doi:10.1006/jmbi.2001.4995 available online at http://www.idealibrary.com on J. Mol. Biol.

(2001) 312, 591±595

COMMUNICATION

Porous Silicon: an Effective Nucleation-inducing

Material for Protein Crystallization
N. E. Chayen1*, E. Saridakis1, R. El-Bahar2 and Y. Nemirovsky2
1
Biological Structure and Protein crystals play a pivotal part in structural genomics, hence there is
Function Section, Division of an urgent requirement for new and improved methodology to aid crystal
Biomedical Sciences, Imperial growth. Considerable effort has been invested in the search of substances
College of Science, Technology (nucleants) that will induce ef®cient nucleation of protein crystals in a
and Medicine, London controlled manner. To date, nucleation has been facilitated mainly by
SW7 2AZ, UK seeding, epitaxy, charged surfaces or mechanical means. A different
2 approach is introduced here, involving the use of a mesoporous material
Department of Electrical
that is likely to constrain protein molecules and thereby encourage them
Engineering, Technion-Israel
to aggregate in crystalline order. Large single crystals were obtained
Institute of Technology, Haifa using porous silicon at conditions that are not suf®cient for spontaneous
32000, Israel
nucleation, for ®ve out of six proteins that were investigated. We propose
that this success is due to the size distribution of pores in the specially
designed porous silicon.
# 2001 Academic Press
Keywords: nucleation; phase diagrams; porous silicon; mesoporous;
*Corresponding author protein crystallization

Nucleation is the necessary ®rst step in the crys-
tallisation process, which in¯uences it decisively.
Consequently, the ability to control it is of primary
importance in crystallization experiments. Nuclea-
tion presents a free energy barrier which must be
overcome in a speci®c way, different from the
supersaturation conditions which subsequently
make crystal growth an energetically favourable
process.1 Formation of nuclei in the bulk of a sol-
ution is a stochastic process where protein mol-
ecules interact until a critical size aggregate is
formed. Any environment that favours a higher
local concentration of macromolecules provides a
potential nucleation point and may lower the
energy barrier for nucleation.
Previous studies attempting to ®nd nucleants
have been undertaken by introducing candidate
substances into crystallization trials in a controlled
manner.2 ± 6 Some have been useful for individual
proteins but none have yet turned out to be of gen-
eral use. Various other attempts to induce nuclea-
tion on irregular or rough surfaces, or surfaces of
special composition (e.g. poly-L-lysine, plastic)
have only proved effective in isolated cases.7,8
Abbreviations used: PEG, polyethylene glycol; MPD,
2-methyl-2,4-pentanediol; HF, hydro¯uoric acid.
E-mail address of the corresponding author:
n.chayen@ic.ac.uk
Due to the lack of general success of these sub-
stances, we decided to take a new approach,
searching for materials with cavities, which might
entrap protein molecules and encourage them to
nucleate and form crystals. Microporous and meso-
porous silicon materials consist of networks of
pores and cages, electrochemically etched on a
crystalline silicon surface.9,10 Porous materials have
recently been highlighted in Nature,11 concerning
their uses in various areas of chemistry, chemical
engineering, semiconductor research and the devel-
opment of physical, chemical and biological sen-
sors, but not for protein crystal growth. We tested
the suitability of a specially made porous silicon
(Figure 1) as a nucleant for protein crystallization.
A thin layer of porous silicon of 15 mm was elec-
trochemically fabricated on a silicon substrate
using a backside aluminium contact. The silicon
substrate was lightly boron-doped p-type single
side polished, so that the average pore size was 5-
10 nm. The dimensions of the pores exhibit a Gaus-
sian distribution, with an estimated standard devi-
ation of 3 nm. A picture of a cross-section of the
porous silicon material can be seen in Figure 1,
exhibiting a cleaved silicon substrate with a thin
porous silicon layer on top of it. The resulting thin
wafers of material were easily cut into pieces of
sub-millimetre dimensions and were immersed in
crystallization solutions. Crystal growth exper-

0022-2836/01/040591±5 $35.00/0 # 2001 Academic Press

592
Figure 1. A cross-section of porous silicon showing
the structure of the material (tree-like); the silicon is
white while the dark areas are pores in the layer. A por-
ous silicon layer was formed on a boron-doped p-type
crystalline silicon bulk with an aluminium backside con-
tact, using an electrochemical cell with HF:ethanol (1:1)
solution, rinsing in ethanol. (The HF aqueous solution
had a concentration of 48 % v/v). The 15 mm thick
layers with 65 % porosity were prepared over ten min-
utes with a current density of 30 mA/cm2. In the cell,
water molecules oxidise the silicon during the electro-
chemical anodization process. The SiO2 layer is sub-
sequently etched by the HF, and the ethanol is used
only to increase the ability of the solution to penetrate
and keep all the volume wet during the electrochemical
process. The samples were stored in ethanol prior to use
as a nucleation substrate. In this way, the formation of a
native silicon oxide was signi®cantly reduced.10
iments were performed using the two major crys-
tallisation methods, namely vapour diffusion hang-
ing drops and microbatch drops dispensed under a
layer of oil.12
As mentioned, nucleation requires very speci®c,
restricted conditions. In order to ®nd these con-
ditions, we experimentally determined a ``working
phase diagram'' for each protein, an example of
which, for catalase, is shown in Figure 2. Such a
phase diagram shows the supersolubility curve,
which is the threshold above which the protein
molecules spontaneously aggregate (as crystals or
amorphous precipitation). The porous silicon was
inserted at conditions just below the supersolubi-
lity curve (shown by arrows in Figure 2). The
supersolubility curves were established using the
IMPAX automated crystallization system, which
dispenses numerous small-volume crystallization
trials by a computer-controlled robot.13 This was
done by screening around published conditions
for these proteins, except in the case of the phyco-
Porous Silicon: a Nucleant for Protein Crystals
Figure 2. A working phase diagram determined using
microbatch for catalase at 18  C, showing in particular
the supersolubility curve. The variables are concen-
trations of protein and polyethylene glycol (PEG 6 K).
The other conditions are as in Table 1. The two arrows
show the conditions (metastable) at which porous silicon
induced nucleation of large crystals in drops which
otherwise remained clear.
biliprotein, which had not been previously crystal-
lised.
The effect of porous silicon was tested on a
variety of proteins: catalase, concanavalin A, lyso-
zyme, a phycobiliprotein, thaumatin and trypsin,
of Stokes' radii ranging from 2 to 5 nm. The
crystallization trials were also chosen to re¯ect a
selection of different common precipitating agents
and a wide range of pH.
Porous silicon was successful in inducing nuclea-
tion at conditions where the solution otherwise
remained clear (metastable), leading to the growth
of large single crystals of diffracting quality in ®ve
of the six proteins tested (Table 1 and Figure 3). In
some cases, crystals grew only on the silicon frag-
ment, leaving the rest of the drop completely clear
(Figure 3(a)). In other cases, one or more crystals
were also growing in other parts of the drop, their
numbers and sizes decreasing with distance from
the main silicon fragment (Figure 3(b) and (c)). It is
possible that some nuclei diffused away from the
nucleation site after their formation, or that nuclea-
tion also took place on extremely small pieces of
the material which had remained loosely attached
to the main fragment after cleavage. However, the
possibility of other mechanisms of facilitation of
nucleation not localized on the nucleant, due for
example to the creation of protein concentration
gradients, cannot be excluded. In order to ensure
that the crystals were actually attached to the
fragments when they were seen to be so, and not
Porous Silicon: a Nucleant for Protein Crystals 593

Figure 3. Crystals growing on and/or in the proximity of porous silicon fragments. (a) Lysozyme. Area shown is
2.5 mm  1.8 mm. (b) Trypsin. Area shown is 2.2 mm  2.0 mm. (c) Phycobiliprotein. Area shown is
3.0 mm  2.3 mm. (d) Phycobiliprotein close-up showing a crystal attached onto a silicon fragment. Area shown is
1.1 mm  1.0 mm. (e) Same crystal still attached after having moved the fragment. Area shown is 1.0 mm  0.8 mm.
The porous silicon coated wafers were broken into small pieces (ca. 0.06 mm2) and placed inside the droplets set at
various conditions below the supersolubility curves of the proteins. Both microbatch and hanging-drop vapour diffu-
sion set-ups were used, with drop volumes ranging from 2 to 5 ml. For the microbatch trials, Terazaki-type plates
were purchased from Nunc (Denmark). The fragments of porous silicon wafer were placed on the bottom of the
plates' wells (depressions); the crystallization drops were dispensed onto the fragments in the wells and covered with
paraf®n oil. For the vapour diffusion trials, Linbro-type crystallization plates contained 1 ml of the reservoir solutions.
The drops and the silicon fragments were dispensed on silanized glass coverslips that were inverted above the wells
and sealed with Apiezon C oil (M&I, Manchester, UK). All the experiments were run at 18  C.

just lying above or below them, we turned the
fragments in the solutions using microtools. In all
cases, the crystals remained attached to them (e.g.
Figure 3(d) and (e)). The porous silicon was not
effective in the case of concanavalin A. Other por-
ous materials, e.g. (alumino) silicates of uniform
pore sizes up to 5 nm (VPI-514 and MCM-4115)
were also investigated in the course of this study
but were not found to in¯uence the nucleation
process.
Sanjoh and co-workers have reported the electro-
statically driven heterogeneous nucleation of a pro-
tein, lysozyme, on micro¯uidic silicon devices. The
inducement or avoidance of nucleation was depen-
dent on the pH of the crystallization solution and
the surface charge of the protein molecules.16,17
Such dependence was not observed in the case of
the porous silicon reported here. We therefore
assume that the mechanical constraint of the mol-
ecules by the pores, and thus the creation of loca-
594 Porous Silicon: a Nucleant for Protein Crystals

Table 1. Conditions under which the proteins crystallized in the presence of porous silicon (no crystals grow in these
conditions in the absence of porous silicon)
Protein Mol. weight (kDa) Metastable conditions used for nucleation with porous silicon Protein conc. (mg/ml)
Lysozyme 14.5 6%(sat) sodium chloride 20 mM sodium citrate (pH 4.6) 36
Trypsin 24 30-32%(sat) ammonium sulphate 100 mM Tris (pH 8.4) 12-20
Catalase 232 5.1-5.4%(w/v) PEG 6K, 5%(v/v) MPD 100 mM Tris (pH 7.5) 11.5-12.5
Thaumatin 22 0.34 M sodium-potassium tartrate 50 mM Pipes (pH 6.8) 16
Phycobiliprotein 126 0.6-0.7 M ammonium sulphate 40 mM Mes (pH 6.1) 1.5 mM 11
dodecyl maltoside
Bovine liver catalase (Cat. no. C-9322), jack bean concanavalin A type IV (C-2010), hen egg-white lysozyme (L-6876), thaumatin
from Thaumatococcus daniellii (T-7638) and porcine pancreas trypsin (T-0134) were purchased from Sigma (Steinheim, Germany). The
phycobiliprotein was prepared and puri®ed in-house (J. Nield et al., unpublished). Polyethylene glycol of mean molecular weight
6000 (PEG 6 K), 2-methyl-2,4-pentanediol (MPD) and the various salts used were also purchased from Sigma. In order to control
nucleation, it is necessary to work in very clean conditions. Hence, the lysozyme, trypsin and thaumatin stock solutions were ®ltered
with a 300,000 Mr cut-off ®lter, and catalase and the phycobiliprotein with a 0.22 mm mesh size ®lter (Ultrafree-MC, Millipore, Bed-
ford, USA) before setting up the experiments. Concanavalin A was not ®ltered because part of the protein appeared to be sticking to
the ®lter.

lised supersaturation maxima, is the principal
mechanism which facilitates the crystallization.
A practical advantage of porous silicon over the
silicon devices is that porous silicon can be added
to the trials as small solid fragments in any stan-
dard crystallization set-up. There is no need to use
specially designed vessels or to modify the crystal-
lization method that has been chosen by the exper-
imenter.
There is increasing pressure to ®nd methods that
will help to produce suitable protein crystals. This
is becoming more and more obvious from pilot
structural genomics projects, which show the suc-
cess rate of getting from clone to structure to be
about 10 %. Production of crystals suitable for X-
ray crystallography is found to be the rate-limiting
step (http://proteome.bnl.gov/progress.html). The
ultimate aim would be to ®nd a ``universal'' nucle-
ant, which would promote crystallization of a high
percentage of the proteins subjected to crystallisza-
tion attempts under a wide range of conditions.
Though this hope may appear somewhat unrealis-
tic, porous silicon induces nucleation of a range of
protein crystals, at conditions of supersaturation
that are favourable to crystal growth, but are
inadequate or insuf®cient for spontaneous nuclea-
tion. The discovery of such agents has threefold
importance. Firstly, they can provide a means to
maximize the chances of obtaining crystals during
initial screening of crystallization conditions (par-
ticularly useful in the structural genomics era). Sec-
ondly, they can be used to grow crystals at
metastable conditions, at which the slower growth
and the lack of excess and secondary nucleation
often provide for growth of larger, better diffract-
ing crystals.1,18 Thirdly, they represent a ®rst step
towards the universal nucleant.
This is the ®rst report of the possibly wide effec-
tiveness of an engineered porous material as a
nucleant for protein crystallization. We believe its
effectiveness is due to the size distribution of its
pores. This distribution (of pore sizes) on the por-
ous silicon surface may be a crucial advantage, in
providing for each of various proteins the pore size
most suited to its molecular diameter and to the
shape of the initial aggregates that it forms. In con-
trast, the other porous materials tested have mini-
mal variation of their mean pore size.
Further research into the use of porous or cage-
structured materials for protein crystallization, if it
goes hand-in-hand with the rapid technical devel-
opments in this ®eld, is likely to lead to the discov-
ery of more widely effective heterogeneous crystal
nucleants, or those which may be more speci®c to
recalcitrant proteins.
Acknowledgements
We thank Professor D.M. Blow for fruitful discussions
concerning this research and Dr J. Nield for providing
the phycobiliprotein. The European Commission EBCI
Initiative (B104-CT98-0086) and the BBSRC are gratefully
acknowledged for ®nancial support.
References
1. Feher, G. & Kam, Z. (1985). Nucleation and growth
of protein crystals: general principles and assays.
Methods Enzymol. 114, 77-112.
2. McPherson, A. & Shlichta, P. (1988). Heterogeneous
and epitaxial nucleation of protein crystals on min-
eral surfaces. Science, 239, 385-387.
3. Chayen, N. E., Radcliffe, J. W. & Blow, D. M. (1993).
Control of nucleation in the crystallization of lyso-
zyme. Protein Sci. 2, 113-118.
4. Blow, D. M., Chayen, N. E., Lloyd, L. F. &
Saridakis, E. (1994). Control of nucleation of protein
crystals. Protein Sci. 3, 1638-1643.
5. Edwards, A. M., Darst, S. A., Hemming, S. A., Li, Y.
& Kornberg, R. D. (1994). Epitaxial growth of pro-
tein crystals on lipid layers. Struct. Biol. 1, 195-197.
6. Punzi, J. S., Luft, J. & Cody, V. (1991). Protein crys-
tal growth in the presence of poly(vinylidene
di¯uoride) membrane. J. Appl. Crystallog. 24, 406-
408.
7. Fermani, S., Falini, G., Minnucci, M. & Ripamonti,
A. (2001). Protein crystallization on polymeric ®lm
surfaces. J. Crystal Growth, 224, 327-334.
Porous Silicon: a Nucleant for Protein Crystals
8. Chayen, N. E. & Saridakis, E. (2001). Is lysozyme
really the ideal model protein? J. Crystal Growth,
232, 262-264.
9. Cullis, A. G., Canham, L. T. & Calcott, P. D. J.
(1997). The structural and luminescence properties
of porous silicon. J. Appl. Phys. 82, 909-965.
10. Halimaoui, A. (1997). Porous silicon formation by
anodisation. In Properties of Porous Silicon (Cahnam,
L., ed.), EMIS Datareviews series, INSPEC no. 18,
pp. 12-22, The Institution of Electrical Engineers,
London.
11. Sakamoto, Y., Kaneda, M., Terasaki, O., Zhao, D. Y.,
Kim, J. M. & Stucky, G., et al. (2000). Direct imaging
of the pores and cages of three-dimensional meso-
porous materials. Nature, 408, 449-453.
12. Chayen, N. E. (1997). The role of oil in macromol-
ecular crystallization. Structure, 5, 1269-1274.
13. Chayen, N. E., Shaw-Stewart, P. D. & Blow, D. M.
(1992). Microbatch crystallization under oil- a new
(Received 14 June 2001; received in revised form 2 August 2001; accepted 6 August 2001)
595
technique allowing many small-volume crystalliza-
tion trials. J. Crystal Growth, 122, 176-180.
14. Davis, M. E., Montes, M., Hathway, P. E., Arhancet,
J. P., Hasha, D. L. & Garces, J. M. (1989). Physico-
chemical properties of VPI-5. J. Am. Chem. Soc. 111,
3919-3924.
15. Chen, C., Li, H. & Davis, M. E. (1993). Studies on
mesoporous materials. I. Synthesis and characteriz-
ation of MCM-41. Microporous Mater, 2, 17-26.
16. Sanjoh, A. & Tsukihara, T. (1999). Spatiotemporal
protein crystal growth studies using micro¯uidic
silicon devices. J. Crystal Growth, 196, 691-702.
17. Sanjoh, A., Akioka, K., Inoue, S., Tsukihara, T. &
Gorti, S. (2000). 8th International Conference on the
Crystallization of Biological Macromolecules. In
Abstracts, p. 32, Sandestin, Florida.
18. Ataka, M. (1993). Protein crystal growth: an
approach based on phase diagram determination.
Phase Transitions, 45, 205-219.
Edited by R. Huber