Cataract-associated mutant E107A of human
γD-crystallin shows increased attraction to
α-crystallin and enhanced light scattering
Priya R. Banerjeea, Ajay Pandea, Julita Patroszb,1, George M. Thurstonc, and Jayanti Pandea,2
a
Department of Chemistry, and bDepartment of Biology, University at Albany, State University of New York, Albany, NY 12222; and cDepartment of
Physics, Rochester Institute of Technology, Rochester, NY 14623
Edited by George B. Benedek, Massachusetts Institute of Technology, Cambridge, MA, and approved November 2, 2010 (received for review September
30, 2010)
Several point mutations in human γD-crystallin (HGD) are now
known to be associated with cataract. So far, the in vitro studies
of individual mutants of HGD alone have been sufficient in providing plausible molecular mechanisms for the associated cataract in
vivo. Nearly all the mutant proteins in solution showed compromised solubility and enhanced light scattering due to altered
homologous γ–γ crystallin interactions. In sharp contrast, here we
present an intriguing case of a human nuclear cataract-associated
mutant of HGD—namely Glu107 to Ala (E107A), which is nearly
identical to the wild type in structure, stability, and solubility properties, with one exception: Its pI is higher by nearly one pH unit.
This increase dramatically alters its interaction with α-crystallin.
There is a striking difference in the liquid–liquid phase separation
behavior of E107A–α-crystallin mixtures compared to HGD–α-crystallin mixtures, and the light-scattering intensities are significantly
higher for the former. The data show that the two coexisting
phases in the E107A–α mixtures differ much more in protein density than those that occur in HGD–α mixtures, as the proportion of
α-crystallin approaches that in the lens nucleus. Thus in HGD–α mixtures, the demixing of phases occurs primarily by protein type
while in E107A–α mixtures it is increasingly governed by protein
density. Analysis of these results suggests that the cataract due
to the E107A mutation could result from the instability caused
by the altered attractive interactions between dissimilar proteins
—i.e., heterologous γ–α crystallin interactions—primarily due to
the change in surface electrostatic potential in the mutant protein.
T
he vertebrate lens contains three families of constituent proteins, the α-, β-, and γ-crystallins, closely packed such that the
total protein concentration in the central region of the lens, i.e.,
nucleus, exceeds 400 mg∕mL (1). Normally, the close protein
packing ensures that short-range order is maintained within the
fiber cell, such that fluctuations in the refractive index that cause
light scattering are minimized, and the lens is transparent (2, 3).
Disruption of short-range order associated with aging, altered
environment, or mutations in crystallins can cause increased light
scattering and cataract.
Cataract-associated mutations in the crystallin genes are
known to occur in all three crystallin families (1, 4) and show a
variety of phenotypes. Over the last decade, we have focused on
the point mutations occurring in human γD-crystallin (HGD), a
member of the γ-crystallin family expressed at high levels along
with γC- and γS-crystallins in the human lens (5). By comparing
the properties of several mutants of HGD in solution with those
of the wild type, we have identified specific changes in the homologous (i.e., γ–γ interactions) that led to the formation of distinct
condensed phases (6, 7) and accounted for the increased light
scattering due to these mutations (8–13). These studies revealed
a class of γ-crystallin mutations in which small changes to the
protein surface were sufficient to bring about severe solubility
changes, without unfolding the protein, and were in direct contrast to other mutations (i.e., other missense mutations, inser574–579 ∣ PNAS ∣ January 11, 2011 ∣ vol. 108 ∣ no. 2
tions, and splice mutations) that are accompanied by alterations
in protein fold and stability (1, 14, 15).
Here we report on yet another cataract-associated mutation in
HGD—the Glu107 to Ala (E107A) mutation—first reported in
2006 by Messina-Baas et al. (16) from genetic association studies.
As in the case of the other mutants we have studied, the E107A
mutant also shows minimal alterations in structure and stability.
However, unlike the mutants described above, we find that the
E107A mutation does not form a distinct condensed phase that
could result in a measurable lowering of solubility—instead, it is
soluble to protein concentrations upwards of 325 mg∕mL just
like HGD at pH 7. These data clearly suggested that despite the
loss of a negative charge following the replacement of Glu107
with Ala, the γ–γ (homologous) interactions among the mutant
protein molecules remain essentially unaltered relative to those
in HGD. As expected, there is a difference in the isoelectric point
(pI) between HGD and E107A with the pI of E107A being
approximately one pH unit higher (8.2 0.1) than that of HGD
(7.2 0.1). Because the net γ–γ interactions are essentially unaltered in E107A relative to HGD, the increase in pI raises the
possibility of altered electrostatic interactions between E107A
and other crystallins—for example, one with a lower pI. A likely
candidate is α-crystallin, because it is negatively charged at physiological pH (17), and it is already known that interactions
among α- and γ-crystallins are fine tuned for optimal packing
and transparency, and small changes in such interactions lead
to instability (18–22).
We therefore examined mixtures containing solutions of αcrystallin and either HGD or E107A, at compositions representing those in the young human lens and approaching those in the
lens nucleus (23). Here we show that all the mixtures containing
E107A and α-crystallin (E107A–α) consistently show higher lightscattering intensities relative to mixtures of HGD and α-crystallin
(HGD–α) of identical composition. Furthermore, as the weight
fraction of α-crystallin (X α ) in E107A–α mixtures is raised incrementally from 0.2 to 0.5 (a value comparable to that in the lens
nucleus) (24), the observed changes in the phase separation temperature (T ph ) and the tie lines of the mixtures are consistent with
an increase in the attraction between E107A and α-crystallin (22).
Preliminary accounts of parts of this work were presented at the Association for Research
in Vision and Ophthalmology annual meeting in Fort Lauderdale, FL, 2009, and the
Biophysical Society annual meeting in San Francisco, 2010.
Author contributions: A.P., G.M.T., and J. Pande designed research; P.R.B., A.P., and
J. Patrosz performed research; P.R.B., A.P., G.M.T., and J. Pande analyzed data; and
P.R.B., A.P., G.M.T., and J. Pande wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
See Commentary on page 437.
1
Present address: Regeneron Pharmaceuticals, 745 Old Sawmill River Road, Tarrytown,
NY 10591.
2
To whom correspondence should be addressed. E-mail: [email protected].
This article contains supporting information online at www.pnas.org/lookup/suppl/
doi:10.1073/pnas.1014653107/-/DCSupplemental.
www.pnas.org/cgi/doi/10.1073/pnas.1014653107
Results
Comparison of the Structures of HGD and E107A. Near- and far-UV
CD studies: In Fig. S1 we compare the CD spectra of HGD and
E107A in the near-UV (Fig. S1A) and far-UV (Fig. S1B) regions.
The near-UV CD data show that the tertiary structures of the
two proteins are virtually indistinguishable, indicating that the
substitution of Glu107 with Ala does not alter the protein fold.
This is also true for the secondary structure represented by the
far-UV spectra. Both proteins have far-UV CD profiles typical of
predominantly β-sheet-containing proteins, with a minor increase
in the negative ellipticity bands for E107A but well within the
differences observed in other highly homologous γ-crystallins
(26). Therefore, we conclude from these studies that the secondary and tertiary structures of HGD do not undergo any significant
change upon mutation.
Fluorescence Studies. Tryptophan fluorescence. The tryptophan
fluorescence emission spectra are shown in Fig. S2A. γ-crystallins
contain four buried tryptophan residues, two in each domain,
which are excellent reporters of the overall structural integrity
of the protein (27). Fig. S2A shows that the fluorescence emission
spectra for HGD and E107A are comparable, and the intensities
are almost equally quenched for both proteins. The λmax values
for HGD and E107A of 327 and 326 nm, respectively, are typical
of tryptophan residues buried within the hydrophobic core. The
two proteins also show nearly identical fluorescence spectra when
unfolded with guanidinium hydrochloride (Gdn.HCl), shown in
Fig. S2A, again attesting to their structural similarity.
Bis-ANS and Nile Red Fluorescence to determine surface hydrophobicity. The surface hydrophobicities of HGD and E107A were de-
Isoelectric Focusing. The first evidence for a clear distinction
between HGD and E107A came from the measurement of the pI.
Fig. 1 shows a representative isoelectric-focusing gel (IEF)
for HGD and E107A with pI values of 7.3 and 8.1, respectively.
Repeat runs yielded an average pI value of 7.2 0.1 for HGD
and 8.2 0.1 for E107A. The increase in pI of approximately
one pH unit confirms that there is an effective increase in overall
positive charge in E107A relative to HGD. Therefore, except
for the increase in pI, E107A appears similar in structure and
stability to HGD.
We also found that the mutant protein could be concentrated
to at least 325 mg∕mL without the appearance of any condensed
phase, just like HGD. Thus it became apparent that the homologous interactions between E107A molecules may not be the
basis of cataract, in contrast to many other γ-crystallin mutants
we reported earlier (8, 12, 13). However, the increase in pI of
the mutant pointed to the possibility of altered interaction with
α-crystallin, which is negatively charged at the physiological pH
(17). Thurston and colleagues (20–22, 29) have been investigating
the interactions between native α- and γ-crystallins and predicted
that the strength of the γ–α interactions in the lens is optimized
for transparency in a very narrow range, and deviations from this
range would lead to instability and cataract. Their work provided
the impetus for us to examine E107A–α interactions, and therefore we compared the thermodynamic properties of a series of
mixtures containing either HGD or E107A and α-crystallin in
termined using two well-tested fluorescent probes, the negatively
charged Bis-ANS and the uncharged Nile Red. Based on the
fluorescence changes (Fig. S2B), the Bis-ANS data show a small
increase in the net surface hydrophobicity for E107A compared
to HGD, while Nile Red does not. Because Bis-ANS is known to
be a more sensitive probe for surface hydrophobicity compared to
Nile Red (28), this difference is not surprising. The slight increase
in surface hydrophobicity is consistent with the substitution of
Glu by Ala in the mutant. We note however that because BisANS is negatively charged and Nile Red is neutral, the difference
in fluorescence could be due to the stronger electrostatic interactions between the negatively charged Bis-ANS and E107A,
which is more positively charged than HGD. Thus the mutant
may only have a marginally higher surface hydrophobicity relative
to the wild-type protein.
Thermal and Chemical Denaturation. Thermal stabilities of the two
proteins were measured by the loss of ellipticity at 290 nm in the
near-UV CD spectrum, in 0.1 M sodium phosphate buffer, pH 7
Banerjee et al.
SEE COMMENTARY
(Fig. S3). The midpoint of the transition corresponds to the melting temperature (T m ) for each protein. Fig. S3 shows that both
proteins unfold in an apparently single step without the formation
of a stable intermediate, and the respective T m values differ by
less than 2 ° (≅79.3 °C for HGD and ≅77.7 °C for E107A). The
apparent reduction in T m is not significant because small variations in T m (2–4 °C) are typical even among members of the
γ-crystallin family in their native state (12, 27). The inset of
Fig. S3 shows chemical unfolding data using Gdn.HCl as a denaturant and recorded as a change in the tryptophan fluorescence
spectra. Because all four tryptophan residues in HGD are significantly quenched, the unfolding process can be reliably monitored
as an increase in the tryptophan fluorescence intensity around
350 nm with a corresponding decrease around 320 nm. The data
in Fig. S3 show also that both wild-type and mutant unfold by an
apparently single-step transition, just as in the case of thermal
unfolding, with midpoint values of 3.54 M and 3.44 M Gdn.HCl
for HGD and E107A, respectively. These data clearly suggest that
the conformational stability of the mutant protein is comparable
to that of the wild type.
Fig. 1. Isoelectric-focusing gel for HGD and E107A. The Glu107 to Ala
mutation causes an increment of approximately 1.0 pH unit in the pI of
the protein.
PNAS ∣
January 11, 2011 ∣ vol. 108 ∣
no. 2 ∣
575
BIOPHYSICS AND
COMPUTATIONAL BIOLOGY
The E107A–α mixture with X α ¼ 0.5 undergoes density-type
phase separation, unlike all the HGD-α mixtures examined. Thus
these data provide direct experimental evidence for the prediction of Dorsaz et al. (22) that an increase in the attractive interactions between the α- and γ-crystallins will make mixtures of
these proteins unstable, tending toward density-driven phase separation, and can thereby contribute to cataract formation. This
work also constitutes a prime example of a cataract-associated
mutation that highlights the importance of heterologous protein–
protein interactions optimized to maintain transparency and how
changes in these interactions can lead to instability and opacity of
the human lens. We note here that the role of heterologous protein–protein interactions between HGD and βB1 crystallins in affecting thermodynamic stability has also been studied recently
(25). In analogy to our present work, one might expect cataractogenic mutations in the future that impact stability primarily
through their effects on, for example, the γ–β interactions.
lens nucleus (30). Thus, the light-scattering efficiencies for the
E107A–α mixtures relative to the HGD–α mixtures at all the
temperatures studied show inherently higher light-scattering profiles. Notably, as we approach the critical temperature for phase
separation in each mixture (<10 °C), the difference in light-scattering efficiencies seems to be dominated by the instability due to
the phase transition, which is clearly observed in Fig. 2B where the
proportion of γ-crystallin is at a maximum in the mixture.
Fig. 2. Light-scattering efficiencies of HGD-α-crystallin (▴) and E107A–αcrystallin mixtures (□) at (A) weight fraction of α-crystallin, X α ¼ 0.0 (pure
proteins), (B) X α ¼ 0.2, (C) X α ¼ 0.35, and (D) X α ¼ 0.5 as a function of
temperature, at a total protein concentration of 265 mg∕mL for pure
proteins and 275 mg∕mL for mixtures.
solution. The data, presented in this report, constitute remarkable developments in the molecular basis for lens opacity due
to a human cataract-associated mutation.
Light-Scattering Efficiencies of HGD–α and E107A–α Mixtures. Light
scattering is intimately associated with cataract disease, and all
cataract-associated mutants of HGD show increased light scattering relative to the normal protein at appropriate concentrations
(8, 12, 13). In Fig. 2 we compare the light-scattering efficiencies,
expressed as excess Rayleigh ratios, of the pure proteins (Fig. 2A)
as well as solutions containing mixtures (either HGD–α or
E107A–α) at three different compositions (X α ¼ 0.2, 0.35, and
0.5) at a total protein concentration of 275 mg∕mL (Fig. 2 B–D).
The light-scattering profiles of the two proteins, HGD and
E107A alone, are nearly identical (Fig. 2A), further attesting to
the similarity between the two proteins. However, in contrast,
the data for the E107A–α mixtures show a nearly 25% increase
in light scattering relative to the HGD–α mixtures at all three compositions (X α ¼ 0.20, 0.35, and 0.50) and temperatures examined
(10–35 °C). The difference in scattering efficiency between the
two remains virtually independent of temperature and composition at X α ¼ 0.5. As stated earlier, in terms of the composition,
the mixtures at X α ¼ 0.5 are comparable to that found in the
Coexistence Curves of Pure HGD and E107A and of Mixtures. Because
coexistence curves provide a measure of the strength of protein–
protein interactions (31), we measured liquid–liquid phase separation in solutions of pure γ-crystallins as well as in E107A–α and
HGD–α mixtures. In Fig. 3 we compare the data for pure HGD
and E107A (Fig. 3A) and several mixtures as a function of the
weight fraction of α-crystallin (Fig. 3B) and total protein concentration (Fig. 3C). Again, as expected, the T ph values for the pure
proteins are nearly identical and fall on the coexistence curve for
HGD published earlier (8). Thus, the data in Fig. 3A confirm that
the magnitude of the like–like (i.e., γ–γ) homologous interactions
in E107A are nearly identical to those in HGD, despite the loss of
a negative charge in E107A.
This behavior changes dramatically when we compare the liquid–liquid phase separation profiles of HGD–α and E107A–α
mixtures. The coexistence curve (Fig. 3B) and the T ph profile
(Fig. 3C) for E107A–α were suppressed relative to those for
HGD–α. In Discussion we show how such a suppression of T ph
is in fact expected to accompany an initial increase in γ–α attraction, based on previous work (22), and how such a change, together
with the increase in light scattering well above T ph and the rotation
of the tie lines toward density–density phase separation (see below) can all be consistent with an increase in E107A–α attraction
over that of HGD–α. In Appendix S1 we construct a mean-field
model to help analyze the qualitative response of each measured
quantity to changes in γ–α attraction strength. The model is also
consistent with the inference of increased γ–α attraction in going
from the HGD–α to the E107A–α mixtures.
Tie Lines and Concentrations and Compositions of the Coexisting
Phases in HGD–α and E107A–α Mixtures. Next we examined the tie
lines that join equilibrium protein compositions of the separated
phases in HGD–α and E107A–α mixtures (Fig. 4). At 20% α-crystallin (X α ¼ 0.2, Fig. 4A), the dense phase contains predominantly
γ-crystallin and the dilute phase contains mostly α-crystallin, and
the tie lines are essentially the same for both mixtures. These data
are consistent with the measured tie lines of mixtures of bovine
α- and γB-crystallins at low X α values (21). As X α increases to
0.35 (Fig. 4B), there is a noticeable counterclockwise rotation
of the tie line for E107A–α, relative to that for HGD–α. The coun-
Fig. 3. (A) Liquid–liquid coexistence curves of pure HGD and E107A in 0.1 M phosphate buffer, pH 7.0, containing 20 mM DTT. The data obtained in the present
study [HGD, (▴), E107A, (□)], are compared with the published HGD data (8) (filled gray circles). (B) Liquid–liquid coexistence curves of HGD–α and E107A–α
mixtures as a function of X α at a constant total protein concentration of approximately 350 mg∕mL, 0.1 M phosphate buffer, pH 7.0, containing 20 mM DTT. The
lines are shown as guides to the eye. (C) Change in the (T ph ) for HGD–α and E107A–α mixtures as a function of total protein concentration at X α ¼ 0.50 in 0.1 M
phosphate buffer, pH 7.0, containing 20 mM DTT. Lines drawn through the data in all the figures are merely guides to the eye.
576 ∣
www.pnas.org/cgi/doi/10.1073/pnas.1014653107
Banerjee et al.
SEE COMMENTARY
terclockwise rotation is much more pronounced when X α is further
increased to 0.5 (Fig. 4C). Here, the tie line for E107A–α is nearly
perpendicular to that of HGD–α, and the dense protein phase is
enriched in both α- and γ-crystallins while the dilute phase is poor
in both α- and γ-crystallins, while HGD–α at X α ¼ 0.5 still shows
more compositional phase separation. Thus as protein composition and concentration approach values comparable to those in
the lens nucleus (X α ¼ 0.5), demixing changes from being predominantly by “protein type” (HGD–α), toward “density-type”
(E107A–α), a phenomenon analyzed by Dorsaz et al. (22). Fig. 4D
shows the assembled tie line results with a backdrop of estimated
refractive index contours. In Fig. S4I we present a three-dimensional view of the phase diagrams of the HGD–α and E107A–α
ternary mixtures. In Fig. S4II we show the orientations of the
tie lines at X α ¼ 0.5 with respect to the refractive index gradient
ð∇nÞ. The E107A–α tie line is approximately 30° from ∇n, while
the HGD–α tie line is 70° away (Fig. 4D and Fig. S4II). The relationship between the rotation of the tie lines, increased attractive
E107A–α interactions, and the possible consequences for increased light scattering from E107A–α mixtures are discussed
below and modeled mathematically in Appendix S1.
Modeling Attractive Interactions Between HGD/E107A and α-crystallin. We have computationally modeled the putative attractive
interactions between E107A and α-crystallin, based on the highresolution crystal structures of HGD (32) and truncated human
αB-crystallin (HAB), respectively (33). These studies point out
that the mutation in E107A could result in a distinct positivepotential surface patch that is likely to show attractive interactions with a large negative-potential patch present on the surface
of HAB. We obtained an electrostatic interaction energy of
−2.04 kcal∕mol for the HGD–HAB interaction and −2.43 kcal∕
mol for the E107A–HAB interaction, representing a change in
interaction energy that is consistent with stronger attractive interactions of E107A with α-crystallin. The details of the methods
used, justification for the use of HAB to model bovine α-crystallin, results of the modeling, and corresponding discussion are
presented in Appendix S2.
Discussion
Our data show that the cataract-associated E107A mutant of
HGD is a novel case where neither structure and stability, nor
the stability of its aqueous solutions with respect to precipitation,
Banerjee et al.
is significantly compromised as a result of the mutation. The
strong similarity in the coexistence curves of the two proteins
indicates that the strength of the γ–γ interaction in the mutant
is nearly identical to that in the wild-type protein. Therefore,
we conclude that the molecular basis for lens opacity in this case
cannot be attributed directly to the altered, homologous γ–γ
interactions in the mutant.
The isoelectric-focusing data confirm an expected primary
difference between the mutant and wild-type proteins by showing
a pI increase of about one pH unit for E107A. Consistent with
this increase, E107A will be more positively charged than HGD
at physiological pH. Mapping the electrostatic potential on the
protein surfaces revealed a positive-potential patch that includes
residues Arg89, Arg115, and Arg169 in E107A, and that is clearly
different from that in the wild type (Fig. S5 A and B).
It is known that α-crystallins are negatively charged at physiological pH (17), and it is only recently that Thurston and colleagues (20–22) have shown comprehensively that heterologous
interactions between the α- and γ-crystallins are so delicately
balanced that even slight (∼0.5 kB T) strengthening or weakening
of such interactions could lead to instability in mixtures containing
these proteins (22). Our comparison of the thermodynamic properties of mixtures of either HGD or E107A with α-crystallin at
physiological pH, and at high concentrations comparable to those
in the eye lens, provides significant experimental evidence to
support their prediction for a cataract-associated mutant of HGD.
We now discuss how the (a) lowered T ph values, (b) enhanced
protein density-type phase separation resulting in rotated tie
lines, and (c) enhanced light scattering in the single-phase region
can be a consequence of E107A–α interactions that are more
attractive than HGD–α interactions. A full quantitative analysis
of the joint light scattering, tie line, and phase boundary data will
require further experimental and theoretical work.
Fig. 3 shows that the liquid–liquid phase boundaries of
E107A–α mixtures were lower than those of HGD–α mixtures
by about 5 °C. Such a reduction in T ph is exactly what would
be expected as the initial response of this system to increased
γ–α attraction, based on the work of Stradner, Dorsaz, and coworkers (20, 22), as we now explain. They performed a molecular
dynamics (MD) simulation to model their small-angle neutron
scattering (SANS) data (20) and applied thermodynamic perturbation theory to study the phase boundaries (21, 22). The γB–α
attraction, for which MD simulations were consistent with SANS,
PNAS ∣
January 11, 2011 ∣ vol. 108 ∣
no. 2 ∣
577
BIOPHYSICS AND
COMPUTATIONAL BIOLOGY
Fig. 4. Tie lines for the coexistence curves showing
the compositions of the two phases of mixtures of
HGD–α or E107A–α: (A) at X α ¼ 0.20, (B) X α ¼ 0.35,
and (C) X α ¼ 0.50, in 0.1 M phosphate buffer, pH 7.0,
containing 20 mM DTT. In D all the tie lines are
shown, along with the rotation of the tie lines for
the E107A–α mixtures (curved arrow). The parallel
dotted lines (in gray) are contours of constant refractive index, and the gradient of refractive index ð∇nÞ
is indicated by a small black arrow. Starting protein
concentrations for each mixture is shown as a solid
circle for HGD–α and a solid cube for E107A–α
mixtures, respectively. The quench temperature for
X α ¼ 0.2 and 0.5 mixtures were 0.5 °C while that
for X α ¼ 0.35 was 0.8 °C.
put these mixtures just beyond the lower edge of a stability zone
that extends from about 0.5 kB T to 1.0 kB T in the γB–α attraction
well-depth (20, 22). For those mixtures, initial increases in the
γB–α attraction are expected to decrease T ph , while rotating tie
lines counterclockwise toward density–density-type liquid–liquid
phase separation (see figures 6, 10, 11, and 14 in ref. 22 and
corresponding discussion). This behavior is consistent with the
present data. Further increases in well-depth, on the order of
0.5 kB T in magnitude, are expected to continue to rotate tie lines
and eventually to lead to increased T ph in a composition-dependent manner (22), as we also show with use of a generalized van
der Waals free energy model in Appendix S1. Because the γB–α
mixture phase diagram (21), is quantitatively quite similar to the
HGD–α mixture phase diagram found here, the starting values
of the interaction parameters for the two systems are likely to
be quite similar. Accordingly, we hypothesize that the contrast
between the HGD–α and the E107A–α phase diagrams results
from the predicted, initially decreased T ph and concomitant
enhancement of density-driven phase separation in response to
increased γ–α attraction.
Consistent with this hypothesis, we note that in the electrostatic modeling work we found a difference in interaction
strengths of approximately 0.4 kcal∕mole or 0.7 kB T that was
more attractive for E107A–α than for HGD–α (Appendix S2). In
view of the fact that angular averaging of patchy interactions as
well as nonelectrostatic interactions will also be of relevance, the
magnitude of the modeled increase in strength of attractive electrostatic interactions between the HGD–α and the E107A–α
cases is quite compatible with the size of approximately 0.5 kB T
predicted by Dorsaz et al. (22) to lead to enhancement of densitytype fluctuations.
The decreased T ph in response to increased γ-α attraction in
ternary mixtures contrasts with binary protein-water mixtures, in
which increased attraction raises T ph . The mean-field model in
Appendix S1 shows how such a decrease can occur in ternary mixtures. Thus, in analogy to the fact that increased T ph signals
increased homologous attraction in binary aqueous γ-crystallin
solutions, the decreased T ph in ternary γ–α mixtures suggests
increased γ–α attraction. The measured tie lines (Fig. 4) showed
that for HGD–α mixtures, the segregation of the two phases is by
protein type and remains so even as the total protein concentration or weight fraction of α-crystallin is increased, consistent with
what is found for bovine γB–α mixtures (21).
In contrast, the distinct compositions of the coexisting phases,
and the dramatic rotation of the slope of the tie lines for the
E107A–α mixture at X α ¼ 0.35 and 0.50 relative to the HGD–α
mixtures, strongly suggest a density-type phase separation as a
result of the increase in γ–α attraction due to the E107A mutation. Density-type phase separation in response to increased γ–α
attraction was predicted in Dorsaz et al. (see especially figures. 10
and 11 and the accompanying text) (22). This is of considerable
significance because the protein composition at the central core
of the eye lens is such that the amounts of the α- and γ-crystallins
are comparable to X α ¼ 0.50. While at X α ¼ 0.2, the paired compositions and the direction of the tie line connecting the coexisting phases at equilibrium are very similar for both HGD–α and
E107A–α, the direction of the tie line is already found to be tilted
toward the density-type phase separation for the E107A–α mixture at X α ¼ 0.35. This is remarkable because it documents a
clear shift in the magnitude of the heterologous protein–protein
interactions between α- and γ-crystallins at the molecular level.
We hypothesize that the susceptibility of the E107A–α mixture
to density-type phase separation can facilitate the formation of
density inhomogeneities in the solution well above the phase
boundary and also contribute to enhanced light scattering that
we have observed, as we explain below.
One of the key reasons that increased E107A–α attraction can
contribute to increased light scattering is that enhanced density
578 ∣
www.pnas.org/cgi/doi/10.1073/pnas.1014653107
fluctuations also enhance refractive index fluctuations and thereby scatter more light than other composition fluctuations. A
mean-field model that includes a detailed rationale for the linkage between E107A–α attraction and enhanced light scattering is
presented in Appendix S1. Although other factors detailed there
are also of importance, one factor is that increased alignment
between the direction of least Gibbs free energy second derivative
and that of the index of refraction gradient ð∇nÞ, leads to increased light scattering (29). Therefore in Fig. 4D, using published refractive index increments for α-crystallin (34) and
γB-crystallin (35), we compare ∇n with the tie lines, which serve
to estimate directions of minimal Gibbs free energy second
derivative nearby in the single-phase region, as explained in
Appendix S1. As shown in Fig. 4 and Fig. S4II, the E107A–α
tie line directions are approximately 31° from ∇n, while the
HGD–α tie lines are 70° from ∇n, qualitatively consistent with
the lower intensity observed in HGD–α mixtures, even though
they are closer to the phase separation (see Appendix S1).
In conclusion, we have presented an intriguing case of a cataract-associated mutant, E107A of human γD-crystallin, which
has comparable γ–γ interactions to those in HGD and maintains
its structure and stability, similar to the other mutants we reported earlier (8, 12, 13). However, in striking contrast to the
other mutants, aqueous solutions of the pure E107A mutant also
do not show either a lowered solubility or any indication of the
formation of a condensed phase. In fact, the data show a significant alteration in the heterologous interaction of the mutant
protein with α-crystallin, relative to the wild type. More precisely,
the dramatic rotation in the tie lines connecting coexisting phases
in the E107A–α-crystallin mixtures, the altered phase boundary
and enhanced light scattering, compared to the HGD–α-crystallin
mixtures, are all consistent with increased attractive E107A–
α-crystallin interactions. This work thus constitutes an example
of a cataract-associated mutant of HGD that shows altered interaction with α-crystallin thereby promoting the thermodynamic
instability and opacity observed in cataract formation.
Materials and Methods
Cloning, Expression and Purification of Recombinant Proteins. Overexpression
and purification of recombinant wild-type HGD has been reported earlier (8).
For the mutant protein, primers were designed and synthesized (MSG
Operon) to substitute Glu107 with Ala. Nucleotide sequence in the resulting
plasmid was confirmed in-house at the Biomolecular Core facility. Expression
and purification of the mutant protein was carried out as for HGD (8). Five
independent preparations of HGD and E107A gave an average mass of
20;608 2 and 20;548 1 Da, respectively, using electrospray ionization
mass spectrometry (ESIMS), performed at the Center for Functional Genomics
at the University at Albany. The results agree with the published mass for
HGD (8) and are consistent with the Glu to Ala mutation in HGD.
Purification of α-crystallin from Young Bovine (Calf) Lenses. Bovine α-crystallin
was extracted from 7 to 10 calf lenses using size-exclusion chromatography
(SEC) in a Sepharose CL-6B column (GE Healthcare), and the proteins were
eluted isocratically at a flow rate of 2 mL∕ min using 0.1 M sodium phosphate
buffer pH 7, containing 0.02% sodium azide as in (21). The chromatography
was repeated two or three times to obtain a single protein peak, which was
then characterized by typical spectroscopic criteria.
Spectroscopic and Static Light-Scattering Studies. Circular Dichroism (CD)
spectra were recorded on a JASCO J-815 spectropolarimeter. Protein concentrations of 0.5 mg∕mL in 0.1 M sodium phosphate buffer (pH 7) were used to
obtain the near-UV CD spectra in a 10 mm path length cuvette and normalized with respect to protein concentration. Far-UV CD spectra were measured
using protein concentrations of 0.1 mg∕mL in a 1.0 mm path length cuvette
and normalized with respect to the concentration of the backbone peptide
bonds.
Tryptophan, Bis-ANS, and Nile Red fluorescence spectra were obtained
using a Horiba Jobin Yvon Fluorolog-3 spectrometer. Excitation wavelengths
of 290 nm, 390 nm and 540 nm were used to measure tryptophan fluorescence emission, and the fluorescence emission following Bis-ANS and Nile
Red binding to the proteins respectively. Excitation and emission slits were
Banerjee et al.
Thermal and Chemical Denaturation of HGD and E107A. Thermal stability of
HGD and E107A was determined using near-UV CD and monitoring the
change in ellipticity at 290 nm as function of temperature. CD spectra were
collected at 5 °C intervals from 50 °C to 65 °C and at every 2 °C from 67 °C to
88 °C. Protein samples were equilibrated at each temperature, and each spectrum was an average of four different scans. The midpoint of the thermal
denaturation profile gave the melting points (T m ) of proteins.
Chemical denaturation due to guanidinium hydrochloride (Gdn.HCl) was
monitored using tryptophan fluorescence emission spectra. The concentration of Gdn.HCl was determined using a refractometer (model r 2 mini, Reich1. Hejtmancik JF (2008) Congenital cataracts and their molecular genetics. Semin Cell Dev
Biol 19:134–149.
2. Benedek GB (1971) Theory of transparency of the eye. Appl Opt 10:459–473.
3. Delaye M, Tardieu A (1983) Short-range order of crystallin proteins accounts for eye
lens transparency. Nature 302:415–417.
4. Andley UP (2007) Crystallins in the eye: Function and pathology. Prog Retin Eye Res
26:78–98.
5. Aarts HJ, Jacobs EH, van Willigen G, Lubsen NH, Schoenmakers JG (1989) Different
evolution rates within the lens-specific beta-crystallin gene family. J Mol Evol
28:313–321.
6. Benedek GB (1997) Cataract as a protein condensation disease: The Proctor lecture.
Invest Ophth Vis Sci 38:1911–1921.
7. Gunton JD, Shiryayev A, Pagan DL (2007) Protein Condensation: Kinetic Pathways to
Crystallization and Disease (Cambridge Univ Press, Cambridge, UK; New York) p 364.
8. Pande A, et al. (2000) Molecular basis of a progressive juvenile-onset hereditary
cataract. Proc Natl Acad Sci USA 97:1993–1998.
9. Stephan DA, et al. (1999) Progressive juvenile-onset punctate cataracts caused by
mutation of the gammaD-crystallin gene. Proc Natl Acad Sci USA 96:1008–1012.
10. Pande A, Gillot D, Pande J (2009) The cataract-associated R14C mutant of human
gamma D-crystallin shows a variety of intermolecular disulfide cross-links: A Raman
spectroscopic study. Biochemistry 48:4937–4945.
11. Kmoch S, et al. (2000) Link between a novel human gammaD-crystallin allele and a
unique cataract phenotype explained by protein crystallography. Hum Mol Genet
9:1779–1786.
12. Pande A, et al. (2001) Crystal cataracts: human genetic cataract caused by protein
crystallization. Proc Natl Acad Sci USA 98:6116–6120.
13. Pande A, et al. (2005) Decrease in protein solubility and cataract formation caused by
the Pro23 to Thr mutation in human gamma D-crystallin. Biochemistry 44:2491–2500.
14. Fu L, Liang JJ (2002) Conformational change and destabilization of cataract gammaCcrystallin T5P mutant. FEBS Lett 513:213–216.
15. Pigaga V, Quinlan RA (2006) Lenticular chaperones suppress the aggregation of
the cataract-causing mutant T5P gamma C-crystallin. Exp Cell Res 312:51–62.
16. Messina-Baas OM, Gonzalez-Huerta LM, Cuevas-Covarrubias SA (2006) Two affected
siblings with nuclear cataract associated with a novel missense mutation in the CRYGD
gene. Mol Vis 12:995–1000.
17. Bera S, Ghosh SK (1998) Interaction of H(+)-ions with alpha-crystallin: Solvent
accessibility of ionizable side chains and surface charge. Biophys Chem 70:147–160.
18. Bettelheim FA, Chen A (1998) Thermodynamic stability of bovine alpha-crystallin in its
interactions with other bovine crystallins. Int J Biol Macromol 22:247–252.
19. Takemoto L, Sorensen CM (2008) Protein-protein interactions and lens transparency.
Exp Eye Res 87:496–501.
20. Stradner A, Foffi G, Dorsaz N, Thurston G, Schurtenberger P (2007) New insight into
cataract formation: Enhanced stability through mutual attraction. Phys Rev Lett
99:198103.
21. Thurston GM (2006) Liquid-liquid phase separation and static light scattering of
concentrated ternary mixtures of bovine alpha and gammaB crystallins. J Chem Phys
124:134909.
22. Dorsaz N, Thurston GM, Stradner A, Schurtenberger P, Foffi G (2009) Colloidal
characterization and thermodynamic stability of binary eye lens protein mixtures.
J Phys Chem B 113:1693–1709.
Banerjee et al.
SEE COMMENTARY
ert). Samples containing increasing concentrations of Gdn.HCl ranging from
0 M to 5.5 M and a final protein concentration of 0.02 mg∕mL were used
for the fluorescence measurements. Each sample was equilibrated for at least
7 h at 25 °C prior to measurement, and the spectra were corrected for the
contribution of the appropriate concentration of Gdn.HCl. The ratio of the
fluorescence intensities at 360 nm and 320 nm were used for data analysis.
Liquid–Liquid Phase Separation-Coexistence Curves and Tie Lines. T ph measurements were made using the cloud-point method (39), and the coexistence
curves for pure HGD and E107A were mapped as described earlier (8). For
tie line measurements in solutions of ternary mixtures containing HGD–α
and E107A–α, the coexisting liquid phases were formed by quenching the
clear mixtures to a temperature below the coexistence curve (21), followed
by gentle centrifugation at 1;000 × g for 50–70 h in a rotor preequilibrated
at the quench temperature. The samples were considered to be at equilibrium when the clouding temperature (T cloud values) of the separated liquid
phases were within 1 °C of the centrifugation temperature and also typically
within 1 °C of each other. Compositions of the separated liquid phases were
determined using SEC.
Note See also ref. 40, which further supports our analysis..
ACKNOWLEDGMENTS. This work was supported by National Institutes of
Health grants EY018249 (G.M.T) and EY010535 (J. Pande). J. Patrosz, a
student of the Honors College at University at Albany was supported in
part by the University at Albany Summer Research Program and received
a Goldwater Scholarship while doing research in the Pande lab.
23. Horwitz J (1993) Proctor Lecture. The function of alpha-crystallin. Invest Ophth Vis Sci
34:10–22.
24. Veretout F, Tardieu A (1989) The protein concentration gradient within eye lens might
originate from constant osmotic pressure coupled to differential interactive properties
of crystallins. Eur Biophys J 17:61–68.
25. Wang Y, Lomakin A, McManus JJ, Ogun O, Benedek GB (2010) Phase behavior of
mixtures of human lens proteins Gamma D and Beta B1. Proc Natl Acad Sci USA
107:13282–13287.
26. Evans P, Bateman OA, Slingsby C, Wallace BA (2007) A reference dataset for circular
dichroism spectroscopy tailored for the betagamma-crystallin lens proteins. Exp Eye
Res 84:1001–1008.
27. Mandal K, Bose SK, Chakrabarti B, Siezen RJ (1985) Structure and stability of gammacrystallins. I. Spectroscopic evaluation of secondary and tertiary structure in solution.
Biochim Biophys Acta 832:156–164.
28. Pande A, Ghosh KS, Banerjee PR, Pande J (2010) Increase in surface hydrophobicity of
the cataract-associated P23T mutant of human gammaD-crystallin is responsible for its
dramatically lower, retrograde solubility. Biochemistry 49:6122–6129.
29. Ross DS, Thurston GM, Lutzer CV (2008) On a partial differential equation method for
determining the free energies and coexisting phase compositions of ternary mixtures
from light scattering data. J Chem Phys 129:064106.
30. Siezen RJ, Fisch MR, Slingsby C, Benedek GB (1985) Opacification of gamma-crystallin
solutions from calf lens in relation to cold cataract formation. Proc Natl Acad Sci USA
82:1701–1705.
31. Liu C, et al. (1996) Phase separation in aqueous solutions of lens gamma-crystallins:
Special role of gamma s. Proc Natl Acad Sci USA 93:377–382.
32. Basak A, et al. (2003) High-resolution X-ray crystal structures of human gammaD
crystallin (1.25 A) and the R58H mutant (1.15 A) associated with aculeiform cataract.
J Mol Biol 328:1137–1147.
33. Laganowsky A, et al. (2010) Crystal structures of truncated alphaA and alphaB crystallins reveal structural mechanisms of polydispersity important for eye lens function.
Protein Sci 19:1031–1043.
34. Pierscionek B, Smith G, Augusteyn RC (1987) The refractive increments of bovine
alpha-, beta-, and gamma-crystallins. Vision Res 27:1539–1541.
35. Fine BM (1994) Light scattering by aqueous protein solutions that exhibit liquid-liquid
phase separation. PhD thesis (Massachusetts Inst of Technology).
36. Yu XC, Margolin W (1998) Inhibition of assembly of bacterial cell division protein FtsZ
by the hydrophobic dye 5,5′-bis-(8-anilino-1-naphthalenesulfonate). J Biol Chem
273:10216–10222.
37. Ruvinov SB, et al. (1995) Ligand-mediated changes in the tryptophan synthase
indole tunnel probed by nile red fluorescence with wild type, mutant, and chemically
modified enzymes. J Biol Chem 270:6357–6369.
38. Schurtenberger P, Augusteyn RC (1991) Structural properties of polydisperse biopolymer solutions: A light scattering study of bovine alpha-crystallin. Biopolymers
31:1229–1240.
39. Broide ML, Berland CR, Pande J, Ogun OO, Benedek GB (1991) Binary-liquid phase
separation of lens protein solutions. Proc Natl Acad Sci USA 88:5660–5664.
40. Dorsaz N, Thurston GM, Stradner A, Schurtenberger P, Foffi G (2011) Phase separation
in binary eye lens protein mixtures. Soft Matter doi:10.1039/C0SM00156B.
PNAS ∣
January 11, 2011 ∣ vol. 108 ∣
no. 2 ∣
579
BIOPHYSICS AND
COMPUTATIONAL BIOLOGY
set to 5 nm for all the measurements. Tryptophan fluorescence spectra were
measured at a protein concentration of 0.02 mg∕mL in 0.1 M phosphate
buffer (pH 7) and corrected by subtracting the contribution of buffer. Fluorescence emission following Bis-ANS and Nile Red binding was measured
using protein concentrations of 0.1 mg∕mL (∼5 μM) and a Bis-ANS or Nile
Red concentration of 100 μM. Stock solutions of Bis-ANS and Nile Red were
prepared in methanol and the final alcohol concentration was maintained
below 7% v∕v when the reagents were mixed with the proteins. Concentrations of Bis-ANS and Nile Red were measured using extinction coefficients of
16.8 mM−1 cm−1 at 385 nm (36) and 45 mM−1 cm−1 at 552 nm (37), respectively. Spectra were corrected by subtracting the contribution of the buffer
containing the appropriate amount of either reagent.
Static light-scattering experiments were performed using a Brookhaven
light-scattering model 200SM goniometer and a temperature-controlled
sample stage, equipped with a 633 nm laser. Data were collected at a scattering angle of 90°. Protein samples were in the concentrations range of
260–270 mg∕mL and Rayleigh ratios were determined as described in ref. 38
using toluene as the standard.
Supporting Information
Banerjee et al. 10.1073/pnas.1014653107
SI Text
Appendix S1. A Mean-Field Model for Protein Mixtures. We use a
mean-field thermodynamic analysis to further examine the connections between increased γ–α attraction, phase separation temperature, tie line direction, and light-scattering intensity. To do so
we construct a mean-field model of the generalized van der Waals
type (1), for the free energy of aqueous mixtures that are concentrated in both γ- and α-crystallin, and apply the mathematical
conditions for phase separation and light scattering to consider
the observed contrasts between E107A–α and HGD–α mixtures.
The model is intended to illustrate the plausibility of the considerations above in a more mathematically specific manner, not to
provide a quantitative model for the data.
The model is the sum of a part kB Tf ðρ1 ;ρ2 Þ that represents the
mixing entropy of hard particles of the same size and shape as
γ- and α-crystallins, and a part uðρ1 ;ρ2 Þ that represents non hardcore interactions. f ðρ1 ;ρ2 Þ could be from an equation of state for
mixtures of hard spheres of different sizes (2, 3), as in the unperturbed part of the first-order perturbation free energy used in (4),
or by forms that also account for molecular shape (5). Here, we
leave f unspecified in order to focus on the role of γ–α attractions.
In line with the regular solution approach to modeling liquid–
liquid phase separation (6, 7), we take uðρ1 ;ρ2 Þ ¼ wγ-γ ϕ1
ð1 − ϕ1 − ϕ2 Þ þ wα-α ϕ2 ð1 − ϕ1 − ϕ2 Þ þ wγ -α ϕ1 ϕ2 , in which ϕ1 ¼
ρ1 ν̄1 and ϕ2 ¼ ρ2 ν̄2 are the volume fractions of γ and α, respectively, and ν̄1 and ν̄2 denote their corresponding partial molecular
volumes, which we take to be constant, together with that of the
solvent. Note that 1 − ϕ1 − ϕ2 is the volume fraction of solvent.
The quantities wγ-γ , wα-α and wγ-α represent average, effective
attractions or repulsions between species and are proportional
to the effective critical temperatures for liquid–liquid phase
separation on the binary solvent-γ and solvent-α axes, and on
a hypothetical γ–α axis without water. Further expressions that
could be used to relate the wi−j to the intermolecular potentials
and the partial pair correlation functions are given in a perturbation theory model for γ–α mixtures (4), but there the composition
dependences of the pair correlation functions are considered, and
yield a model for uðρ1 ;ρ2 Þ that better reflects the short-range of
the interprotein attractions than the simpler form used here.
We take δwγ-α ¼ wγ-α − ðwγ -γ þ wα-α Þ to represent effective
solvent-mediated γ–α interactions; note that increased γ-α attraction decreases wγ-α and hence δwγ -α . We take wα-α ¼ 0 because
α-crystallin solutions are modeled well by the hard-sphere Carnahan–Starling equation of state (8, 9) and at physiological pH
α-crystallin solutions do not phase separate; see for example
refs. 9 and 10. Thus we model
g ¼ G∕V kB T
¼ f ðρ1 ;ρ2 Þ þ ðwγ-γ ∕kB TÞϕ1 ð1 − ϕ1 Þ þ ðδwγ-α ∕kB TÞϕ1 ϕ2 : [S1]
In Eq. S1 g is the Gibbs free energy G per unit volume V , per unit
thermal energy kB T.
To apply the model, we note that the present data show the
critical temperatures and therefore the wγ -γ values for E107Asolvent and HGD-solvent solutions to be nearly identical.
Second, the data show E107A and HGD to be similar in folding,
so f will also be identical for E107A–α and HGD–α mixtures.
Thus, differences between the mixtures will be modeled solely
by differing δwγ-α , or equivalently by changes in wγ -α , given our
assumptions.
Banerjee et al. www.pnas.org/cgi/doi/10.1073/pnas.1014653107
In the following we take gijk to denote a mixed partial derivative of g that is taken i times with respect to ρ1 , j times with respect
to ρ2 , and k times with respect to T. We treat the buffer as a single
component and assume that all partial molecular volumes are
constant. It is convenient to note that under these assumptions
differentiation with respect to ρ1 , while holding ρ2 fixed (with
solvent volume fraction remaining 1 − ϕ1 − ϕ2 ) gives a result that,
when multiplied by (1∕ν̄1 ), is the same as that from differentiation
with respect to ϕ1 , while holding ϕ2 fixed.
The spinodal boundary between unstable and metastable
regions satisfies
g200 g020 − ðg110 Þ2 ¼ 0.
[S2]
Substituting Eq. S1 into Eq. S2, we obtain
ðf 200 − ð2ν̄21 wγ -γ ∕kB TÞÞðf 020 Þ − ðf 110 þ ðν̄1 ν̄2 δwγ-α ∕kB TÞÞ2 ¼ 0.
[S3]
The left-hand side of Eq. S3 is quadratic in δwγ-α and thus clearly
nonmonotonic in that quantity. Because wγ-γ is the same for
E107A and HGD, and at a given composition the derivatives
of f are also the same for the E107A–α and HGD–α systems,
Eq. S3 is qualitatively consistent with the findings of Dorsaz
et al. (4) that the stability of the mixtures is expected to be nonmonotonic in the strength of effective, solvent-mediated heterologous, γ–α attractions, δwγ -α . This follows because δwγ-α values
that are too negative or too positive decrease the left-hand side
and therefore enhance instability.
Such nonmonotonic dependence of stability on free energy
parameters is well-known from other ternary mixture contexts,
for example in the case of the regular solution model (6). To
understand this implication further, note that in the term in
Eq. S2 from which the present nonmonotonic dependence originates, −ðg110 Þ2 , the quantity g110 represents the coefficient of a
saddle-like term in the free energy, that has concave down directions that promote instability for either sign of nonzero values
of g110 ; broadly speaking, here these directions correspond to
compositional and density-type phase separation.
Eq. S2 implies that there is a composition direction along
which the second directional derivative of the free energy
vanishes,that of the eigenvalue 0 eigenvector of the Hessian
g200 g110
. That vector is perpendicular to both row
H ρ ½g ¼
g110 g020
vectors of H and therefore has slope ðdρ2 ∕dρ1 Þsp;0 ¼ −g110 ∕
g020 ¼ −g200 ∕g110 .
The direction ðdρ2 ∕dρ1 Þsp;0 has special significance at a critical
point because its value there ðdρ2 ∕dρ1 Þcr;0 is that to which nearby,
short, tie line slopes are asymptotic. Thus, through this connection, the observed directions of the tie lines near such a critical
point can be taken to estimate the value of g110 ∕g020 there. Note
that in Fig. S4II, by including the measured cloud point temperatures for both of the X ¼ 0.5 mixtures, indicates that both observed tie lines must indeed lie rather close though somewhat
below corresponding critical loci, so that considerations of this
paragraph are expected to apply.
More specifically, for the present model we find that
ðdρ2 ∕dρ1 Þsp;0 ¼ −ðf 110 þ ðν̄1 ν̄2 δwγ -α ∕kB TÞÞ∕f 020 :
[S4]
1 of 5
Eq. S4 suggests that increasing γ–α attractive interactions and
therefore decreasing δwγ-α will tend to increase asymptotic tie
line slopes near a critical point in the ternary concentration
triangle, consistent with the data on E107A–α and HGD–α mixtures shown in Fig. 4. However, it should be noted that the connection between δwγ -α and tie line slopes near a critical point is
not direct, because changing δwγ -α can change critical point locations and thereby change relevant values of f 110 and f 020 in Eq. S4.
From a physical point of view, the rotation of tie lines toward
density-type phase separation can be understood by considering
this phenomenon as a much less extreme version of a hypothetical
case of stable complex formation due to attractive interactions
between the two dissimilar proteins, in which case the tie lines
would rotate still further so as to nearly become rays through
the pure solvent vertex of the composition triangle.
The Rayleigh ratio in excess of solvent, ΔR, can also be
expressed in terms of g, for small scattering vector magnitudes
(10, 11). For constant partial specific volumes, one can show (11)
ΔR ¼ ðπ 2 ∕λ4 Þ∇ρ ε · H ρ ½g−1 · ∇ρ ε
[S5]
where ∇ρ ε is the gradient (ϵ100 , ϵ010 ) of the dielectric constant ϵ,
H ρ ½g−1 is the inverse of the Hessian of g with respect to ρ1 and
ρ2 , and λ is the wavelength in vacuum of the incident light, assumed
polarized normal to the scattering plane, as for the present data.
Briefly, Eq. S5 implies that small values of the directional second
composition derivatives of g increase light scattering intensity,
the more so when they are aligned with the dielectric constant gradient vector; a precise formulation of this statement may be found in
(11). Using Eq. S1 in Eq. S5 we obtain
ΔR ¼ ðπ 2 ∕λ4 Þ½fε2100 ; ε2010 ; − 2ε100 ε010 g
· fg200 ; g020 ; g110 g∕ðg200 g020 − ðg110 Þ2 Þ
¼ ðπ 2 ∕λ4 Þ½fε2100 ; ε2010 ; − 2ε100 ε010 g
· ff 200 − ð2ν̄12 wγ -γ ∕kB TÞ; f 020 ; f 110
þ ðν̄1 ν̄2 δwγ -α ∕kB TÞg∕ðg200 g020 − ðg110 Þ2 Þ:
[S6]
We now consider the implications of Eq. S6 for the contrast
between E107A–α and HGD–α mixtures. First, at a given temperature and composition in the single phase region, the closer HGD–α
spinodal suggests that the right-hand side denominator will in general be smaller for HGD-α mixtures. Therefore the observed increase in light scattering of E107A–α mixtures suggests that the
corresponding numerator of ΔR would now be larger for
E107A–α mixtures than for HGD–α mixtures. To see what can
be inferred for δwγ -α , note first the dot product in the numerator
and that ε100 and ε010 are positive. Becuase wγ-γ is the same for
E107A and HGD, increased light scattering intensity observed
for E107A–α mixtures implies that δwγ-α is smaller for E107A–α
than for HGD–α mixtures, again consistent with increased γ–α attractions in E107A–α mixtures.
In summary, the model is consistent with (a) nonmonotonic
dependence of phase separation temperature on δwγ -α (Eq. S3),
(b) rotation of tie lines toward density–density phase separation
implying increased γ–α attractions (Eq. S4), and (c) linkage
between increased γ–α attractions and the observed, increased
light scattering intensity (Eq. S6).
However, in light of (i) the well-established role of short-range
attractions in determining the phase boundaries of protein solutions (12), (ii) the success of molecular dynamics and perturbation theory that include short-range attraction in modeling
neutron scattering data for γ- and α-crystallin mixtures (4, 13),
together with (iii) the importance of aeolotopic, or anisotropic
interactions in affecting the single-protein component phase
boundaries (14), it is important to note that the present model
Banerjee et al. www.pnas.org/cgi/doi/10.1073/pnas.1014653107
assumes functional forms for free energy that need modification
in order to account for short-range, aeolotopic attractions. It can
nevertheless be a guide for expectations as to the phenomena that
can result from increased γ–α attractions.
Appendix S2. A Computational Protein Model for Attractive Interactions Between HGD/E107A and Human αB-crystallin (HAB). Electro-
static surface potential was calculated using the Delphi
software package, v.4 (15) for HGD, E107A and HAB as shown
in Fig. S5. The E107A structure was derived from the HGD structure (16) (PDB ID: 1hk0), by making the Ala substitution at position 107, followed by energy minimization using the DeepView
PDB viewer (17). The HAB structure was modeled using the
recently reported structure of truncated HAB (18). Even though
bovine α-crystallin is an oligomer of the αA- and αB-crystallin
chains present in a 3∶1 ratio, αB-crystallin is known to decorate
the surface of the oligomer while αA-crystallin predominantly
resides within the oligomer core (19, 20). Therefore, we have
chosen to use HAB as the representative of α-crystallin in these
studies. Thus the choice of the HAB structure instead of the
bovine protein is based on the availability of the recent structure
of the human protein. This choice is reasonable due to the overall
strong sequence identity (which exceeds 94% between the two
proteins), with a total identity in the residues of direct interest
for this study. The interaction of HGD and E107A with HAB
was investigated using the AUTODOCK program (21). Pymol
graphics software (22) was used to map the electrostatic potential
on the protein surface.
Fig. S5 shows the electrostatic potential mapped on the surface
of HGD (Fig. S5. A) and E107A (Fig. S5.B). The surface of E107A
shows a distinct positive-potential patch (blue patch around Ala
107) which also includes arginines 89, 115, and 169 and which
can interact effectively with the negatively charged α-crystallin.
A similar patch in HGD (Fig. S5A) includes these arginines as well,
but the negatively charged Glu107 in HGD disrupts the positivepotential.
Fig. S5 also shows the electrostatic potential on the surface of
HAB (PDB ID: 3L1G) with a negative-potential patch in red,
that is likely to show attractive interactions with the opposite
potential on the surface of HGD or E107A. The prominent large
negative patch on the surface of HAB (Fig. S5C) includes residues Glu87, 88, 105, 106, Asp109, and Glu110. We assume this
large negative-potential patch as the likely site of interaction
between HAB and HGD/E107A. Although there are other negative-potential patches on the surface of HAB, they are much
smaller in size, and have not been modeled in this work.
The surface electrostatic potential maps on these proteins
(Fig. S5) guided us in deciding the sites of interaction for our
modeling studies. These studies were carried out to obtain a
working structural model to estimate changes in the attractive
electrostatic interaction energies for the HGD–α and E107A–α
interactions. We used the AUTODOCK program (21), which
is widely used to “dock” a protein with another protein or ligand.
The program also estimates the free energies of interaction for
the best fit in the docking experiments. We set the γ-crystallin
(either HGD or E107A) as the “protein” and HAB as the “ligand.” This selection enabled us to set the grid box dimensions
(which specify the binding site on the protein) in such a way that it
included the positive patch on E107A, and an identical location
on HGD but without the positive patch. All the residues of HGD
or E107A were considered to be rigid. For the ligand, namely
HAB, we used a six-residue HAB fragment, which contained
residues Glu105 to Glu110 extracted from the HAB crystal structure (18), and which represented the negatively charged patch on
HAB. The docked conformations for HGD–α and E107A–α are
shown in Fig. S6 A and B, respectively. Alternate views of the
docking results are presented in Figs. S6 C and D that clearly
highlight the greater overlap of Arg 115 in E107A with Glu105
2 of 5
in HAB (Fig. S6D), which is not the case in HGD (Fig. S6C). This
best fit configuration yields an electrostatic interaction energy of
−2.04 kcal∕mol for the HGD–HAB interaction and −2.43 kcal∕
mol for the E107A–HAB interaction, representing a change in
interaction energy that is consistent with stronger binding of
E107A with α-crystallin.
1. Barker JA, Henderson D (1976) What is a liquid—understanding states of matter. Rev
Mod Phys 48:587–671.
2. Boublik T (1970) Hard-sphere equation of state. J Chem Phys 53:471–472.
3. Mansoori GA, Carnahan NF, Starling KE, Leland TW (1971) Equilibrium thermodynamic
properties of mixture of hard spheres. J Chem Phys 54:1523–1525.
4. Dorsaz N, Thurston GM, Stradner A, Schurtenberger P, Foffi G (2009) Colloidal
characterization and thermodynamic stability of binary eye lens protein mixtures.
J Phys Chem B 113:1693–1709.
5. Gibbons RM (1970) Scaled particle theory for mixtures of hard convex particles. Mol
Phys 18:809–816.
6. Meijering JL (1950) Segregation in regular ternary solutions. 1. Philips Res Rep
5:333–356.
7. Furman D, Dattagupta S, Griffiths RB (1977) Global phase-diagram for a 3-component
model. Phys Rev B 15:441–464.
8. Carnahan NF, Starling KE (1969) Equation of state for nonattracting rigid spheres.
J Chem Phys 51:635–636.
9. Veretout F, Delaye M, Tardieu A (1989) Molecular basis of eye lens transparency. Osmotic pressure and X-ray analysis of alpha-crystallin solutions. J Mol Biol 205:713–728.
10. Thurston GM (2006) Liquid-liquid phase separation and static light scattering of
concentrated ternary mixtures of bovine alpha and gammaB crystallins. J Chem Phys
124:134909.
11. Ross DS, Thurston GM, Lutzer CV (2008) On a partial differential equation method
for determining the free energies and coexisting phase compositions of ternary
mixtures from light scattering data. J Chem Phys 129:064106.
12. Lomakin A, Asherie N, Benedek GB (1996) Monte Carlo study of phase separation in
aqueous protein solutions. J Chem Phys 104:1646–1656.
13. Stradner A, Foffi G, Dorsaz N, Thurston G, Schurtenberger P (2007) New insight into
cataract formation: Enhanced stability through mutual attraction. Phys Rev Lett
99:198103.
14. Lomakin A, Asherie N, Benedek GB (1999) Aeolotopic interactions of globular
proteins. Proc Natl Acad Sci USA 96:9465–9468.
15. Rocchia W, Alexov E, Honig B (2001) Extending the applicability of the nonlinear
Poisson-Boltzmann equation: Multiple dielectric constants and multivalent ions.
J Phys Chem B 105:6507–6514.
16. Basak A, et al. (2003) High-resolution X-ray crystal structures of human gammaD
crystallin (1.25 A) and the R58H mutant (1.15 A) associated with aculeiform cataract.
J Mol Biol 328:1137–1147.
17. Kaplan W, Littlejohn TG (2001) Swiss-PDB viewer (Deep View). Brief Bioinform
2: 195–197.
18. Laganowsky A, et al. (2010) Crystal structures of truncated alphaA and alphaB crystallins reveal structural mechanisms of polydispersity important for eye lens function.
Protein Sci 19:1031–1043.
19. Asselbergs FA, Koopmans M, Van Venrooij WJ, Bloemendal H (1978) Posttranslational
assembly of lens alpha-crystallin in the reticulocyte lysate and in Xenopus laevis
oocytes. Eur J Biochem 91:65–72.
20. Aquilina JA, et al. (2005) Subunit exchange of polydisperse proteins: Mass
spectrometry reveals consequences of alphaA-crystallin truncation. J Biol Chem
280: 14485–14491.
21. Morris GM, et al. (1998) Automated docking using a Lamarckian genetic algorithm
and an empirical binding free energy function. J Comput Chem 19:1639–1662.
22. DeLano WL (2008) The Pymol Molecular Graphics System.
Fig. S1. (A) Near-UV CD spectra of HGD and E107A in 0.1 M phosphate buffer, pH 7.0. Protein concentration, 0.5 mg∕mL. (B) Far-UV CD spectra of HGD and
E107A in 5 mM phosphate buffer, pH 7.0. Protein concentration, 0.1 mg∕mL.
Fig. S2. (A) Tryptophan fluorescence emission spectra of folded and unfolded HGD and E107A (excitation at 290 nm) in 0.1 M phosphate buffer, pH 7.0,
protein concentration, 0.02 mg∕mL. Folded HGD (—), E107A (⋯⋯); unfolded with guanidinium hydrochloride (Gdn.HCl), HGD (-▴-), E107A (-□-). (B) Fluorescence spectra of HGD (—) and E107A (⋯⋯) with Bis-ANS (excitation at 390 nm) in 0.1 M phosphate buffer, pH 7.0, protein concentration, 0.1 mg∕mL, and BisANS concentration, 100 μM. (Inset) Fluorescence spectra of HGD and E107A with Nile Red in 0.1 M sodium phosphate buffer, pH 7; excitation, 540 nm, protein
concentration, 0.1 mg∕mL, and Nile Red concentration, 100 μM.
Banerjee et al. www.pnas.org/cgi/doi/10.1073/pnas.1014653107
3 of 5
Fig. S3. Thermal stability profiles for HGD and E107A in 0.1 M sodium phosphate buffer, pH 7.0, measured as change in ellipticity at 290 nm. (Inset) Gdn.HClinduced unfolding of HGD and E107A in 0.1 M sodium phosphate buffer, pH 7.0. Unfolding monitored by the ratio of the emission intensities at 320 nm and
360 nm, with excitation at 290 nm.
Fig. S4. (I) Mixtures of E107A and α-crystallin (dark blue, B) have lower cloud point temperatures than mixtures of HGD and α-crystallin (red, A), while solutions
of pure HGD and E107A have nearly identical coexistence curves (purple). Assembled coexistence curves are plotted above the ternary α/γ/buffer concentration
triangle. (II) A 3D view of the direction of the measured tie lines (on lower blue triangle) and the differing principal fluctuation directions they suggest (on
upper blue triangle) for the nearby, single phase region. E107A/α-crystallin fluctuations are projected to be much more closely aligned with the gradient vector
of refractive index than those for HGD/α-crystallin mixtures at X a ¼ 0.50 (see text). The tops of the vertical red and blue lines denote the T ph of HGD–α and
E107A–α mixtures, respectively.
Fig. S5. Electrostatic potential mapped on the structures of (A) HGD, (B) E107A, and (C) Human αB-crystallin (HAB). The corresponding color scale is shown at
the bottom. The mutation site in HGD is shown in the boxed region (Glu107 to Ala). A large part of the negative-potential patch on the surface of HAB which
was used for docking studies has been circled in green. See text for details.
Banerjee et al. www.pnas.org/cgi/doi/10.1073/pnas.1014653107
4 of 5
Fig. S6. Docked conformations showing a model peptide fragment of HAB interacting with (A) HGD and (B) E107A. HAB residues are labeled in red and HGD/
E107A residues are labeled in white. The mutation site in E107A is shown in yellow. Different views of the same conformations are shown in C and D to stress
that Arg115 in HGD is not interacting with the HAB peptide (C), while in E107A the same residue is interacting with Glu105 of the HAB peptide (D).
Banerjee et al. www.pnas.org/cgi/doi/10.1073/pnas.1014653107
5 of 5