Subunit Exchange of Small Heat Shock Proteins

αA-Crystallin, a member of the small heat shock protein (sHsp) family, is a large multimeric protein composed of 30–40 identical subunits. Its quaternary structure is highly dynamic, with subunits capable of freely and rapidly exchanging between oligomers. We report here the development of a fluorescence resonance energy transfer method for measuring structural compatibility between αA-crystallin and other proteins. We found that Hsp27 and αB-crystallin readily exchanged with fluorescence-labeled αA-crystallin, but not with other proteins structurally unrelated to sHsps. Truncation of 19 residues from the N terminus or 10 residues from the C terminus of αA-crystallin did not significantly change its subunit organization or exchange rate constant. In contrast, removal of the first 56 or more residues converts αA-crystallin into a predominantly small multimeric form consisting of three or four subunits, with a concomitant loss of exchange activity. These findings suggest residues 20–56 are essential for the formation of large oligomers and the exchange of subunits. Similar results were obtained with truncated Hsp27 lacking the first 87 residues. We further showed that the exchange rate is independent of αA-crystallin concentration, suggesting subunit dissociation may be the rate-limiting step in the exchange reaction. Our findings reveal a quarternary structure of αA-crystallin, consisting of small multimers of αA-crystallin subunits in a dynamic equilibrium with the oligomeric complex.

␣A-Crystallin, a member of the small heat shock protein (sHsp) family, is a large multimeric protein composed of 30 -40 identical subunits. Its quaternary structure is highly dynamic, with subunits capable of freely and rapidly exchanging between oligomers. We report here the development of a fluorescence resonance energy transfer method for measuring structural compatibility between ␣A-crystallin and other proteins. We found that Hsp27 and ␣B-crystallin readily exchanged with fluorescence-labeled ␣A-crystallin, but not with other proteins structurally unrelated to sHsps. Truncation of 19 residues from the N terminus or 10 residues from the C terminus of ␣A-crystallin did not significantly change its subunit organization or exchange rate constant. In contrast, removal of the first 56 or more residues converts ␣A-crystallin into a predominantly small multimeric form consisting of three or four subunits, with a concomitant loss of exchange activity. These findings suggest residues 20 -56 are essential for the formation of large oligomers and the exchange of subunits. Similar results were obtained with truncated Hsp27 lacking the first 87 residues. We further showed that the exchange rate is independent of ␣A-crystallin concentration, suggesting subunit dissociation may be the rate-limiting step in the exchange reaction. Our findings reveal a quarternary structure of ␣A-crystallin, consisting of small multimers of ␣A-crystallin subunits in a dynamic equilibrium with the oligomeric complex.
The functions of ␣A-crystallin, ␣B-crystallin, and Hsp27 are closely related to each other. All three proteins can serve as chaperones in preventing the aggregation of denatured proteins (29 -38). The expression of ␣B-crystallin and Hsp27 is inducible by stress such as heat shock or exposure to transition metals (4, 39 -42). Interestingly, the intracellular location of both proteins change upon heat shock, moving from the cytoplasm to the peri-nuclear region (43)(44)(45)(46). They have also been shown to suppress cell death due to thermal and oxidative insults (47)(48)(49)(50)(51)(52)(53)(54)(55) and act as novel regulators of apoptosis in L929 fibroblast cells (56). Although the exact mechanism of protection against stress and cell death is still unknown, current evidence suggests ␣B-crystallin and Hsp27 may perform similar physiological functions.
Like many other members of the small heat shock protein family, ␣-crystallin and Hsp27 exist as high molecular mass complexes consisting of a large number of subunits (8,57). However, the size of these complexes can change depending on a number of physical and chemical parameters such as temperature, pH, and ionic strength, as well as the phosphorylation state of the protein (58 -66). Using fluorescence resonance energy transfer (FRET) to monitor subunit-subunit interaction, we have previously shown that the subunits of ␣A-crystallin can freely and reversibly exchange between oligomers (67). In this study we further demonstrate by this method that ␣A-crystallin can form a reversible hetero-oligomeric complex with Hsp27 and ␣B-crystallin, but not with other proteins unrelated to the sHsp family. In addition, we have identified the N-terminal region of ␣A-crystallin and Hsp27 to be essential for subunit exchange and oligomerization. Based on the results of our studies, we further propose an exchange mechanism that may explain the oligomeric organization of ␣A-crystallin subunits.

EXPERIMENTAL PROCEDURES
Materials-Lucifer Yellow iodoacetamide (LYI) and 4-acetamido-4Ј-((iodoacetyl)amino) stilbene-2,2Ј-disulfonic acid (AIAS) were purchased from Molecular Probes, Eugene, OR. Rhodanese was obtained from Sigma and used in the experiments without further purification. Restriction enzymes and Taq polymerase were purchased from New England Biolabs and Promega, respectively. Escherichia coli strain BL21DE3 and the pET 20bϩ expression vector were obtained from Novagen. Purified ␤-crystallin, ␥-crystallin, and human recombinant ␣B-crystallin were generous gifts from Dr. Joseph Horwitz, University of California, Los Angeles, CA. * This work was supported in part by Grants EY05895 (to B. K.-K. F.) and EY12018 (to H. S. M) from NEI, National Institutes of Health and by a grant from the Wong Fund (to B. K.-K. F). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Recipient of Predoctoral Training Grant EY07026 from NEI, National Institutes of Health, as well as a postdoctoral fellowship from the Bank of America Giannini Foundation.
Construction of ␣A-crystallin and Hsp27 Mutants-cDNA constructs of ␣A-crystallin and Hsp27 mutants were prepared by PCR amplification using the appropriate primers. The PCR reactions were carried out in a 25-l volume containing 10 mM Tris-HCl, pH 8.3, 50 mM KCl, 1.5 mM MgCl 2 , 50 M dNTP, 1.0 M primers, 2.5 units of Taq polymerase, and 100 ng of ␣A-crystallin or Hsp27 cDNA as a template. Amplification was performed for 25 cycles with conditions for denaturation at 94°C for 1 min, annealing at 54°C for 2 min, and extension at 72°C for 3 min. The PCR products were gel-purified and ligated into the pET 20b ϩ expression vector. The correct constructs were confirmed by DNA sequencing.
Expression and Purification of Wild-type and Truncated ␣A-Crystallin and Hsp27-BL21DE3 cells harboring the pET 20b ϩ ␣A-crystallin or Hsp27 constructs were grown in 500 ml of LB broth to a cell density between 0.6 -1.0 optical density at 600 nm and induced with 0.5 mM isopropyl-␤-D-thiogalactopyranoside for 3 h. Cells were harvested by centrifugation at 4,000 ϫ g for 10 min; resuspended in 20 ml of ice-cold buffer containing 0.1 M NaCl, 2 mM EDTA, 50 mM Tris, pH 7.9; and lysed by sonication. The cell lysates were clarified by centrifugation at 15,000 ϫ g for 30 min to remove the particulates, followed by filtration through a 0.2-m filter. Polyethyleneimine was then added to the filtrate with rapid stirring to form a 0.12% solution. After incubation on ice for 2 min, the mixture was centrifuged at 15,000 ϫ g for 10 min to remove the precipitated DNA. The clear supernatant containing the soluble recombinant protein was applied onto a Mono-Q column (Amersham Pharmacia Biotech) pre-equilibrated with 0.1 M NaCl, 20 mM Tris, pH 8.5, and eluted using a linear gradient of 0.1-1 M NaCl in the same buffer. Fractions containing the recombinant protein were pooled, concentrated by ultrafiltration, and further purified by Superose 6 gel filtration chromatography (Amersham Pharmacia Biotech) equilibrated with 100 mM NaCl, 50 mM sodium phosphate, pH 7.5.
All the wild-type and truncated proteins of ␣A-crystallin and Hsp27 were purified by the procedure described above except ␣A-crystallin-(56 -173), ␣A-crystallin- , and ␣A-crystallin-(85-173), which required an additional step following the Mono-Q ion exchange chromatography. In this case, the fractions containing mutant protein were pooled, concentrated, and adjusted to 1 M ammonium sulfate, pH 7.0. The protein solution was then applied to a phenyl-Sepharose HP column containing a 1-ml bed volume pre-equilibrated with 1 M ammonium sulfate, 50 mM sodium phosphate, pH 7.0, and eluted using a linear gradient of 1.0 -0.1 M ammonium sulfate in the same buffer. Fractions containing these mutant proteins were pooled, concentrated, and further purified by gel filtration chromatography as described earlier.
Labeling of Recombinant ␣A-crystallin and Hsp27-␣A-crystallin was labeled with AIAS or LYI under the conditions as described previously (67). The specific labeling of the single cysteine residue was confirmed by titration with 5,5Ј-dithiobis(2-nitrobenzoic acid). AIASlabeled ␣A-crystallin-(56 -173) was prepared using the same conditions as the wild type, with the exception that the purified truncated protein was labeled with AIAS for 3 h at room temperature. LYI-labeled ␣Acrystallin-(56 -173) was prepared similarly. However, the reaction time was extended for an additional 1 h at 37°C. Hsp27 was labeled with AIAS in a reaction mixture containing 1 mg/ml recombinant protein, 3.2 mM AIAS, 100 mM NaCl, 20 mM MOPS, pH 7.9 for 3 h at 37°C. Hsp27 was labeled with 8.4 mM LYI for 9 h at 37°C with the same buffer conditions. Unreacted AIAS or LYI were separated from the labeled proteins on a G-25 Sephadex desalting column equilibrated with Buffer A (100 mM NaCl, 50 mM sodium phosphate, pH 7.5). The final protein preparations were separated on SDS-PAGE, and the absence of free fluorescent label was confirmed by fluorescence imaging of the unstained gels. In some preparations, gel filtration of the fluorescent ␣A-crystallin under denaturing condition was also employed to assure the covalent attachment of the label to the protein.
Measurement of Quaternary Structural Compatibility-Proteins with a quaternary structure compatible with ␣A-crystallin were determined by a FRET method based on reversibility of the subunit exchange reaction. ␣A-crystallin mixtures containing an equal amount of donor and acceptor was prepared by incubating LYI-labeled ␣A-crystallin (20 M) and AIAS-labeled ␣A-crystallin (20 M) in Buffer A at 37°C for 4 h. The complete mixing of the two subunit populations was confirmed by monitoring the quenching of the emission intensity of AIAS fluorescence at 415 nm. A 100-l aliquot of the test protein at concentration specified in the figure legend was then added to an equal volume of fluorescent ␣A-crystallin in Buffer A. At various time periods, 40 l of the protein mixture was removed, diluted 50-fold with the same buffer, and the AIAS emission intensity at 415 nm were determined. The rate was calculated from the equation defined as the transfer rate constant, and F(t) and F(0) correspond to the emission intensity at time ϭ t and 0, respectively. The constants, A and B, were determined using the initial condition where A ϩ B ϭ 1 at t ϭ 0 and the final value of A ϭ F(ϱ)/F(0) at t ϭ ϱ. It is important to note that since F(ϱ)/F(0) Ͼ 1, the constant B is a negative number, in agreement with the decrease in energy transfer between donors and acceptors as the fluorescent ␣A-crystallin is diluted by the test protein.
The rate constant was determined by non-linear regression analysis of the data using the Biomedical Statistical Package Program.
Analytical Methods-Protein concentrations were determined by Coomassie Blue binding (68) using ␥-globulin as a standard. SDS-PAGE of proteins was performed by the method of Laemmli (69). The concentrations of LYI and AIAS were determined from their absorption spectra using molar extinction coefficients of 13,000 cm Ϫ1 M Ϫ1 at 435 nm and 35,000 cm Ϫ1 M Ϫ1 at 335 nm, respectively. The oligomeric structure of wild-type ␣A-crystallin and its mutant proteins were determined by gel filtration on a Superose 6 column using thyroglobulin, bovine ␥-globlulin, chicken ovalbumin, equine carbonic anhydrase, equine myoglobin fragment, and bovine pancreatic trypsin inhibitor as standards. The far UV circular dichroism spectra of wild-type ␣A-crystallin and ␣Acrystallin-(56 -173) were recorded in a cuvette with a 0.2-mm path length at room temperature using a Jasco J-600 spectropolarimeter. The secondary structure was estimated using the self-consistent method of Sreerama and Woody (70).

RESULTS
Quantitative Determination of Subunit Exchange of ␣A-crystallin by FRET-We have previously reported the application of FRET to monitor the exchange of ␣A-crystallin subunits (67). In this earlier study, we labeled recombinant ␣A-crystallin with AIAS, which served as an energy donor, and LYI, which serves as an energy acceptor. Upon mixing of the two populations of labeled ␣A-crystallin, we observed a time-dependent decrease in AIAS fluorescence at 415 nm, indicating an exchange reaction that brings the two labeled ␣A-crystallin subunits closer to each other. We further showed that subunit exchange is reversible, as revealed by the rapid recovery of the AIAS emission intensity as the fluorescent ␣A-crystallin was diluted with unlabeled ␣A-crystallin. We surmised that the recovery of donor fluorescence can be utilized to monitor the exchange of ␣A-crystallin with other proteins having similar structure and biophysical properties. Fig. 1 shows a more extensive study of diluting the mixture of AIAS-labeled and LYI-labeled ␣A-crystallin with unlabeled recombinant ␣A-crystallin. To prepare the fluorescent ␣A-crystallin mixture, we incubated an equal amount of AIAS-labeled and LYI-labeled ␣A-crystallin together at 37°C for at least 4 h. Measurement of the donor quenching indicated that the subunits from these two populations of ␣A-crystallin were completely scrambled. When unlabeled ␣A-crystallin were added to the fluorescent ␣A-crystallin mixture, we observed a time-dependent increase in AIAS emission intensity at 415 nm ( Fig. 1,  upper panel). This result indicated that the fluorescent ␣Acrystallin subunits were rapidly exchanging with the unlabeled ␣A-crystallin, resulting in a greater separation between the AIAS-labeled and LYI-labeled ␣A-crystallin subunits. Moreover, the final level of AIAS emission intensity is proportional to the amount of unlabeled ␣A-crystallin added (Fig. 1, lower  panel). The intensity reaches a maximum at a molar ratio of approximately 10 unlabeled ␣A-crystallin to 1 AIAS-labeled ␣A-crystallin. Under this condition, the distance between AIAS-labeled and LYI-labeled ␣A-crystallin was so far apart that there was no observable energy transfer between donors and acceptors. This quantifiable relationship between the donor emission intensity and amount of added ␣A-crystallin demonstrated the feasibility of using this simple dilution method to identify proteins structurally compatible with ␣A-crystallin.
Effect of ␣A-crystallin Concentration-We found that the transfer rate constant, as determined by fitting the data shown in Fig. 1 to the exponential function F(t)/F(0) ϭ A ϩ Be Ϫkt , was independent of the ␣A-crystallin concentration (data not shown). The average value in the range of 2.5-100 M ␣Acrystallin was 6.36 Ϯ 0.47 ϫ 10 Ϫ4 s Ϫ1 . This interesting observation was independently confirmed by measuring the subunit exchange between two populations of fluorescent ␣A-crystallin. Fig. 2 shows that the exchange rate, as determined by measuring the decrease in AIAS emission at 415 nm 15 min after mixing different concentrations of AIAS-labeled ␣A-crystallin and LYI-labeled ␣A-crystallin, is relatively constant over a 90-fold difference in ␣A-crystallin concentration.
Exchange of ␣A-crystallin Subunits with Other Lens Crystallins-We have examined the interaction of ␣A-crystallin with other proteins in the lens using the dilution method. We found that the addition of either ␣-crystallin purified from bovine lens or recombinant ␣B-crystallin to premixed AIAS-labeled and LYI-labeled ␣A-crystallin resulted in a time-dependent increase of donor fluorescence (Fig. 3). The transfer rate constant between fluorescent ␣A-crystallin and unlabeled ␣B-crystallin was 6.3 ϫ 10 Ϫ4 s Ϫ1 , which is identical to the rate constant between fluorescent ␣A-crystallin with unlabeled ␣A-crystallin at the same concentration. In contrast, ␤-crystallin and ␥-crystallin did not exchange with ␣A-crystallin, suggesting that there was no measurable interaction between these proteins in their native state (Fig. 3). This result suggests the existence of a common structural interface that allows an unrestricted exchange between the subunits of ␣Aand ␣B-crystallin.
Mapping the Region of ␣A-crystallin Essential to Subunit Exchange-We have previously reported the expression of fulllength ␣A-crystallin cDNA in E. coli BL21DE3 cells and shown that the recombinant protein retains its chaperone activity, multimeric organization and subunit exchange properties (67). In this study we used the same expression system to produce a number of truncation mutants in order to determine which region of ␣A-crystallin is essential for subunit exchange. We found that most mutants were expressed in the soluble fraction at levels as high as 40% of the total cell proteins. The high level  of expression allowed the purification of the truncated ␣Acrystallin to near homogeneity by successive Mono-Q ion exchange chromatography and gel filtration chromatography (Fig. 4, upper panel). We further characterized the size distribution of the truncation mutants by gel filtration (Fig. 4, lower  panel). Our results indicated that the removal of up to 10 amino acids from the C terminus did not significantly change the subunit organization of ␣A-crystallin. These C-terminal truncation mutants form oligomers with an average molecular mass of 700 kDa. Similarly, ␣A-crystallin-(20 -173) retained its high molecular mass oligomeric structure, as indicated by an average molecular mass of about 560 kDa. In contrast, ␣Acrystallin-(56 -173), ␣A-crystallin-(65-173), and ␣A-crystallin-(85-173) were predominantly smaller in size, with average molecular masses of 43, 60, and 48 kDa, respectively (Fig. 4,  lower panel). In addition, the elution profiles of ␣A-crystallin-(56 -173) reveals a narrow size distribution (Fig. 5, upper panel), which is markedly different from the broad size distribution of wild-type ␣A-crystallin (8,67). A similar narrow size distribution was also observed for ␣A-crystallin-(65-173) and ␣Acrystallin-(85-173) (data not shown). The discrete size distribution and apparent molecular mass of 43 kDa indicates that ␣A-crystallin-(56 -173) exists as small multimers consisting of three or four subunits.
The organization of ␣A-crystallin-(56 -173) into small multimers is not due to denaturation or marked changes of the protein secondary structure. The far UV circular dichroism spectra of wild-type ␣A-crystallin and ␣A-crystallin-(56 -173) were very similar (Fig. 5, lower panel). ␣A-crystallin-(56 -173) shows a pronounced minimum at 217 nm, which is indicative of a protein with a high content of ␤-sheet. Using the self-consistent method of analysis described by Sreerama and Woody (70), the secondary structure of ␣A-crystallin-(56-173) was estimated to consist of 41% ␤-sheet, 16% ␤-turn, and 11% ␣-helix. This result is comparable to the secondary structure of wild-type ␣A-crystallin (39% ␤-sheet, 16% ␤-turn, and 13% ␣-helix) reported earlier (71) and to our own determination. These results suggest that the removal of the N-terminal region of ␣A-crystallin only changes its multimeric subunit organization.
Is there a correlation between subunit exchange and oligomerization? To answer this important question, we tested the subunit exchange property of all the truncated ␣A-crystallin using the dilution method as described in the last section. Fig. 6 shows that both ␣A-crystallin-(20 -173) and ␣Acrystallin-(1-163) readily and freely exchanged with the fluorescent ␣A-crystallin subunits, as indicated by the recovery of donor emission intensity. The rate constants were 5.3 and 6.3 ϫ 10 Ϫ4 s Ϫ1 , respectively, which are similar to the rate constant of 6.3 ϫ 10 Ϫ4 s Ϫ1 for the unlabeled ␣A-crystallin at the same concentration (Fig. 4, lower panel). In contrast, ␣Acrystallin-(56 -173) did not exchange with fluorescent ␣A-crystallin subunits (Fig. 6), suggesting that the fundamental exchange unit is a tetramer or smaller multimers. If this hypothesis is correct, then the ␣A-crystallin-(56 -173) subunits of the smaller multimers should not exchange with each other; Fig. 7 shows that this is indeed the case. When two populations of ␣A-crystallin-(56 -173) labeled with AIAS or LYI were mixed together at 37°C, the fluorescence intensity of the mixture remains constant over a period of 60 min.
Formation of a Reversible Complex between ␣A-crystallin and Hsp27 Subunits-We have also examined the structural compatibility between ␣A-crystallin and Hsp27. Like ␣A-crystallin, recombinant Hsp27 subunits forms a large oligomeric complex with an average molecular mass of 680 kDa (Fig. 4,  lower panel). It also contains a single cysteine at amino acid residue 137 that can be labeled with LYI or AIAS. Calculation based on the molar extinction coefficient of LYI at 335 nm revealed an average of 1 mol/mol of Hsp27 subunit. We found Lower panel, sizes and transfer rate constants (k) of ␣A-crystallin, Hsp27, and their truncation mutants. The oligomer size was determined by gel filtration chromatography, and the transfer rate constant was measured by the dilution method as described under "Experimental Procedures." that fluorescent labeling of Hsp27 did not lead to a major change in its molecular mass or conformation, suggesting that the chemical modification did not markedly alter its conformation (data not shown).
We first tested the exchange of subunits between ␣A-crystallin and Hsp27 by mixing AIAS-labeled ␣A-crystallin with LYI-labeled Hsp27. We observed a time-dependent quenching of AIAS fluorescence, indicating an intermixing of subunits between Hsp27 and ␣A-crystallin oligomers (Fig. 8, upper panel). The exchange was also completely reversible, as the addition of unlabeled ␣A-crystallin to the fluorescent hetero-oligomer resulted in a complete return of donor fluorescence (Fig.  8, lower panel). This result demonstrated an unrestricted exchange of subunits between ␣A-crystallin and Hsp27 oligomers and validated the use of the dilution method to monitor the formation of the hetero-oligomer.
Is the N-terminal region of Hsp27 also important for oligomerization? To answer this question, we prepared an Nterminal truncation mutant of Hsp27 with amino acid sequence corresponding to ␣A-crystallin-(65-173) and then determined its size distribution and subunit exchange activity. Interestingly, Hsp27-(88 -206) appears to be able to assemble into a much larger oligomeric complex than ␣A-crystallin-(65-173), with an average molecular mass of 180 kDa (Fig. 4, lower  panel). Nevertheless, similar to the truncation of ␣A-crystallin, the removal of the N-terminal region of Hsp27 impaired subunit exchange (Fig. 9).

DISCUSSION
In this study we have developed a method to monitor the formation of hetero-oligomeric complexes between ␣A-crystallin and other structurally compatible proteins. Our method is based on the observed decrease in FRET between AIAS-labeled and LYI-labeled ␣A-crystallin upon the integration of a foreign protein into the oligomer. We found that the subunits of Hsp27 and ␣B-crystallin readily exchanged with ␣A-crystallin (Figs. 3 and 8), suggesting that their quaternary structures are compatible. In contrast, proteins that are not members of the sHsp family, such as ␤-crystallin, ␥-crystallin, and rhodanese, were unable to integrate into the ␣A-crystallin oligomeric complex. ␤-Crystallin and ␥-crystallin are composed of predominantly ␤-pleated sheets (8). While ␥-crystallin is a monomer, ␤-crystallin can exist in multimeric forms (8). The absence of interaction between ␣A-crystallin and these proteins indicates that subunit exchange is not a property of any multimeric ␤-pleated sheet proteins.
Our study indicated that the formation of a hetero-oligomeric complex between various forms of sHsp is not only feasible, but is probably quite common. What then is the importance of this type of hetero-oligomeric complex in vivo? ␣-Crystallin is composed of a 3:1 ratio of ␣A-crystallin to ␣B-crystallin in human lens. It has been shown that this complex has greater thermal stability than either protein alone (72). Thus, one potentially important function of hetero-oligomers is the enhanced effectiveness in protecting cells from damage and stress. This is of paramount importance in the lens, where the lens proteins are constantly subjected to UV light, oxidative stress, and other external insults. Interestingly, Hsp27 is also expressed in the lens (15). It would be interesting to find out whether or not it also forms a complex with ␣-crystallins and further contributes to the overall stability of these lens proteins.
Hetero-oligomers containing ␣B-crystallin and Hsp27 have also been shown to be present in non-oncogenic adenovirustransformed rat kidney cells and in human pectoral muscle (73,74). However, contrary to the increased thermal stability observed in ␣-crystallin, the hetero-oligomer of ␣B-crystallin and Hsp27 was disrupted when cells were treated at 44°C for 12 min (75). We found that heat treatment at 44°C did not dissociate the preformed complex of recombinant ␣A-crystallin and Hsp27 (data not shown), suggesting other cellular mechanisms such as phosphorylation or interaction with other proteins may modulate the dissociation of the Hsp27-␣B-crystallin complex when the cell is under stress conditions.
A hallmark of ␣-crystallin and other sHsp is their tendency to form large oligomeric structures. The organization of ␣Acrystallin subunits into larger oligomers most likely involves both ionic and hydrophobic interactions, since we have observed a decrease in the exchange rate under either very low salt or very high salt conditions (67). Our truncation study further suggests that sequence 20 -56 is essential for the for- short N-terminal region and no C-terminal tail exist as discrete tetramers (77). Similarly, yeast Hsp12.6, which is related to Hsp12.2 and Hsp12.3, is monomeric (78). Taken together, these findings support the hypothesis that the N-terminal region is important for oligomerization.
Our findings and those of others have suggested that the overall quaternary structure of the oligomer is variable and dynamic (67, 79 -81) and is shaped by a continuous exchange of subunits (67,81). The absence of subunit exchange between the small ␣A-crystallin-(56 -173) multimers further implies that the fundamental exchange unit is likely to be a small multimer composed of three or four subunits; otherwise, the ␣Acrystallin-(56 -173) subunits, which apparently can interact to form multimers, should be able to exchange with each other and with the ␣A-crystallin oligomers. This finding is consistent with the observed conversion of ␣A-crystallin into tetramers in sodium deoxycholate (82) and the reversible dissociation of Hsp27 to a smaller multimeric form following phosphorylation (64 -66) or dilution of the protein to very low concentrations (79).
How do subunits of ␣A-crystallin assemble into a stable exchange unit? Most sHsp contain a stretch of relatively homologous sequence of approximately 100 residues termed the "␣-crystallin domain." This region has been postulated to mediate subunit-subunit interaction (83). Recently, the structure of a 16.5-kDa sHsp from Methanococcus jannaschii has been determined at 2.9-Å resolution (84). The oligomer was shown to be composed of 24 subunits, with each subunit in the complex making extensive contacts with other subunits through hydrogen bonds, hydrophobic contacts, and ionic interactions. Most of the contact sites are located within the ␣-crystallin domain, which consists primarily of anti-parallel ␤-strands. Similarly, a ␤-sheet structure consisting of three anti-parallel ␤-strands has also been identified in the same region of ␣A-crystallin and Hsp27 by site-directed spin-labeling (85,86). Moreover, residues 109 -120 of ␣A-crystallin and 133-144 of Hsp27 have been shown to form contacts between equivalent strands from neighboring subunits (83,86). Taken together, these recent findings suggest that some of the residues within the ␣-crystallin domain participate in subunit-subunit interaction. Perhaps these contact sites are also involved in stabilizing the quaternary structure of the small multimers.
Our findings further suggest that the exchange unit of ␣Acrystallin are capable of freely associating and dissociating to form large oligomeric complexes. There are two potential mechanisms for the observed exchange reaction. The first mechanism involves the interchanging of the exchange unit as a result of collision between oligomers. The second mechanism is based on the continuous dissociation and reassociation of the exchange unit. Our data indicated that the transfer rate constant is independent of ␣A-crystallin concentration (Figs. 1 and 2), implying that the exchange of subunits is not due to the collision of the oligomers, but is most likely by a dissociation mechanism. If this mechanism is correct and the dissociation of the exchange unit is rate-limiting, as implied by the high activation energy of 60 kcal/mol (67), then the apparent dissociation rate constant (k off ) can be determined by measuring the recovery of AIAS emission intensity after diluting the preformed fluorescent ␣A-crystallin mixture with a large pool of unlabeled ␣A-crystallin (Fig. 1). We have obtained a k off value of 6.3 ϫ 10 Ϫ4 s Ϫ1 under this condition. Interestingly, the dissociation rate constants of many biological processes, such as receptor binding (87), subunit assembly and dissociation (88), protein-protein interaction (89), and antibody-antigen recognition (90), are all in the same range.
We have also shown that the Hsp27 subunit can freely integrate into the ␣A-crystallin oligomeric complex (Fig. 8). Our truncation study further suggests that the N-terminal onethird of the Hsp27 molecule also contains the structure essential for subunit exchange activity (Figs. 6 and 9). Sequence comparison between ␣A-crystallin, Hsp27, and other members of the sHsp family indicates residues 18 -28 and 40 -46 of ␣A-crystallin are highly conserved. It is tempting to speculate that these residues may form an integral part of a common structural surface that allows unrestricted interfacing between various forms of sHsp. Site-directed mutagenesis of these residues in these regions would be an excellent way to test this hypothesis.
␣A-crystallin has been shown to possess chaperone-like property that binds unfolded proteins and prevents their aggregation during denaturation (29 -38). A putative substrate binding domain of ␣A-crystallin has been mapped to the sequence cor-responding to residues 71-88 (91,92). Within this region, residues 76 -80 are in a buried environment (85). We hypothesize that the area of interaction with denatured protein is this hydrophobic crevice located near the N-terminal interface between. The access to this pocket may involve a transient exposure through rearrangement of the exchange unit.