Antifreeze protein hydration waters: Unstructured unless bound to ice


COMMENTARY

Antifreeze protein hydration waters: Unstructured
unless bound to ice
Sean M. Marksa and Amish J. Patela,1

How do fish, insects, and other organisms survive in
frigid polar environments? They do so with the help of
remarkable molecules known as antifreeze proteins
(AFPs), which suppress freezing and associated cell
death despite being present at concentrations of less
than 1 wt % (1). In contrast, automotive antifreeze
needs roughly 20 to 50 wt % of the additive to function
(2). AFPs are able to suppress freezing at such low
concentrations because, unlike antifreeze, they do
not rely on altering the inherent structure of water;
instead, they bind to nascent ice nuclei and prevent
them from growing (3). Thus, being able to recognize
and preferentially bind ice in a vast excess of water is
the key to AFP function (1). However, in the absence of
any chemical differences between water and ice, how
do AFPs discriminate between them? Moreover, both
water and ice are composed of a tetrahedral network of
hydrogen bonds, so even the structural differences be-
tween them are subtle. Indeed, how AFPs are able to
perform what has been touted as one of the most chal-
lenging molecular-recognition tasks in all of biology has
long been a source of amazement and intrigue (1, 4). In
PNAS, Hudait et al. (5) clarify important aspects of the
ice-recognition puzzle by using molecular simulations
to study TmAFP, a hyperactive insect AFP.

What makes this puzzle even more fascinating is
that a wide array of organisms, ranging from bacteria
to insects and fishes, has independently evolved
AFPs that display substantial differences in their
sequences, structures, and ice-binding sites (IBS) (6,
7). In other words, there is not one, but a diversity of
motifs that can confer AFPs with their ice-binding
abilities. What, then, are the characteristic features
of IBS, and how do they enable AFPs to bind ice?
Early suggestions focused on hydrophilic moieties,
which could hydrogen-bond with ice; the rationale
was that if the hydrophilic groups on the IBS were
spaced to match the lattice spacing of ice, the bind-
ing between the AFP and ice would be particularly
favorable (1, 8). In other words, a complementarity
in the spacing of hydrophilic groups on the AFP
surface and the crystal planes of ice was believed

to enable AFPs to bind, and integrate themselves
into, the ice lattice.

However, as numerous AFPs from diverse organ-
isms were discovered, and their crystal structures
solved, it became clear that hydrophobic residues
featured prominently on the IBS of AFPs (7). Even hy-
drophilic residues, such as threonine, had hydropho-
bic groups (−CH3), which protruded outward from the
IBS (Fig. 1B). The presence of hydrophobic groups on
the IBS suggested that although hydrogen bonding
and lattice matching could help AFPs bind ice, they
were not sufficient by themselves. It also raised the
question that, given that hydrophobes do not interact
favorably with water or with ice, what role might they
have in facilitating the binding of AFPs to ice? The
hydration waters of a small hydrophobic group are
constrained by having to hydrogen-bond around the
hydrophobe, which is unfavorable from an entropic
standpoint, so it was suggested that the release of
such constrained waters from hydrophobic groups
on the IBS might facilitate AFP–ice binding (1).

Thus, both hydrophilic and hydrophobic groups
seem to be important in conferring AFPs with an
affinity for ice. Mutagenesis studies have further shown
that a single mutation, whether it replaces a hydro-
phobic group with a hydrophilic one or vice versa,
can result in the loss of AFP function (4). Such seem-
ingly whimsical behavior has also been observed in
molecular-simulation studies, which probe the “ice
philicity” of extended surfaces by interrogating their
propensities for nucleating ice. In particular, Patey
and coworkers found that kaolinite, a clay mineral,
is only able to nucleate ice if it has the right amount
of flexibility (9); Molinero and coworkers observed
that the roughness of a graphitic surface influences
its ice philicity (10); and Michaelides and coworkers
showed that neither lattice matching (to ice) nor sur-
face hydrophilicity alone is sufficient to confer ice
philicity, but that appropriate values of both are
needed (11). Collectively, these studies suggest that
the affinity of an AFP surface for ice depends not only
on its ability to hydrogen-bond with ice but also on

aDepartment of Chemical & Biomolecular Engineering, University of Pennsylvania, Philadelphia, PA 19104
Author contributions: S.M.M. and A.J.P. wrote the paper.
The authors declare no conflict of interest.
Published under the PNAS license.
See companion article on page 8266.
1To whom correspondence should be addressed. Email: amish.patel@seas.upenn.edu.
Published online August 6, 2018.

8244–8246 | PNAS | August 14, 2018 | vol. 115 | no. 33 www.pnas.org/cgi/doi/10.1073/pnas.1810812115

C
O

M
M

E
N
T
A
R
Y

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 

http://crossmark.crossref.org/dialog/?doi=10.1073/pnas.1810812115&domain=pdf
http://www.pnas.org/site/aboutpnas/licenses.xhtml
mailto:amish.patel@seas.upenn.edu
www.pnas.org/cgi/doi/10.1073/pnas.1810812115


many other properties, such as its curvature, roughness, and
flexibility, in a nontrivial manner (9–11) (Fig. 1C).

Although no clear unifying features appear to enable AFPs to
preferentially bind ice, could the affinity of an AFP for ice be
reflected in certain properties of its hydration waters? Using mo-
lecular simulations of diverse AFPs, Sharp and coworkers showed
that the orientational distribution of water molecules in the vicinity
of IBS display certain key differences, which persist across differ-
ent classes of AFPs (12–14); Nutt and Smith reported a subtle
increase in tetrahedral order in the hydration shell of a hyperactive
insect AFP (15); and both molecular simulations and terahertz
spectroscopy measurements have reported slower relaxation dy-
namics of the AFP hydration waters (16–18). In each case, the
waters in the AFP hydration shell seem to, in one way or another,
bear some resemblance to those in ice. Do such seemingly ice-like
waters enable AFPs to bind to ice and, if so, how?

The crystal structure of a bacterial AFP, MpAFP, solved by
Davies and coworkers, provided an important clue by presenting,
for the first time, a clear molecular picture of the AFP–ice binding
interface (6, 19). Although AFP hydration waters can be resolved
in the X-ray structure, they are usually perturbed by protein–pro-
tein contacts in the crystal and tend not to be characteristic of
waters that mediate AFP–ice binding. However, MpAFP crystal-
lizes in a unique manner with its IBS well removed from any pro-
tein–protein contacts, so that the AFP–ice interface is clearly
visible in the crystal structure (6, 19). In what Davies and coworkers
called the anchored clathrate (AC) motif, the AFP uses both hy-
drophobic and hydrophilic groups on its IBS to bind ice (19). In the
AC motif, water molecules adopt a highly ordered structure, form-
ing a clathrate-like shell around the hydrophobic groups of the IBS
and participating in hydrogen bonds with its hydrophilic groups
(Fig. 1D). Based on the clear evidence of structured waters at the
AFP–ice interface, and the fact that the IBS hydration waters are
somewhat ice-like (in solution), it was suggested that a highly
structured layer of clathrate-like waters exists at the IBS of some
AFPs, which then enables the proteins to recognize and bind ice
(6, 19).

Molinero and coworkers explore this tantalizing suggestion in
a series of recent articles (5, 20, 21). In ref. 20, the authors studied
a number of hyperactive AFPs derived from insects and found that
AFP binding to ice can occur not only through the AC motif but
also through other similar motifs, wherein structured waters
bridge the AFP surface and ice. Given the diversity of IBS ob-
served across different AFPs, this result makes sense. Although
both experiments and simulations highlight the presence of struc-
tured waters at the AFP–ice interface, does this structuring

precede binding, and, if so, could the preordering of waters by
a surface be used as a signature of its affinity for ice? To answer
these questions, Hudait et al. (5) chose to study TmAFP, a hyper-
active insect AFP, which displays the best lattice matching with
hexagonal ice and may thus be the most likely to structure waters
into the AC motif in solution. Using molecular simulations with
three different water models, and multiple ways of characterizing
the structuring of IBS hydration waters, the authors conclusively
show that the IBS does not display ice-like or AC-like order in
solution; rather, the AC motif forms only after the AFP has moved
next to the ice surface and aligned itself in an orientation that is
optimal for binding to ice (5). These results suggest that a preor-
dering of their hydration waters is not needed for ice recognition
by AFPs. Support for these findings is also provided by NMR
studies, wherein a fast (subnanosecond) exchange of waters is
observed between the hydration shell of TmAFP and the bulk
(22); by contrast, AC waters ought to relax much more slowly, akin
to those in ice. More support comes from earlier work by Molinero
and coworkers, which showed that poly(vinyl alcohol), a flexible
polymer that does not preorder water, nevertheless binds ice (21).
As more AFPs belonging to different classes are studied, and as
even more AFPs are discovered, it will be interesting to see if an
AFP capable of preordering its hydration waters will be found.

Ice-nucleating proteins (INPs) represent another class of proteins
that are ice-philic; although they are structurally and chemically
similar to AFPs, INPs are much bigger in size and can even cluster,
thereby providing large surfaces capable of nucleating ice. Although
TmAFP does not structure its waters in solution, might the larger
INPs or their clusters be able to do so? To answer this question,
Hudait et al. (5) use model AFP-like surfaces spanning a range of
sizes; INP monomers are represented by rectangular surfaces with a
fixed nanoscopic width and increasing lengths, whereas INP clusters
are mimicked using square surfaces. The authors find that large
rectangular surfaces do not structure their hydration waters in solu-
tion, but that as the size of the square surfaces is increased, they
become more effective at ordering water. These findings suggest
that it is not easy for surfaces to structure their hydration waters into
ice-like or clathrate-like order; only large extended surfaces, like the
ones assembled by INP clusters, may be successful in doing so. To
facilitate experimental validation of their findings, the authors also
used their simulations to predict vibrational spectra of the OH
stretch of water; however, they find that neither infrared nor Raman
spectra can discriminate between AC waters at the AFP–ice interface
and the hydration waters of the AFP in solution (5).

Despite the remarkable progress made in understanding AFPs,
their most sought-after secrets remain elusive for now—both what

B D
Flexibility 

Hydrogen 
bonding 

Lattice 
matching 

C

Hydro-
phobicity

Affinity 
for Ice 

A

water 

ice 

Fig. 1. (A) AFPs have evolved to recognize and bind ice in a vast excess of water. (B) The IBS of typical AFPs display both hydrophilic (blue)
and hydrophobic (red) groups (7). (C) The affinity of a surface for ice depends sensitively on a number of surface properties, including its ability to
hydrogen-bond with ice, correspondence with the ice lattice, flexibility, and the presence of nonpolar groups (4, 9–11). Adapted from ref. 24. (D, Left)
Although TmAFP binds ice through the AC motif, (D, Right) it does not preorder water in solution. Adapted with permission from ref. 5.

Marks and Patel PNAS | August 14, 2018 | vol. 115 | no. 33 | 8245

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 



the molecular signatures of surface ice philicity are and how they
enable AFPs to bind ice continue to mystify. However, the work of
Hudait et al. (5) shows that the preordering of waters in solution,
whether in the AC motif or otherwise, is not necessary for AFPs to
bind ice. Importantly, the authors demonstrate how molecular sim-
ulations can build upon the wealth of knowledge provided by struc-
tural studies, and how they are beginning to make meaningful
contributions in furthering our understanding of AFPs. With such
synergistic advances, we are hopeful that it will not be long before
the secrets of AFPs are uncovered. A fundamental understanding of
ice philicity could also have far-reaching practical implications and
may lead to the discovery of synthetic molecules and materials that

are even more potent at binding ice than AFPs; such materials
would find use in diverse contexts, ranging from the preservation
of organs for transplant to increasing the freeze tolerance of crop
plants and the transportation of frozen foods (23).

Acknowledgments
A.J.P. thanks Kim Sharp for introducing him to AFPs and for numerous insightful
discussions. This work was supported by National Science Foundation Grants
DMR 1720530, CBET 1652646, and CHE 1665339, Charles E. Kaufman Founda-
tion Grant KA2015-79204, American Chemical Society Petroleum Research Fund
Grant 56192-DNI6, and Alfred P. Sloan Research Foundation Fellowship FG-2017-
9406 (all to A.J.P.) and by a Computational Science Graduate Fellowship from the
Department of Energy (to S.M.M.).

1 Jia Z, Davies PL (2002) Antifreeze proteins: An unusual receptor–ligand interaction. Trends Biochem Sci 27:101–106.
2 Kelland MA (2006) History of the development of low dosage hydrate inhibitors. Energy Fuels 20:825–847.
3 Knight C, DeVries A (2009) Ice growth in supercooled solutions of a biological “antifreeze”, afgp 1–5: An explanation in terms of adsorption rate for the
concentration dependence of the freezing point. Phys Chem Chem Phys 11:5749–5761.

4 Prabhu N, Sharp K (2006) Protein–solvent interactions. Chem Rev 106:1616–1623.
5 Hudait A, et al. (2018) Preordering of water is not needed for ice recognition by hyperactive antifreeze proteins. Proc Natl Acad Sci USA 115:8266–8271.
6 Sharp KA (2011) A peek at ice binding by antifreeze proteins. Proc Natl Acad Sci USA 108:7281–7282.
7 Davies PL (2014) Ice-binding proteins: A remarkable diversity of structures for stopping and starting ice growth. Trends Biochem Sci 39:548–555.
8 Devries AL, Lin Y (1977) Structure of a peptide antifreeze and mechanism of adsorption to ice. Biochim Biophys Acta 495:388–392.
9 Zielke SA, Bertram AK, Patey GN (2016) Simulations of ice nucleation by kaolinite (001) with rigid and flexible surfaces. J Phys Chem B 120:1726–1734.

10 Lupi L, Hudait A, Molinero V (2014) Heterogeneous nucleation of ice on carbon surfaces. J Am Chem Soc 136:3156–3164.
11 Fitzner M, Sosso GC, Cox SJ, Michaelides A (2015) The many faces of heterogeneous ice nucleation: Interplay between surface morphology and hydrophobicity.

J Am Chem Soc 137:13658–13669.
12 Gallagher KR, Sharp KA (2003) Analysis of thermal hysteresis protein hydration using the random network model. Biophys Chem 105:195–209.
13 Yang C, Sharp KA (2004) The mechanism of the type III antifreeze protein action: A computational study. Biophys Chem 109:137–148.
14 Yang C, Sharp KA (2005) Hydrophobic tendency of polar group hydration as a major force in type I antifreeze protein recognition. Proteins 59:266–274.
15 Nutt DR, Smith JC (2008) Dual function of the hydration layer around an antifreeze protein revealed by atomistic molecular dynamics simulations. J Am Chem Soc

130:13066–13073.
16 Xu Y, Gnanasekaran R, Leitner DM (2012) Analysis of water and hydrogen bond dynamics at the surface of an antifreeze protein. J At Mol Opt Phys 2012:1–6.
17 Meister K, et al. (2013) Long-range protein-water dynamics in hyperactive insect antifreeze proteins. Proc Natl Acad Sci USA 110:1617–1622.
18 Meister K, et al. (2014) Observation of ice-like water layers at an aqueous protein surface. Proc Natl Acad Sci USA 111:17732–17736.
19 Garnham CP, Campbell RL, Davies PL (2011) Anchored clathrate waters bind antifreeze proteins to ice. Proc Natl Acad Sci USA 108:7363–7367.
20 Hudait A, Odendahl N, Qiu Y, Paesani F, Molinero V (2018) Ice-nucleating and antifreeze proteins recognize ice through a diversity of anchored clathrate and ice-

like motifs. J Am Chem Soc 140:4905–4912.
21 Naullage PM, Lupi L, Molinero V (2017) Molecular recognition of ice by fully flexible molecules. J Phys Chem C 121:26949–26957.
22 Modig K, Qvist J, Marshall CB, Davies PL, Halle B (2010) High water mobility on the ice-binding surface of a hyperactive antifreeze protein. Phys Chem Chem Phys

12:10189–10197.
23 Zeng H, Wilson LD, Walker VK, Ripmeester JA (2006) Effect of antifreeze proteins on the nucleation, growth, and the memory effect during tetrahydrofuran

clathrate hydrate formation. J Am Chem Soc 128:2844–2850.
24 The Dill Research Group (2018) Energy landscapes. Available at dillgroup.org/#/landscapes. Accessed July 9, 2018.

8246 | www.pnas.org/cgi/doi/10.1073/pnas.1810812115 Marks and Patel

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 

http://dillgroup.org/#/landscapes
www.pnas.org/cgi/doi/10.1073/pnas.1810812115