Biochimica et Biophysica Acta 1842 (2014) 2216–2229

Contents lists available at ScienceDirect

Biochimica et Biophysica Acta

j ourna l homepage: www.e lsev ie r .com/ locate /bbad is

Cataract-causing mutation R233H affects the stabilities of βB1- andβA3/βB1-crystallins with different pH-dependence

Yi-Bo Xi, Wei-Jie Zhao, Xiao-Tong Zuo, Harry Christianto Tjondro, Jing Li, An-Bang Dai,Sha Wang 1, Yong-Bin Yan ⁎State Key Laboratory of Biomembrane and Membrane Biotechnology, School of Life Sciences, Tsinghua University, Beijing 100084, China

Abbreviations: ADCC, autosomal dominant coanilinonaphthalene-8-sulfonate; BSA, bovine serum albDTT, dithiothreitol; Emax, maximum emission wavelengtguanidine hydrochloride; IPTG, isopropyl-1-thio-β-D-galadynamics; SEC, size-exclusion chromatography; SDS, sodiSDS-polyacrylamide gel electrophoresis; Tm, themidpointtransition; WT, wild type⁎ Corresponding author. Tel.: +86 10 6278 3477; fax: +

E-mail address: ybyan@tsinghua.edu.cn (Y.-B. Yan).1 Current address: Boston Children's Hospital, Harvar

02115, USA.

http://dx.doi.org/10.1016/j.bbadis.2014.07.0220925-4439/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 8 June 2014Received in revised form 22 July 2014Accepted 23 July 2014Available online 30 July 2014

Keywords:β-CrystallinCataract-causing mutationMolecular dynamic simulationIon-paring networkProtein aggregationStabilizing mutation

Disease-causingmutations can be stabilizing or destabilizing. Missensemutations of structural residues are gen-erally destabilizing, while stabilizing mutations are usually linked to alterations in protein functions. Stabilizingmutations are rarely identified in mutations linked to congenital cataract, a disease caused by the opacificationof the lens. In this research,we found that R233Hmutation had little impact onβB1-crystallin structure, solubilityand thermal stability under neutral solution pH conditions. The mutation increased βB1 stability against guani-dine hydrochloride-induced denaturation, suggesting that Arg233might be a functional residue. Further analysisindicated that the R233Hmutation did not affect the formation of βA3/βB1 heteromer, but significantly reducedheteromer stability against heat- and guanidine hydrochloride-induced denaturation. The R233H mutationnegatively affected the thermal stabilities and aggregatory propensities of βB1 and βA3/βB1 with different pH-dependence, implying that the protonation of His side chains during acidification played a regulatory role incrystallin stability and aggregation. Molecular dynamic simulations indicated that Arg233 is one of the residuesforming an inter-subunit ion-pairing networkwith intrinsically dynamic nature. Based on these observations, weproposed that the highly dynamic ion-pairing network contributed to the tradeoff amongβB1 solubility, stability,aggregatory propensity and function of protecting βA3.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

In normal cells, the misfolded or aberrant proteins are eitherrefolded or degraded by the quality-control system [1]. When theamount of misfolded proteins exceeds the degradative capacity of thecells, the misfolded proteins or their fragments are prone to depositinto intracellular or extracellular aggregates, which may interfere withthe normal cellular functions [2]. The inappropriate deposition of pro-tein aggregates has been associated with many diseases such asAlzheimer's disease, Parkinson's disease, prion disease, type II diabetesand cataract [3–6]. Unlike neurodegenerative diseases in which theprotofibrils or oligomers are assumed to be the cause of pathology [6],cataract is a disease directly linked to the appearance of large protein

ngenital cataract; ANS, 1-umin; CD, circular dichroism;h of Trp fluorescence; GdnHCl,ctopyranoside; MD, molecularum dodecyl sulfate; SDS-PAGE,temperature of protein thermal

86 10 6277 2245.

d Medical School, Boston, MA

aggregates that scatter the visible light [5]. Cataract, the opacificationof the lens resulting in loss of vision, is a highly prevalent protein aggre-gation diseaseworldwide. Cataract can be induced bymany factors and,among them, genetic disorders and aging are the most important riskfactors [7]. Heredity cataracts account for 8.3–25% of the congenital cat-aracts in children [8]. Inherited mutations may result in isolated andnonsyndromic cataract or systemic anomalies in which cataract is partof the syndrome [9,10]. Compared to congenital cataract, age-relatedcataract is more prevalent and seems to be unavoidable in aging pop-ulations due to the lack of protein turnover in the highly differentiat-ed lens fiber cells [11]. The prevalence of cataract increases verysharply after 50 years of age, and the incidence is about 88% forpeople older than 75 [12]. Interestingly, the study in female twinsdemonstrated that heritability accounts for about half of the casesof age-related nuclear cataract, implying that generic variationsplay an important role not only in congenital cataract but also inage-related cataract [13].

Many genes have been identified to associatewith inherited cataract[8,14]. About half of the genetic disorders have been identified incrystallin genes [8], which is consistent with the proposal that theshort-range ordered packing and gradient distribution of crystallinsare important for the transparency and refractive index of the lens(reviewed in [15]). Although crystallins have been termed as lens-specific genes due to their extremely high expression levels in

2217Y.-B. Xi et al. / Biochimica et Biophysica Acta 1842 (2014) 2216–2229

vertebrate lens, most members in the crystallin family are also widelyexpressed in the other tissues [16,17]. Consequently, crystallins havealso been associated with non-cataract abnormities such as tumorigen-esis, various myopathies, neurological disorders, retinal and vasculardiseases [17–22].

According to their elution positions in the profile of size-exclusionchromatography (SEC), crystallins can be classified into three catego-ries: α-, β- and γ-crystallins [15]. Among them, α-crystallins containtwomembers (αA andαB), both of which are small heat shock proteinswith molecular chaperone-like activity [23]. β/γ-Crystallins are solublestructural proteinswith a conserved two-domain fold, and each domaincontains two Greek-key motifs (Fig. 1A). Despite the high conservationin structural fold, β- and γ-crystallins have quite different oligomericstates. γ-Crystallins are exclusively monomeric, while β-crystallinsexist as homomers or heteromers using the acidic (βA1/A3, βA2 andβA4) or basic (βB1, βB2 andβB3) β-crystallins as the building elements.The basic β-crystallins have been proposed to protect the relatively un-stable acidic β-crystallins in the heteromers [24–26]. Among the sevenhuman β-crystallins, βB1 is unique for its long N-terminal extension,which has been proposed to regulate the size distributions of β-crystallin oligomers [27] and act as an intramolecular chaperone to facil-itate the action of α-crystallins [28].

The importance of β/γ-crystallins in lens functions is supported bythe identification of at least 50 mutations in families with congenitalcataract [14]. Most pathogenic mutations associated with familiaraggregation diseases are destabilizing, and a single-residue mutationusually changes the ΔGfolding value by 0.5–5 kcal/mol, which is of thesamemagnitude as ΔGfolding [29]. Thus the mutation increases the pop-ulation of partially unfolded proteins, which usually have an increasedpropensity to aggregate [2]. The same mechanism has also been identi-fied in many β/γ-crystallin mutations [5]. Particularly, nuclear cataract-linked β/γ-crystallin mutations have been proposed to be Greek-keymotif breakers [30]. Crystallins are required to maintain not only life-long stability but also high solubility in the normal lens. Thusmutationswith little impact on protein structure and stability can also lead to con-genital cataract by decreasing the solubility and promoting aggregationunder high protein concentration conditions [30–34]. The effects ofβ-crystallin mutations may be more complicated than γ-crystallin

Fig. 1. (A) Crystal structure of the truncated human βB1-crystallin (PDB ID: 1OKI). The two subuand C-termini. In the crystal structure, salt bridges are formed between the side chains of Arg(http://www.pymol.org/). (B) Sequence alignment of the C-terminal domain of β/γ-crystalliand blue arrows, respectively. The γ-, acidic and basic β-crystallins are shown in boxes with diidues at the C-terminus are colored in red and blue, respectively. The sequences used for alighuman γD (NP_008822), human γS (NP_060011), human βA4 (NP_001877), human βA2(NP_000487), humanβB1 (NP_001878), ratβB1 (NP_037068), chickenβB1 (NP_989511), bovinβB1 (NP_775338) and pig βB1 (NP_001072149), from top to bottom, respectively. All sequenc

mutations since β-crystallins assemble into homomers or heteromers.Moreover, β-crystallins also have a high expression in some non-lenstissues, implying that β-crystallins may have important physiologicalfunctions other than acting as structural proteins in the lens [16,17,19,35–37]. Besides those mechanisms shared by γ-crystallin mutations[38–41], β-crystallin mutations have been proposed to be able todissociate the homomers [42–45], weaken the heteromer assembly[42,44,45] or disrupt the dynamic oligomeric equilibrium [38].

In theory, a single-residue mutation can be either stabilizing ordestabilizing. Mutations of structural residues are usually destabilizing,while those of functional residues may be stabilizing or neutral due tothe optimization of stability-function tradeoff during evolution [46].Thus disease-causing mutations with a stabilizing effect are expectedto provide us clues in protein functions. To our knowledge, the only sta-bilizing mutation characterized thus far in crystallins is S129R in βB1.S129R has been shown to increase the stability and decrease theaggregatory propensity of βB1 homomer [45]. If assuming that theprotecting effect of βB1 on acidic β-crystallins is one of the functionsof βB1, Ser129 is more likely to be a functional residue contributed toheteromer assembly and stability [42,45]. In this research, we proposedthat Arg233was another functional residue in βB1 based on the dissim-ilar effects of the R233Hmutation on βB1 and βA3/βB1. The R233Hmu-tation has recently been associated with nuclear autosomal dominantcongenital cataract (ADCC) [47]. In the crystal structure of βB1 [48],Arg233 locates adjacent to the last β-strand in Greek-key motif IV ofβB1 (Fig. 1B). The salt bridges formed between the side chains ofArg233 and Asp169 stabilize the subunit interface in the homodimericstructure [33]. This arginine residue is fully conserved in basicβ-crystallins but not in acidic β-crystallins or γ-crystallins, implyingthat it may have specific roles in basic β-crystallins. The results here-in showed that under neutral pH conditions, the R233Hmutation didnot significantly affect the structure and stability of βB1 homomerbut greatly destabilized the βA3/βB1 heteromer. Furthermore, thecomplicated pH-dependence of the mutational effects on proteinstability and aggregatory propensity strongly suggested that solu-tion pH was an important regulatory factor in cataract formation,particularly for congenital cataracts caused by residues substitutedby His.

nits in the dimer are distinguished by cyan and pink, respectively. N and C represent theN-233 and Asp169 to stabilize the subunit interface. The structure was rendered by PyMolns. The residues equivalent to Asp169 and Arg233 in βB1-crystallin are indicated by redfferent colors. The negatively charged residues around Asp169 and positively charged res-nment are: human γA (NP_055432), human γB (NP_005201), human γC (NP_066269),(NP_476434), human βA1/A3 (NP_005199), human βB3 (NP_004067), human βB2eβB1 (NP_776951),mouseβB1 (NP_076184), guinea pigβB1 (NP_001166549), zebrafishes are numbered relative to the starting methionine of each protein (Met1).

2218 Y.-B. Xi et al. / Biochimica et Biophysica Acta 1842 (2014) 2216–2229

2. Materials and methods

2.1. Materials

Ultra-pure guanidine hydrochloride (GdnHCl), bovine serumalbumin(BSA), EDTA, isopropyl-1-thio-β-D-galactopyranoside (IPTG), sodiumdodecylsulfate (SDS), dithiothreitol (DTT) and 1-anilinonaphthalene-8-sulfonate (ANS) were Sigma products. All the other chemicals werelocal products of analytical grade.

2.2. Protein expression and purification

The pET28a plasmids containing thewild type (WT)human βB1 andβA3 genes (CRYBB1 and CRYBA3, respectively) were constructed as thatdescribed previously [42]. Site-directed mutagenesis for R233H wasperformed using the following primers: forward, 5′-GCGTCGCCTGCATGACAAGCAGT-3′ and reverse, 5′-ACTGCTTGTCATGCAGGCGACGC-3′,respectively. The recombinant pET28a plasmids containing the WT ormutated βB1 gene were transformed into Escherichia coli Rosetta(DE3) for the production of the recombinant proteins. Details regardingthe overexpression and purification of βB1 and βA3 were the same asthose describedpreviously [26,42]. In brief, the expression of the recom-binant proteins was induced by 0.1 mM IPTG, and then the cells weregrown in the Luria–Bertani medium for 4 h at 37 °C. The His-tagged re-combinant proteinswere collected from the supernatant of cell extractsby a Ni-NTA affinity column. Then the proteins were purified using aHiload 16/600 Superdex 200 prep-grade column. The protein concen-trationwas determined by the Bradfordmethod using BSA as a standard[49]. For spectroscopic and denaturation studies, the proteins were di-luted into 20 mM phosphate buffered saline (PBS buffer) containing20 mM sodium phosphate, 150 mM NaCl, 0.5 mM EDTA, 0.5 mM DTT,pH 7.0, with a protein concentration of 0.2 mg/ml. For pH-dependentstudies, the proteins were prepared in 100 mM citric acid buffer con-taining 100 mM citric acid and 200 mM sodium phosphate with pHadjusted by citric acid stock solutions.

2.3. Preparation of βA3/βB1 heteromer

The βA3/βB1 heteromer was prepared using the method describedelsewhere [26]. In brief, the stock solutions of βA3 and βB1 were pre-pared in the PBS or citric acid buffers. After mixing the stock solutionsof homomers with a molar ratio of 1:1, the heteromer solutions wereequilibrated at 37 °C for 4 h before use.

2.4. Circular dichroism (CD) spectroscopy

The CD spectra of proteinswere recorded on a Jasco J-715 spectropo-larimeter using a spectral resolution of 0.2 nm at 25 °C. The CD sampleswere prepared in PBS buffer diluted in 40× distilled water to minimizethe interference of salt on the low wavelength CD signals. The proteinconcentration and pathlength were 0.2 mg/ml and 1 mm for the far-UV CD, while 1 mg/ml and 10 mm for the near-UV CD, respectively.For each sample, the CD spectrum was the average of three repetitions.The resultant spectra were obtained by subtracting the correspondingcontrol (spectrum of the buffer).

2.5. Fluorescence spectroscopy

All fluorescence spectra were recorded on an F-2500 fluorescencespectrophotometer using a protein concentration of 0.2 mg/ml. Theslit width was 5 nm for both excitation and emission. The excitationwavelengthswere 295 nmand380nm for the intrinsic Trpfluorescenceand extrinsic ANS fluorescence, respectively. The extrinsic probe ANSwas added to the protein solutions with a molar ratio of 75:1 (probe:protein) and incubated for 30 min in the dark before measurements.

Parameter A of the intrinsic Trp fluorescence was calculated by dividingthe fluorescence intensity at 320 nm by that at 365 nm [50].

2.6. Protein aggregation

Protein aggregation was monitored by turbidity or resonanceRayleigh light scattering. Turbidity is sensitive to large aggregates,while light scattering can reflect the existence of soluble small aggre-gates [51]. Turbidity of the solutionswas determined by the absorbanceat 400 nm (A400) using an Ultraspec 4300 pro UV/Visible spectropho-tometer. Resonance Rayleigh light scattering was measured on anF-2500fluorescence spectrophotometer using an excitationwavelengthof 295 nm and recorded at 90° as described previously [51].

2.7. Size-exclusion chromatography (SEC)

The SEC analysis was performed using a Superdex 200HR 10/300 GLor a Superdex 75 10/300 GL column equipped on an ÄKTA fast proteinliquid chromatography as described previously [44]. In brief, the SECcolumn was pre-equilibrated using PBS or citric acid buffer for about10 column volumes of mobile phase. About 100 μl protein solutionswere injected into the column and run at a flow rate of 0.4 ml/min at4 °C. The dynamic oligomeric equilibrium of β-crystallin homomerswas analyzed by ranging the protein concentration from 0.2 to2 mg/ml. The formation of β-crystallin heteromers was checked byanalyzing the protein samples equilibrated for 0 h or 2 h at 37 °C aftermixing of the pre-equilibrated stock solutions of βA3 and βB1homomers. The homomers were pre-equilibrated in the citric acid buff-er with pH 5, 6 or 7 overnight. The pH-dependence was analyzed bypre-equilibration of the column using the citric acid buffer with agiven pH, and then the elution profiles of the proteins were obtained.

2.8. Protein solubility

Protein solubility was determined using the microconcentrationmethod as that described elsewhere [38]. In brief, the proteins wereconcentrated by centrifugation using a microminiature 10K MWCOMillipore concentrator at 12,800 g and 4 °C. After the concentrationreached its maximum, the solubility was determined by measuringthe concentration of the proteins in the soluble fractions. All solubilityexperiments were repeated for three times, and the data were present-ed as average ± standard error.

2.9. Protein thermal denaturation

The equilibrium thermal transition curves were obtained by heatingthe protein samples continuously from 26 °C to 82 °C. The temperaturewas controlled by a water-bath connected to the instruments. Trp fluo-rescence, light scattering and turbidity data were collected every 2 °Cafter a 2 min equilibration. The volume of protein solutions was 0.6 mland the protein concentration was 0.2 mg/ml. The cuvette was sealedto avoid the possible volume change during heating. The aggregation ki-netics was measured by heating the protein solutions continuously at agiven temperature and recording the turbidity data every 2 s. The kinet-ic parameters of thermal aggregation were obtained by fitting the time-course data by a first-order kinetics using the following equation [52]:

A ¼ Alim 1− expð−kagg t−t0ð Þ� �

; ð1Þ

where A is the A400 value at a time t, Alim is the A400 value at the infinitetime, kagg is the rate constant, t is the heating time at a given tempera-ture and t0 is the lag time. The Alim × k value, which reflects the initialvelocity of aggregation, was also calculated using the best-fitted kineticparameters.

2219Y.-B. Xi et al. / Biochimica et Biophysica Acta 1842 (2014) 2216–2229

2.10. Acid-induced denaturation

The protein stock solutions were mixed with the citric acid bufferwith various pH values ranging from pH 2 to 8. The final pH value waschecked after mixing. The samples were equilibrated for at least30 min before spectroscopic measurements. The transition curveswere obtained by measuring the pH-dependence of CD, Trp fluores-cence, ANS fluorescence, light scattering and turbidity of the proteinsolutions.

2.11. GdnHCl-induced denaturation

The equilibrium denaturation was carried out by incubating theβ-crystallins in PBS buffer containing various concentrations of GdnHClranging from0 to 6Movernight at 25 °C. After incubation, the structuralchanges were monitored by CD, Trp fluorescence, light scattering, tur-bidity and ANS fluorescence. The final protein concentration was0.2 mg/ml for equilibrium unfolding studies. The fully denatured sam-ples used for kinetic refolding studies were prepared by incubating5 mg/ml proteins in citric acid buffer (pH 5, 6 and 7) containing 4 MGdnHCl overnight at ambient temperature. Kinetic refoldingwas initiat-ed by a 1:40 fastmanual dilution of the 4MGdnHCl-denatured proteinsinto the citric acid buffer. The final GdnHCl and protein concentrationswere 0.1 M and 0.125 mg/ml, respectively. The time-course turbiditydata were collected every 2 s immediately after the initiation of kineticrefolding. After 1 h refolding, the protein samples were centrifuged at12,800 g and the proteins in the supernatant and precipitation fractionswere analyzed by 12.5% SDS-PAGE.

2.12. Molecular dynamic (MD) simulations

MD simulations for the WT and mutated βB1-crystallins were per-formed using the same method as that described elsewhere [45]. Inbrief, the crystal structure of the truncated human βB1-crystallin (PDBID: 1OKI) [48] was used as the template structure to create the startingstructure of the R233Hmutant. To mimic the effect of pH on the chargestates of His, the simulations were performed using starting structurescreated with either neutral (HSE) or positively charged (HSP) His resi-dues. The simulations were carried out in a water box with 150 mMNaCl added by VMD [53]. The MD simulations were performed usingNAMD 2.8 [54] at 300 K (27 °C) and 1 atm. The time step for simulationwas 2 fs. After 10 ns simulation, the structures were used for furtheranalysis. The free energies of subunit interactions were calculatedusing the VMD plugins NAMD Energy. The structures were analyzedand rendered using PyMol (The PyMol Molecular Graphics System,Version 0.99rc6, Schrödinger, LLC, http://www.pymol.org/).

3. Results

3.1. Effect of the R233H mutation on βB1 structure by biophysicalexperiments

The R233H mutant behaved similar to the WT βB1 with most ofthe recombinant proteins existed in the soluble fractions whenoverexpressed in E. coli cells at 37 °C (data not shown). The purifiedproteins were used for spectroscopic experiments to investigate the ef-fect of themutation onβB1 secondary structure by far-UVCDand tertia-ry structure by near-UV CD and Trp fluorescence. As shown in Fig. 2, thealmost superimposed spectra indicated that the mutation did not affectβB1 secondary and tertiary structures. The hydrophobic exposure of theproteins was detected by the extrinsic fluorescence probe ANS, whichbecomes pronouncedly fluorescent with a blue-shift in the emissionwavelength when bound to the exposed hydrophobic sites of proteins[55]. The mutation slightly decreased the ANS fluorescence of βB1, sug-gesting that themutant had fewer ANS-accessible sites. The effect of theR233H mutation on the oligomeric equilibrium of βB1 was studied

using SEC analysis by varying protein concentrations. The mutant hada relatively larger elution volume in the SEC profiles when comparedto theWT protein, suggesting that the mutation might slightly increasethe dissociation constant of the dimeric protein. Thus the biophysicalanalysis above indicated that the R233H mutation had very limited im-pacts on βB1 structures.

3.2. Effect of the R233H mutation on βB1 stability against GdnHCl-induceddenaturation

The transition curves of βB1 denaturation induced by GdnHCl weresimilar to those reported previously [26]. The GdnHCl concentration-dependent changes of Parameter A, which is a sensitive monitor of thecharacteristics of the shape and position of the Trp fluorescence [50],are shown in Fig. 3. At GdnHCl concentrations below 4M, the transitionobserved by Parameter A has been proposed to be best fitted by a three-state model involving a monomeric intermediate appeared at ~1.6 MGdnHCl: native state (N)↔intermediate (I)↔unfolded state (U) [26].The mutation did not affect the process at GdnHCl concentrationsabove 1.6 M as evidenced by the superimposed transition curves fromParameter A as well as the other techniques (data not shown). Thusthe mutation has no impact on the I↔U transition but influenced theN↔I transition. This is consistent with the structural analysis thatArg233 mainly contributes to the stabilization of monomer–monomerinterface in the dimer (Fig. 1). Surprisingly, themutation slightly shiftedthe N↔I transition to a higher GdnHCl concentration, suggesting thatthe R233H mutation was a stabilizing one. A quantitative analysis ofthe changes in ΔGfolding using the samemethod as that described previ-ously [26] indicated that ΔGfolding for the N↔I transition was increased4 kcal/mol by the R233H mutation, while no difference was observedfor the I↔U transition.

3.3. Effect of the R233H mutation on the acid denaturation of βB1

The imidazole side-chain in free His has a pKa value of approximate-ly 6. This means that the imidazole group is hydrophobic at physiologi-cal relevant pH. Under acidic conditions with pH values below 6, theprotonation of the imidazole ring will introduce a positive charge andthe hydrophobic side-chain of His will become hydrophilic. Consideringthat the alteration of the charge state of His233 might affect the behav-ior of the R233H mutant, acid denaturation was performed to investi-gate the pH-dependent structural changes for both proteins (Fig. 4).Decreasing solution pH only induced minor changes in the ellipticityof the far-UV CD and the maximum emission wavelength (Emax) of Trpfluorescence, suggesting that βB1 was quite resistant to acidification.This phenomenon is quite different from βB2, which loses most of itsnative structure at extremely acidic pH [40]. Nonetheless, at pH below5, an abrupt increase in the ANS fluorescence could be observed. Thestep-wise transition of ANS fluorescence also suggested that the acid de-naturation of βB1 might involve stable unfolding intermediate(s). Thisinference was confirmed by the appearance of aminor amount of aggre-gates at around pH 4.5 by turbidity measurements and the dramatic in-crease of the light scattering at around pH 3.5. Thus the step-wiseincrease of ANSfluorescencemight be caused by the accumulation of dif-ferent off-pathway oligomers/aggregates since ANS has the ability tobind with aggregates induced by low-concentration denaturants [56].The mutant behaved similarly to the WT protein for most transitioncurves shown in Fig. 4. The most significant difference was the changesin Parameter A at pH around 5. Few changes were observed for theParameter A values of the WT βB1 between pH 4 and 7.5. As for theR233H mutant, the maximum value appeared at pH 5 and small shiftsin pH significantly altered the Parameter A value. This observation sug-gested that the change in His233 charge state could modify the localstructures of the R233H mutant, which further influenced the popula-tions of various Trp fluorophores.

Fig. 2. Effect of the R233H mutation on βB1 structure probed by biophysical methods. (A) Far-UV CD. The inset of panel A shows the SDS-PAGE analysis of the purifiedWT and mutatedproteins. (B) Near-UV CD. (C) Intrinsic Trp fluorescence. (D) Extrinsic ANS fluorescence. (E) SEC analysis. (F) Protein concentration-dependence of the elution volumes. The proteinswereprepared in 20mMPBS buffer. The protein concentrationswere 0.2 mg/ml for far-UV CD and fluorescencemeasurements and 1 mg/ml for near-UV CD experiments. The SEC analysiswasperformed using a Superdex 200HR 10/300 GL column.

2220 Y.-B. Xi et al. / Biochimica et Biophysica Acta 1842 (2014) 2216–2229

3.4. Effect of the R233H mutation on the pH-dependence of βB1 solubility

Cataract-causingmutations can lead to disease simply by decreasingthe solubility of crystallins [15,31–34], thus the solubility of theWT andmutated proteins was determined using the microconcentration meth-od (Fig. 5). At neutral pH, both proteins were highly soluble with a sol-ubility of about 250 mg/ml. The solubility of both proteins increasedduring acidification,while themutant had a significantly higher solubil-ity than the WT protein (P b 0.01) under mild acidic conditions.

Fig. 3. Effect of the R233H mutation on βB1 denaturation induced by GdnHCl. The transi-tion curves of protein equilibrium unfoldingwere obtained by calculating the Parameter Avalues of the Trp fluorescence spectra. The raw data were fitted using the samemethod asthat described previously [26]. Similar transition curves could be also obtained by moni-toring the Emax values of Trp fluorescence (data not shown). The inset shows the turbidityof each sample. No aggregates could be detected under our denaturation conditions. Forreference, the symbol of triangle shows the A400 value of 0.2 mg/ml βB1 solution heatedat 70 °C.

3.5. Effect of the R233H mutation on the pH-dependence of βB1 thermalstability

The results shown in Fig. 4 indicated that the R233Hmutationmightinduce local structural changes of βB1. To further investigate the impactof these structural changes on βB1 stability, thermal denaturation of theproteins was conducted at various pH values (Fig. 6). The transitioncurves of R233H during equilibrium thermal denaturation were almostsuperimposed to those of theWT protein at pH 7 and pH below 4. Inter-estingly, both proteins did not aggregate at high temperatures at pHbelow 4. However at mildly acidic pH, the curves of themutant showeda divergence from those of the WT βB1. The midpoints of thermalunfolding (Tm) for the mutant were 7.1 °C and 8.7 °C lower than thoseof theWT protein when incubated at pH 5 and 6, respectively. The mu-tational effect on βB1 thermal aggregation was weaker than that onthermal unfolding. The mutation decreased the starting temperatureof aggregation by 2–3 °C at pH 5 and 6. Thus the protonation of the im-idazole side-chain of His233 in the mutant did not stabilize the protein,

but was deleterious to the thermal stability of R233H. Below pH 4, thedeleterious effect of His233 protonationwas eliminated due to the over-all structural changes of both proteins.

Fig. 4.Acid-induced unfolding of theWT βB1 and R233Hmutant. (A) Ellipticity at 215 nm from far-UV CD. (B)Maximumemissionwavelength of Trpfluorescence (Emax). (C) Parameter Aof intrinsic Trp fluorescence spectrum. (D) Intensity at 470 nm of extrinsic ANS fluorescence. (E) Turbidity monitored by the absorbance at 400 nm (A400). (F) Resonance Rayleigh lightscattering excited at 295 nm. Theprotein sampleswere prepared in citric acid bufferwith a protein concentration of 0.2 mg/ml. In panel B, the Emax value of the 4MGdnHCl-denaturedβB1is shown as an indicator of the fully denatured state. In panel E, the A400 value of 0.2 mg/ml βB1 heated at 70 °C is shown as an indicator of serious aggregation (the triangle symbol).

2221Y.-B. Xi et al. / Biochimica et Biophysica Acta 1842 (2014) 2216–2229

The equilibrium thermal denaturation results indicated that themutant aggregated at a lower temperature than the WT protein undermildly acidic conditions. In theory, the decrease in the starting temper-ature of aggregation can be caused by either a decrease in stability or anincrease in the aggregatory propensity. To elucidate which mechanismis true, the kinetic aggregation process was studied by incubating theproteins at 70 °C or 75 °C continuously. The kinetic data recorded at75 °C are shown in Fig. 6E, and similar behaviors were observed for pro-teins aggregated at 70 °C (data not shown). The mutant aggregatedabout 3-fold faster than the WT protein at pH 7. This suggested that

Fig. 5. pH-dependence of βB1 solubility. The proteins were prepared in citric acid bufferwith various pH values. The solubility of the proteins was determined by themicroconcentration method with three repetitions. ** represents P b 0.01. The solubilitywas also significantly different (P b 0.01) between neutral and acidic solutions.

the thermally denatured mutant had stronger aggregatory propensitythan the WT protein at neutral pH since the thermal stability of themutant was similar to that of theWT protein. Lowering solution pH de-creased the maximum turbidity values but increased the aggregationrates of both proteins. However, the change of the mutant was milderthan that of theWT protein, and thus a slower aggregation rate was ob-served for the mutant at pH 5. Thus the R233Hmutation had dissimilareffects on βB1 thermal stability and aggregation propensity. At neutralpH, the mutation did not affect βB1 stability but increased the aggrega-tion rate at high temperatures. However under mildly acidic conditions,themutation decreased both the thermal stability and aggregatory pro-pensity of βB1.

3.6. Effect of the R233H mutation on βB1 structure by MD simulations

In the crystal structure of truncated βB1 [48], the salt bridgesbetween Arg233 and Asp169 contribute to the stabilization of subunitinterface in the dimer (Fig. 1A). Previously it has been shown that thecontribution of a single salt bridge is 3–5 kcal/mol for ΔGfolding [57].Unexpectedly, the disruption of the two Arg233–Asp169 salt bridgesdid not significantly affect βB1 thermal stability, but increased ΔGfolding

for the N↔I transition during GdnHCl-induced unfolding at neutral pH.Under mildly acidic conditions, the mutation decreased βB1 thermalstability but increased solubility. To explore the underlying structuralbasis, MD simulations were performed by including either the neutralimidazole ring of His (HSE) or the positively charged His (HSP) tomimic the charge states of His at neutral or acidic pH, respectively.The overall structure of βB1 was not significantly affected by theR233H mutation (Fig. 7A), which is consistent with the spectroscopicresults (Fig. 2). For both proteins, electrostatic interactions played adominant role in the stability of βB1 dimer interface (Fig. 7B). To

Fig. 6. pH-dependence of βB1 thermal stability. (A) Transition curvesmonitored by Parameter A of Trp fluorescence. The raw data were fitted by a two-state (N↔U)model and the fitteddata are shown as solid (WT) anddashed lines (R233H). (B) Changes in light scattering duringβB1 thermal denaturation. Thedata recorded at pH4were similar to those at pH3 except fora higher light scattering value (data not shown). (C) Equilibrium thermal aggregation of theWT andmutatedβB1 proteinsmonitored by turbidity. (D) Themidpoint of thermal transition(Tm) obtained by fitting the data in panel A. (E) Thermal aggregation kinetics of the WT and mutated βB1 proteins at 75 °C. (F) pH-dependence of the aggregation rate constants (kagg)obtained by fitting the data in panel E by Eq. (1). The other kinetic parameters are shown in Supplemental Fig. S1.

2222 Y.-B. Xi et al. / Biochimica et Biophysica Acta 1842 (2014) 2216–2229

avoidmisleading, the simulationswere performed twice using the samestarting structures. A minor deviation could be observed in the bindingfree energy originated from electrostatic interactions for the WT pro-tein, while a large discrepancy was observed for the R233H mutant.For both cases, the RMSD and binding energy fromhydrophobic interac-tions reached equilibrium, suggesting that the modeling dispersionmight be caused by the dynamic nature of βB1 subunit interface. Thehigher dissociation propensity of the mutated protein was alsoverified by the SEC analysis shown in Fig. 2E.

When the hydrophobic side chains (HSE) of His residues werereplaced by the positively charged ones (HSP), no significant changeswere observed for the overall structures of both proteins (Fig. 7A). Thepositions of most Trp fluorophores except for Trp237 maintainedunchanged in the modeled structures (Fig. 7C), coinciding with theminor changes in Emax of the Trp fluorescence during acid denaturation(Fig. 4). However, there were minor alterations in Parameter A of themutant at mildly acidic pH. Trp237 is located at the C-terminal exten-sion of βB1 and is close to the mutation site. Although the protonationof His233 in the mutant might perturb the local microenvironmentaround Trp237, the alteration in the positions of Trp237wasmore likelyto be caused by the intrinsic flexibility of the C-terminal extension ofβB1 (Supplemental Fig. S2). It is worth noting that the MD simulationswere performed based on the crystal structure of truncated βB1 lackingthe first 54 AA off the N-terminus and the last 15 AA off the C-terminus.However, the in vitro experiments were performed using full-lengthproteins, and thus the impact of the long N- and C-terminal extensionson Trp fluorophores could not be ruled out from MD simulations.

The effect of the R233Hmutation on the surface charge potentials ofβB1 is shown in Fig. 7D. Although the analysis is based on truncatedproteins, the results clearly indicated that both of the mutation and

His protonation affected the surface charge distributions. Unexpectedly,it seems that R233H-HSP possessed more surface charges. One possiblereason is that the mutation disrupted the ion-pairing network and re-leased the charged residues to the surface of the molecule (Fig. 8, alsosee Supplemental Fig. S3). The inter-subunit ion-pairing network isformed by the positive charge cluster at the C-terminus (Arg230,Arg231, Arg233 and Lys235) and the negative charge cluster in theC-terminal domain (Glu117, Glu148, Asp169 and Asp170). Amongthese charged residues, Arg230 and Arg231 reside at the last β-strandof Greek-key motif IV, while the other residues are at loops orC-terminal extensions. Thus most of these charged residues are poten-tially flexible, which is supported by the dispersion between the tworounds of MD simulations. The high flexibility of these charged residuesresulted in a highly dynamic ion-pairing network. Except for the stableion-pairs formed between Arg230 and Glu117, most ion-pairs weremore likely to be formed randomly between the positively and nega-tively charged residues. The probabilities of various ion-pairs involvingArg233 were calculated by analyzing the 1000 structures obtained bythe two rounds of simulations of WT-HSE. Arg233 had a probability of77% to form either intra- or inter-subunit ion-pair(s). Among them,the probability was 36%, 30% and 11% for Arg233(A)–Asp170(B),Arg233(B)–Glu148(B) and Arg233(B)–Asp170(B), respectively. Thesubstitution of Arg233 by His fully abolished the possibility ofion-pairing formation at neutral pH. The protonation of His233 in themutant could partially rescue the ion-pairing network. However,the ion-pairing network was weakened by the mutation under both ofthe neutral and acidic pH conditions. Thus the highly dynamic natureof the ion-pairing network around Arg233might provide βB1 the struc-tural basis of tolerance to single-point mutations. Meanwhile, therelease of the charged residues by either mildly acidic conditions or

Fig. 7. Structural insight into the effects of the R233H mutation and protonation of His side-chains on βB1 structure by MD simulations. (A) A comparison of the modeled structures be-tween theWT protein and R233Hmutantwith neutral (HSE) or positively charged (HSP) His side-chains. (B) Effect of the R233Hmutation on themonomer–monomer binding energy ofβB1. The simulationswere run twice using the same starting structure, and labeled as Round-1 and Round-2, respectively. The energies of electrostatic and hydrophobic interactionswerecalculated separately. For clarity, the total energies, which are the sum of energies from electrostatic and hydrophobic interactions, are not shown. (C) A comparison of the structures ofR233H with hydrophobic or hydrophilic His side-chains. The positions of the His residues are indicated by spheres, while the Trp fluorophores are shown by the stick model. For clarity,only the Trp residues adjacent to the mutation sites are labeled. (D) Surface electrostatic potentials of the WT and mutated proteins. Surfaces in blue, red and white represent positivelycharged, negatively charged and hydrophobic, respectively.

2223Y.-B. Xi et al. / Biochimica et Biophysica Acta 1842 (2014) 2216–2229

mutations might increase the numbers of surface charges and therebyincrease protein solubility (Fig. 5).

3.7. Effect of the R233H mutation on βA3/βB1 heteromer formation

β-Crystallins are capable to form heteromers in vivo [27], and somecataract-causingmissense mutations can affect heteromer formation orstability [38,42,45]. Thus SEC and spectroscopic methods were used toexplorewhether theR233Hmutation impairedβA3/βB1heteromer for-mation. As shown in Fig. 9, βA3 and βB1 could assemble into a stableheteromer eluted earlier than βA3 or βB1. The subunit exchange wasnearly completed after 2 h incubation at 37 °C. The elution volumes ofβA3 and βB1 decreased along with the decline of solution pH, whilethe shift of βB1 was larger than that of βA3. The formation of βA3/βB1heteromer was strongly dependent on solution pH as revealed by theincrease of non-associated βA3 and βB1 along with acidification. Theeffect of the R233H mutation on heteromer formation was negligible,implying that Arg233 was not a key residue involved in heteromerrecognition and association. This conclusion was also verified by theidentical Emax values of Trp fluorescence under all pH conditions studiedhere (Fig. 9D).

3.8. Effect of the R233H mutation on βA3/βB1 folding

GdnHCl-induced denaturationwas performed to investigate wheth-er the R233H mutation in βB1 affects βA3/βB1 heteromer stability. TheWT heteromer began to unfold at around 1.4 M GdnHCl, while theheteromer containing the mutated βB1 quickly lost its native structureunder low GdnHCl concentrations (Fig. 10A). It is worth noting thatβA3/R233H was even less stable than βA3, which has been shown tomaintain stable atGdnHCl concentrations below1.0M [26]. Consideringthat R233H homomer was more stable than the WT βB1, the extremeinstability of βA3/R233H suggested that the subunit interactions in theheteromer were different from those in the homomers. Fitting of thetransition curves using the method reported previously [26] indicatedthat the R233H mutation destabilized βA3/βB1 by 6.1 kcal/mol for theN↔I transition, but did not affect the I↔U transition. The opposingeffects of the R233H mutation on the stabilities of βB1 and βA3/βB1suggested that Arg233 played a much more important structural rolein the heteromer than in the homomer.

Aggregation during kinetic refolding was used to examine whetherthe R233H mutation interfered with the correct folding of βB1 andβA3/βB1 (Fig. 10B). Consistent with the previous observations [26],βA3 aggregated immediately after diluting the denatured proteins in

Fig. 8. Dynamic ion-pairing network at the subunit interface of βB1 dimer. (A) Ion-pairing network formed by positively charged residues at the C-terminus and negatively charged res-idues aroundAsp169 inWT-HSE. Different equilibrated stateswere found for the two rounds of simulations although both of them reached a stable equilibriumafter 10 ns simulations. Theion-pairing networks of WT-HSP are shown in panel A of Supplemental Fig. S3. Residues from the same subunit are labeled in the same color. (B) Time-course changes of the energies ofelectrostatic interactions between Arg233 and adjacent ion-pairing partners during MD simulations. A or B in the brackets denotes the residue in subunit A or B, respectively. (C) Ion-pairing network at the subunit interface of R233H with protonated or non-protonated His residues. For clarity, Arg230 and Arg231 are not labeled. The results of the other round of sim-ulations are shown in panel B of Supplemental Fig. S3. (D) Time-course changes of the energies of electrostatic interactions between His233 and adjacent ion-pairing partners duringMDsimulations. The protonation of His233 could partially rescue the disruption effect of the R233H mutation on the ion-pair network.

2224 Y.-B. Xi et al. / Biochimica et Biophysica Acta 1842 (2014) 2216–2229

the refolding buffer, while βB1 could protect βA3 to avoid the unpro-ductive pathway during co-refolding. The behaviors of the R233H mu-tant were the same as the WT βB1, implying that Arg233 contributedlittle to the folding and assembly of both βB1 and βA3/βB1.

3.9. Effect of the R233H mutation on the pH-dependence of βA3/βB1thermal stability

Equilibrium thermal denaturation studies indicated that the R233Hmutation in βB1 significantly reduced the thermal stability of βA3/βB1under all pH conditions (Fig. 11). When monitored by Parameter A ofTrp fluorescence, the Tm values of βA3/βB1 were decreased by 6.9 °C,5.3 °C and 3.5 °C when heated in solutions with pH values of 7, 6 and5, respectively. Meanwhile, the mutation also promoted βA3/βB1 ther-mal aggregation by lowering the starting temperature of aggregation by4–6 °C. Compared with the results of βB1 homomer shown in Fig. 6, thedestabilizing effect of the R233Hmutationwasmuchmore pronouncedfor the βA3/βB1 heteromer under neutral pH conditions. Under acidicconditions, the impairment caused by the mutation was strengthenedfor the homomer, but weakened for the heteromer.

Thermal aggregation kinetics indicated that the aggregation rateconstant of βA3 decreased along with acidification (panels E and F inFig. 11), which was different from the behavior of βB1 or βA3/βB1. AtpH 7 and 6, the aggregation rate of βA3 could be successfully sloweddown by the existence ofβB1 in the heteromer. At pH5, the aggregationrate of βA3/βB1 heteromer was slightly higher than that of βA3homomer, which might be caused by the opposing pH-dependence ofkagg for βA3 and βB1 and/or the partially dissociation of the heteromerduring acidification. The R233H mutation greatly impaired the protec-tion effect of βB1 on βA3 thermal aggregation at pH 7 and 6, but

behaved similar to the WT protein at pH 5. The strong pH-dependenceof the mutational effect on βB1 and βA3/βB1 aggregation kineticssuggested that the C-terminus of βB1 might be involved in βB1 andβA3/βB1 thermal aggregation.

4. Discussion

Proteins are marginally stable [58]. The marginal stability presum-ably results from the natural optimization of the tradeoff between pro-tein stability and function during evolution [46,59]. Mutations of thestructural residues usually destabilize the proteins and/or promoteaggregation by accumulating misfolded polypeptides under normal orstressed conditions. On the other hand, the polar or charged functionalresidues are generally embedded in the hydrophobic clefts of themolecule and therefore thermodynamically unfavorable [46]. Thus mu-tations of functional residues may lead to an increase in stability. Iden-tifications of the stability-beneficial disease-causing mutations willdefinitely provide us new information regarding protein functions inthe cells. Considering that cataract is a protein aggregation disease, itis not surprising that most cataract-causing β/γ-crystallin mutationsare found to decrease either crystallin stability or solubility [5,30]. Previ-ously, we have identified a stability-beneficial mutation S129R in βB1[42]. S129R has further been shown to impair the protection effect ofβB1 on βA3 and increase the sensitivity to proteases of both βB1 andβA3/βB1 [42,45]. In this research, we identified another stability-beneficialmutation inβ/γ-crystallins, R233H inβB1 associatedwith nu-clear ADCC. Themechanistic investigations herein provided not only thepossible molecular mechanism underlying ADCC caused by the R233Hmutation, but also novel insights into the solubility–stability–functiontradeoff of βB1 during evolution.

Fig. 9. pH-dependence of βA3/βB1 heteromer formation. (A) SEC analysis of homomers and heteromers at pH 7. (B) SEC analysis of homomers and heteromers at pH 6. (C) SEC analysis ofhomomers and heteromers at pH 5. (D) Intrinsic Trp fluorescence of the βA3/βB1 heteromers. The SEC analysis was performed using heteromer samples incubated for 0 h or 2 h at 37 °Caftermixing theβA3 and βB1 stock solutions. The SEC profiles of β-crystallin homomers are also shown. To achieve a better peak resolution, a Superdex 75 10/300 GL columnwas used forthe SEC analysis. The protein concentration was 1 mg/ml and 0.2 mg/ml for the SEC analysis and fluorescence experiments, respectively. The maximum emission wavelength of Trpfluorescence (Emax) had a minor red-shift along with acidification, and the positions of Emax are indicated by dotted lines.

2225Y.-B. Xi et al. / Biochimica et Biophysica Acta 1842 (2014) 2216–2229

4.1. The ion-pairing network at the subunit interface of βB1 is intrinsicallydynamic

Previous structural study has proposed that the inter-subunit saltbridges between Arg233 and Asp169 in βB1 as well as the equivalentresidues in βB2 may facilitate the oligomerization of domain-swappedβ-crystallin homomers (Fig. 1A) [48]. Unexpectedly, no significantalterations induced by the R233Hmutation were observed in the struc-ture, solubility and stability of βB1 under neutral conditions, in whichHis233 in the mutant is hydrophobic. These observations suggestedthat the Arg233–Asp169 salt-bridges were not critical to homomer for-mation although both residues are fully conserved in basic β-crystallins(Fig. 1B). According to the previous structural study [48] and the MDsimulation study herein, we proposed that the highly dynamic natureof the ion-pairing network at the subunit interface might be an intrinsicmechanism for β-crystallins to get tolerant to random mutations. Asshown in Figs. 7 and 8, electrostatic interactions are the predominantforce to stabilize βB1 dimer. The most intense ion-pairing was formedamong eight conserved residues in each subunit including four positive-ly charged ones and four negatively charged ones. The overall structureof the proteins was well maintained under all simulating conditions.The reliability of theMD simulation results was also verified by the con-sistency in the formation of stable ion-pairs between Arg230 andGlu117 under all simulation rounds. However, the ion-pairing statuswas quite different for the two rounds of simulations of both the WTand mutated proteins. A close inspection of the ion-pairing network

during simulation suggested that most interactions between the posi-tively charged and negatively charged residues might be random orhighly dynamic. For example, Arg233 could interact with Glu148,Asp169 or Asp170 located at either the same or adjacent subunit. Con-sequently, the substitution of Arg233 by the hydrophobic His residuedid reduce the number of ion-pairs, but had little impact on the stabilityof the highly dynamic ion-pairing network.

One of the possible roles of the dynamic ion-pairing network mightbe the maintenance of the dynamic oligomeric equilibrium. In the lens,β-crystallins form a mixture of various homomers and heteromers viadomain-swapping [27]. Domain-swapping has been shown to be oneof the mechanisms of protein aggregation/fibrilization under nativeconditions [60]. Thus the highly dynamic ion-pairing network allowsβ-crystallin oligomers to keep a proper dissociation constant andavoid misassembly into large aggregates. Very recently we haveshown that the replacement of Arg188 in βB2, which is equivalent toArg230 in βB1, by His destroys the monomer–monomer interactionbut strengthens the dimer–dimer association [38]. The dimer–dimer in-terface of βB2 possesses a similar ion-pairing network with that at themonomer–monomer interface of βB1 [38,48]. The unusual stabilizationof the dynamic ion-pairing network perturbs the natural equilibriumamong various oligomeric states, whichmay be the origin of the defectscaused by the R188H mutation in βB2 [38,40].

The other possible role of the dynamic ion-pairing networkmight bethe tradeoff between solubility and stability. Preferred solvent–proteininteractions will definitely increase protein solubility. The formation of

Fig. 10. Unfolding and refolding of βA3/βB1 heteromer induced by GdnHCl.(A) Equilibrium unfolding of βA3/βB1 probed by Parameter A. The raw data were fittedby a three-statemodel and the fitted data are shown as lines. The inset shows the changesin ANS fluorescence during GdnHCl-induced unfolding monitored by the absorbance at400 nm. No aggregation was observed for all samples (data not shown). (B) Aggregationduring the kinetic refolding of βA3, βB1 and βA3/βB1. The proteins were fully denaturedby4MGdnHCl overnight at ambient temperature. The refoldingwas initiated by fastman-ual mixing of the denatured proteins with the refolding buffer, and then the turbidity wasrecorded every 2 s. Aggregation of βA3 appeared immediately after dilution, while theother proteins did not aggregate. The inset shows the SDS-PAGE analysis of soluble andinsoluble fractions of the proteins refolded 1 h after dilution.

2226 Y.-B. Xi et al. / Biochimica et Biophysica Acta 1842 (2014) 2216–2229

an ion-pairing networkwill significantly reduce the number of solvatedresidues, but enhance stability. A comparison of the simulated struc-tures of βB1 with HSE or HSP suggested that the number of ion-pairsin the networkwas reduced by acidic conditions. Consequently, a signif-icantly enhancement of protein solubilitywas observed for both theWTand mutated βB1 (Fig. 5), accompanied with a dramatically decrease inthermal stability (Fig. 6). Under neutral conditions, the R233H mutanthad the same solubility and stability as the WT βB1 since Arg233 inβB1 is desolvated. However under acidic conditions, the protonationof His233 together with the less intact ion-pairing network helped themutant to have an even higher solubility and concomitantly lowerstability.

4.2. Arg233 is a functional residue crucial to βA3/βB1 heteromer

The R233H mutation increased the ΔGfolding value of βB1 whendenatured by GdnHCl. The stability enhancement by themutation indi-cated that Arg233 formed destabilizing salt bridges with the acidic res-idues. Destabilizing salt bridges have been foundwidely in proteins andthe destabilization has been proposed to be caused by the relative largerdesolvation penalty than the bridge energy [61]. Thus the full

conservation of this Arg residue in basic β-crystallins was from the re-quirement of function but not stability. In vitro experiments haveshown that native acidic and basic β-crystallin can spontaneously ex-change subunits and form heteromers [62–65]. The acidic β-crystallinsare generally unstable and prone to undergo the misfolding pathway,while the basic β-crystallins can stabilize the acidic ones and assisttheir folding [24,26,28,66]. If assuming the protection effect is one ofthe functions of the basic β-crystallins, the behavior of the R233Hmuta-tion suggested that Arg233 inβB1wasmore likely to be a functional res-idue rather than a structural residue, or alternatively, a structuralresidue for the heteromer but not for the homomer. In supporting thisinference, the heteromer stability was dramatically decreased duringboth GdnHCl- and heat-induced denaturation under neutral pH condi-tions (Figs. 10 and 11). The large dissociation constant of heteromersat pH 5 suggested that electrostatic interactions might be the mainsource of heteromer formation and stability. Thus Arg233 in βB1might form inter-molecular ion-pair with the acidic residues in βA3,probably the equivalent acidic residues in the ion-pairing networkshown in Fig. 8 due to the highly conserved domain structures ofβ-crystallins. Therefore the protonation of His233 in the βB1 mutantcould partially rescue the defective functional role of bindingwith βA3, unlike its deleterious effect on βB1 stability under acidicconditions.

β-Crystallin heteromers have long been known to be energeticallyfavorable [62–65]. From the above proposal of the intrinsically dynamicion-pairing network in βB1, it could be deduced that the ion-pairingnetwork was tighter in the βA3/βB1 heteromer. However, the factthat the mutation did not affect heteromer association suggested thatArg233 contributed little to the recognition between βA3 and βB1 aswell as heteromer formation. Consistently, the R233H mutant was thesame effective as the WT βB1 in protecting βA3 aggregation duringkinetic refolding. The N-terminal extension of βB1 has been shown toplay a major role in βA3/βB1 heteromer formation as well as the self-refolding of βB1 or the co-refolding of βA3/βB1 [25,28]. Thus it seemsthat the N-terminal extension and C-terminal ion-pairing network ofβB1 played distinct roles in heteromer assembly and stabilization.That is, the N-terminal extension was crucial to the recognition processand thus failure in recognition would lead to protein misfolding andaggregation. The C-terminal ion-pairing network might be importantto the stabilization of the heteromer and consequently, impairment ofthe interactions would result in instability.

4.3. The C-terminus regulates β-crystallin aggregation

Previously it has been shown that the most destabilizing mutationsin γS-crystallin are not always the most aggregation-enhancing ones[67]. Herein we found that under neutral pH conditions, the R233Hmutant thermally aggregated much more rapidly and seriously al-though the R233H mutation had no impact on βB1 thermal stability.Meanwhile, the mutation destabilized βB1 at pH 5, but the mutanthad a lower aggregatory propensity than the WT protein. Thus theaggregatory propensity was not directly linked to protein stabilityalthough the destabilized mutations usually facilitated protein aggrega-tion under a relative mild condition. Very recently it has been shownthat the thermal aggregation of γD-crystallin involved most of the na-tive structures in the C-terminal domain but not the last β-strand [68].The pH-dependent alternations of the aggregatory propensity of βB1suggested that the charge state of His233 in the mutant modulatedthe packing of the core β-sheet structures in the aggregates. Higherhydrophobicity of the residue at position 233 resulted in largeraggregatory propensity. Similarly, it has been proposed recently thatthe protonation of His188 (equivalent to residue 230 in βB1) signifi-cantly accelerates both the amorphous aggregation and amyloid-likefibril formation of the R188H mutant of βB2 [40]. Thus it seems thatthe residues involved in the ion-pairing network, particularly those

Fig. 11. pH-dependence ofβA3/βB1 thermal stability. (A) Equilibrium thermal denaturationmonitored by ParameterA of Trpfluorescence. The rawdatawerefitted by a two-state (N↔U)model and the fitted data are shown as lines. (B) Equilibrium thermal denaturation monitored by light scattering. (C) Equilibrium thermal aggregation monitored by turbidity. (D) Themidpoint of thermal transition (Tm) obtained by fitting the data shown in panel A. (E) Thermal aggregation kinetics at 75 °C. (F) Aggregation rate constants (kagg) obtained by fittingthe data shown in panel E by Eq. (1). The other kinetic parameters are shown in Supplemental Fig. S4.

2227Y.-B. Xi et al. / Biochimica et Biophysica Acta 1842 (2014) 2216–2229

positively charged residues at the C-terminus, played a regulatory rolein β-crystallin aggregation.

4.4. Solution pH is an important factor triggering protein aggregationduring cataractogenesis

It has long been known that protein structure, function, folding andaggregation are strongly influenced by solution pH (reviewed in [69]).In addition to protein denaturation induced by extreme pH,mild chang-es of solution pH around the physiological value also influence proteinfunction and stability. The intracellular pH (pHi) of lens is neutral inthe cortex and acidic in the nucleus, while pHi can drop down to ~6when exposing to acidic solutions [70]. Lens has long been known tobe sensitive to intracellular acidification [70], and many physiologicalprocesses such as the action of gap junctions [71] are modulated bypHi. The results herein further indicated that pHi might also be an im-portant factor triggering lens protein aggregation, particularly for con-genital cataract caused by missense mutations substituted by His. Theimpact of pHi on crystallin aggregation might also play a role in muta-tions of the non-crystallin proteins that contribute to pHi homeostasis.In theory, solution pH determines the protonation status of His, whichfurther affects the surface charges, intramolecular and intermolecularinteractions. The change in surface charges alters protein solubility,the perturbation of intramolecular interactions affects protein stability,while the alternation in intermolecular interactions influences the be-havior of protein aggregation. Thus solution pHmayaffect the behaviorsof various crystallins aswell as theirmutants differentially. For example,the thermal aggregation rate constant of βB1 and βA3 exhibited anopposing pH-dependence (Figs. 6 and 11). The R233H mutation didnot affect βB1 aggregation during acidification at the physiological

temperature (Fig. 4), but the R188H mutation significantly promotesβB2 aggregation [40]. There are many cataract-linked mutations in-volved the substitutions of conserved residues by His, such as R116Hmutation in αA [72], R188H mutation in βB2 [38], and R58H mutationin γD [73]. We proposed that acidification might also play a role in thecataractogenesis caused by these mutations.

4.5. Conclusion

The molecular mechanism underlying cataract caused by inheritedmutationsmay be very complicated although cataract is directly associ-ated with protein aggregation. Although the present study was per-formed by in vitro protein assays using extreme pH and temperatureconditions, our results clearly indicated that the R233H mutation al-tered the biophysical properties. It is worth noting that the humanlens is one of the tissues with the highest protein concentration [15].The behavior of the R233H mutation might be more serious in such acrowded microenvironment since molecular crowding generally accel-erates protein association [74]. Moreover, although extreme pH valueshad been tested in this research, a significant alteration in the behaviorsof theWT andmutated βB1 and βA3/βB1was found to occur at aroundpH 6, which is a physiological/pathological-relevant pH in the lens nu-cleus [70]. Thus the strong pH dependence of the R233H mutantsmight be associatedwith the nuclear phenotype of ADCC in the patients.

Missense mutations can be stabilizing or destabilizing. The mecha-nisms of mutations at the structural residues are, to some extent, easilyelucidated since these mutations are generally destabilizing. As for thestabilizing or neutral mutations, there is a high probability that thesecorresponding residues are functional. The results in this research sug-gested that Arg233 in βB1 was a functional residue contributing to the

2228 Y.-B. Xi et al. / Biochimica et Biophysica Acta 1842 (2014) 2216–2229

stabilization of the βA3/βB1 heteromer but not the βB1 homomer.Based on the results herein we proposed that the ion-pairing networkaround Arg233 was intrinsically highly dynamic, which might be im-portant to the understanding of the tradeoff among solubility, stabilityand aggregatory propensity under themild acidic conditions in the lens.

Acknowledgments

This study was supported by funds from the Ministry of Science andTechnology of China (2010CB912402) and the State Key Laboratory ofBiomembrane and Membrane Biotechnology.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx.doi.org/10.1016/j.bbadis.2014.07.022.

References

[1] E.M. Sontag, W.I. Vonk, J. Frydman, Sorting out the trash: the spatial nature ofeukaryotic protein quality control, Curr. Opin. Cell Biol. 26 (2014) 139–146.

[2] C.M. Dobson, Protein folding and misfolding, Nature 426 (2003) 884–890.[3] M. Stefani, C.M. Dobson, Protein aggregation and aggregate toxicity: new insights

into protein folding, misfolding diseases and biological evolution, J. Mol. Med. 81(2003) 678–699.

[4] G.B. Benedek, Cataract as a protein condensation disease: the Proctor Lecture, Invest.Ophthalmol. Vis. Sci. 38 (1997) 1911–1921.

[5] K.L. Moreau, J.A. King, Protein misfolding and aggregation in cataract disease andprospects for prevention, Trends Mol. Med. 18 (2012) 273–282.

[6] F. Chiti, C.M. Dobson, Protein misfolding, functional amyloid, and human disease,Annu. Rev. Biochem. 75 (2006) 333–366.

[7] L. Robman, H. Taylor, External factors in the development of cataract, Eye (Lond) 19(2005) 1074–1082.

[8] A. Shiels, J.F. Hejtmancik, Genetics of human cataract, Clin. Genet. 84 (2013)120–127.

[9] J. Graw, The genetic and molecular basis of congenital eye defects, Nat. Rev. Genet. 4(2003) 876–888.

[10] J. Francois, Syndromes with congenital cataract, Am J. Ophthalmol. 52 (1961)207–238.

[11] S. Bassnett, Lens organelle degradation, Exp. Eye Res. 74 (2002) 1–6.[12] B.E. Klein, R. Klein, K.E. Lee, Incidence of age-related cataract over a 10-year interval:

the Beaver Dam Eye Study, Ophthalmology 109 (2002) 2052–2057.[13] C.J. Hammond, H. Snieder, T.D. Spector, C.E. Gilbert, Genetic and environmental

factors in age-related nuclear cataracts in monozygotic and dizygotic twins, N.Engl. J. Med. 342 (2000) 1786–1790.

[14] A. Shiels, T.M. Bennett, J.F. Hejtmancik, Cat-Map: putting cataract on the map, Mol.Vis. 16 (2010) 2007–2015.

[15] H. Bloemendal, W. de Jong, R. Jaenicke, N.H. Lubsen, C. Slingsby, A. Tardieu, Ageingand vision: structure, stability and function of lens crystallins, Prog. Biophys. Mol.Biol. 86 (2004) 407–485.

[16] J. Piatigorsky, Lens crystallins and their genes: diversity and tissue-specific expres-sion, FASEB J. 3 (1989) 1933–1940.

[17] U.P. Andley, Crystallins in the eye: function and pathology, Prog. Retin. Eye Res. 26(2007) 78–98.

[18] J. Graw, Genetics of crystallins: cataract and beyond, Exp. Eye Res. 88 (2009)173–189.

[19] J.S. Zigler Jr., C. Zhang, R. Grebe, G. Sehrawat, L. Hackler Jr., S. Adhya, S. Hose, D.S.McLeod, I. Bhutto, W. Barbour, G. Parthasarathy, D.J. Zack, Y. Sergeev, G.A. Lutty, J.T. Handa, D. Sinha, Mutation in the βA3/A1-crystallin gene impairs phagosomedegradation in the retinal pigmented epithelium of the rat, J. Cell Sci. 124 (2011)523–531.

[20] R. Kannan, P.G. Sreekumar, D.R. Hinton, Novel roles for α-crystallins in retinalfunction and disease, Prog. Retin. Eye Res. 31 (2012) 576–604.

[21] P. Chen, W. Ji, F.Y. Liu, H.Z. Tang, S. Fu, X. Zhang, M. Liu, L. Gong, M. Deng, W.F. Hu, X.H. Hu, X.W. Chen, Z.L. Li, X. Li, J. Liu, D.W. Li, Alpha-crystallins and tumorigenesis,Curr. Mol. Med. 12 (2012) 1164–1173.

[22] V. Prokosch, M. Schallenberg, S. Thanos, Crystallins are regulated biomarkers formonitoring topical therapy of glaucomatous optic neuropathy, PLoS One 8 (2013)e49730.

[23] J. Horwitz, α-Crystallin can function as a molecular chaperone, Proc. Natl. Acad. Sci.U. S. A. 89 (1992) 10449–10453.

[24] O.A. Bateman, R. Sarra, S.T. van Genesen, G. Kapp, N.H. Lubsen, C. Slingsby, Thestability of human acidic β-crystallin oligomers and hetero-oligomers, Exp. EyeRes. 77 (2003) 409–422.

[25] K. Srivastava, R. Gupta, J.M. Chaves, O.P. Srivastava, Truncated human βB1-crystallinshows altered structural properties and interaction with human βA3-crystallin,Biochemistry 48 (2009) 7179–7189.

[26] S. Wang, X.-Y. Leng, Y.-B. Yan, The benefits of being β-crystallin heteromers:βB1-crystallin protects βA3-crystallin against aggregation during co-refolding,Biochemistry 50 (2011) 10451–10461.

[27] M.S. Ajaz, Z. Ma, D.L. Smith, J.B. Smith, Size of human lens β-crystallin aggregatesare distinguished by N-terminal truncation of βB1, J. Biol. Chem. 272 (1997)11250–11255.

[28] X.-Y. Leng, S. Wang, N.-Q. Cao, L.-B. Qi, Y.-B. Yan, The N-terminal extension ofβB1-crystallin chaperones β-crystallin folding and cooperates with αA-crystallin,Biochemistry 53 (2014) 2464–2473.

[29] T. Alber, Mutational effects on protein stability, Annu. Rev. Biochem. 58 (1989)765–798.

[30] V.P.R. Vendra, G. Agarwal, S. Chandani, V. Talla, N. Srinivasan, D. Balasubramanian,Structural integrity of the Greek key motif in βγ-crystallins is vital for central eyelens transparency, PLoS One 8 (2013) e70336.

[31] P. Evans, K. Wyatt, G.J. Wistow, O.A. Bateman, B.A. Wallace, C. Slingsby, The P23Tcataract mutation causes loss of solubility of folded γD-crystallin, J. Mol. Biol. 343(2004) 435–444.

[32] A. Pande, K.S. Ghosh, P.R. Banerjee, J. Pande, Increase in surface hydrophobicity ofthe cataract-associated P23T mutant of human γD-crystallin is responsible for itsdramatically lower, retrograde solubility, Biochemistry 49 (2010) 6122–6129.

[33] A. Basak, O. Bateman, C. Slingsby, A. Pande, N. Asherie, O. Ogun, G.B. Benedek, J.Pande, High-resolution X-ray crystal structures of human γD crystallin (1.25 Å)and the R58H mutant (1.15 Å) associated with aculeiform cataract, J. Mol. Biol.328 (2003) 1137–1147.

[34] J.J. McManus, A. Lomakin, O. Ogun, A. Pande, M. Basan, J. Pande, G.B. Benedek,Altered phase diagram due to a single point mutation in human γD-crystallin,Proc. Natl. Acad. Sci. U. S. A. 104 (2007) 16856–16861.

[35] B. Ma, T. Sen, L. Asnaghi, M. Valapala, F. Yang, S. Hose, D.S. McLeod, Y. Lu, C. Eberhart,J.S. Zigler Jr., D. Sinha, βA3/A1-Crystallin controls anoikis-mediated cell death inastrocytes by modulating PI3K/AKT/mTOR and ERK survival pathways through thePKD/Bit1-signaling axis, Cell Death Dis. 2 (2011) e217.

[36] M. Valapala, S. Hose, C. Gongora, L. Dong, E.F. Wawrousek, J. Samuel Zigler Jr., D.Sinha, Impaired endolysosomal function disrupts Notch signalling in optic nerveastrocytes, Nat. Commun. 4 (2013) 1629.

[37] T. Liedtke, J.C. Schwamborn, U. Schroer, S. Thanos, Elongation of axons during regen-eration involves retinal crystallin βB2 (crybb2), Mol. Cell. Proteomics 6 (2007)895–907.

[38] K. Zhang, W.-J. Zhao, X.-Y. Leng, S. Wang, K. Yao, Y.-B. Yan, The importance of thelast strand at the C-terminus in βB2-crystallin stability and assembly, Biochim.Biophys. Acta 1842 (2014) 44–55.

[39] M.A. Reddy, O.A. Bateman, C. Chakarova, J. Ferris, V. Berry, E. Lomas, R. Sarra, M.A.Smith, A.T. Moore, S.S. Bhattacharya, C. Slingsby, Characterization of the G91delCRYBA1/3-crystallin protein: a cause of human inherited cataract, Hum. Mol.Genet. 13 (2004) 945–953.

[40] Y.-B. Xi, K. Zhang, A.-B. Dai, S.-R. Ji, K. Yao, Y.-B. Yan, Cataract-linkedmutation R188Hpromotes βB2-crystallin aggregation and fibrillization during acid denaturation,Biochem. Biophys. Res. Commun. 447 (2014) 244–249.

[41] B.F. Liu, J.J. Liang, Interaction and biophysical properties of human lens Q155*βB2-crystallin mutant, Mol. Vis. 11 (2005) 321–327.

[42] K.J. Wang, S. Wang, N.-Q. Cao, Y.-B. Yan, S.Q. Zhu, A novel mutation in CRYBB1 asso-ciated with congenital cataract-microcornea syndrome: the p.Ser129Arg mutationdestabilizes the βB1/βA3-crystallin heteromer but not the βB1-crystallin homomer,Hum. Mutat. 32 (2011) E2050–E2060.

[43] J. Xu, C. Wong, X. Tan, H. Jing, G. Zhou, W. Song, Decreasing the homodimerinteraction: a common mechanism shared by the ΔG91 mutation and deamidationin βA3-crystallin, Mol. Vis. 16 (2010) 438–444.

[44] J. Xu, S. Wang, W.-J. Zhao, Y.-B. Xi, Y.-B. Yan, K. Yao, The congenital cataract-linkedA2V mutation impairs tetramer formation and promotes aggregation of βB2-crystallin, PLoS One 7 (2012) e51200.

[45] S. Wang, W.-J. Zhao, H. Liu, H. Gong, Y.-B. Yan, Increasing βB1-crystallin sensitivityto proteolysis caused by the congenital cataract-microcornea syndrome mutationS129R, Biochim. Biophys. Acta 1832 (2013) 302–311.

[46] M.A. DePristo, D.M. Weinreich, D.L. Hartl, Missense meanderings in sequence space:a biophysical view of protein evolution, Nat. Rev. Genet. 6 (2005) 678–687.

[47] K.J. Wang, B.B. Wang, F. Zhang, Y. Zhao, X. Ma, S.Q. Zhu, Novel β-crystallin genemutations in Chinese families with nuclear cataracts, Arch. Ophthalmol. 129(2011) 337–343.

[48] R.L.M.v. Montfort, O.A. Bateman, N.H. Lubsen, C. Slingsby, Crystal structure oftruncated human βB1-crystallin, Protein Sci. 12 (2003) 2606–2612.

[49] M.M. Bradford, A rapid and sensitive method for the quantitation of microgramquantities of protein utilizing the principle of protein-dye binding, Anal. Biochem.72 (1976) 248–254.

[50] K.K. Turoverov, S.Y. Haitlina, G.P. Pinaev, Ultra-violet fluorescence of actin. Determi-nation of native actin content in actin preparations, FEBS Lett. 62 (1976) 4–6.

[51] G.-J. He, A. Zhang, W.-F. Liu, Y. Cheng, Y.-B. Yan, Conformational stability and multi-state unfolding of poly(A)-specific ribonuclease, FEBS J. 276 (2009) 2849–2860.

[52] B.I. Kurganov, Kinetics of protein aggregation. Quantitative estimation of thechaperone-like activity in test-systems based on suppression of protein aggregation,Biochem. Mosc. 67 (2002) 409–422.

[53] W. Humphrey, A. Dalke, K. Schulten, VMD: visual molecular dynamics, J. Mol. Graph.14 (1996) 33–38 (27-38).

[54] J.C. Phillips, R. Braun, W. Wang, J. Gumbart, E. Tajkhorshid, E. Villa, C. Chipot, R.D.Skeel, L. Kale, K. Schulten, Scalable molecular dynamics with NAMD, J. Comput.Chem. 26 (2005) 1781–1802.

[55] M. Cardamone, N.K. Puri, Spectrofluorimetric assessment of the surface hydropho-bicity of proteins, Biochem. J. 282 (Pt 2) (1992) 589–593.

[56] O.I. Povarova, I.M. Kuznetsova, K.K. Turoverov, Differences in the pathways of pro-teins unfolding induced by urea and guanidine hydrochloride: molten globulestate and aggregates, PLoS One 5 (2010) e15035.

2229Y.-B. Xi et al. / Biochimica et Biophysica Acta 1842 (2014) 2216–2229

[57] D.E. Anderson, W.J. Becktel, F.W. Dahlquist, pH-Induced denaturation of proteins: asingle salt bridge contributes 3–5 kcal/mol to the free energy of folding of T4lysozyme, Biochemistry 29 (1990) 2403–2408.

[58] C.N. Pace, The stability of globular proteins, Crit. Rev. Biochem. 3 (1975) 1–43.[59] M. Soskine, D.S. Tawfik, Mutational effects and the evolution of new protein

functions, Nat. Rev. Genet. 11 (2010) 572–582.[60] F. Chiti, C.M. Dobson, Amyloid formation by globular proteins under native condi-

tions, Nat. Chem. Biol. 5 (2009) 15–22.[61] S. Kumar, R. Nussinov, Salt bridge stability in monomeric proteins, J. Mol. Biol. 293

(1999) 1241–1255.[62] M.P. Chan, M. Dolinska, Y.V. Sergeev, P.T. Wingfield, J.F. Hejtmancik, Association

properties of βB1- and βA3-crystallins: ability to form heterotetramers, Biochemis-try 47 (2008) 11062–11069.

[63] J.F. Hejtmancik, P.T. Wingfield, C. Chambers, P. Russell, H.C. Chen, Y.V. Sergeev, J.N.Hope, Association properties of betaB2- and betaA3-crystallin: ability to formdimers, Protein Eng. 10 (1997) 1347–1352.

[64] Y.V. Sergeev, J.F. Hejtmancik, P.T.Wingfield, Energetics of domain-domain interactionsand entropy driven association of β-crystallins, Biochemistry 43 (2004) 415–424.

[65] C. Slingsby, O.A. Bateman, Quaternary interactions in eye lens β-crystallins: basicand acidic subunits of β-crystallins favor heterologous association, Biochemistry29 (1990) 6592–6599.

[66] B.F. Liu, J.J. Liang, Protein–protein interactions among human lens acidic and basicbeta-crystallins, FEBS Lett. 581 (2007) 3936–3942.

[67] W.D. Brubaker, J.A. Freites, K.J. Golchert, R.A. Shapiro, V. Morikis, D.J. Tobias, R.W.Martin, Separating instability from aggregation propensity in γS-crystallin variants,Biophys. J. 100 (2011) 498–506.

[68] S.D. Moran, T.O. Zhang, M.T. Zanni, An alternative structural isoform in amyloid-likeaggregates formed from thermally denatured human γD-crystallin, Protein Sci. 23(2014) 321–331.

[69] E.Y. Chi, S. Krishnan, T.W. Randolph, J.F. Carpenter, Physical stability of proteins inaqueous solution: mechanism and driving forces in nonnative protein aggregation,Pharm. Res. 20 (2003) 1325–1336.

[70] R.T. Mathias, G. Riquelme, J.L. Rae, Cell to cell communication and pH in the froglens, J. Gen. Physiol. 98 (1991) 1085–1103.

[71] M.S. Nielsen, L.N. Axelsen, P.L. Sorgen, V. Verma, M. Delmar, N.H. Holstein-Rathlou,Gap junctions, Compr. Physiol. 2 (2012) 1981–2035.

[72] F. Gu, W. Luo, X. Li, Z. Wang, S. Lu, M. Zhang, B. Zhao, S. Zhu, S. Feng, Y.-B. Yan, S.Huang, X. Ma, A novel mutation in AlphaA-crystallin (CRYAA) caused autosomaldominant congenital cataract in a large Chinese family, Hum. Mutat. 29 (2008)769.

[73] E. Héon, M. Priston, D.F. Schorderet, G.D. Billingsley, P.O. Girard, N. Lubsen, F.L.Munier, The γ-crystallins and human cataracts: a puzzle made clearer, Am. J.Hum. Genet. 65 (1999) 1261–1267.

[74] N.A. Chebotareva, B.I. Kurganov, N.B. Livanova, Biochemical effects of molecularcrowding, Biochemistry (Mosc) 69 (2004) 1239–1251.