lab Chip 9, 661–668
25 Reymann, A-C. et al. (2010) Nucleation geometry governs ordered actin
networks structures. Nat. Mater. 9, 827–832
26 Pinot, M. et al. (2012) Confinement induces actin flow in a meiotic
cytoplasm. Proc. Natl. Acad. Sci. U.S.A. 109, 11705–11710
27 Reymann, A-C. et al. (2012) Actin network architecture can determine
myosin motor activity. Science 336, 1310–1314
28 Thoresen, T. et al. (2011) Reconstitution of contractile actomyosin
bundles. Biophys. J. 100, 2698–2705
29 Shelanski, M.L. (1973) Chemistry of the filaments and tubules of brain.
J. Histochem. Cytochem. 21, 529–539
30 Holy, T.E. et al. (1997) Assembly and positioning of microtubule asters
in microfabricated chambers. Proc. Natl. Acad. Sci. U.S.A. 94, 6228–
6231
31 Laan, L. et al. (2012) Cortical dynein controls microtubule dynamics to
generate pulling forces that position microtubule asters. Cell 148, 502–
514
32 Nedelec, F. et al. (1997) Self-organization of microtubules and motors.
Nature 389, 305–308
33 Pinot, M. et al. (2009) Effects of confinement on the self-organization of
microtubules and motors. Curr. Biol. 19, 954–960
34 Cosentino Lagomarsino, M. et al. (2007) Microtubule organization in
three-dimensional confined geometries: evaluating the role of elasticity
through a combined in vitro and modeling approach. Biophys. J. 92,
1046–1057
35 Emsellem, V. et al. (1998) Vesicle deformation by microtubules: a phase
diagram. Phys. Rev. E 58, 4807–4810
36 Fygenson, D. et al. (1997) Mechanics of microtubule-based membrane
extension. Phys. Rev. Lett. 79, 4497–4500
37 Karsenti, E. and Vernos, I. (2001) The mitotic spindle: a self-made
machine. Science 294, 543–547
Review
38 Heald, R. et al. (1996) Self-organization of microtubules into bipolar
spindles around artificial chromosomes in Xenopus egg extracts.
Nature 382, 420–425
39 Halpin, D. et al. (2011) Mitotic spindle assembly around RCC1-coated
beads in Xenopus egg extracts. PLoS Biol. 9, e1001225
40 Gaetz, J. et al. (2006) Examining how the spatial organization of
chromatin signals influences metaphase spindle assembly. Nat. Cell
Biol. 8, 924–932
41 Dinarina, A. et al. (2009) Chromatin shapes the mitotic spindle. Cell
138, 502–513
42 Bornens, M. et al. (1989) The cortical microfilament system of
lymphoblasts displays a periodic oscillatory activity in the absence of
microtubules: implications for cell polarity. J. Cell Biol. 109, 1071–1083
43 Bailly, E. et al. (1991) The cortical actomyosin system of cytochalasin D-
treated lymphoblasts. Exp. Cell Res. 196, 287–293
44 Paluch, E. et al. (2005) Cortical actomyosin breakage triggers shape
oscillations in cells and cell fragments. Biophys. J. 89, 724–733
45 Cavalcanti-Adam, E.A. et al. (2007) Cell spreading and focal adhesion
dynamics are regulated by spacing of integrin ligands. Biophys. J. 92,
2964–2974
46 Schvartzman, M. et al. (2011) Nanolithographic control of the spatial
organization of cellular adhesion receptors at the single-molecule level.
Nano Lett. 11, 1306–1312
47 Vignaud, T. et al. (2012) Reprogramming cell shape with laser nano-
patterning. J. Cell Sci. 125, 2134–2140
48 Lehnert, D. et al. (2004) Cell behaviour on micropatterned substrata:
limits of extracellular matrix geometry for spreading and adhesion. J.
Cell Sci. 117, 41–52
49 Vianay, B. et al. (2010) Single cells spreading on a protein lattice adopt
an energy minimizing shape. Phys. Rev. Lett. 105, 3–6
50 Levina, E.M. et al. (2001) Cytoskeletal control of fibroblast length:
experiments with linear strips of substrate. J. Cell Sci. 114, 4335–4341
51 Picone, R. et al. (2010) A polarised population of dynamic microtubules
mediates homeostatic length control in animal cells. PLoS Biol. 8,
e1000542
52 Vartanian, K.B. et al. (2008) Endothelial cell cytoskeletal alignment
independent of fluid shear stress on micropatterned surfaces. Biochem.
Biophys. Res. Commun. 371, 787–792
53 Terenna, C.R. et al. (2008) Physical mechanisms redirecting cell
polarity and cell shape in fission yeast. Curr. Biol. 18, 1748–1753
54 Bergert, M. et al. (2012) Cell mechanics control rapid transitions
between blebs and lamellipodia during migration. Proc. Natl. Acad.
Sci. U.S.A. 666, 1–7
55 James, J. et al. (2008) Subcellular curvature at the perimeter of
micropatterned cells influences lamellipodial distribution and cell
polarity. Cell Motil. Cytoskeleton 65, 841–852
56 Parker, K.K. et al. (2002) Directional control of lamellipodia extension
by constraining cell shape and orienting cell tractional forces. FASEB
J. 16, 1195–1204
57 Brock, A. et al. (2003) Geometric determinants of directional cell
motility revealed using microcontact printing. Langmuir 19, 1611–
1617
58 Rossier, O.M. et al. (2010) Force generated by actomyosin contraction
builds bridges between adhesive contacts. EMBO J. 29, 1055–1068
59 Tan, J.L. et al. (2003) Cells lying on a bed of microneedles: an approach
to isolate mechanical force. Proc. Natl. Acad. Sci. U.S.A. 100, 1484–
1489
60 Rape, A.D. et al. (2011) The regulation of traction force in relation to cell
shape and focal adhesions. Biomaterials 32, 2043–2051
61 Kilian, K. et al. (2010) Geometric cues for directing the differentiation
of mesenchymal stem cells. Proc. Natl. Acad. Sci. U.S.A. 107, 4872–
4877
62 Hu, S. et al. (2004) Mechanical anisotropy of adherent cells probed by a
3D magnetic twisting device. Am. J. Physiol. Cell Physiol. 287, C1884–
C1191
63 Khatau, S.B. et al. (2009) A perinuclear actin cap regulates nuclear
shape. Proc. Natl. Acad. Sci. U.S.A. 106, 19017–19022
64 Bray, M-A. et al. (2008) Sarcomere alignment is regulated by myocyte
shape. Cell Motil. Cytoskeleton 651, 641–651
65 Xu, J. et al. (2011) Effects of micropatterned curvature on the motility
and mechanical properties of airway smooth muscle cells. Biochem.
Biophys. Res. Commun. 415, 591–596
Trends in Cell Biology December 2012, Vol. 22, No. 12
66 Théry, M. et al. (2006) Cell distribution of stress fibres in response to
the geometry of the adhesive environment. Cell Motil. Cytoskeleton 63,
341–355
67 Tseng, Q. et al. (2011) A new micropatterning method of soft substrates
reveals that different tumorigenic signals can promote or reduce cell
contraction levels. Lab Chip 11, 2231–2240
68 Bischofs, I.B. et al. (2008) Filamentous network mechanics and active
contractility determine cell and tissue shape. Biophys. J. 95, 3488–
3496
69 Théry, M. et al. (2006) Anisotropy of cell adhesive microenvironment
governs cell internal organization and orientation of polarity. Proc.
Natl. Acad. Sci. U.S.A. 103, 19771–19776
70 Wu, J. et al. (2011) Effects of dynein on microtubule mechanics and
centrosome positioning. Mol. Biol. Cell 22, 4834–4841
71 Hale, C.M. et al. (2011) SMRT analysis of MTOC and nuclear
positioning reveals the role of EB1 and LIC1 in single-cell
polarization. J. Cell Sci. 124, 4267–4285
72 Zhu, J. et al. (2010) finding the cell center by a balance of dynein and
myosin pulling and microtubule pushing: a computational study. Mol.
Biol. Cell 21, 4418–4427
73 Dupin, I. et al. (2009) Classical cadherins control nucleus and
centrosome position and cell polarity. J. Cell Biol. 185, 779–786
74 Camand, E. et al. (2012) N-cadherin expression level modulates
integrin-mediated polarity and strongly impacts on the speed
and directionality of glial cell migration. J. Cell Sci. 125, 844–
857
75 Lombardi, M.L. et al. (2011) The interaction between nesprins and sun
proteins at the nuclear envelope is critical for force transmission
between the nucleus and cytoskeleton. J. Biol. Chem. 286, 26743–
26753
76 Pitaval, A. et al. (2010) Cell shape and contractility regulate
ciliogenesis in cell cycle-arrested cells. J. Cell Biol. 191, 303–312
77 Mahmud, G. et al. (2009) Directing cell motions on micropatterned
ratchets. Nat. Phys. 5, 606–612
78 Kushiro, K. et al. (2012) Modular design of micropattern geometry
achieves combinatorial enhancements in cell motility. Langmuir 28,
4357–4362
79 Yoon, S-H. et al. (2011) A biological breadboard platform for cell
adhesion and detachment studies. Lab Chip 11, 3555–3562
80 Csucs, G. et al. (2007) Locomotion of fish epidermal keratocytes
on spatially selective adhesion patterns. Cell Motil. Cytoskeleton 64,
856–867
81 Doyle, A.D. et al. (2009) One-dimensional topography underlies
three-dimensional fibrillar cell migration. J. Cell Biol. 184, 481–
490
82 Pouthas, F. et al. (2008) In migrating cells, the Golgi complex and the
position of the centrosome depend on geometrical constraints of the
substratum. J. Cell Sci. 121, 2406–2414
83 Maiuri, P. et al. (2012) The world first cell race. Curr. Biol. 22, R673–
R675
84 Théry, M. et al. (2007) Experimental and theoretical study of mitotic
spindle orientation. Nature 447, 493–496
85 Fink, J. et al. (2011) External forces control mitotic spindle positioning.
Nat. Cell Biol. 13, 771–778
86 Samora, C.P. et al. (2011) MAP4 and CLASP1 operate as a safety
mechanism to maintain a stable spindle position in mitosis. Nat. Cell
Biol. 13, 1040–1050
87 Kiyomitsu, T. and Cheeseman, I.M. (2012) Chromosome- and spindle-
pole-derived signals generate an intrinsic code for spindle position and
orientation. Nat. Cell Biol. 14, 311–317
88 Minc, N. et al. (2011) Influence of cell geometry on division-plane
positioning. Cell 144, 414–426
89 Minc, N. and Piel, M. (2012) Predicting division plane position and
orientation. Trends Cell Biol. 22, 193–200
90 Huang, S. et al. (2005) Symmetry-breaking in mammalian cell cohort
migration during tissue pattern formation: role of random-walk
persistence. Cell Motil. Cytoskeleton 61, 201–213
91 Burute, M. and Thery, M. (2012) Spatial segregation of cell–cell and
cell–matrix adhesions. Curr. Opin. Cell Biol. 24, 628–636
92 Tseng, Q. et al. (2012) Spatial organization of the extracellular matrix
regulates cell-cell junction positioning. Proc. Natl. Acad. Sci. U.S.A.
109, 1506–1511
681
Review
93 McCain, M.L. et al. (2012) Cooperative coupling of cell-matrix and cell-
cell adhesions in cardiac muscle. Proc. Natl. Acad. Sci. U.S.A. 109,
9881–9886
94 Serra-Picamal, X. et al. (2012) Mechanical waves during tissue
expansion. Nat. Phys. 8, 628–634
95 Wan, L.Q. et al. (2011) Micropatterned mammalian cells exhibit
phenotype-specific left-right asymmetry. Proc. Natl. Acad. Sci.
U.S.A. 108, 12295–12300
682
Trends in Cell Biology December 2012, Vol. 22, No. 12
96 Desai, R.A. et al. (2009) Cell polarity triggered by cell-cell adhesion via
E-cadherin. J. Cell Sci. 122, 905–911
97 Reffay, M. et al. (2011) Orientation and polarity in collectively
migrating cell structures: statics and dynamics. Biophys. J. 100,
2566–2575
98 Mullins, R.D. (2010) Cytoskeletal mechanisms for breaking cellular
symmetry. Cold Spring Harb. Perspect. Biol. 2, a003392
 Review
Special Issue – Synthetic Cell Biology
Directing the assembly of spatially
organized multicomponent tissues
from the bottom up
Jennifer S. Liu and Zev J. Gartner
Department of Pharmaceutical Chemistry, University of California San Francisco, San Francisco, CA 95108, USA
The complexity of the human body derives from numer-
ous modular building blocks assembled hierarchically
across multiple length scales. These building blocks,
spanning sizes ranging from single cells to organs, in-
teract to regulate development and normal organismal
function but become disorganized during disease. Here,
we review methods for the bottom-up and directed
assembly of modular, multicellular, and tissue-like con-
structs in vitro. These engineered tissues will help refine
our understanding of the relationship between form and
function in the human body, provide new models for the
breakdown in tissue architecture that accompanies dis-
ease, and serve as building blocks for the field of regen-
erative medicine.
Investigating the relationship between tissue form and
function with in vitro engineered tissues
Humans contain trillions of cells spanning over 200 spe-
cialized subtypes. This complex cellular community grows
within a web of extracellular matrix (ECM) to form the
tissues and organs that perform the numerous functions of
our bodies. Cells, tissues, and organs constitute a hierar-
chy of structures spanning tens of microns to meters, in
which the arrangement of building blocks at one scale
forms the building block for the next. A major goal of cell
and developmental biology is to delineate how the form of
the body – or the composition and physical organization of
its building blocks – affects function at the level of tissues,
organs, or the whole organism [1]. However, this remains a
challenging goal because direct and general methods for
controlling the relative spatial position of cells in tissues
and organs do not exist.
One powerful approach for elucidating fundamental
principles relating form to function is an engineering
strategy that builds tissue-like structures ex vivo. Like
studies in model organisms, tissue-engineering strategies
can incorporate genetically modified cellular building
blocks. Unlike studies using model organisms, however,
tissue-engineering strategies are also compatible with the
use of primary or immortalized human cells, advanced
imaging techniques, and techniques that control the
non-cellular components of the microenvironment. Most
Corresponding author: Gartner, Z.J. (zev.gartner@ucsf.edu).
Keywords: bottom up; programmed assembly; tissue engineering;
cell–cell interactions; paracrine signaling.
0962-8924/$ – see front matter ß 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.tcb.2012.09.004 Trends in Cell Biology, December 2012, Vol. 22, No. 12
tissue engineering approaches are ‘top down’ and require
the use of patterned substrates, molds, or ECM scaffolds to
assist cells in finding their appropriate positions and dif-
ferentiation states within a tissue. In principle, a 3D
scaffold of ECM of the precise composition and organiza-
tion can provide all the necessary structural and microen-
vironmental cues to direct the organization of individual
cells into a functional tissue or organ, as evidenced by
recent experiments using decellularized organs [2]. How-
ever, de novo construction of scaffolds with the requisite
level of detail at all length scales is not currently possible.
As a consequence, tissue reconstruction starting from cells
or cell aggregates remains challenging, because mixtures
of dissociated cells do not typically reconstitute complex
tissue structures or functions without pre-organization
into the correct 3D geometry. Therefore, additional means
of controlling the spatial organization of cells or groups of
cells will facilitate tissue engineering.
Bottom-up or synthetic approaches are emerging as
valuable and alternative means to more prevalent top-
down approaches for pre-organizing groups of cells into
tissue-like structures. Bottom-up approaches are distinct
from top-down approaches in that they link together sim-
plified building blocks to generate objects that are struc-
turally organized at larger length scales [3]. Directing the
assembly of building blocks from the bottom up may pro-
vide enhanced control over the relative spatial arrange-
ment of cells in engineered tissues when used together
with currently available top-down approaches. In addition
to the advantages of the top-down tissue engineering
strategies outlined above, bottom-up methods have several
other desirable features. First, they are inherently modu-
lar, allowing for the simple replacement of specific cells or
nodes in a network of interacting cells, tissues, or organs
(Box 1). This feature makes bottom-up engineering attrac-
tive as a versatile method for incorporating multiple cell
types into tissues as well as for building different tissue
types or states (for example, functional or pathological) by
interchanging building blocks. Furthermore, these meth-
ods are inherently scalable; many nearly identical tissue
constructs can be prepared without the need for complex or
specialized scaffolds. Finally, bottom-up approaches are
ideally suited for studying the direct interactions between
individual building blocks. Recent research has highlight-
ed the importance of interactions between heterogeneous
683
Review
Box 1. The hierarchical organization of a modular organ: the breast
The human breast contains an organized hierarchy of structures built
up from modular units, from nanometer-sized proteins of the
basement membrane to micron-sized cells to millimeter-sized tissues
[80]. The bilayered epithelium of the mammary gland, for example,
has two principle building blocks: luminal epithelial and myoepithe-
lial cells. Considerable heterogeneity exists even within each of these
cell types. For example, subpopulations of luminal cells express
estrogen and progesterone receptors. When stimulated, these cells
release growth factors triggering the growth of their neighbors. In
addition, luminal and myoepithelial cells play distinct functional
roles, serving to secrete and pump milk, respectively (Figure Ia).
These cellular building blocks are organized into ducts and acini that
are further supported by fibroblasts that synthesize and reside in a
collagenous ECM. Endothelial cells provide additional support for
these structures through a meshwork of capillaries delivering
nutrients, facilitating circulation of lymphocytes, and relaying hor-
mones from distant organs (Figure Ib). Ducts and acini are further
organized into terminal ductal lobular units (TDLUs) that are
surrounded by a secondary and specialized ECM containing a denser
collagenous matrix and beds of adipocytes that add additional form
to the organ (Figure Ic) [80]. Finally, TDLUs are organized into
multiple lobes that drain into large ducts, together delivering milk to
the nipple [82] (Figure Id). Although the overall architecture of the
gland is drastically remodeled over the course of a woman’s lifetime,
the relative position of the different cell types with respect to each
other and the modular organization of the organ remain constant in
healthy tissue.
Endocrine signals
(a) (b)
Growth
factors
ER+
ER- Epithelium
ER-
Fibroblast
10–5 10–4
Figure I. The modular and hierarchical organization of the human mammary gland. (a) Individual glandular epithelial cells exchange signals with each other and the
basement membrane. (b) Epithelial cells of the ducts and acini also exchange signals with the surrounding lobular stroma. (c) Ducts and acini are organized into
terminal ductal lobular units (TDLUs) that are embedded in a second type of collagenous extracellular matrix (ECM) that also contains many adipocytes. (d) The entire
organ is integrated with the rest of the body to mediate its function in delivering milk during breastfeeding. Adapted from [1].
cell types on tissue behaviors, whether the interactions
occur within an epithelium [4–7], between the epithelium
and surrounding stroma [8], or even between cells in
different organ systems [9]. Although spatially organizing
multiple heterogeneous cellular interactions can be chal-
lenging using top-down tissue engineering approaches, a
multiplicity of interacting partners can be systematically
incorporated using a modular, bottom-up approach.
This review focuses on bottom-up and directed-assem-
bly approaches that utilize predefined building blocks to
construct spatially defined multicellular structures con-
taining more than one cell type. The goal of these methods
is to mimic the cellular heterogeneity and physical ar-
rangement of the modular repeating units found in mam-
malian tissues (Figure 1) by directing their assembly from
simpler building blocks. This review will not focus on other
techniques that control the spatial organization of tissues
through organ printing, microscale technologies, genetic or
optogenetic techniques, directed differentiation of stem cell
684
Trends in Cell Biology December 2012, Vol. 22, No. 12
For this modular and hierarchically organized tissue to function,
mechanical, chemical, and electrical signals must be detected by
individual cells and transmitted across each hierarchy of the organ to
synchronize cellular behaviors. Single epithelial cells integrate
chemical and mechanical cues from the basement membrane, a
specialized ECM, to maintain cell polarity and secrete milk [83].
Epithelial cells are also mechanically coupled with their neighbors
and distant tissues via the cytoskeleton and the ECM, respectively
[84,85]. The cytosols of epithelial cells are chemically and electrically
linked through gap junctions [86], and an array of secreted and
membrane-localized signaling proteins coordinate tissue homeosta-
sis and function across the epithelium. Additionally, secreted
endocrine factors link the mammary gland with stromal adipocytes,
the nervous system, and other reproductive organs [87]. Processes
such as breastfeeding require not only coordination between multiple
cell types and modules within the breast but also between sensory
cells in the nervous system, which further synchronize actions in
distant organs to those in the breast. Although it is appreciated that
overlapping signals across this hierarchy of building blocks are
required for the higher order function of the breast and all other
tissues, the structural organization of building blocks that serves as
the foundation for the proper exchange of intercellular signaling
events remains challenging to study. By spatially pre-organizing cells,
tissues and organs, bottom-up and directed tissue engineering
strategies aim to control and understand the exchange of signals
between building blocks at all levels of structural hierarchy in the
human body.
Stroma
(c) (d) Nervous
system
TDLU
Mammary
gland
Nipple
10–3 10–1m
TRENDS in Cell Biology
or progenitor sources, or synthetically engineered genetic
circuits. The reader is directed to several recent reviews for
discussion of these topics [10–15].
Directing the bottom-up assembly of tissues using
single cells as building blocks
Control over the relative position of single cells provides
the fine resolution necessary for probing interactions be-
tween neighboring cells in a tissue. This level of spatial
resolution is required for recreating stem cell niches [16] or
when rebuilding cell–cell connections found in fully differ-
entiated tissues [17]. In some cases, a group of cells has the
ability to self-assemble into specific structures at these
length scales. Townes and Holtfreter famously found that
dissociated cells from amphibian embryos would aggregate
and self-sort into germ layers without outside intervention
[18]. This strategy occasionally allows a multiplicity of cell
types to self-organize in wells or in hanging drops [19–21].
However, isolated mixtures of cells of multiple types do not
Review
(a) (b)
Figure 1. Modular functional units in mammalian organs. (a) Human skeletal muscle cross section. (b) Human mammary terminal ductal lobular unit (TDLU) cross-section.
Arrow indicates extralobular stroma; arrowhead indicates lobular stroma. (c) Pig liver cross-section showing repeating lobules. (d) Mouse embryonic kidney. Reproduced
with permission from [80], Pathpedia, Werning, S. (2007) (http://calphotos.berkeley.edu/cgi/img_query?seq_num=223971&one=T), and [81].
always spontaneously organize into structures that mimic
their tissue of origin without the aid of external ECM
scaffolds. In the absence of these external positioning cues,
bottom-up and directed-assembly techniques may be used
to spatially position cells in relation to each other at the
microscale.
DNA-programmed assembly is a recently developed ap-
proach to direct the organization of multicellular structures
in vitro with single-cell resolution [22]. Key to this approach
is the covalent or non-covalent remodeling of the adhesive
properties of the cell surface with single-stranded DNA
(ssDNA). ssDNA is linked to the cell surface by several
means. In one approach, cells are first cultured in the
presence of an azide-modified monosaccharide that is incor-
porated into cell surface glycans. The accessible azides react
with chemically modified oligonucleotides by Staudinger
ligation [23] or [1,3]-dipolar cycloaddition [22] to covalently
attach the DNA to the glycocalyx. In an alternative ap-
proach, N-hydroxysuccinimide-modified DNA is added to
cell suspensions, where it reacts covalently with free lysines
on the cell surface [24]. Lastly, lipid–DNA conjugates are
added to culture medium where they passively partition into
the cell membrane for non-covalent cell surface modification
[25–27]. Labeling or binding other interacting biomolecules
to cell surfaces will also direct the programmed assembly of
cells, though these are often limited to only single pairs of
interacting molecules [28–31].
Mixing different cell populations labeled with comple-
mentary ssDNA strands or molecules directs the formation
of heterogeneous microtissues, whereas changing the ratios
of labeled populations can be used to achieve discrete mul-
ticellular arrangements (Figure 2a). This strategy has been
used to recapitulate synthetic paracrine signaling networks
[22] and to mimic immune cell homing to sites of inflamma-
tion [29]. More recently, a DNA-mediated programmed
assembly approach was used to investigate the conse-
quences of cell-to-cell variability among mammary epitheli-
al cells during the dynamic process of morphogenesis [32].
Wild-type (WT) MCF10A mammary epithelial cells and
derivatives with elevated Ras activation were labeled with
ssDNA and combinatorially assembled to form homoge-
neous and mosaic microtissues of defined compositions.
Mosaic epithelial aggregates assembled from single Ras-
activated cells and WT MCF10A neighbors displayed emer-
gent behaviors: Ras-activated cells were basally extruded or
led motile multicellular protrusion that directed the motility
of the surrounding WT cells across tens to hundreds of
Trends in Cell Biology December 2012, Vol. 22, No. 12
(c) (d)
TRENDS in Cell Biology
microns. Neither phenotype was observed in aggregates
homogeneous for Ras-activated cells nor in aggregates as-
sembled from single WT MCF10A cells with Ras-activated
neighbors (Figure 2b,c). Because this method controls the
initial architecture of aggregates, the emergent phenotypes
could be directly attributed to the underlying cell-to-cell
variability in Ras activity.
DNA-programmed assembly of cellular building blocks
enables the study of cell–cell interactions in the context of
multicellular structures with precise spatial arrange-
ments. Because essentially any cell type can be modified
with ssDNA using the various DNA-labeling methods out-
lined above and a nearly unlimited set of orthogonal DNA
sequences is available, DNA-programmed assembly can be
used to reconstitute various complex heterotypic cell–cell
interactions for study. Importantly, unlike in genetic engi-
neering, the DNA used to program cellular interactions is
temporary and degrades rapidly at 37 8C to leave unmodi-
fied interacting cells [24,25]. This technique closely
apposes interacting cell surfaces and is ideally suited for
the study of multicellular circuits that operate over short
distances. Interactions amenable to study with this ap-
proach include those mediated by electrical and chemical
signals through gap junctions [33], short-range mechanical
signals coupled to the cytoskeleton [34], juxtacrine signal-
ing such as through the Notch pathway [35], and short-
range paracrine signaling such as through the Wnt and
Hedgehog pathways [36,37]. However, because products of
microscale cell–cell assembly have been limited to around
100 microns in diameter, cell–cell interactions that occur
between neighboring or distant tissues or tissue compart-
ments have evaded study using this strategy alone. Access
to these larger structures might be achieved by combining
programmed assembly with top-down techniques or by
using the products of programmed assembly themselves
as building blocks for further elaboration.
Directing the bottom-up assembly of tissues using cell
sheets and aggregates as building blocks
Many tissues and organs contain repetitive subunits com-
prising groups of cells with dimensions of hundreds of
microns to a millimeter. Such structures include pancre-
atic islets, lymph nodes, and the lobules of the breast and
liver. Most cells within these repeating units are fully
differentiated and structurally integrated with their neigh-
bors and the ECM. Therefore, approximating these units
with cell aggregates that contain fully formed cell–cell
685
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

(a) Assemble Purify Culture

Label cell lines aggregates aggregates microssues

DNA ‘a’
‘a’
Gene

Gen DNA ‘b’

e ‘b’

0.5 h 6h 12 h 18 h 24 h
(b)

Normal H2B-GFP

Extrusion

Protrusion

(c)
Normal Extrusion Protrusion

WT WT ns ns
Ras Ras
***
***

WT Ras
0
0

10

20

30

40
20

40

60

80

10

20

30

40
10

% Frequency
TRENDS in Cell Biology

Figure 2. DNA-programmed assembly. (a) Fluorescent epithelial cells labeled with complementary single-stranded DNA (ssDNA) (or other interacting molecules) are
brought together through molecular recognition. Mixing cell populations 1:50 results in discrete multicellular aggregates that can be purified using fluorescence activated
cell sorting. Assembled aggregates are then cultured in laminin-rich extracellular matrix (ECM) to form polarized microtissues. Genetically distinct input cells can be
incorporated to build mosaic aggregates. (b) Mosaic microtissues assembled from single histone H2B-green fluorescent protein (GFP)-expressing MCF10AT cells, which
express low levels of H-RasV12, and wild type MCF10A neighbors display emergent behaviors that are not observed in homogeneous assemblies (scale bar, 10 mm). (c)
Quantification of the emergent behaviors in homogeneous and mosaic aggregates. Mean values of greater than 500 observations are displayed, with error bars
representing the standard deviation of the mean. WT, wild type; Ras, Ras-activated MCF10neoT. Reproduced with permission from [32].

junctions may provide a means of constructing tissues at
larger length scales than can be achieved from assembly of
single cells alone.
Spherical cell aggregates of the appropriate size have
been engineered using top-down molding techniques and
then assembled within microwells into larger structures of
686
various shapes and sizes [38–41]. Such an approach has
been used to combine aggregates of human fibroblasts and
rat hepatoma cells. When precultured together in spher-
oids, these two cell types self-sort into an inner core of
fibroblasts surrounded by hepatoma cells. When directed
to assemble in rectangular molds, these precultured
 Review Trends in Cell Biology December 2012, Vol. 22, No. 12

CD31 Merged
(a) (d) No. 2 EC - Fb - Fb (e) No. 2-1-3Dy No. 2-2-3Dy

(b) 20 µm 20 µm

No. 3-1-3Dy No. 3-2-3Dy

No. 3 Fb - EC - Fb

20 µm 20 µm

No. 4-1-3Dy No. 4-2-3Dy

No. 4
Fb - Fb - EC

(c)
20 µm 20 µm

No. 5 Co - Co - Co No. 5-1-3Dy No. 5-2-3Dy

20 µm 20 µm

TRENDS in Cell Biology

Figure 3. Directed assembly of cell aggregates. (a) Preformed, spheroidal cell aggregates (100 mm in diameter) assembled in trough-shaped wells fused into rod-shaped
structures over 24 hours. (b) Precultured heterogeneous spheroids with a core of human fibroblasts (red) and surrounding rat hepatoma (green) cells cultured in wells for
24 hours. (c) Confocal image shows that the assembled spheroids retained the inner fibroblast core during spheroid fusion (scale bars, 200 mm). (d) Different combinations
of three cell sheet layers are stacked to make heterogeneous tissues of endothelial (green) and fibroblast (pink) cells. (e) After 3 days, endothelial sheets formed vessels
(outlined with indirect CD31 staining in green) when stacked within or under fibroblast sheets (actin staining in red in merged images; scale bars, 20 mm). Reproduced with
permission from [40] and [51].

spheroidal building blocks fuse into a rod but maintain the
fibroblast cores for at least 24 hours [40] (Figure 3a–c).
Directed assembly with globular cell aggregates (Table 1)
has been demonstrated only with modules of a single type,
though the same basic strategy should be applicable to
modules made of different cell types. However, the overall
architecture in large assemblies of cell aggregates may
require further structural support from the microenviron-
ment to be stable over the long term [39].
Cell sheets have also been used as building blocks and can
be stacked manually to build macroscopic tissue structures.
Sheets are cultured in suspension or released using ther-
mal- or ion-sensitive coatings [42–44]. Single sheets have
been made from various cell types, including fibroblasts,
endothelial cells, and epithelial cells [45,46]. Cell–cell adhe-
sions and secreted ECM components are maintained in
lifted sheets [47,48], which can also promote tissue-like
functions that are not observed with dissociated cells [49].
3D tissue constructs comprising various cell types can be
prepared by manually stacking different cell sheets. For
instance, myoblast cell sheets intercalated with either
layers, lines, or dissociated human umbilical vein endothe-
lial cells (HUVECs) [50–52] (Figure 3d,e) formed vascular
networks in vitro that could integrate with host vasculature
in vivo when grafted subcutaneously into rats [52].
By starting with multicellular building blocks with an
established geometry, directed assembly of cell aggregates
of different cell types can be used to design and spatially
position tissue-sized networks of interacting cells. These
higher-order structures are amenable to further elabora-
tion with layers of individual cells. For example, pancreatic
islets isolated from rodents were labeled with lipid-modi-
fied DNA and surrounded with mammalian cells bearing
complementary DNA to provide a barrier function to the
sensitive cell aggregates [53]. In principle, such a strategy
could also introduce components of the surrounding tissue
stroma to provide additional trophic support between
modular units. The modular organization of assembled

Table 1. Features, advantages, and opportunities for directed assembly approaches

Building block Major advantage Size of products (relevant Current applications Challenges and opportunities
biological structures)
Cell Single cell spatial resolution 10–100 mm (groups of Study of cell–cell interactions Assembling more complex tissues;
cells, niches) across short length scales; Incorporation of products into
Study of heterotypic cell–cell multiscale tissues;
interactions; Control of product architecture
Drug screening over time
Cell aggregate Mature cell–cell junctions 100 mm to 1 mm Study of cell aggregation; Control of product architecture
or sheet (functional tissue units) Study of interactions between over time;
cell populations; Nutrient delivery and/or
Production of engineered and vascularization
functional tissues
Cell-laden Incorporation of specialized 100 mm to 1 mm Study of paracrine signaling; Matching matrix properties to
hydrogel ECM or ECM-like material; (functional tissue units Generation of tissue engineering in vivo tissue;
Easily interfaced with or multicomponent constructs Nutrient delivery and/or
microscale engineering tissues) vascularization;
approaches Responsive or smart hydrogels

687
Review Trends in Cell Biology December 2012, Vol. 22, No. 12

building blocks may also be stratified with exogenous
ECM-like components to mimic tissue architectures of
increasing complexity.
Directing the bottom-up assembly of tissues using cell-
laden hydrogels as building blocks
The minimal functional unit of many tissues comprises
small groups of structurally organized cells embedded in
specialized ECM. Collections of these functional units are
further organized in 3D space within additional ECM that is
frequently of a different composition. The terminal ductal
lobular units (TDLUs) and the intervening adipocyte-rich
regions of the mammary gland exemplify this organization
(Box 1). These structures should be modeled with explicit
inclusion of spatially segregated ECM-like materials in
building blocks, particularly when matrix components are
critical to cell and tissue function or when the cells cannot
generate the volume and composition of the ECM found in
biological tissues on their own. Fortunately, many natural
and synthetic polymers are available for designing ECM-
like hydrogels with specific structural and mechanical prop-
erties to mimic different tissues within the human body
[54,55]. Although ECM-like materials can be incorporated
into tissues by mixing microspheres of hydrogel with indi-
vidual cells [56], cell-laden hydrogel modules with dimen-
sions of hundreds of microns to millimeters are closer in size
to the ECM-encapsulated repeating units found in many
tissues in vivo. Therefore, these may be a more tractable
building block for constructing interactions between repeat-
ing units within tissues or organs.
The straightforward packing of modular cell-laden
hydrogel units into a defined space can direct their assembly
into larger structures. Such an approach was used with
building blocks of cell-laden and collagen-based cylinders
covered with an endothelial cell layer [57,58]. The
authors assembled multiple endothelium-coated cylinders
containing rat myocardiocytes onto a planar nylon filter
then added temporary alginate glue. The resulting modular
tissue formed a sheet-like structure that contracted on
electrical stimulation [57]. Collagen hydrogels containing
different cell types have also been assembled into a linear
array using microfluidics to constrain hydrogel orientation
[59].
In contrast to the random but constrained assembly of
subunits, directed-assembly schemes can guide the forma-
tion of a multiplicity of hydrogel building blocks into more
defined structures. Assembling submillimeter-sized poly-
ethylene glycol (PEG)-based hydrogels in hydrophobic me-
dia, for example, promotes the self-association of the
hydrophilic objects. Hydrogel building blocks containing
two different cell types can be prepared by sequential
photolithography steps that meld two concentric rings of
cell-laden PEG hydrogel together into a single unit
(Figure 4a). Additional assembly of these two-component
building blocks into linear arrays generated tubes contain-
ing an inner layer of endothelial cells surrounded by
smooth muscle cells that mimicked the architecture of
vasculature [60] (Figure 4b). Another way to generate
spatially defined heterogeneous tissues using this direct-
ed-assembly method is to mix two different populations of
hydrogels containing different cell types; changing the
ratio of these two hydrogel monomers can bias the compo-
sition of the end products [61]. Nanofibers added to hydro-
gel monomers can provide additional mechanical coupling
between modules, facilitating cell differentiation and tis-
sue function [62]. An additional level of control over the
interface between two populations of hydrogel-embedded
cells can be added by using gels with lock-and-key designs
(Figure 4c,d) to access precise architectures with controlled
stoichiometries of building blocks [61] (Figure 4e,f).
Similar to cellular building blocks, hydrogels can also be
labeled with ssDNA or other biomolecules to program their

(a) (b) (c) (d)

(e) (f) (g)

100 µm

TRENDS in Cell Biology

Figure 4. Directed assembly of cell-laden hydrogels. (a) Donut-shaped hydrogels with an inner hydrogel ring loaded with endothelial cells (green) surrounded by an outer
ring loaded with smooth muscle cells (red) were made using sequential photolithography steps. (b) Side view of tubular structure formed after sequential assembly of
hydrogel units from a (scale bars, 100 mm). (c,d) Lock-and-key (cross- and rod-shaped) hydrogels stained with fluorescent dextran were made using photolithography. (e,f)
Lock-and-key hydrogels loaded with fluorescent murine fibroblast cells assembled through self-association in a hydrophobic medium (scale bars, 200 mm). (g) 3D volume
reconstruction of DNA-directed hydrogel assemblies. Spherical hydrogels bearing green fluorescent tracking beads and labeled with single-stranded DNA (ssDNA) were
bound to a microarray template. A second layer of hydrogels loaded with red beads and labeled with complementary DNA was assembled onto the first population (scale
bar, 100 mm). Reproduced with permission from [60,61,63].

688
Review
assembly. Such an approach avoids the necessity to fabri-
cate hydrogel units of complex shape but requires addi-
tional chemical labeling steps either before or after
fabrication. A two-step approach was used to label uniform,
spherical PEG-based hydrogel units with streptavidin,
followed by modification with biotinylated ssDNA [63].
Additional control of building block positioning was dem-
onstrated by annealing the ssDNA-coated hydrogels onto
microarray templates or by directed assembly of two popu-
lations of hydrogels bearing complementary ssDNA on
their surfaces (Figure 4g). Recent reports suggest that
additional control over interfacial interactions between
hydrogel building blocks using non-biological molecules
could add further complexity to tissues engineered at this
length scale [54,64].
Due to advances in top-down fabrication techniques,
hydrogel building blocks of various sizes, shapes, and com-
positions are readily designed to mimic architectures and
interfaces observed in tissues in vivo. The density of cells
within the hydrogel units can also be controlled, though
achieving tissue-like cell densities may require extended
periods of culture [59]. Because cells are physically con-
strained within the building blocks, cell-laden hydrogels
may be especially appropriate for studying the exchange
of soluble factors between different cell populations [56].
Moreover, hydrogel compositions can be adjusted using
biologically derived or synthetic polymers [65,66] designed
to match the physical and chemical properties of ECM of
specific tissues, or to probe the consequence of changing
ECM properties on cellular behaviors. Such control would be
especially useful for modeling tissues that contain multiple
compartments with different matrices and cell types, such
as between neighboring TDLUs in the breast. Finally, cell-
laden hydrogels of various levels of complexity can be
implanted in vivo. In one example, gel-encapsulated cells
coated with an endothelial layer show enhanced survival
and differentiation, as well as a favorable host response to
growth factors secreted by the encapsulated cells [67–69].
Considerations
The techniques outlined here are designed to direct the
assembly of cells into a specific position relative to other
cells in the context of a multicellular tissue. To form a
functional tissue, however, the cells incorporated into
building blocks must also retain the capacity to interact
properly with their neighbors and to deposit and remodel
their own ECM. Unfortunately, established cell lines prop-
agated in 2D culture are frequently used as building blocks
in tissue engineering applications. In these cell lines,
tissue-specific gene expression patterns and cell surface
proteins necessary for directing heterotypic cell–cell inter-
actions are often downregulated or lost [70,71]. In fact,
even primary cells can begin to lose markers of differenti-
ation after brief growth on tissue-culture plastic [72,73].
Therefore, care must be taken to validate that the cell
types used in an engineered tissue retain expression of
appropriate markers of differentiation and the ability to
interact with neighboring cell populations.
Optimization of other components in the microenviron-
ment may also be required for assembled cells to transition
to tissue-like organization [74]. For instance, ECM
Trends in Cell Biology December 2012, Vol. 22, No. 12
components and soluble factors in tissues can be extreme-
ly dynamic [75,76] and may differ substantially from
formulations typically used in culture. Because the chem-
ical, mechanical, and structural organization of ECM can
affect tissue and cell behaviors [77,78], use of smart
materials with adjustable and controllable properties,
such as encapsulating hydrogels, could allow for dynamic
control of the ECM and aid the proper morphogenesis of
the engineered construct [79]. Additionally, secreted
molecules can be integrated into the design of tissue
modules for gradual release [22,68,69] or the necessary
secreted factors could be delivered and controlled exter-
nally such as through microfluidics or photo-uncaging.
Concluding remarks
The applications of bottom-up strategies for directing the
assembly of specific tissue architectures are still in their
infancy, but they have potential to address important
questions relating cellular organization to the coordination
of multicellular behaviors. At the spatial resolution of
single cells, directed assembly strategies will aid the study
of stem cell niches from multiple tissue types, with respect
to both their structure and composition. These methods
will also benefit the study of cell-to-cell variability in gene
expression or pathway activation in normal and diseased
tissues. Finally, these methods will provide a means of
studying the coordination of cellular behaviors during
processes such as morphogenesis and tissue repair. At
the larger spatial resolution of cell aggregates and whole
tissues, directed-assembly strategies will impact the field
of regenerative medicine as well as the study of mechanical
and chemical coupling of cell groups, particularly between
epithelial cells and the components of the stroma.
Numerous opportunities exist for improving the direct-
ed assembly of tissues (Table 1). One avenue of interest is
in the combination of the various approaches described
above to direct the hierarchical assembly of tissues with
spatial precision across multiple length scales, spanning
that of single cells to full organs. Some progress along these
lines has been made [53] or would be a logical extension of
existing work [63]. Another opportunity will be in directing
the assembly of anisotropic or asymmetric tissue struc-
tures. Finally, introducing genetic circuits into modules to
control interactions between neighboring cells or tissues
will provide additional mechanisms for engineering the
processes of development and morphogenesis. We antici-
pate that progress towards these challenges, better inte-
gration with top-down engineering techniques, and new
applications will be forthcoming in this exciting area.
Acknowledgments
The authors would like to acknowledge support for a portion of the work
covered in this review by DOD grant W81XWH-10-1-1023 to Z.J.G., by
grant P50 GM081879 to the UCSF Center for Systems and Synthetic
Biology, and by NSF grant DGE-0648991 to J.S.L.
References
1 Nelson, C.M. and Bissell, M.J. (2005) Modeling dynamic reciprocity:
engineering three-dimensional culture models of breast architecture,
function, and neoplastic transformation. Sem. Cancer Biol. 15, 342–352
2 Ott, H.C. et al. (2008) Perfusion-decellularized matrix: using nature’s
platform to engineer a bioartificial heart. Nat. Med. 14, 213–221
689
Review
3 Elbert, D.L. (2011) Bottom-up tissue engineering. Curr. Opin.
Biotechnol. 22, 674–680
4 Hogan, C. et al. (2009) Characterization of the interface between
normal and transformed epithelial cells. Nat. Cell Biol. 11, 460–467
5 Leung, C.T. and Brugge, J.S. (2012) Outgrowth of single oncogene-
expressing cells from suppressive epithelial environments. Nature 482,
410–413
6 Eisenhoffer, G.T. et al. (2012) Crowding induces live cell extrusion to
maintain homeostatic cell numbers in epithelia. Nature 484, 546–549
7 Johnston, L.A. (2009) Competitive interactions between cells: death,
growth, and geography. Science 324, 1679–1682
8 Engelhardt, J.J. et al. (2012) Marginating dendritic cells of the tumor
microenvironment cross-present tumor antigens and stably engage
tumor-specific T cells. Cancer Cell 21, 402–417
9 DeNardo, D.G. et al. (2009) CD4(+) T cells regulate pulmonary
metastasis of mammary carcinomas by enhancing protumor
properties of macrophages. Cancer Cell 16, 91–102
10 Mironov, V. et al. (2009) Organ printing: tissue spheroids as building
blocks. Biomaterials 30, 2164–2174
11 Kaji, H. et al. (2011) Engineering systems for the generation of
patterned co-cultures for controlling cell-cell interactions. Biochim.
Biophys. Acta 1810, 239–250
12 Toettcher, J.E. et al. (2011) The promise of optogenetics in cell biology:
interrogating molecular circuits in space and time. Nat. Methods 8, 35–
38
13 Miesenbock, G. (2011) Optogenetic control of cells and circuits. Annu.
Rev. Cell Dev. Biol. 27, 731–758
14 Wu, S.M. and Hochedlinger, K. (2011) Harnessing the potential of
induced pluripotent stem cells for regenerative medicine. Nat. Cell
Biol. 13, 497–505
15 Lenas, P. et al. (2009) Developmental engineering: a new paradigm for
the design and manufacturing of cell-based products. Part I: from
three-dimensional cell growth to biomimetics of in vivo
development. Tissue Eng. Part B: Rev. 15, 381–394
16 Losick, V.P. et al. (2011) Drosophila stem cell niches: a decade of
discovery suggests a unified view of stem cell regulation. Dev. Cell
21, 159–171
17 Desai, R.A. et al. (2009) Cell polarity triggered by cell-cell adhesion via
E-cadherin. J. Cell Sci. 122, 905–911
18 Townes, P.L. and Holtfreter, J. (1955) Directed movement and selective
adhesion of embryonic amphibian cells. J. Exp. Zool. 128, 53–120
19 Kunz-Schughart, L.A. et al. (2006) Potential of fibroblasts to regulate the
formation of three-dimensional vessel-like structures from endothelial
cells in vitro. Am. J. Physiol. Cell Physiol. 290, C1385–C1398
20 Wenger, A. et al. (2005) Development and characterization of a
spheroidal coculture model of endothelial cells and fibroblasts for
improving angiogenesis in tissue engineering. Cells Tissues Organs
181, 80–88
21 Foty, R. (2011) A simple hanging drop cell culture protocol for
generation of 3D spheroids. J. Vis. Exp. 51, 2720
22 Gartner, Z.J. and Bertozzi, C.R. (2009) Programmed assembly of 3-
dimensional microtissues with defined cellular connectivity. Proc.
Natl. Acad. Sci. U.S.A. 106, 4606–4610
23 Chandra, R.A. et al. (2006) Programmable cell adhesion encoded by
DNA hybridization. Angew. Chem. Int. Ed. Engl. 45, 896–901
24 Hsiao, S.C. et al. (2009) Direct cell surface modification with DNA for
the capture of primary cells and the investigation of myotube formation
on defined patterns. Langmuir 25, 6985–6991
25 Selden, N.S. et al. (2012) Chemically programmed cell adhesion with
membrane-anchored oligonucleotides. J. Am. Chem. Soc. 134, 765–768
26 Teramura, Y. et al. (2010) Control of cell attachment through polyDNA
hybridization. Biomaterials 31, 2229–2235
27 Liu, H. et al. (2011) Membrane anchored immunostimulatory
oligonucleotides for in vivo cell modification and localized
immunotherapy. Angew. Chem. Int. Ed. Engl. 50, 7052–7055
28 Liu, X. et al. (2011) Targeted cell-cell interactions by DNA nanoscaffold-
templated multivalent bispecific aptamers. Small 7, 1673–1682
29 Zhao, W. et al. (2011) Mimicking the inflammatory cell adhesion
cascade by nucleic acid aptamer programmed cell-cell interactions.
FASEB J. 25, 3045–3056
30 Dutta, D. et al. (2011) Synthetic chemoselective rewiring of cell
surfaces: generation of three-dimensional tissue structures. J. Am.
Chem. Soc. 133, 8704–8713
690
Trends in Cell Biology December 2012, Vol. 22, No. 12
31 Hamon, M. et al. (2011) Avidin-biotin-based approach to forming
heterotypic cell clusters and cell sheets on a gas-permeable
membrane. Biofabrication 3, 034111
32 Liu, J.S. et al. (2012) Programmed assembly of mosaic cell aggregates
reveals the consequences of cell-to-cell variability in Ras activity on the
collective behavior of mammary epithelial cells. Cell Rep. http://
dx.doi.org/10.1016/j.celrep.2012.08.037
33 Locke, D. (1998) Gap junctions in normal and neoplastic mammary
gland. J. Pathol. 186, 343–349
34 Eyckmans, J. et al. (2011) A hitchhiker’s guide to mechanobiology. Dev.
Cell 21, 35–47
35 Sprinzak, D. et al. (2010) Cis-interactions between Notch and Delta
generate mutually exclusive signalling states. Nature 465, 86–90
36 Taipale, J. and Beachy, P.A. (2001) The Hedgehog and Wnt signalling
pathways in cancer. Nature 411, 349–354
37 Takebe, N. et al. (2011) Targeting cancer stem cells by inhibiting Wnt,
Notch, and Hedgehog pathways. Nat. Rev. Clin. Oncol. 8, 97–106
38 Tejavibulya, N. et al. (2011) Directed self-assembly of large scaffold-
free multi-cellular honeycomb structures. Biofabrication 3, 034110
39 Livoti, C.M. and Morgan, J.R. (2010) Self-assembly and tissue fusion of
toroid-shaped minimal building units. Tissue Eng. Part A 16, 2051–
2061
40 Rago, A.P. et al. (2009) Controlling cell position in complex heterotypic
3D microtissues by tissue fusion. Biotechnol. Bioeng. 102, 1231–1241
41 Rago, A.P. et al. (2009) Encapsulated arrays of self-assembled
microtissues: an alternative to spherical microcapsules. Tissue Eng.
Part A 15, 387–395
42 Jean, J. et al. (2011) Bioengineered skin: the self-assembly approach. J.
Tissue Sci. Eng. S5, 001
43 Haraguchi, Y. et al. (2012) Fabrication of functional three-dimensional
tissues by stacking cell sheets in vitro. Nat. Protoc. 7, 850–858
44 Zahn, R. et al. (2012) Ion-induced cell sheet detachment from standard
cell culture surfaces coated with polyelectrolytes. Biomaterials 33,
3421–3427
45 Yang, J. et al. (2007) Reconstruction of functional tissues with cell sheet
engineering. Biomaterials 28, 5033–5043
46 Labbe, B. et al. (2011) Cell sheet technology for tissue engineering: the
self-assembly approach using adipose-derived stromal cells. Methods
Mol. Biol. 702, 429–441
47 Nishida, K. et al. (2004) Functional bioengineered corneal epithelial
sheet grafts from corneal stem cells expanded ex vivo on a temperature-
responsive cell culture surface. Transplantation 77, 379–385
48 Ohashi, K. et al. (2007) Engineering functional two- and three-
dimensional liver systems in vivo using hepatic tissue sheets. Nat.
Med. 13, 880–885
49 Sekine, H. et al. (2011) Cardiac cell sheet transplantation improves
damaged heart function via superior cell survival in comparison with
dissociated cell injection. Tissue Eng. Part A 17, 2973–2980
50 Tsuda, Y. et al. (2007) Cellular control of tissue architectures using a
three-dimensional tissue fabrication technique. Biomaterials 28, 4939–
4946
51 Asakawa, N. et al. (2010) Pre-vascularization of in vitro three-
dimensional tissues created by cell sheet engineering. Biomaterials
31, 3903–3909
52 Sasagawa, T. et al. (2010) Design of prevascularized three-dimensional
cell-dense tissues using a cell sheet stacking manipulation technology.
Biomaterials 31, 1646–1654
53 Teramura, Y. et al. (2010) Microencapsulation of islets with living cells
using polyDNA-PEG-lipid conjugate. Bioconjug. Chem. 21, 792–796
54 Seliktar, D. (2012) Designing cell-compatible hydrogels for biomedical
applications. Science 336, 1124–1128
55 Correia, A.L. and Bissell, M.J. (2012) The tumor microenvironment is a
dominant force in multidrug resistance. Drug Resist. Updat. 15, 39–49
56 Scott, E.A. et al. (2010) Modular scaffolds assembled around living cells
using poly(ethylene glycol) microspheres with macroporation via a non-
cytotoxic porogen. Acta Biomater. 6, 29–38
57 Leung, B.M. and Sefton, M.V. (2010) A modular approach to cardiac
tissue engineering. Tissue Eng. Part A 16, 3207–3218
58 McGuigan, A.P. and Sefton, M.V. (2006) Vascularized organoid
engineered by modular assembly enables blood perfusion. Proc.
Natl. Acad. Sci. U.S.A. 103, 11461–11466
59 Bruzewicz, D.A. et al. (2008) Fabrication of a modular tissue construct
in a microfluidic chip. Lab Chip 8, 663–671
Review
60 Du, Y. et al. (2011) Sequential assembly of cell-laden hydrogel
constructs to engineer vascular-like microchannels. Biotechnol.
Bioeng. 108, 1693–1703
61 Du, Y. et al. (2008) Directed assembly of cell-laden microgels for
fabrication of 3D tissue constructs. Proc. Natl. Acad. Sci. U.S.A.
105, 9522–9527
62 Dvir, T. et al. (2011) Nanowired three-dimensional cardiac patches.
Nat. Nanotechnol. 6, 720–725
63 Li, C.Y. et al. (2011) DNA-templated assembly of droplet-derived PEG
microtissues. Lab Chip 11, 2967–2975
64 Zheng, Y. et al. (2012) Switching of macroscopic molecular recognition
selectivity using a mixed solvent system. Nat. Commun. 3, 831
65 Van Vlierberghe, S. et al. (2011) Biopolymer-based hydrogels as scaffolds
for tissue engineering applications: a review. Biomacromolecules 12,
1387–1408
66 Jabbari, E. (2011) Bioconjugation of hydrogels for tissue engineering.
Curr. Opin. Biotechnol. 22, 655–660
67 Gupta, R. and Sefton, M.V. (2011) Application of an endothelialized
modular construct for islet transplantation in syngeneic and
allogeneic immunosuppressed rat models. Tissue Eng. Part A 17,
2005–2015
68 Butler, M.J. and Sefton, M.V. (2012) Cotransplantation of adipose-
derived mesenchymal stromal cells and endothelial cells in a modular
construct drives vascularization in SCID/bg mice. Tissue Eng. Part A
18, 1628–1641
69 Vallbacka, J.J. and Sefton, M.V. (2007) Vascularization and improved
in vivo survival of VEGF-secreting cells microencapsulated in HEMA-
MMA. Tissue Eng. 13, 2259–2269
70 Birgersdotter, A. et al. (2005) Gene expression perturbation in vitro–a
growing case for three-dimensional (3D) culture systems. Semin.
Cancer Biol. 15, 405–412
71 Petersen, O.W. et al. (1992) Interaction with basement membrane
serves to rapidly distinguish growth and differentiation pattern of
normal and malignant human breast epithelial cells. Proc. Natl.
Acad. Sci. U.S.A. 89, 9064–9068
72 Chaffer, C.L. et al. (2011) Normal and neoplastic nonstem cells can
spontaneously convert to a stem-like state. Proc. Natl. Acad. Sci.
U.S.A. 108, 7950–7955
73 Lacorre, D.A. et al. (2004) Plasticity of endothelial cells: rapid
dedifferentiation of freshly isolated high endothelial venule
Trends in Cell Biology December 2012, Vol. 22, No. 12
endothelial cells outside the lymphoid tissue microenvironment.
Blood 103, 4164–4172
74 Rivron, N.C. et al. (2009) Tissue assembly and organization:
developmental mechanisms in microfabricated tissues. Biomaterials
30, 4851–4858
75 Maller, O. et al. (2010) Extracellular matrix composition reveals
complex and dynamic stromal-epithelial interactions in the
mammary gland. J. Mammary Gland Biol. Neoplasia 15, 301–318
76 Muller, P. and Schier, A.F. (2011) Extracellular movement of signaling
molecules. Dev. Cell 21, 145–158
77 Gudjonsson, T. et al. (2002) Normal and tumor-derived myoepithelial
cells differ in their ability to interact with luminal breast epithelial
cells for polarity and basement membrane deposition. J. Cell Sci. 115,
39–50
78 Lu, P. et al. (2012) The extracellular matrix: a dynamic niche in cancer
progression. J. Cell Biol. 196, 395–406
79 Samchenko, Y. et al. (2011) Multipurpose smart hydrogel systems. Adv.
Colloid Interface Sci. 168, 247–262
80 Damassa, D.A. et al. (1996) Purification and characterization of the sex
hormone-binding globulin in serum from Djungarian hamsters. Comp.
Biochem. Physiol. B: Biochem. Mol. Biol. 113, 593–599
81 Davies, J.A. (2006) A method for cold storage and transport of viable
embryonic kidney rudiments. Kidney Int. 70, 2031–2034
82 Nelson, C.M. and Bissell, M.J. (2006) Of extracellular matrix, scaffolds,
and signaling: tissue architecture regulates development, homeostasis,
and cancer. Annu. Rev. Cell Dev. Biol. 22, 287–309
83 Streuli, C.H. et al. (1991) Control of mammary epithelial differentiation:
basement membrane induces tissue-specific gene expression in the
absence of cell-cell interaction and morphological polarity. J. Cell
Biol. 115, 1383–1395
84 Paszek, M.J. et al. (2005) Tensional homeostasis and the malignant
phenotype. Cancer Cell 8, 241–254
85 Guo, C.L. et al. (2012) Long-range mechanical force enables self-
assembly of epithelial tubular patterns. Proc. Natl. Acad. Sci. U.S.A.
109, 5576–5582
86 McLachlan, E. et al. (2007) Connexins and gap junctions in mammary
gland development and breast cancer progression. J. Membr. Biol. 218,
107–121
87 Brisken, C. and O’Malley, B. (2010) Hormone action in the mammary
gland. Cold Spring Harb. Perspect. Biol. 2, a003178
691