Molecular Vision 2004; 10:857-866 <http://www.molvis.org/molvis/v10/a103/> Received 8 May 2004 | Accepted 15 November 2004 | Published 16 November 2004

Download Reprint

Probing α-crystallin structure using chemical cross-linkers and mass spectrometry

James J. Peterson,¹ Malin M. Young,² Larry J. Takemoto¹
 
 

¹Kansas State University, Department of Biology, Manhattan, KS; ²Sandia National Laboratories, Biosystems Research Department, Livermore, CA

Correspondence to: Larry J. Takemoto, PhD, Department of Biology, Rm 231 Ackert Hall, Kansas State University, Manhattan, KS, 66506; Phone: (785)532-6811; email: takemlj@ksu.edu

Abstract

Purpose: Alternatives to X-ray crystallographic techniques are needed to probe the structure of the hetero-oligomeric lens protein α-crystallin. We utilized mass spectrometry for 3 dimensional analysis (MS3D) to study the quaternary structural characteristics of this important lens protein and molecular chaperone.

Methods: We have employed two types of chemical cross-linkers to probe key protein-protein and protein-solvent interactions of α-crystallin using MS3D. Native α-crystallin was exposed to 3,3'-dithiobis[sulfosuccinimidyl propionate] (DTSSP) and the common fixative, formaldehyde. The reaction products were denatured and enriched in cross-linked and modified species using size exclusion chromatography. Tryptic digests of these fractions were purified using reverse phase HPLC and analyzed by both electrospray and matrix assisted laser desorption mass spectrometry. Comprehensive spectra for each C₁₈ fraction were screened for ions with mass unique to each chemical treatment and candidate sequences matching the experimental data were assigned using MS3D "Links" and "ASAP" software. Selected ions were sequenced by collision induced dissociation.

Results: Peptides including residues 164-175 of αB-crystallin and residues 1-99 of αA-crystallin were modified by formaldehyde and partially hydrolyzed DTSSP. Peptides containing modified lysines 11, 78, and 99 of αA-crystallin were sequenced and the modified amino acids identified. In addition, ions corresponding to intramolecular and/or intermolecular cross-links were assigned a sequence based on two criteria. First, the mass values observed were unique to a single cross-linking experiment and were not present in a control where no cross-linker was utilized. Second, two unique ions detected from different cross-linking experiments were correlated in that the structures assigned to the masses were equivalent apart from the structure of the cross-linker. One such correlation was found involving lysine₁₂₁, within the "highly conserved α-crystallin domain" of αB-crystallin, cross-linked to either lysine₁₁ or lysine₉₉ of αA-crystallin. Another two independent correlations involving lysine₇₂ of αB-crystallin were found that indicate cross-linking of two subunits of αB-crystallin through this same residue.

Conclusions: Sequences of peptides modified by partially hydrolyzed DTSSP and formaldehyde provide experimental evidence for models of α-crystallin quaternary structure that suggest a similar tertiary fold for both αA-crystallin and αB-crystallin. Analogous to multiple phosphorylations along the N-terminus of αB-crystallin, our data indicate that the same region of αA-crystallin, up to and including lysine₉₉ is also relatively accessible to modification despite its hydrophobicity. Mass correlation between experiments using different reagents suggests that cross-linking occurred between N-termini of adjacent subunits of αB-crystallin in the native complex in support of the amphiphilic, toroidal, or "open micelle" models. In addition, multiple cross-links involving lysine₁₂₁ of the so called "dimer interface" region within the "highly conserved α-crystallin domain" indicate that this region is a site of inter-subunit contacts in the native context.

Introduction

Native α-crystallin [1], by far the most abundant protein found in the lens, is a hetero-oligomeric polydisperse complex about 700-800 kDa in size [2]. The oligomer is composed of about 35-40 noncovalent subunits in a 3:1 mixture of two highly homologous 20 kDa proteins, αA-crystallin and αB-crystallin, respectively. Tertiary structure is thought to be similar to that of other small heat shock proteins (sHSPs) where the crystal structure has been solved [3,4]. For example, the "β sandwich fold" of the region collectively known as the "highly conserved α-crystallin domain" is thought to be present in all sHSPs [1].

The need for alternative techniques to explore the tertiary and dynamic quaternary structure of α-crystallin is acute. Once thought to have only structural and optical functions in the lens, αB-crystallin is a molecular chaperone [5] and has been found in many other tissues [6,7]. Moreover, the role of αB-crystallin in several neurological pathologies in addition to cataract, including Alzheimer's disease [8], and recent findings that crystallins undergo fibril formation characteristic of amyloid pathology [9,10] argue for the accelerated development and application of new techniques to study both native and disease states of this important protein.

The polydisperse nature of α-crystallin is thought to have been the main obstacle to acquisition of high resolution X-ray crystallograghic data to determine its structure [1]. Over many years several models for the tertiary and quaternary structure of α-crystallin have been proposed using a variety of analytical techniques [11-18]. Many sought to reconcile the amphiphilic and hetero-oligomeric nature of α-crystallin with both its optical and molecular chaperone properties. Wistow et al. [13] suggested that in order to interact with many other proteins (e.g., as a chaperone), α quaternary structure may be highly adaptable and dynamic. Indeed, Vanhoudt et al. [2] and Abgar et al. [17] suggest that in terms of the size of the chaperone complex, a dramatic reorganization occurs after aggregation of the chaperoned target protein is prevented. Cryoelectron microscopic (CEM) studies by Haley et al. [18] suggested that total α-crystallin structure is roughly spherical, but relatively unsymmetric and variable in comparison to other sHSPs where higher symmetry was clearly demonstrated using CEM. The data were presented with deference to spin labeling data that provide strong evidence for at least one "dimer interface" between recombinant subunits involving the "highly conserved α-crystallin domain" [19-21]. This region of conserved sequence homology among sHSPs was first reported by Ingolia et al. [22] by comparing mammalian α-crystallin to Drosophila sHSPs. Haley et al. [18] hypothesized that the range of structural variation among N- and C-terminal extensions flanking the conserved domain of different sHSPs results in a corresponding range of order and symmetry of quaternary structure among sHSPs. Clearly, the question of α-crystallin quaternary and tertiary structure is still quite unsettled and lends itself to alternative experimental approaches.

Recent advances in mass spectrometers have encouraged development of Mass spectrometry for 3 dimensional analysis (MS3D) to probe protein structure [23-26]. In this technique, native proteins or protein complexes are modified with and/or without cross-linking in solution, enriched in modified species, and cleaved by chemical or enzymatic means into constituent peptides. The mass of these peptides are then determined. New ions arising from chemically modified protein are indicative of peptides that are either solvent exposed and modified without cross-linking (i.e., decorated) or in proximity to another modifiable residue from the same or other interacting protein and cross-linked. Here we used ester and formaldehyde cross-linkers in a study of α-crystallin using MS3D and its associated software to probe the solvent accessible regions, and interaction sites of total native bovine α-crystallin.

Methods

Isolation of α-crystallin

Two bovine fetal lenses were homogenized in 4 ml of chilled size exclusion buffer (0.1 M Na₂SO₄ and 60 mM sodium phosphate [pH 7.0]) in a glass dounce homogenizer, and centrifuged at 14,000 RPM on an Eppendorf 5402 centrifuge for 30 min at 4 °C. The supernatant was stored on ice, and 100 μl aliquots were injected into a Tosoh Biosep TSK3000 HPLC size exclusion column operating at 22 °C at a flow rate of 1 ml/min and monitoring absorbance at 280 nm. Total native oligomeric α-crystallin complex was taken from the leading edge of the first fraction to elute, from 5.5-6 min as previously reported [27]. The α-crystallin oligomeric complexes prepared in this manner have been previously characterized [28], and were similar in size to those reported by other investigators [29]. This fraction was dialyzed for 12 h against six changes of PBS (0.15 M NaCl and 25 mM sodium phosphate [pH 7]) at 4 °C. Protein concentrations were measured according to the Bradford method using bovine serum albumin as the standard [30]. Consistent with previous findings [31], analysis of the α-crystallin preparation by reverse phase chromatography under denaturing conditions demonstrated that the molar ratio of αA:αB-crystallin subunits was approximately 3:1 (results not shown).

Cross-linking reactions

Preliminary experiments using 5:1, 10:1, and 200:1 M excess of 3,3'-dithiobis[sulfosuccinimidyl propionate] (DTSSP; Pierce Chemical, Rockford, IL) cross-linker to α-crystallin were performed to determine optimal conditions for cross-linking (data not shown). A molar ratio of 5:1 was selected in order to minimize nonspecific cross-links at the expense of cross-link yield. Similarly, a ratio of 5,700:1 formaldehyde to α-crystallin was chosen in order to achieve the same approximate extent of cross-linking at the same temperature and duration of reaction as judged by SDS-PAGE. Immediately prior to cross-linking, protein was diluted to 6.4 μM (160 nM oligomeric complex) in 0.15 M NaCl and 25 mM sodium phosphate (pH 7.7). Protein (25 mg; 1.25 μmol) was reacted with DTSSP and 6 mg protein (300 nmol) was reacted with formaldehyde. A stock 20 mM DTSSP solution was prepared on ice in 0.15 M NaCl, 25 mM sodium phosphate (pH 6.0) immediately before reaction with protein (<60 s). A five fold excess of DTSSP was added. Alternatively, a 5,700 fold excess of formaldehyde was added directly from a 37% solution (Fisher catalog number F79-500). Finally, a control reaction was carried out with 6 mg of protein where buffer containing no cross-linker was added and all subsequent purification and analytical steps were applied to this control in the same way as was done for each cross-linking reaction. Reactions were shaken in an incubator at 25 °C for 1 h. After 1 h, reactions were quenched on ice using 2 ml or 8 ml of 1 M TRIS (pH 7.7). After 30 min on ice a small aliquot of each completed and quenched reaction was analyzed by 10% SDS-PAGE under both reducing and non-reducing conditions (with and without β-mercaptoethanol) and the mixtures were dialyzed against 5 changes of water (1 l/change) at 4 °C.

Enrichment of cross-linked species

Cross-linked and dialyzed protein was evaporated to about 100 μl and dissolved in 6 M guanidinium hydrochloride (Gn-HCl; Sigma, St Louis, MO), 25 mM sodium phosphate (pH 7.0) to a concentration of about 7 mg/ml. These solutions were fractionated by size exclusion HPLC as above except using 5 M Gn-HCl/25 mM sodium phosphate as the running buffer. Four fractions bounded by retention times of 3.8, 4.5, 6.5, 8.0, and 9.5 min were collected, dialyzed against 2 changes of water (4 l/change), and concentrated to a volume of 250-400 μl. A small aliquot (10 μl) of each of these concentrated samples was mixed with 10 μl 2X SDS-PAGE non-reducing sample application buffer and analyzed by SDS-PAGE.

Trypsin digestions

The concentrated size exclusion fractions collected from 4.5-6.5 min were dissolved by careful application of a dense layer of 6 M Gn-HCl to the bottom of the reaction tube (to dissolve any solid present), then diluted with water, and 1 M (NH₄)₂CO₃ (pH 8.5) to final concentrations of 0.1 M (NH₄)₂CO₃, 0.1 M Gn-HCl, and 0.5-1.3 mg/mL protein. The solution was warmed to 37 °C and trypsin was added at a ratio of 20:1 substrate:trypsin. Digestions were carried out for 18 h in an orbital shaker/incubator at 100 RPM. Digestion was monitored at 215 nm by analytical scale C₁₈ HPLC using a Vydac (St Louis, MO) 4.6x250 mm 300 Å pore size 5 μm particle size reverse phase column fitted with a C₁₈ cartridge system from Phenomenex (Torrance, CA). The following gradient was utilized for both digestion progress and separation of tryptic peptides: Solvent A: 0.1% (v/v) trifluoroacetic acid (TFA) in water; Solvent B: 0.083% (v/v) TFA in acetonitrile. After injection (t=0), a mixture of 5% solvent B and 95% solvent A was held until t=5 min. Solvent B was then ramped linearly up to 63.3% at t=40 min. Digestion was judged complete after 18 h at 37 °C by monitoring the conversion of undigested protein eluting at 37 min to tryptic peptides eluting from 5 to 41 min by HPLC. After digestion, peptides were separated using the same gradient and column and 1.5 min fractions were collected, pooled, and lyophilized to dryness.

Mass spectrometry

Dried fractions from C₁₈ HPLC were dissolved in 1% (v/v) formic acid/49% (v/v) water/50% (v/v) acetonitrile and lyophilized again. Final samples were dissolved in 25-60 μl of 0.1% (v/v) formic acid/49.9% (v/v) water/50% (v/v) acetonitrile and injected into the source of a Bruker Esquire 3000 mass spectrometer using a 20 μl injection loop in line with a Cole-Parmer 74900 series injection pump fitted with a 100 μl Hamilton syringe and operating at a flow rate of 2 μl/min. Nebulizer gas pressure and source temperature were 10 PSI and 325 °C, respectively. The scan range was set to scan mass to charge ratios (m/z) from 50-3,000 m/z and scan resolution to 13,000 m/z per second. Mass spectra were compiled and saved for each tryptic digestion fraction for each reaction condition. After construction of the native and crosslink assignment map (see below) selected ions were submitted to the University of Massachusetts Proteomic Facility for sequencing by collision induced dissociation (CID) using a matrix assisted laser desorption quadrupole ion trap (MALDI-QIT) mass spectrometer.

Construction of the native, modified, and cross-linked peptide map

For every ion detected above an arbitrary intensity unit (1x10⁴), native, modified, and/or cross-linked peptide sequences were assigned using "Links" software (version 1.0) from Sandia National Laboratories [23]. Given the mass, charge state of the ion, information about the cross-linker, the cleavage enzyme or reagent used, and amino acid modifications, "Links" calculates the theoretical structure for a given mass. Modifications without cross-linking were easily assigned for DTSSP using existing software. In the case of formaldehyde, older "Automated Sequence Assignment Program" (ASAP) software (version 1.0) from the same web site was used. Since ASAP only considers intramolecular cross-links, a de novo protein with αA-crystallin fused to the C-terminus of αB-crystallin (αB-crystallin has two N-terminal lysines) was entered into the protein field. To search for crosslinks only, experimental mass and charge were entered into the input field of ASAP. However, the current version of ASAP does not search for modifications without crosslinking (i.e., "decorations") where the peptide has been chemically modified by formaldehyde (e.g., as a Schiff base [Δm=12] or hydroxylmethyl group [Δm=30]). In addition, to simultaneously detect possible decorations with possible cross-links, formaldehye decorations (Schiff base and hydroxymethyl) and combinations of these decorations (e.g., Δm=12, one Schiff base; 24, two Schiff bases; 30, one hydroxylmethyl; 42, one Schiff base and one hydroxymethyl) were subtracted from the experimental mass data using Excel before running the program. In this way, ions reported by the software to be native tryptic peptides could then be assigned the subtracted modification, and ions reported by the program to be cross-linked could then be assigned the designation "cross-linked and decorated". Since this software does not yet accommodate the relatively relaxed cross-linking specificity of formaldehyde where non-lysine residues (e.g., Arg, His, Tyr, Trp) can contribute to the cross-link [32], reactions resulting from formaldehyde treatment were only assigned if each component of the cross-link contained at least one eligible lysine (e.g., lysines N-terminal to the site of cleavage by trypsin must be unmodified).

For all assignments an error window of 200 ppm was entered into the software query and the following amino acid modifications were considered; (1) pyroglutamic acid modifications of N-terminal glutamines, (2) deamidations of asparagine and glutamine residues, (3) oxidation of methionine, and (4) acetylation of initiating methionines for both αA-crystallin and αB-crystallin. All plausible assignments for modifications and cross-links for each reagent were tabulated in Excel, including the retention time analyzed, experimental mass, charge, and theoretical mass of the assignment. Correlations between different cross-linker experiments were tabulated using the Excel finder function under the edit menu to find all instances of a given sequence assigned.

Results

Chemical cross-linking and enrichment of modified products

Cross-linking reaction conditions were optimized to result in detectible cross-linking of α-crystallin for both cross-linkers as shown by reducing and nonreducing SDS-PAGE (Figure 1). Protein bands corresponding to unmodified or decorated monomer were by far the major products. Under reducing conditions, oligomers resulting from cross-linking using DTSSP were cleaved into constituent monomer components as expected. Under these standard reducing conditions, formaldehyde cross-linking gave three relatively well resolved bands corresponding to cross-linking combinations resulting from two proteins (αA-crystallin and αB-crystallin) with slightly different mobility on SDS-PAGE. Two polypeptides with mobility approximately corresponding to the 20 kDa marker were resolved for all experiments using these conditions when less protein was loaded on the gel (results not shown). The use of the chaotropic agent, Gn-HCl, and size exclusion HPLC was effective in eliminating much of the unmodified native protein from digestion and subsequent analysis as shown by SDS-PAGE analysis of the fractions collected (Figure 2).

Analysis of tryptic peptides by mass spectrometry

Complete digestion was monitored by HPLC (Figure 3). After 18 h at 37 °C, the peak with retention time of 37 min co-eluting with starting material could not be diminished by further addition of trypsin. Peptides with mass corresponding to sequences αA_22-49 and αB_1-11 were detected in this fraction by both ESI and MALDI-TOF mass spectrometry, but no remaining protein was detected (data not shown). For the remaining digestion products, ions corresponding to the mass of every possible fully digested tryptic peptide were readily detected for both αA-crystallin and αB-crystallin within 200 ppm error. For αA-crystallin, eight different peptides containing one missed cleavage were detectable during the entire course of analysis (all cross-linking and control reactions), and three peptides containing two missed cleavages. For αB-crystallin, seven peptides containing one missed cleavage and one peptide containing two missed cleavages were detectable. Several amino acid modifications not related to cross-linking were detected. Specifically, masses corresponding to N-terminal glutamines converted to pyroglutamic acid, deamidated glutamines and asparagines, and oxidized methionines were detected.

Assignment of sequences corresponding to experimental ion mass

Several ions corresponding to chemical modification without crosslinking (i.e., decorated) or modification with cross-linking were assigned sequences using MS3D software, and those confirmed by CID sequencing (Figure 4) and/or mass correlation (i.e., where the differences in the masses of the ions were accounted for by the differences in the masses of the cross-linkers) are summarized in Table 1. Lysine₁₁ was correlated in that it was decorated by DTSSP, and formaldehyde. Among these N-terminal peptides from αA-crystallin, in both cases where an amide side chain (Asn or Gln) was present the deamidated species was also observed at a different retention time and sequenced separately by CID (data not shown).

Several ions detected by mass spectrometry were found to be correlated; the assigned structures from different cross-linking experiments were identical except for the cross-linking moiety. Data summarized in Table 1 suggests that the C-terminus of αB-crystallin was extensively modified at one or more of the lysines αB-Lys₁₆₆, αB-Lys₁₇₄, or αB-Lys₁₇₅. In contrast, semihydrolized adducts with masses corresponding to decoration of internal lysines αA₁₄₅, αB₈₂, αB₉₀, αB₉₂, αB₁₀₃, αB₁₂₁, and αB₁₅₀ of the conserved α-crystallin domain were not detected for DTSSP within 200 ppm. The results suggest that these lysine residues are located in regions of αA-crystallin and αB-crystallin molecules that are not solvent exposed.

A number of cross-links were assigned and correlated (Table 1, Figure 5). A correlation for intrapeptide cross-linking of the C-terminus of αB-crystallin was detected for both DTSSP and a combination of carboxymethylation and methylene bridge cross-linking by formaldehyde (Δm=42). In addition to this predictable crosslink, there appeared to be homodimer cross-linking of lysine αB-Lys₇₂ between different subunits of αB-crystallin. A correlation between DTSSP and formaldehyde occurred for this crosslink in two instances where the outcome of enzymatic cleavage can result in two different structures. An intrasubunit crosslink between αB-Lys₁₅₀ and αB-Lys₁₆₆ was indicated by correlation. Similarly, correlation was found for heterodimer cross-linking of αB-Lys₁₂₁ to αA-Lys₉₉. In addition, a parent ion was partially sequenced (loss of acetylated Met and Arg) for the ion assigned to αB_117-123 cross-linked to αA-Lys₁₁. The mass of the parent ion is about 1 Da heavier than theoretical as expected for a deamidated peptide containing Gln₆ to Glu₆. This particular Glu also belongs to a decorated peptide (αA-crystallin_1-12) where both native and deamidated peptides were isolated at different retention times and sequenced separately by CID.

Comparison of standard diester and formaldehyde cross-linking

The extent of cross-linking was comparable for all reagents as estimated by SDS-PAGE (Figure 1). However, a total of 1,252 ions were detectable as a result of reaction with formaldehyde followed by trypsin digestion compared to 498 ions from reaction with DTSSP using the same minimal signal to noise ratio to distinguish peaks from noise. Sequence assignments corresponding to modification and cross-linking sites for formaldehyde were frequently distinct from those of DTSSP, as expected from the broader range of amino acids modifiable by formaldehyde [32]. This trend was reflected in a number of unique masses assignable only to peptides modified and/or cross-linked by formaldehyde (data not shown). In contrast, in most cases involving cross-linking or decoration with DTSSP, a corresponding crosslink or decoration was also assignable with at least one possible mass shift attributable to modification or cross-linking by formaldehyde.

Discussion

We have taken advantage of vast improvements in the resolution, accuracy, and sensitivity of mass spectrometers and the computer based software evolving with this modern instrumentation to renew investigation of α-crystallin structure using chemical modifications. Bi- and multifunctional reagents are ideal for this purpose because of the wealth of information such modifications bring. For example, a single hydrophilic reagent such as DTSSP can be used both to "decorate" residues exposed to a hydrophilic environment and to cross-link residues involved in nearest neighbor interactions. Four equally important structural clues arise directly from the identification of these modifications. First, lack of any modification using the hydrophilic DTSSP and the much smaller geminal diol, formaldehyde, suggests the region is completely buried and/or inaccessible. Second, modification exclusively by formaldehyde indicates intermediate solvent accessibility. Third, decoration by both reagents to the exclusion of cross-linking indicates the region is highly exposed and solvent accessible. Finally, cross-linking indicates the presence of specific intermolecular subunit interactions taking place.

Cross-linking conditions for this experiment were chosen to obtain a minimal detectable amount of cross-linking in order to avoid the possibility of unnatural structural perturbation caused by the reaction. It was expected that the majority of the tryptic peptides modified by cross-linkers that contain two hydrolysable functional groups would be decorated rather than cross-linked. This is, in fact, what was found.

The first significant finding of our analyses concerns the high solvent exposure of the N-terminus of αA-crystallin. Our mass sequencing data shows that the N-terminus of αA-crystallin (about residues 1-99) was extensively decorated despite the relative hydrophobicity of this domain. The correlation data from Table 1 also indicates that the C-terminal tail of αB-crystallin (about residues 164-175) was both decorated and internally cross-linked. These predictable modifications serve as a positive control since the C-terminal tail of αB-crystallin is generally considered to be highly soluble, flexible, and accessible. In addition, the C-terminus contains three lysines in close proximity in the primary sequence (Figure 5). The data in Table 1 suggests that both decoration by DTSSP and a combination of Schiff base and hydroxymethyl modes of formaldehyde modification were present. At the same time, masses corresponding to peptides with DTSSP decorations of lysines_100-163 of either protein were not observed. Often, possible formaldehyde decorations were observed in this region but no correlation was possible because the tentative sequence assignments did not contain at least one lysine per chain (data not shown).

How do our data showing high solvent exposure of the N-terminus of αA-crystallin relate to current hypotheses concerning α-crystallin structure and function? Chemical modifications of α-crystallin have been summarized by Groenen et al. [33]. The primary sequences of αA-crystallin and αB-crystallin are almost 60% homologous, and may have similar 3 dimensional structures [34]. They are also thought to occupy equivalent positions in the native complex [35,36]. If the three dimensional structures are indeed similar it is not likely that the N-terminus of either subunit is buried. In vivo and/or in vitro phosphorylation has been observed for αB-Ser₁₉, αB-Ser₄₃, and αB-Ser₄₅ [33,37]. A model for crystallin quaternary structure analogous to that of other chaperone proteins has been proposed [14] that considers both the accessibility and hydrophobicity of the N-terminal domain (about residues 1-70) [38] in the context of chaperone activity. This "cylindrical bilayer" model is similar to the 'open micelle' model proposed by Groth-Vasselli et al. [39], but accommodates protein interaction about the outside of the toroidal bilayer where hydrophobic N-terminal residues would necessarily be exposed to water. Both models consider the hollow region indicated by the CEM data but this data also suggests greater variability and relative lack of symmetry in the native complex than is hypothesized [18]. By "opening" the model to accommodate accessibility it was implied that unfolded proteins can be accommodated by the model without some mode of dynamic adaptation. The situation may be more complex than theorized. Recently, Vanhoudt et al. [2] have demonstrated a dramatic reorganization of α-crystallin peptide subunits and increase of the size of the complex during chaperone activity.

The second form of information resulting from this study concerned a number of cross-links that we assigned due to mass correlation and a partial sequence. For example, masses corresponding to intrapeptide cross-linking of the C-terminus of αB-crystallin were both expected and observed for both types of cross-linkers in addition to decoration. Homodimer cross-linking involving αA-Lys₇₂ may indicate a parallel mode of β-sheet interface involving a self complimentary motif of the type suggested by Farnsworth et al. [40]. The cross-link between αB-Lys₁₅₀ and αB-Lys₁₆₆ is relatively ambiguous in terms of structural information in the native context, since it is impossible to discern if it is the result of intra- or intermolecular cross-linking. The peptide common to both heterodimer cross-links, αB_117-123 (EFHRKYR), is a tryptic fragment belonging to a region of αB-crystallin that is nearly identical to an anti-parallel β-sheet 'dimer interface', αA_109-120 (ISREFHRRYRLISRE), discovered by Berengian et al. [19-21]. These authors used spin labeling techniques to study α-crystallin protein interactions in a non-native context, on full length and truncated recombinant homodimer constructs of the conserved α-crystallin domain. It is possible that this subunit interface may also be utilized as an intersubunit interface in the native context. If αA-crystallin and αB-crystallin do indeed have similar 3 dimensional structures, one would expect an analogous crosslink at or near position 121 of αA-crystallin, except that in the wild type bovine αA-crystallin homolog of our study an arginine replaces lysine at position 121 (Figure 5). The closest lysine in the primary sequence of the αA-crystallin chain is αA-Lys₉₉, which is one of the residues we are proposing to be cross-linked to αB-Lys₁₂₁.

To corroborate our findings regarding protein interactions of the α-crystallins, we searched the literature for alternative techniques and conducted experiments using a different technology. In a separate study, we investigated synthetic peptides of αA-crystallin and αB-crystallin that interact with α-crystallin using a novel surface plasmon resonance technique developed in this laboratory. Among peptides that interacted with α-crystallin, αB_110-123, which contains αB-Lys₁₂₁, and N-terminal peptides αA_22-49, and αA_41-56, which contain no lysines, were comparable and by far the strongest observed (unpublished results). A recent comprehensive survey of synthetic human αB-crystallin peptides that bind to both αA-crystallin and αB-crystallin has been conducted using peptide pins in ELISA format. Human peptides αB_41-58, αB_73-82, and αB_131-142 were found to bind most strongly to human αA-crystallin and αB-crystallin [41]. Other than N-terminal and conserved domain peptides in general, there appears to be little consensus for any one particular locus of peptide interaction at this point. We suppose that a higher resolution probe may be helpful.

We have provided evidence through correlation that analogous formaldehyde cross-links and decorations are likely to be found which are similar to that of the proven conventional cross-linker, DTSSP. The converse was far less frequently observed; some tentative sequence assignments (e.g., in the conserved region) were unique to formaldehyde modification. The current resolution of MS3D as a structural tool is limited by the length (about 24.5 Å, or 6 amino acids) of ester reagents typically used in these studies [26]. The DTSSP cross-linking reagent used in the present study contains succinimidylester groups separated by a spacer arm of approximately 12 Å. Moreover, the relative importance of any given protein interaction is limited by the exclusive specificity for primary amines. Three obvious advantages of formaldehyde are; (1) it is the smallest cross-linker, (2) it exhibits relaxed specificity, and thus, can probe a greater number of important interactions, and (3) it is a common, inexpensive biological fixative that can perfuse rapidly into cells and tissues and as such, represents a high resolution alternative to yeast-2-hybrid studies of cellular protein interactions in vivo. The chief disadvantages in the present study proved to be; (1) the larger number and types of cross-links and decorations created, and (2) the lack of software that considers a more complex ensemble of modifications and cross-links. Such software will need to consider a greater, but finite number of cross-links and decorations. Based upon the results of Table 1, formaldehyde with Δm=24, was able to crosslink Lys₁₂₁ of αB-crystallin with Lys₁₁ of αA-crystallin, suggesting a very short distance no greater than approximately 2 Å between these two residues.

Using MS3D, we provide the first evidence that the hydrophobic N-terminal domain of αA-crystallin is highly accessible in the native complex. Correlation data from different cross-linkers supports a model where the N-termini and "highly conserved α-crystallin domain" of both α-crystallin chains participate in protein interactions of a homo-oligomeric and hetero-oligomeric nature.These sequences from both chains contain different regions that contain high numbers of either hydrophobic and hydrophilic residues, so it is impossible at this time to speculate whether the nature of the interaction(s) between monomeric subunits involves primarily hydrophobic or charge-charge interactions [42]. Nor is it possible at this time to identify the exact boundaries of the protein interfaces involved in α/α interactions, and whether significant changes in conformation occur upon subunit interactions involving large sites of protein-protein recognition [43]. However, the results of the present study have suggested regions of the α-crystallin monomers that are involved in subunit-subunit interactions, and it is anticipated that continued development of the MS3D technology will lead to more specific identification of α-crystallin regions involved in these important subunit interactions.

Acknowledgements

We would like to thank Sandia National Laboratories for technical assistance with assignment of structures and operation of sequence assignment software. We greatly appreciate helpful discussions from Dr. John Tomich, Dr. Takeo Iwamoto, and Dr. John Leszyk, regarding mass spectrometry instrumentation and the excellent technical and logistical support of Eric McConkey and Kyra Albin. This research was supported by a grant (RO1 EY02932) to LT from the National Eye Institute.

References

1. Horwitz J. Alpha-crystallin. Exp Eye Res 2003; 76:145-53.

2. Vanhoudt J, Abgar S, Aerts T, Clauwaert J. Native quaternary structure of bovine alpha-crystallin. Biochemistry 2000; 39:4483-92.

3. Kim KK, Kim R, Kim SH. Crystal structure of a small heat-shock protein. Nature 1998; 394:595-9.

4. van Montfort RL, Basha E, Friedrich KL, Slingsby C, Vierling E. Crystal structure and assembly of a eukaryotic small heat shock protein. Nat Struct Biol 2001; 8:1025-30.

5. Horwitz J. Alpha-crystallin can function as a molecular chaperone. Proc Natl Acad Sci U S A 1992; 89:10449-53.

6. Bhat SP, Nagineni CN. alpha B subunit of lens-specific protein alpha-crystallin is present in other ocular and non-ocular tissues. Biochem Biophys Res Commun 1989; 158:319-25.

7. Head MW, Hurwitz L, Goldman JE. Transcription regulation of alpha B-crystallin in astrocytes: analysis of HSF and AP1 activation by different types of physiological stress. J Cell Sci 1996; 109 (Pt 5):1029-39.

8. Lowe J, McDermott H, Pike I, Spendlove I, Landon M, Mayer RJ. alpha B crystallin expression in non-lenticular tissues and selective presence in ubiquitinated inclusion bodies in human disease. J Pathol 1992; 166:61-8.

9. Meehan S, Berry Y, Luisi B, Dobson CM, Carver JA, MacPhee CE. Amyloid fibril formation by lens crystallin proteins and its implications for cataract formation. J Biol Chem 2004; 279:3413-9.

10. Goldstein LE, Muffat JA, Cherny RA, Moir RD, Ericsson MH, Huang X, Mavros C, Coccia JA, Faget KY, Fitch KA, Masters CL, Tanzi RE, Chylack LT Jr, Bush AI. Cytosolic beta-amyloid deposition and supranuclear cataracts in lenses from people with Alzheimer's disease. Lancet 2003; 361:1258-65.

11. Tardieu A, Laporte D, Licinio P, Krop B, Delaye M. Calf lens alpha-crystallin quaternary structure. A three-layer tetrahedral model. J Mol Biol 1986; 192:711-24.

12. Augusteyn RC, Koretz JF. A possible structure for alpha-crystallin. FEBS Lett 1987; 222:1-5.

13. Wistow G. Possible tetramer-based quaternary structure for alpha-crystallins and small heat shock proteins. Exp Eye Res 1993; 56:729-32.

14. Carver JA, Aquilina JA, Truscott RJ. A possible chaperone-like quaternary structure for alpha-crystallin. Exp Eye Res 1994; 59:231-4.

15. Farnsworth PN, Frauwirth H, Groth-Vasselli B, Singh K. Refinement of 3D structure of bovine lens alpha A-crystallin. Int J Biol Macromol 1998; 22:175-85.

16. Smulders RH, van Boekel MA, de Jong WW. Mutations and modifications support a 'pitted-flexiball' model for alpha-crystallin. Int J Biol Macromol 1998; 22:187-96.

17.Abgar S, Vanhoudt J, Aerts T, Clauwaert J. Study of the chaperoning mechanism of bovine lens alpha-crystallin, a member of the alpha-small heat shock superfamily. Biophys J 2001; 80:1986-95.

18. Haley DA, Bova MP, Huang QL, Mchaourab HS, Stewart PL. Small heat-shock protein structures reveal a continuum from symmetric to variable assemblies. J Mol Biol 2000; 298:261-72.

19. Berengian AR, Bova MP, Mchaourab HS. Structure and function of the conserved domain in alphaA-crystallin. Site-directed spin labeling identifies a beta-strand located near a subunit interface. Biochemistry 1997; 36:9951-7.

20. Mchaourab HS, Berengian AR, Koteiche HA. Site-directed spin-labeling study of the structure and subunit interactions along a conserved sequence in the alpha-crystallin domain of heat-shock protein 27. Evidence of a conserved subunit interface. Biochemistry 1997; 36:14627-34.

21. Berengian AR, Parfenova M, Mchaourab HS. Site-directed spin labeling study of subunit interactions in the alpha-crystallin domain of small heat-shock proteins. Comparison of the oligomer symmetry in alphaA-crystallin, HSP 27, and HSP 16.3. J Biol Chem 1999; 274:6305-14.

22. Ingolia TD, Craig EA. Four small Drosophila heat shock proteins are related to each other and to mammalian alpha-crystallin. Proc Natl Acad Sci U S A 1982; 79:2360-4.

23. Young MM, Tang N, Hempel JC, Oshiro CM, Taylor EW, Kuntz ID, Gibson BW, Dollinger G. High throughput protein fold identification by using experimental constraints derived from intramolecular cross-links and mass spectrometry. Proc Natl Acad Sci U S A 2000; 97:5802-6.

24. Rappsilber J, Siniossoglou S, Hurt EC, Mann M. A generic strategy to analyze the spatial organization of multi-protein complexes by cross-linking and mass spectrometry. Anal Chem 2000; 72:267-75.

25. Bennett KL, Kussmann M, Bjork P, Godzwon M, Mikkelsen M, Sorensen P, Roepstorff P. Chemical cross-linking with thiol-cleavable reagents combined with differential mass spectrometric peptide mapping--a novel approach to assess intermolecular protein contacts. Protein Sci 2000; 9:1503-18.

26. Schilling B, Row RH, Gibson BW, Guo X, Young MM. MS2Assign, automated assignment and nomenclature of tandem mass spectra of chemically crosslinked peptides. J Am Soc Mass Spectrom 2003; 14:834-50.

27. McFall-Ngai MJ, Ding LL, Takemoto LJ, Horwitz J. Spatial and temporal mapping of the age-related changes in human lens crystallins. Exp Eye Res 1985; 41:745-58.

28. Boyle D, Gopalakrishnan S, Takemoto L. Localization of the chaperone binding site. Biochem Biophys Res Commun 1993; 192:1147-54.

29. Koretz JF, Augusteyn RC. Electron microscopy of native and reconstituted alpha crystallin aggregates. Curr Eye Res 1988; 7:25-30.

30. Bradford MM. A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal Biochem 1976; 72:248-54.

31. Delcour J, Papaconstantinou J. A change in alpha-crystallin subunit composition in relation to cellular differentiation in adult bovine lens. Biochem Biophys Res Commun 1970; 41:401-6.

32. Metz B, Kersten GF, Hoogerhout P, Brugghe HF, Timmermans HA, de Jong A, Meiring H, ten Hove J, Hennink WE, Crommelin DJ, Jiskoot W. Identification of formaldehyde-induced modifications in proteins: reactions with model peptides. J Biol Chem 2004; 279:6235-43.

33. Groenen PJ, Merck KB, de Jong WW, Bloemendal H. Structure and modifications of the junior chaperone alpha-crystallin. From lens transparency to molecular pathology. Eur J Biochem 1994; 225:1-19.

34. Augusteyn RC, Putilina T, Seifert R. Quenching of tryptophan fluorescence in bovine lens proteins by acrylamide and iodide. Curr Eye Res 1988; 7:237-45.

35. Hendriks W, Weetink H, Voorter CE, Sanders J, Bloemendal H, de Jong WW. The alternative splicing product alpha Ains-crystallin is structurally equivalent to alpha A and alpha B subunits in the rat alpha-crystallin aggregate. Biochim Biophys Acta 1990; 1037:58-65.

36. Siezen RJ, Bindels JG, Hoenders HJ. The quaternary structure of bovine alpha-crystallin. Chemical crosslinking with bifunctional imido esters. Eur J Biochem 1980; 107:243-9.

37. Kamei A, Hamaguchi T, Matsuura N, Masuda K. Does post-translational modification influence chaperone-like activity of alpha-crystallin? I. Study on phosphorylation. Biol Pharm Bull 2001; 24:96-9.

38. Liang JN, Li XY. Interaction and aggregation of lens crystallins. Exp Eye Res 1991; 53:61-6.

39. Groth-Vasselli B, Kumosinski TF, Farnsworth PN. Computer-generated model of the quaternary structure of alpha crystallin in the lens. Exp Eye Res 1995; 61:249-53.

40. Farnsworth PN, Singh K. Self-complementary motifs (SCM) in alpha-crystallin small heat shock proteins. FEBS Lett 2000; 482:175-9.

41. Ghosh JG, Clark JI. Identification of Functional Domains in the sHSP, Human B Crystallin, Using Protein Pin Arrays and Homology Modeling. ARVO Annual Meeting; 2003 May 4-9; Fort Lauderdale (FL).

42. Sheinerman FB, Norel R, Honig B. Electrostatic aspects of protein-protein interactions. Curr Opin Struct Biol 2000; 10:153-9.

43. Lo Conte L, Chothia C, Janin J. The atomic structure of protein-protein recognition sites. J Mol Biol 1999; 285:2177-98.