Identification of Subunit-Subunit Interaction Sites in αA-WT Crystallin and Mutant αA-G98R Crystallin Using Isotope-Labeled Cross-Linker and Mass Spectrometry

Cataract is characterized by progressive protein aggregation and loss of vision. α-Crystallins are the major proteins in the lens responsible for maintaining transparency. They exist in the lens as highly polydisperse oligomers with variable numbers of subunits, and mutations in α-crystallin are associated with some forms of cataract in humans. Because the stability of proteins is dependent on optimal subunit interactions, the structural transformations and aggregation of mutant proteins that underlie cataract formation can be understood best by identifying the residue-specific inter- and intra-subunit interactions. Chemical crosslinking combined with mass spectrometry is increasingly used to provide structural insights into intra- and inter-protein interactions. We used isotope-labeled cross-linker in combination with LC-MS/MS to determine the subunit–subunit interaction sites in cataract-causing mutant αA-G98R crystallin. Peptides cross-linked by isotope-labeled (heavy and light forms) cross-linkers appear as doublets in mass spectra, thus facilitating the identification of cross-linker–containing peptides. In this study, we cross-linked wild-type (αA-WT) and mutant (αA-G98R) crystallins using the homobifunctional amine-reactive, isotope-labeled (d0 and d4) cross-linker–BS2G (bis[sulfosuccinimidyl]glutarate). Tryptic in-solution digest of cross-linked complexes generates a wide array of peptide mixtures. Cross-linked peptides were enriched using strong cation exchange (SCX) chromatography followed by both MS and MS/MS to identify the cross-linked sites. We identified a distinct intermolecular interaction site between K88 — K99 in the β5 strand of the mutant αA-G98R crystallin that is not found in wild-type αA-crystallin. This interaction could explain the conformational instability and aggregation nature of the mutant protein that results from incorrect folding and assembly.


Introduction
Crystallins (a, b and c) are the major water-soluble proteins of the lens and are responsible for its transparency. The structural interactions of crystallins and a-crystallin chaperone activity are critical to lens transparency. a-Crystallins (800-1000 kDa) constitute 40% of the lens crystallins and are made up of two subunits: aAand aB-crystallins (20 kDa each). Chaperone activity of acrystallin prevents the aggregation of lens proteins [1][2][3]. Loss of protein stability and a-crystallin chaperone activity causes changes in the lens architecture, leading to protein aggregation, increasing lens opacity and, ultimately, cataract development. The formation of high-molecular weight aggregates is the hallmark of cataract development, in both hereditary and age-related cataract.
Hereditary, or congenital, cataract results from mutations in crystallin genes. Several aA-crystallin mutants, such as R116C, R21L, R12C, R54C and R49C, are known to cause congenital cataract [4][5][6]. G98R mutation in the aA-crystallin subunit is associated with a dominant, progressive total cataract that starts in the teenage years [7]. The mutation of G to R introduces charge and bulkiness to the a-crystallin domain of the protein to which the mutation is localized and thus disrupts the immunoglobulin fold. The mutation results in misfolded and destabilized protein, with altered secondary and tertiary structure, increased oligomeric size and a propensity to aggregate [8][9][10][11].
Crystallins belong to the family of small heat shock proteins, a monomer of which is characterized by a central a-crystallin domain of ,90 amino acids, flanked by an N-terminal arm and a C-terminal extension [11,12]. a-Crystallins assemble as polydisperse hetero-oligomers comprising a variable number of aA and aB subunits, with a constant exchange of subunits. The structural assembly of a-crystallin and therefore its chaperone function are largely regulated by appropriate interactions between the neighboring subunits [12], but identifying the subunit interaction regions by structural studies has been hampered by the large mass and polydisperse nature of the protein. The available structures of crystallins are often comprised of truncated proteins or portions removed to increase the solubility or crystallization property [13]. Still largely a mystery is the molecular basis for stability of the wild-type protein and instability and aggregation of mutant proteins in congenital cataract and the interfacial interactions that contribute to the specific characteristics of the wild-type and mutant protein.
Chemical cross-linking combined with mass spectrometry offers a promising tool for studying the structural aspects of the proteins and multiprotein complexes [14][15][16]. The technique is used to provide a topological map of multi-protein complexes and interactions sites between protein subunits as well as proteinpeptide complexes. The linkage sites identified in a protein are used in structural modeling to elucidate protein conformational folds [15,17]. Isotope -labeled cross-linkers, available as a mixture of ''heavy'' and ''light'' forms, are particularly attractive in these studies. Both forms cross-link lysine's in close proximity within and between proteins. After digestion of cross-linked species, the peptides modified by ''heavy'' and ''light'' forms carry a unique isotopic signature in mass spectra, facilitating identification of the cross-linked peptides among a large excess of non-cross-linked peptides [18,19].
Cross-linkers are available with varied reactivities and spacer length. The most common cross-linkers are amine-reactive Nhydroxysuccinimide (NHS) esters. In this study, we used an isotope-labeled (d 0 and d 4 ) homobifunctional NHS ester, bis-(sulfosuccinimidyl) glutarate (BS 2 G) cross-linker, in combination with mass spectrometry to define and differentiate subunit interaction sites in native and a mutant crystallin. NHS esters form a stable amide or imide bond with the primary amines in lysine and in the N-termini of proteins, in the process releasing the NHS or sulfo-NHS group [20]. The deuterated (d 4 ) and nondeuterated (d 0 ) forms of the cross-linker were used in a 1:1 ratio, and the cross-linked peptides appear as distinct doublet with a 4.025 amu mass difference in the mass spectra [21]. Using this  approach we have identified a unique cross-linked site in the mutant aA-G98R-crystallin. This site is not found in wild-type protein. The location of the cross-linked site explains the conformational difference and therefore the abnormal interactions that might be responsible for aggregation of the mutant protein.

Materials
The cross-linking reagent Bis[Sulfosuccinimidyl] glutarate) (BS 2 G-d 0 and d 4 ) was from Proteochem, Inc (Denver, CO). Protease (Bovine) Trypsin (sequencing grade) was from G Biosciences (St Louis, MO). All other chemicals were of the highest grade commercially available.
aA-WT and aA-G98R crystallins Human recombinant aA-WT and aA-G98R crystallins were expressed and purified as described earlier [22]. Briefly, both recombinant proteins were expressed in E coli BL21(DE3) pLysS cells (Invitrogen, Carlsbad, CA). E coli extracts were lysed and centrifuged. Supernatants were processed for aA-WT crystallin purification, and insoluble pellets were processed for aA-G98R crystallin purification. The purification was achieved by gelfiltration (Superdex G-200) and anion-exchange chromatography (Q-Sepharose Fast Flow ion-exchange column). SDS-PAGE and mass spectrometry were used to determine the purity and molecular mass of the purified proteins.

Cross-linking of proteins
aA-G98R-Crystallin and aA-WT crystallin (35 mM) in 50 mM phosphate buffer (pH 7.4), 150 mM NaCl (final volume 250 ml), were incubated separately at 37uC for 30 min prior to the addition of homobifunctional cross-linker BS 2 Gd 0 /d 4 (bis[sulfosuccinimydyl]glutarate). The deuterated and non-deuterated forms of the cross-linker were prepared as 1 M stock solution in DMSO. Prior to the addition to proteins, a 1:1 ratio of deuterated and nondeuterated form was prepared. Preliminary studies were performed using 1:10, 1:20, 1:50 and 1:100 (protein to cross-linker ratio) to determine the optimal conditions that gives good crosslinking yield. To avoid nonspecific and excessive cross-linking, a 1:20 molar ratio was chosen and found to be optimal under the experimental conditions. A 20-fold molar excess of cross-linker mixture was added to aA-WT and aA-G98R protein separately. The samples were kept in ice for 1 hr. The cross-linking reaction was terminated by adding Tris (50 mM final concentration). After incubation at room temperature for 10 min, the samples were passed through 10 kDa centrifugal filters (Centricon) to remove unreacted cross-linker. Aliquots of the reaction mixture were run on 4-20% SDS-PAGE.

In-solution digestion of cross-linked proteins
In-solution digestion was carried out with a modified procedure described earlier [23]. In order to make the proteolytic sites more accessible to trypsin, the cross-linked protein samples (aA-WT and aA-G98R) were solubilized in 50 mM Tris buffer containing 8 M urea and 4 mM DTT for 2 h and then 12.5-fold diluted (final urea and DTT concentration of 0.64 M and 0.32 mM respectively) by adding trypsin digestion buffer (0.2 M ammonium bicarbonate, pH 7.9). Trypsin (1.75 mg) (sequencing grade Trypsin, G. Biosciences) was added to 175 mg of cross-linked proteins. The trypsintreated mixtures were incubated overnight at 37uC.

Enrichment of cross-linked peptides
The tryptic digests of cross-linked aA-WT and aA-G98R protein were off-line fractionated using SCX-Stage tips (Thermo Scientific), following the manufacturer's protocol. Peptides were eluted from the SCX column using a stepwise gradient of ammonium acetate. Eluted fractions were pooled and used for LC-MS analysis.

Mass spectrometry analysis
SCX-enriched peptide mixtures (1 ml) were separated on an Agilent HPLC chip (43 mm, Zorbax C18 Chip) and directly coupled to Agilent 6520 Accurate-Mass Quadrupole time-of-flight (Q-TOF) LC/MS. Elution was done using gradients with initial conditions: 3% B to 10% B over 1 min, 10% B to 40% B over 22.5 min, 40% B to 90% B over 1 min, hold at 90% B for 5 min and back to initial conditions at 3% B for 4 min, with a total run time of 35.5 min. Solvent A is 0.1% formic acid in water, solvent B is 99.9% acetonitrile and 0.1% formic acid. Flow rate was maintained at 600 nL/min. MS spectra of the eluting peptides were acquired in the range of 295-2500 m/z, 2 spectra/sec. For each cycle of MS scan (3.1 sec), the five most abundant peptides (.2500 counts) with a charge state of two, three or higher were selected and subjected to N2-induced CID peptide fragmentation (MSMS, 7-2500 m/z, 2 spectra/sec). Two internal reference mass compounds (methyl stearate 299.29 and hexasis [1H, 1H, 4Hhexafluorobutyloxy] phosphazine 1221.99) were used to recalibrate mass spectra during acquisition. Identification of cross-linked peptides LC-MS raw data were processed into deconvoluted peptide peak lists with monoisotopic mass using ''find compounds by molecular feature'' algorithm of Mass Hunter Qualitative Analysis Software B.02.00. Theoretical mass lists of the cross-linked peptides were generated using GPMAW (General Protein Mass Analysis for Windows) version 9.20 (Lighthouse Data, Odense, Denmark). Search parameters were enzyme, Trypsin; crosslinker, BS 2 G; maximum missed cleavages, two; cross-linkable amino acids, N-terminus and lysine; mass limit, 8000; and no cleavage at modified residues. The in-silico-generated mass lists were compared with the mass lists generated by mass spectrometry with a mass tolerance of 10 ppm. Potential cross-linked peptide ions were chosen for fragmentation by tandem MS/MS. Interpretation of the MS/MS spectra was done manually following de novo sequencing.

Cross-linking of proteins
To identify the subunit-subunit interaction sites in aA-WT and aA-G98R crystallins, deuterated and non-deuterated form of a homobifunctional cross-linker BS 2 G were used to cross-link the proteins (Figure 1). BS 2 G reacts with protein N-terminal a-amino and lysine e-amino groups. The analytical strategy used to characterize the intra-and inter-subunit interaction sites in aA-WT and aA-G98R crystallin is illustrated in Figure 2. To identify the initial inter-subunit interactions that initiate aggregation, we pre-incubated the aA-WT and G98R proteins at 37uC for 30 min, before adding the cross-linker. Mild thermal stress at 37uC induces the process of aggregation in the mutant G98R protein [9,10]. After incubation, cross-linker was added in 10-, 20-, 50-and 100fold molar excess to determine the optimal cross-linker to protein ratio. The reaction mixtures were kept in ice for 1 hr to suppress subunit exchange during crosslinking. After termination of crosslinking reactions and removal of unreacted cross-linker by filtration, an aliquot of the sample was boiled and run in SDS-PAGE ( Figure 3). We did not observe sufficient cross-linking in samples containing 1:10 protein to cross-linker ratio (data not shown). At proteins to crosslinker ratios of 1:20 and above, crosslinking of aA-WT crystallin was characterized by the appearance of ladder of bands corresponding to dimer, trimer, tetramer, etc. (Figure 3, Lane 5 and Figure S1A). With aA-WT, the extent of cross-linking and the gel pattern appeared the same at all protein to cross-linker ratios and at all-time points of cross-linking reaction tested. G98R crystallin, being an aggregation-prone mutant protein, formed high mass aggregates visible on top of the gel at all protein to cross-linker ratios tested (Figure 3, Lane 3 and Fig.  S1B). For further analysis, a molar ratio of 1:20 was chosen in order to minimize nonspecific cross-links.
The cross-linked samples were digested in solution with trypsin prior to reversed-phase LC separation and MS/MS analysis. Trypsin cleavage results in peptides with a basic residue at the carboxy terminus, and if the basic lysine residue is cross-linked, it results in missed cleavage [24]. Protease digestion results in complex mixtures of mono-linked, loop-linked and interlinked peptides ( Figure 2B). Among these the desired cross-linked product will be present in low abundance amidst the large excess of unwanted, unmodified peptide [19]. Strong cation exchange (SCX) chromatography has been successfully used to enrich the . Lysine and N-termini cross-links within and between subunits in proteins are shown as connecting lines. Theoretical mass lists of the cross-linked peptides are generated using GPMAW software. In GPMAW, lysines in aA-WT and aA-G98R crystallin are selected and tryptic cleavage products are generated. Since trypsin will not cleave at modified lysines (involved in cross-links), two missed cleavage sites is set as a parameter. A m/z mass difference of 4 amu is set as a criterion to create mass lists corresponding to cross-linked peptide pairs. The mass lists generated by GPMAW are compared with experimental mass lists obtained by LC-MS analysis of tryptic digests of cross-linked aA-WT and aA-G98R (Table 1). Based on the match, a list of potential cross-linked peptides is created. Sequences of the corresponding cross-linked sites in proteins are identified in-silico and represented as a map. The links in the map represent crosslink sites within and between subunits in aA-WT and aA-G98R crystallins, respectively (Table 1). doi:10.1371/journal.pone.0065610.g004 cross-linked peptides [25][26][27]. SCX chromatography is based on the charge difference between the cross-linked and non-crosslinked peptides. Cross-linked peptides have a higher charged state than non-cross-linked peptides. Thus, the cross-linked peptides can be eluted from SCX chromatography using high salt concentrations [19,26,28]. Therefore, in our experiments we enriched the targeted cross-linked peptides using SCX-Stage tips. LC-MS analysis of enriched peptide mixtures from SCX chromatography was done and the mono-isotopic deconvoluted m/z values of 500 high intensity peaks were selected for comparison with theoretical mass lists generated by GPMAW software. Lysines in the proteins were chosen as reaction sites of the cross-linker in GPMAW. The cross-linking reagent and the potential amino acids in the proteins that can be cross-linked were defined. aA-WT has seven lysines, which can potentially be modified by the cross-linker.
Peptides cross-linked with heavy forms (d 4 ) appear with a characteristic mass shift of 100.045 amu, and peptides cross-linked with light forms (d 0 ) show a mass shift of 96 amu. Thus the crosslinked peptides appear as doublet mass signals with a mass difference of 4.025 Da, 2.0123 Da and 1.341 Da for mono, double and triple protonated forms, thus facilitating identification by mass spectrometry. BS 2 G-d 0 and d 4 cross-linked ion pairs with the same retention time and a delta m/z of 4.0247 were used as filters to identify and generate the list of peptide doublets in LC-MS data after comparison with GPMAW data (Table 1). A map of linkage sites between subunits in aA-WT crystallin and aA-G98R crystallin was developed based on the match (Figure 4). In aA-WT crystallin, the major site of cross-linking is at K88, and only one linking site is seen at K99. In contrast, in mutant aA-G98R crystallin, the major interaction point shifts to K99, and only one linking site is seen at K88. This difference in linkage sites is significant as the structural arrangement and conformation of WT and mutant crystallins are different [10,29], and therefore different cross-linked products are expected. Cross-linked peptide ion pairs were subjected to MS/MS for identification of cross-linked sites. Except for one, other cross-linked peptides identified in MS were not amenable to MS/MS analysis due to low abundance and changes in their ionization potential [19]. Figure 5A shows the  Table 1. aA-WT and aA-G98R crystallin cross-links identified based on increase in mass corresponding to peptides cross-linked with deuterated and non-deuterated reagents.  Figure 5B). This interaction site is attributed to a conformational rearrangement within the mutant protein due to replacement of neutral glycine to positively charged arginine.

Discussion
G98R-mutated aA-crystallin exhibits structural and functional differences from aA-WT crystallins [9,10]. These include secondary and tertiary structural perturbations resulting in larger oligomeric size, decreased stability, altered chaperone ability and folding defects in the mutant protein, as reported earlier [9,10]. Increased aggregation propensity of the mutant protein underlies the molecular basis for the lens turbidity and cataract formation. Subunit interactions determine the self-assembly and organization of native proteins. Therefore, it is imperative that these interactions be altered in order for mutant proteins to have increased oligomeric size and aggregation propensity. [9,30,31]. We hypothesized that determining the subunit interaction sites in aA-WT crystallin and aA-G98R crystallin would provide key insights into the molecular basis for the structural conversions leading to the crystallin aggregation. We used chemical cross linking combined with mass spectrometry to identify the subunit interaction regions in aA-WT and aA-G98R crystallins. Such an approach has been successfully used to identify subunit-subunit interactions in native a-crystallin [32,33], phage proteins [34], quaternary domain interactions in Hsp90 chaperones [35], interaction sites in soluble aggregates of monoclonal antibody [36] and sHSP21 and substrate interactions [37], etc.
We report hitherto unreported inter-subunit crosslinks in acrystallin domain of aA-WT and G98R crystallins. Our crosslinking studies reveal that majority of the inter-subunit crosslinking is clustered in the K88 region in aA-WT and in the K99 region in mutant aA-G98R protein ( Figure 4). K99 in aA-WT crystallin is solvent exposed and is not proximal to any other amino group, as has been shown using DTSSP cross-linker [32,33]. In agreement with these studies, we observed no intersubunit interactions involving K99 in aA-WT, but did note intersubunit interactions in the K99 region in mutant G98R protein.
The b-sandwich assembly comprising b2-b9 strands in the acrystallin domain [38] forms an anti-parallel (AP) interface comprising the dimer, which forms the basic assembly unit for the higher order oligomerization of the wild-type protein [39][40][41]. The oligomerization involves the formation of various inter-chain interactions involving ion pairs between topologically equivalent residues [8,36]. G98R mutation in aA-crystallin introduces a charged amino acid, which could have resulted in the gain of ion pairs in the interface not seen in WT proteins. The cross-link observed in G98R mutant proteins between K88-K99 in the b5 strand could have resulted from the close proximity of the two lysines as a result of a conformational change different from the WT protein. Such new and altered interfacial interactions in G98R could have affected the subunit exchange dynamics and the structural organizations required for protein stability and lens transparency. Our findings support the earlier view that mutant proteins have increased and different interfacial interactions not found in wild-type proteins [11,29,42] and has uncovered the specific changes in a-crystallin domain. Further analysis of subunit Table 1. Cont. arrangements in proteins could help in constructing a structural model and therefore the long-range consequences of such mutations.
In conclusion, the results reveal a new, previously unknown interaction site between G98R subunits. The difference in the cross-linking pattern between the aA-WT and G98R crystallin likely reflects the different oligomerization of the proteins due to altered subunit interaction regions. Our studies demonstrate the use of chemical cross-linking and mass spectrometry as a tool for expanding our understanding of the interactions and conformational changes in mutant proteins that contribute to their aggregation.