Kinetic and thermodynamic control of protein adsorption

J. Satulovsky, M. A. Carignano, and I. Szleifer

PNAS 2000;97;9037-9041; originally published online Jul 25, 2000;
doi:10.1073/pnas.150236197
This information is current as of February 2007.

Online Information High-resolution figures, a citation map, links to PubMed and Google Scholar,
& Services etc., can be found at:
www.pnas.org/cgi/content/full/97/16/9037
This article has been cited by other articles:
www.pnas.org/cgi/content/full/97/16/9037#otherarticles
E-mail Alerts Receive free email alerts when new articles cite this article - sign up in the box
at the top right corner of the article or click here.
Rights & Permissions To reproduce this article in part (figures, tables) or in entirety, see:
www.pnas.org/misc/rightperm.shtml
Reprints To order reprints, see:
www.pnas.org/misc/reprints.shtml

Notes:
Kinetic and thermodynamic control of
protein adsorption
J. Satulovsky, M. A. Carignano, and I. Szleifer*
Department of Chemistry, Purdue University, West Lafayette, IN 47907
Communicated by Stuart A. Rice, University of Chicago, Chicago, IL, May 22, 2000 (received for review October 22, 1999)
Control of nonspecific protein adsorption is very important for the
design of biocompatible and biomimetic materials as well as drug
carriers. Grafted polymer layers can be used to prevent protein
adsorption. We have studied the molecular factors that determine
the equilibrium and kinetic control of protein adsorption by
grafted polymer layers. We find that polymers that are not at-
tracted to the surface are very effective for kinetic control but not
very good for equilibrium reduction of protein adsorption. Poly-
mers with attractions to the surface show exactly the opposite
behavior. The implications for molecular design of biocompatible
materials also are discussed in this paper.
P 冋
rotein adsorption plays a major role in a variety of important
biological-related processes. Biocompatible materials are
required to eliminate, or largely reduce, the adsorption of blood
proteins, for example, to avoid surface-induced thrombosis (1,
2). In recent years special attention has been given to the
modification of protein–surface interactions using grafted poly-
mers (3, 4, 5, 6). The idea, borrowed from many years of research
in colloidal stabilization (7) and mimicking one of the roles of
polysaccharides in cell membranes (8), is to build a steric barrier
to the proteins by the presence of the polymer layer (9, 10). In
practice, however, there is still no systematic way of modifying
surfaces that are successful in preventing protein adsorption.
This is because of the complexity involved in the process caused
by the interplay between changing average structure of the
polymer–protein layer, strong protein–surface attractions, and
polymer–surface interactions (11). For example, liposomes†
formed by mixtures of lipid and lipid-polymer have shown
increased longevity in the bloodstream and are now being used
as drug delivery systems (12, 13). It is believed that the role of
the polymers tethered to the liposome interface is to present a
steric barrier to approaching proteins from the blood. However,
the amount of polymer on the surface is rather small, and
experimental observations of the same polymer grafted on
hydrophobic surfaces, at the same surface coverage, show that
although the amount of protein adsorbed is lower than in the
absence of polymer, there is still a significant amount of protein
adsorption (14).
In this paper we demonstrate that the molecular factors that
determine the ability of the polymer layer to prevent protein
adsorption are different for the kinetic control as compared with
the equilibrium case. We will demonstrate that our theoretical
predictions can quantitatively reproduce experimental observa-
tions of protein adsorption isotherms. We will show that sur-
face–polymer interactions play a central role in the ability of the
polymer layer to prevent protein adsorption. However, the role
of those interactions is different for kinetic as compared with
equilibrium control. Our findings enable us to rationalize why
the surface-modified liposomes have an increased longevity but
polymers on hydrophobic surfaces are not as effective to prevent
protein adsorption. Finally, our theoretical predictions can be
used to decide the type of surface modification necessary for the
rational design of biomaterials.
Consider a protein solution of density ␳b that is put into
contact with a surface. The surface has polymers that are
chemically attached to it at one of their ends. The presence of the
surface induces a gradient in the chemical potential of the
protein. This gradient arises from the sudden inhomogeneous
environment that appears in the direction perpendicular to the
surface. The proteins in the close vicinity of the surface feel the
bare attractive interaction of the surface as well as the repulsions
induced by the polymer molecules. The balance between these
interactions will result in a final equilibrium density profile and
amount of protein adsorbed.
The time evolution from the homogeneous system to the new
equilibrium induced by the presence of the modified surface can
be described (in the free-draining limit) with the help of a
diffusion equation (15–18) of the form
册
⭸ ␳ pro 共z;t兲 ⭸ ⭸
⫽ D pro ␳ 共z;t兲 ␮ pro 共z;t兲 , [1]
⭸t ⭸z pro ⭸z
where we have assumed that the diffusion in the xy plane is faster
than the adsorption. This assumption is justified for the relatively
low surface densities that we treat here (19). The time-
dependent local chemical potential is defined by ␮pro(z;t) ⫽
␦W
, where W is the free-energy density of the system. This
␦␳pro(z;t)
free energy includes the contribution of the grafted polymer
molecules, the solvent, and the surface, under the condition of
the fixed protein concentration profile given by ␳pro(z;t) at each
time t.
The adsorption kinetics can be described with the help of Eq.
1, assuming that the time scale of the diffusion of the proteins is
much slower than that of the polymer and solvent rearrange-
ment. This is a physically motivated approximation, because the
solvent molecules (water) are much faster than the protein
motion, and the rearrangements of the polymer molecules
involve only local moves, which, for the flexible chains of interest
here, are much faster than the typical motion of the larger
protein molecules.
The free energy of the system is determined by using a
molecular mean-field theory. The predictions of the theory for
equilibrium systems have been shown to be in excellent agree-
ment with experimental observations for conformational and
thermodynamic properties of tethered polymer layers (20, 21) as
well as for the adsorption of proteins on hydrophobic surfaces
(11, 14). The basic idea of the theory is to treat each molecule
with all of its intramolecular and surface interactions exactly
taken into account (within the chosen model system). The
intermolecular interactions are considered within a mean-field
approximation. The mean field is determined by the average
Abbreviations: pdf, probability distribution function; EO, ethylene oxide; PEO, polyethyl-
ene oxide.
*To whom reprint requests should be addressed. E-mail: igal@purdue.edu.
†Liposomes
are closed spherical bilayers formed by lipid molecules with varying radii from
BIOPHYSICS
10 nm to micrometers.
The publication costs of this article were defrayed in part by page charge payment. This
article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C.
§1734 solely to indicate this fact.
Article published online before print: Proc. Natl. Acad. Sci. USA, 10.1073兾pnas.150236197.
Article and publication date are at www.pnas.org兾cgi兾doi兾10.1073兾pnas.150236197
PNAS 兩 August 1, 2000 兩 vol. 97 兩 no. 16 兩 9037–9041
properties of all of the molecules in the mixture, and thus the
theory is in essence a self-consistent approach.
We write the free energy of the system in complete analogy
with the equilibrium approach (22) but with all of the quantities
now assumed to be time-dependent. The explicit expression for
the time-dependent free energy density (per unit area) reads
␤W ⫽ ␴ 冉冘␣
P共 ␣ ;t兲lnP共 ␣ ;t兲 ⫹ ln␴ ⫹ 冊 冕 ␾ s 共z;t兲ln␾ s 共z;t兲dz ⫹
冕 ␳ pro 共z;t兲关ln␳ pro 共z;t兲 ⫹ ␤ U ps 共z兲兴dz,
where the first two terms are the contribution from the tethered
polymers, with the first being the conformational entropy of the
chains, with P(␣;t) being the time-dependent probability distri-
bution function (pdf) of chain conformations and the second
being the tethered polymers two-dimensional translational en-
tropy. The third term is the solvent entropy, with ␾s(z;t) being the
volume fraction of solvent at distance z from the surface at time
t. The last two terms correspond to the translational entropy of
the proteins and the protein–surface interaction, respectively,
Ups(z) is the bare surface–protein interaction. This interaction is
taken from ref. 23 where the potential between lysozyme and a
hydrophobic surface was calculated by using atomistic potentials.
Note that the solvent and protein contributions are integrated
over z to account for all of the molecules.
Eq. 2 assumes that the intermolecular attractions between all
of the components in the system are the same. This is actually the
model that we successfully used to compare experimental ob-
servations and the predictions of the theory in early work (14)
and in the comparisons shown below. The excellent agreement
between the predictions of the theory and the experimental
observations supports the assumption.
Eq. 2 does not include the repulsive interactions between the
molecules. The hard-core repulsions are accounted for by pack-
ing constraints in which the available volume to the molecules,
at each distance z from the surface, is filled by polymer, protein,
or solvent molecules. This constraint reads
␴具ng共z;t兲典v0 ⫹ 冕 ␳pro共z⬘;t兲v共z;z⬘兲dz⬘ ⫹ ␾s共z;t兲 ⫽ 1, for all z, [3]
where the first term is the volume fraction of polymer at z, with
具ng(z;t)典dz ⫽ ⌺␣P(␣;t)ng(z,␣;t)dz being the average number of
polymer segments of volume v0 at time t at distance z from the
surface. The second term is the volume fraction of proteins with
v(z;z⬘)dz being the volume that the proteins at distance z⬘ from
the surface contribute to z. The last term is the volume fraction
of solvent.
The assumption of solvent and polymer equilibration in the
time scale of protein motion implies that the time-dependent pdf
of tethered polymer configurations and the solvent-density
profile are those that minimize the free energy (Eq. 2) subject
to the packing constraint (Eq. 3) for fixed protein density profile.
Thus, at each time t the density of proteins calculated from the
diffusion equation (Eq. 1) is given as input to the free energy and
this free energy is minimized subject to the packing constraints.
This minimization is done by introducing time-dependent La-
grange multipliers, ␤␲(z;t). The pdf of chain conformations is
P共 ␣ ;t兲 ⫽
1
q共t兲
exp ⫺ 冋冕 册
␤␲(z;t)v 0 n g 共z, ␣ ;t兲dz ,
9038 兩 www.pnas.org
[2]
Fig. 1. Theoretical predictions for the amount of protein adsorbed as a
function of the surface coverage of EO oligomers (full lines). The experimental
observations (symbols) are from Prime and Whitesides (24, 25), where they
measured the thickness of the adsorbed film (in nm), which is proportional to
the surface density of protein, as a function of composition of a mixed
self-assembled film. The mixed films are composed by SC10OH and SC11(EO)n,
with n ⫽ 6 and 17, on gold surfaces. The relationship between the experi-
mentally observed thickness and the protein density was fitted so that at zero
composition the predictions and the experimental observations agree. The
parameters used in the calculations are from refs. 11 and 22.
where q(t) is the time-dependent normalization constant. For
the solvent-density profile, the expression is
␾ s 共z;t兲 ⫽ exp[⫺␤␲(z;t)v 0 ]. [5]
The numerical values of the Lagrange multipliers are obtained
by replacing the expression for the pdf, Eq. 4, and the solvent-
density profile, Eq. 5, into the constraint equation, Eq. 3, where
the protein contribution is given as an input obtained from the
diffusion equation, Eq. 1. Once the Lagrange multipliers are
obtained, one can calculate the time-dependent protein chemical
potential that is needed for the time evolution of the adsorption.
This calculation is done by differentiating the free-energy den-
sity to obtain
␤␮ pro 共z;t兲 ⫽ ln␳ pro 共z;t兲 ⫹ ␤ U ps 共z兲 ⫺ 冕 ␤␲ 共z⬘;t兲v共z⬘;z兲dz⬘,
[6]
which is readily obtained once the Lagrange multipliers are
known. Thus, the time evolution of the system is obtained in the
following way. At time t ⫽ 0, a tethered polymer layer in
equilibrium with pure solvent is brought into contact with a
protein solution. The initial density profile of proteins enables
the calculation of ␲(z;0) from which ␮pro(z;0) is obtained by use
of Eq. 6. From it, the diffusion equation (Eq. 1) is iterated a time
step. The new density profile of proteins now is given as input for
the calculation of the new Lagrange multipliers from which the
chemical potential profile is obtained and the time evolution is
further iterated. This procedure is continued until the new
equilibrium condition is achieved in which ␮pro(z;t) ⫽ ␮pro.eq for
all z.
Eq. 6 shows that the motion of the proteins toward the surface
[4]
contains three contributions. The first one is the purely diffusive
motion that is found in any ideal system because of the gradient
in composition. The second arises from the force exerted to the
Satulovsky et al.
Fig. 2. (A) Amount of proteins adsorbed, in molecules per Å2, as a function of time. The dashed lines correspond to the equilibrium amount of protein adsorbed.
The lines correspond to the following number of EO units: black, no polymer; red, n ⫽6; green, n ⫽15; blue, n ⫽ 25; and orange and magenta, n ⫽ 50. The magenta
curve corresponds to a hydrophobic surface. (B) The potential of mean force, in kJ兾mol, as a function of the distance from the surface for: black, bare
protein–surface interaction; and red, n ⫽ 50 before the beginning of the adsorption. The green and blue curves are for n ⫽ 50 at the times marked by the arrows
in A. (C) Maximal steric repulsion for n ⫽ 50 as a function of time. Note the changes in the shape and strength of the potential as the adsorption process takes
place. All of the results correspond to a grafted density of ␴ ⫽ 0.0012Å⫺2. Higher surface coverage of polymer will result in slower kinetics but qualitatively similar
behavior.

proteins by the surface. These two terms do not depend on the
intermolecular interactions and will lead to a fast approach of
the proteins to the surface. We refer to them as the purely
diffusive motion (see e.g., Fig. 2 and discussion thereafter). The
last term is the one arising from the intermolecular (protein–
polymer and protein–protein) interactions. This term will be
responsible for steric barriers and it is expected to lead to
regimes in which the motion is dominated by kinetic barriers.
The interplay between the diffusion and barrier crossing to
determine the total adsorption process is discussed in detail
below.
Fig. 1 shows the predicted equilibrium adsorption isotherms
for lysozyme and fibrinogen on surfaces with short oligomeric
polyethylene oxide (PEO) together with the experimental ob-
servations of Prime and Whitesides (24, 25). The calculations
were carried out by full minimization of the free energy (ther-
modynamic equilibrium), i.e., with respect to the protein degrees
of freedom as well, as explained in detail in refs 14 and 22. The
agreement between the theory and the experimental observa-
tions is excellent. The adsorption isotherms show that as the
surface coverage of polymer increases, the amount of protein
adsorbed decreases. There is a surface coverage of oligomers
above which there is zero protein adsorption. The role of the
short chains is basically to fill the space close to the surface,
reducing the amount of available surface for the adsorbing
proteins (11, 22, 24, 25). The most important point to note is that
these chains are short and that the surface coverage at which the
adsorption goes to zero is very large. Those surface coverages are
an order of magnitude larger than are typical surface coverages
obtained by grafting PEO-2,000 or PEO-5,000‡. Further, the
calculations assume that the ethylene oxide (EO) oligomers are
not attracted to the surface. In contrast, the ability of long-
grafted PEO to reduce protein adsorption on modified hydro-
phobic glass surfaces was found to be closely linked to the fact
that the EO segments are strongly attracted to the surface and
compete with the proteins for adsorption sites (14).
How good is the equilibrium assumption? To this end, we solve
the diffusion Eq. 1 with the free energy obtained from the same
theoretical approach that was used in the equilibrium predictions
of Fig. 1. Fig. 2A shows the time evolution of the amount of
protein adsorbed on the surface for a variety of PEO chain
lengths. All of the profiles but one correspond to PEOs that are
not attracted to the surface; we discuss the case of the attractive
surface later. There are several interesting features that can be
observed from the profiles. First, as the chain length of the
grafted polymers increases up to 25 segments (corresponding
close to PEO-1,000), the equilibrium amount of protein ad-
sorbed decreases. However, for PEO-2,000, the amount of
protein adsorbed at equilibrium is almost identical to that of
PEO-1,000. The equilibrium amount of protein adsorbed is
independent of polymer chain length once the thickness of the
layer is larger than the protein size (22).
Second, the kinetics of protein adsorption is qualitatively
similar for all chain lengths up to PEO-1,000. There are two
qualitatively different regimes. For short times, the adsorption is
very fast, and it is controlled by the diffusion of the proteins to
the surface, i.e., the first two terms of Eq. 6. The second regime
is a much slower one, and it is controlled by the repulsive barrier
that the proteins already adsorbed on the surface and the grafted
polymers present to the approaching proteins, i.e., the motion is
dominated by the last term in Eq. 6. The total time to reach
equilibrium in these cases supports the equilibrium assumption
used to obtain Fig. 1.
Third, for PEO-2,000, there is no fast regime, i.e., there is no
diffusion-controlled adsorption. At this molecular weight and for
the surface coverage shown, the polymer chains present a steric
BIOPHYSICS

barrier that causes the adsorption to be controlled by the kinetics

of barrier crossing, Fig. 2B.
‡PEO-M, also called PEG-M for polyethylene glycol, is a chain of molecular weight M. Thus,

the number of segments in the chain is given by M兾45. We refer here to the commonly used
The dependence of the time scales for adsorption on polymer
experimental molecular weights. The exact number of segments used in the calculations molecular weight also can explain the success of pegolated
is given in the figure legends. liposomes having increased longevity in the bloodstream. Typ-

Satulovsky et al. PNAS 兩 August 1, 2000 兩 vol. 97 兩 no. 16 兩 9039

Fig. 3. (Left) Schematic representation of the ability of grafted polymers to prevent protein adsorption. For polymers attracted to the surface, there is no kinetic
barrier but the protein competes with the polymer for adsorption sites. Polymers that are not attracted to the surface present a large steric barrier but not very
good thermodynamic prevention because of the ability of the protein to deform the polymer layer. (Right) The curves on the right show the calculated average
shapes of two neighboring polymers as a function of time. The shape is defined here (20) by the lateral average radius of gyration of the chains as a function
ical molecular weights used in those systems are PEO-2,000 to
PEO-5,000. As we can see from Fig. 2 for PEO-2,000, the chains
are long enough that the adsorption is completely controlled by
the kinetic barriers presented by the polymers, and the time scale
to achieve equilibrium is much longer. Actually, the system only
reaches equilibrium in many hours. Moreover, the calculations
presented here are for lysozyme, which is a relatively small
protein as compared to fibrinogen. Thus, for fibrinogen the time
scale for adsorption is going to be much longer, because its large
size results in very large repulsions caused by the interactions
with the polymers before reaching the surface§.
The time-dependent adsorption for the case of PEO-2,000
with attractive interactions with the surface (hydrophobic sur-
faces) shows two important differences with the nonattractive
case. First, the equilibrium amount of adsorbed proteins is much
lower for the hydrophobic surface as compared with the same
molecular weight for nonattractive surfaces. Namely, grafted
polymers that are attracted to the surface are more effective in
reducing protein adsorption. Second, the time scale for adsorp-
tion, both for the initial adsorption and to reach equilibrium, are
orders of magnitude faster for surfaces with attractive interac-
tions with the polymers. The reason for the faster kinetics is that
the polymers have almost all of their segments close to the
surface and thus present no steric barrier to the approaching
proteins up to the point of close vicinity with the surface. The
reason for being better at preventing the equilibrium adsorption
is that the proteins need to compete for adsorption sites with the
polymers. The short time scale for adsorption on hydrophobic
surfaces suggests that the recent comparisons (14) with exper-
imental observations assuming equilibrium are probably very good.
We now discuss in some detail the origin and characteristics
of the repulsive barriers that the proteins feel in the different
cases. First, let us consider the case of no grafted polymer or
§Thebare potential of fibrinogen with surfaces is not known and therefore it is very hard
to provide a reliable estimate for the potential of mean force. The calculations presented
here use the same attractive contribution between the surface and fibrinogen that was
proposed in ref. 14. This interaction is only the attractive strength at contact.
polymer molecular weights below PEO-2,000. For the surface
coverage considered here, we see that in all of the cases there is
a diffusion-controlled regime of the adsorption kinetics. This
finding implies that the proteins approaching the surface feel a
purely attractive interaction potential with no barriers. The
dramatic change in slope in the kinetics is attributable to
the presence of a barrier in the potential of mean force that the
proteins approaching the surface feel¶. For the hydrophobic
surfaces, the same effect is observed but up to much longer
molecular weights (see above). For the PEO-2,000 with no
attraction with the surface, the polymers present a kinetic barrier
even in the absence of any adsorbed proteins (Fig. 2 B and C).
The presence of the barrier implies that there is no diffusion-
controlled regime. Furthermore, any molecular weight above
2,000 will show a stronger barrier and thus the time scale of the
initial adsorption will be even longer.
The main conclusion from this work is that the best surface
modifiers for the kinetic control of protein adsorption are
polymers that are not attracted to the adsorbing surface. How-
ever, the best type of polymers for the thermodynamic control
of protein adsorption are those that are attracted to the surface.
The molecular picture that emerges is summarized in Fig. 3,
where a cartoon is shown of how the different types of polymers
kinetically or thermodynamically prevent the protein adsorption,
along with the calculated variation of the shape of the grafted
polymers as a function of time. The complexity involved in the
adsorption process can be seen in the time-dependent average
shape of the polymer molecules. The kinetics of protein adsorp-
tion on surfaces with grafted polymers involves the competition
between the strong bare attractive interactions between the
surface and the proteins and the conformational entropy loss
that the grafted polymers have to pay to accommodate the
proteins. This process involves a constantly varying environment,
¶The potential of mean force here refers to the interaction (as a function of the distance
from the surface) of a protein with the surface averaged over all of the configurations of
the polymer molecules, solvent, and other proteins in the system, which is given exactly by
the last term in Eq. 6.
and therefore the kinetic behavior cannot be characterized in
simple terms. We find that the kinetic barriers for protein
adsorption vary constantly in time, and the whole potential is a
complicated function that varies with distance and time as a
result of the constant rearrangement of the polymer–protein
layer (see Fig. 2 B and C).
We have presented a detailed study of the equilibrium and
kinetic adsorption of proteins on surfaces with grafted polymers.
The extension of the molecular theory for tethered polymer
layers to dynamical systems enables the study of kinetic processes
over several orders of magnitude in time without compromising
molecular detail. Note that although this approach does not
provide the detail of molecular dynamics simulations, it enables
the study of the dynamic processes over 15 orders of magnitude
in time, thus bridging the gap between microseconds and days in
complex systems with the ability to follow changes on average
molecular properties.
Biocompatible materials can be in contact with the blood-
stream for relatively short times (e.g., drug carriers) or very long
times (e.g., artificial organs). The type of surface modifier that
1. Horbett, T. A. (1993) Cardiovasc. Pathol. 2, 137S–148S.
2. Salzman, E. W., Merrill, E. W. & Hent, K. C. (1994) in Hemostasis and
Thrombosis, eds. Colman, R. W., Hirsh, J., Marder V. J. & Salzman, E. W.
(Lippincott, Philadelphia), pp. 1469–1485.
3. Milton Harris, J. & Zalipsky S., eds. (1997) Poly(ethylene glycol): Chemistry and
Biological Applications (Am. Chem. Soc., Washington, DC).
4. Andrade, J. D., ed. (1985) Surface and Interfacial Aspects Of Biomedical
Polymers (Plenum, New York).
5. Torchilin, V. P. & Trubetskoy, V. S. (1995) Adv. Drug Delivery Rev. 16, 141–155.
6. Needham, D., McIntosh, T. J., Simon, S. A. & Zhelev, D. (1998) Curr. Opin.
Colloid Interface Sci. 3, 511–517.
7. Napper, D. H. (1983) Polymeric Stabilization of Colloidal Dispersions (Aca-
demic, New York).
8. Holland, N. B., Yongxing, Q., Ruegsegger, M. & Marchant, R. E. (1998) Nature
(London) 392, 799–801.
9. Ikada, Y. (1984) Adv. Polym. Sci. 57, 104–140.
10. Jeon, S. I., Lee, J. H., Andrade, J. D. & de Gennes, P. G. (1991) J. Colloid
Interface Sci. 142, 149–158.
Satulovsky et al.
is necessary for protein rejection in each of these cases is
different. In systems where it is necessary to control protein
adsorption over finite time scales, it is probably best to use
relatively dense polymer layers with long polymers that are not
attracted to the surface of the biomaterial. This scenario should
provide the best kinetic control. On the other hand, for materials
that need to be in contact with the bloodstream for years, the
optimal type of polymers may be one that is attracted to the
surfaces, then providing thermodynamic control. It is important
to mention that very high surface coverage of grafted polymers
is extremely hard to obtain experimentally (14, 26). Thus,
alternative approaches may be needed, for example, mixtures of
polymers that present optimal kinetic and thermodynamic con-
trol under experimental realizable conditions. The theory used
here provides quantitative guidelines on how to build those
optimal layers.
This work is supported by the National Science Foundation through
Grant CTS-9624268. I.S. is a Camille Dreyfus Teacher兾Scholar.
11. Szleifer, I. (1997) Curr. Opin. Solid State Mater. Sci. 2, 337–344.
12. Lasic, D. & Martin, F., eds. (1995) Stealth Liposomes (CRC, Boca Raton, FL).
13. Fisher, L. & Lasic, D. (1998) Curr. Opin. Colloid Interface Sci. 3, 509–510.
14. McPherson, T., Kidane, A., Szleifer, I. & Park, K. (1998) Langmuir 14, 176–186.
15. Fraaije, J. G. E. M. (1993) J. Chem. Phys. 99, 9202–9212.
16. Hasegawa, R. & Doi, M. (1997) Macromolecules 30, 3086–3089.
17. Kondepudi, D. & Prigogine, I. (1998) Modern Thermodynamics: From Heat
Engines to Dissipative Structures (Wiley, New York).
18. Marconi, U. M. B. & Tarazona, P. (1999) J. Chem. Phys. 110, 8032–8044.
19. Ravichandran, S. & Talbot, J. (2000) Biophys. J. 78, 110–120.
20. Szleifer, I. & Carignano, M. A. (1996) Adv. Chem. Phys. 94, 165–260.
21. Szleifer, I. (1996) Curr. Opin. Colloid Interface Sci. 1, 416–423.
22. Szleifer, I. (1997) Biophys. J. 72, 595–612.
23. Lee, S. J. & Park, K. (1994) J. Vac. Sci. Technol. 12, 1–7.
24. Prime, K. L. & Whitesides, G. M. (1991) Science 252, 1164–1167.
25. Prime, K. L. & Whitesides, G. M. (1993) J. Am. Chem. Soc. 115, 10714–10721.
26. Halperin, A., Tirrell, M. & Lodge, T. P. (1992) Adv. Pol. Sci. 100, 31–71.
BIOPHYSICS
PNAS 兩 August 1, 2000 兩 vol. 97 兩 no. 16 兩 9041