Role of ATP on the Interaction of α-Crystallin with Its Substrates and Its Implications for the Molecular Chaperone Function*

ATP plays a significant role in the function of molecular chaperones of the large heat shock protein families. However, its role in the functions of chaperones of the small heat shock protein families is not understood very well. We report here a study on the role of ATP on the structure and function of the major eye lens chaperone α-crystallin. Our in vitro study shows that at physiological temperature, ATP induces the association of α-crystallin with substrate proteins. The association process is reversible and low affinity in nature with unit binding stoichiometry. 4,4′-Dianilino-1,1′-binaphthyl-5,5-disulfonic acid, dipotassium salt, binding studies show that ATP induces the exposure of additional hydrophobic sites on α-crystallin, but no appreciable enhancement of the same was observed for the substrate protein γ-crystallin or carbonic anhydrase. An equilibrium unfolding study reveals that ATP at 3 mgm concentration stabilizes the α-crystallin structure by 4.5 kJ/mol. The compactness induced by ATP makes it more resistant to tryptic cleavage. ATP-induced association of chaperone α-crystallin with substrate enhanced its aggregation prevention ability and also enhanced the refolding yield of lactate dehydrogenase from the unfolded state. Our results suggest that the binding of ATP to α-crystallin and not its hydrolysis is required for all these effects, as replacement of ATP by its nonhydrolyzable analogue adenosine-5′-O-(3-thiotriphosphate), tetralithium salt, reproduced all the results faithfully. The implication of the ATP-induced reversible protein-protein association at physiological temperatures on the functional role of α-crystallin in vivo is discussed.

␣-Crystallin is the major protein component of the vertebrate eye lens and is composed of two subunits, ␣A and ␣B, 20 kDa each having 57% sequence homology between them. Both subunits associate to form a large oligomer with an average molecular mass of 800 kDa. It is a key member of the small heat shock protein (sHSP) 1 superfamily having a conserved core "␣-crystallin domain" (1,2). Although ␣Aand ␣B-crystallin are expressed at high levels in the lens, both proteins have been identified in small amounts in non-lens tissues as well (3)(4)(5). An important finding that sheds light on the understanding of the function of ␣-crystallin has been the discovery of its molecular chaperone-like function a decade ago (6,7). ␣-Crystallin, like other sHSPs, acts like a molecular chaperone in preventing the aggregation of other proteins (8 -10). An intense research activity on the structure and function of ␣-crystallin has grown since then (11)(12)(13)(14)(15)(16)(17)(18)(19)(20), and the subject matter has been extensively reviewed in recent years (21)(22)(23).
Despite extensive studies, a number of issues relating to its chaperone function still remain unclear. One such issue concerns the role of ATP in the chaperone function of ␣-crystallin (16, 17, 24 -26). It is known that ATP concentration in lens exceeds 6 mM, one of the highest found among various tissues (27). Many well known classical chaperones of the large heat shock protein family, including GroEL, DnaK, etc., function in an ATP-dependent manner (28). Hydrolysis of ATP plays a crucial role in the chaperone-mediated folding of substrate proteins. However, such an activity has been less well documented in the ␣-crystallin system (23). Thus, ␣-crystallin has been reported to have autophosphorylation activity by some workers (29,30), whereas others failed detect significant ATPase activity (23). Some spectroscopic (16,24) and proteolytic (31) studies suggested interaction between ATP and ␣-crystallin with concomitant conformational changes, but the nature of the changes and its functional implications are far from clear. According to a totally different viewpoint existing in the literature, most sHSPs having the conserved ␣-crystallin domain have no dependence on ATP for their functions (22).
It has already been established that the yield of several substrates refolded in vitro in the presence of ␣-crystallin somewhat increased when ATP was present in the system (16,17,25,26), but mechanistic details are not known. ATP enhanced the yield of ␣-crystallin-mediated refolding of citrate synthase (17,25,26), xylose reductase (16), and luciferase (25). Although some workers suggest that ATP binding enhances the chaperone function of ␣-crystallin (17), others suggest that ATP destabilizes the chaperone-substrate complex much the same way as in GroEL system (26). But there are conflicting views as to the requirement of ATP hydrolysis for these various schemes (16,17).
In this paper, we present results that show that ATP exposes hydrophobic sites on the surface of ␣-crystallin, which triggers reversible association with substrates at physiological temperature. ATP stabilizes the structure of ␣-crystallin and makes it less sensitive to proteases. The reversible association enhances the chaperone activity of ␣-crystallin and can also explain the enhancement in the ␣-crystallin-mediated reactivation yield of unfolded enzymes.

Materials
Bovine eye lenses were obtained from a local slaughterhouse and were stored at Ϫ70°C. Sephacryl S-300 HR, Sephacryl S-400-HR, carbonic anhydrase (CA), insulin, DNase, lysozyme, isopropyl ␤-D-thiogalactoside, and SDS were all purchased from Sigma. Dithiothreitol (DTT), trypsin, and all buffer salts (tris, phosphate, etc.) were from Sisco Research Laboratories, India. Bis-ANS was obtained from Molecular Probes. Phenylmethylsulfonyl fluoride was obtained from Merck. All other chemicals used in this study were of analytical grade.
Preparation of Bovine ␣ L -and ␥-Crystallin ␣ L -Crystallin was purified from bovine eye lenses as described earlier (12). Briefly, eye lenses were homogenized on ice in 10 mM Tris-HCl, pH 7.2, containing 100 mM NaCl, 1 mM EDTA, 0.02% sodium azide, and 0.2 mM phenylmethylsulfonyl fluoride and then centrifuged at 4°C at 12,000 rpm for 45 min. The supernatant fraction was loaded onto a Sephacryl S300-HR column (95 ϫ 1.5 cm). The first peak contained high (␣ H -) crystallin and the second peak low (␣ L ) crystallin. Further purification was done by using a Sephacryl S400-HR column. SDS-PAGE of purified ␣ L -crystallin showed a single band corresponding to 20 kDa. ␥-Crystallin corresponded to the fifth peak (␤ H -and ␤ L -crystallins were the third and fourth peaks, respectively) of the Sephacryl S300 column. This was further purified through Sephacryl S100-HR (95 ϫ 1.5 cm) column at 25°C. The purified protein solution became turbid on heating at 60°C. ␥-Crystallin was extensively dialyzed against ammonium bicarbonate buffer (24 g/liter, pH 8.0), lyophilized, and stored at 4°C. SDS-PAGE showed a single band at 20 kDa.

␣Aand ␣B-Crystallin
Plasmid DNA of human ␣A-crystallin construct in pAED4 vector was a gift from Dr. W. W. de Jong of Katholic University, The Netherlands. Plasmid for human ␣B-crystallin in pET20bϩ expression vector was provided as a gift by Dr. J. Horwitz of the Jules Stein Eye Institute, Los Angeles. For overexpression, both the plasmids were separately introduced into Escherichia coli strain BL21-DE3. Cultures were grown in LB medium at 37°C with isopropyl ␤-D-thiogalactoside induction. Cells were centrifuged, subjected to freeze-thaw treatment, and then extracted with DNase and lysozyme. Proteins were dialyzed in 20 mM Tris, pH 7.2, containing 0.5 mM EDTA and 0.5 mM DTT (buffer A), concentrated in an Amicon stirred cell, applied to a DEAE-anion exchange column, and eluted with linear 0 -0.5 M NaCl gradients. ␣Aand ␣B-crystallin fractions were then applied to Sephacryl S300-HR size exclusion column (1.5 ϫ 90 cm) and eluted with buffer A containing 0.1 M NaCl. Main peak fractions were concentrated and dialyzed against buffer A or 50 mM phosphate buffer, pH 7.2, containing 0.5 mM DTT and stored in aliquots at Ϫ70°C. SDS-PAGE of both ␣Aand ␣B-crystallin showed a single band around 20 kDa. Concentration of recombinant proteins was determined spectrophotometrically by measuring absorbance at 280 nm by using extinction coefficients of 0.83 and 0.95 (mg/ml) Ϫ1 cm Ϫ1 for ␣Aand ␣B-crystallin, respectively (32). Molar concentration of all ␣-crystallins was always expressed in subunit basis.

ATP-induced Association of Substrate Proteins with ␣-Crystallins
Membrane Filtration Method-We incubated 12.5 M ␣A-, ␣B-, or ␣ L -crystallin with 2-25 M bovine ␥-crystallin or CA in 50 mM phosphate buffer, pH 7.2, in the absence and presence of 3 mM ATP at 25, 37, 45, and 55°C for 1 h. No protein aggregation was observed under these conditions. The solutions were cooled back to 25°C for 1 h. Unbound substrate was then separated by centrifugation at 4000 ϫ g through a 100-kDa Microcon (Amicon) filter membrane. The amount of substrate remaining associated with ␣-crystallin was calculated from the concentration of total substrate and free substrate; both were determined by the Bradford assay using BSA as standard.
Gel Filtration Chromatography-Since the presence of ATP interfered with the 280-nm UV peak of the substrate proteins obtained with the analytical TSK-GEL G3000SW XL column, we used FITC-labeled ␥-crystallin for these experiments. The FITC labeling method was described earlier (33) where we also showed that covalent tagging of FITC to bovine ␥-crystallin did not change its gel filtration property or the chaperone function of ␣-crystallin. The mixture of ␣-crystallin and ␥-crystallin (12.5 M each in 50 mM phosphate buffer, pH 7.2) with and without 3 mM ATP, after incubation at 37°C for 1 h followed by cooling to 25°C as mentioned above, was injected (20 l) into the column (30 cm ϫ 7.8 mm, 5 m). The column was run at 0.5 ml/min in the equilibrating buffer (50 mM phosphate, pH 7.2) that contained 3 mM ATP for ATP-containing samples only, and the chromatogram was monitored by the FITC signal at 494 nm. The area of the peaks was calculated by using the HPLC instrument software (Waters).

Bis-ANS Binding by ␣-Crystallin in the Presence of ATP
␣-Crystallin (2.5 M) in 50 mM phosphate buffer, pH 7.2, was incubated at 37°C for 1 h in the absence and presence of 3 mM ATP in a 3-ml fluorimeter cuvette placed inside a Hitachi 4500 spectrofluorometer maintained at the incubation temperature by using a water bath. The solution was titrated with 300 M bis-ANS by adding a small aliquot at a time. After each addition, the solution was stirred magnetically for 1 min, and fluorescence emission spectrum was recorded between 450 and 550 nm by using 390 nm as the excitation wavelength. The excitation and emission bandpasses were 5 nm each. In order to analyze the data according to the Scatchard equation, a titration of 0.2 M bis-ANS by 10.0 mg/ml ␣-crystallin or 4.0 mg/ml ␥-crystallin was performed. The reverse titration data were used to obtain the quantitative relationship between fluorescence intensity change and bound bis-ANS, according to Cardamone and Puri (34).

Enhanced Structural Stability of ␣-Crystallin in the Presence of ATP
The enhancement in structural stability of ␣-crystallin by ATP was measured by comparing the trypsin digestibility of ␣〈-crystallin in the absence and presence of 3 mM ATP. ␣A-crystallin (1 mg/ml in 50 mM phosphate buffer, pH 7.2) was incubated with trypsin at 50:1 and 100:1 (w/w) ratios at 37°C. Aliquots were withdrawn at different digestion times, and the reaction was stopped immediately by adding soybean trypsin inhibitor. SDS-PAGE of the ␣〈-crystallin digest was performed under reducing conditions in a Bio-Rad Mini-PROTEAN 3 electrophoresis set up by using linear 8 -16% gradient polyacrylamide gel. Silver staining was used to detect the bands. Gels were scanned in a densitometer for quantitative analysis.
The enhancement of the stability of ␣〈-crystallin in the presence of ATP was also determined by equilibrium chemical denaturation experiments. ␣〈-crystallin solutions (0.1 mg/ml in 50 mM phosphate buffer, pH 7.2) were incubated at 37°C in various urea concentrations in the range 0 -7 M for 18 h. Tryptophan fluorescence spectra of all solutions were taken in the 300 -400-nm region using 295 nm as excitation wavelength and 5 nm each for excitation and emission bandpass. The equilibrium unfolding profile was fitted according to the three-state model as described by Das et al. (35).

Chaperone Activity Assays of ␣-Crystallin in the Presence of ATP
The chaperone-like activity of recombinant human ␣B-crystallin in the absence and in the presence of 3 mM ATP was studied by using DTT-induced aggregation of insulin (12). Briefly, 0.35 mg/ml insulin in 50 mM phosphate buffer, pH 7.2, in the absence and presence of 0.175 mg/ml ␣B-crystallin was preincubated with 3 mM ATP at 37°C for 1 h. Aggregation was initiated by adding freshly prepared DTT to a final concentration of 20 mM, and the apparent absorbance at 400 nm was monitored at the kinetic mode using a Shimadzu UV-2401PC spectrophotometer maintained at 37°C. Assays were also done using recombinant ␣A-and bovine ␣ L -crystallin.
Chaperone activity was also assayed by measuring the ␣-crystallinmediated refolding of LDH, a homotetrameric enzyme from its fully unfolded state. LDH was denatured in 6 M GdnHCl for 8 h at 25°C at a concentration of 1 M. Refolding of the enzyme was initiated by diluting the denatured LDH 100-fold in a refolding buffer, pH 7.2, consisting of 50 mM phosphate, 10 mM magnesium acetate, 5 mM DTT, 30 M ␣B-crystallin, and 3 mM ATP or ATP␥S. Control experiments without ␣B-crystallin and/or ATP were also done. The enzyme concentration was 10 nM during refolding. The activity of refolded enzyme was assayed by adding 20 l of refolding mixture to 580 l of refolding buffer mentioned above containing 0.1 mM NADH and 0.4 mM sodium pyruvate preincubated at 37°C and by measuring the decrease at A 340 with time.

ATP-induced Binding between Chaperone ␣-Crystallin and Its Substrate
Monitoring of Binding by Membrane Filtration-When a mixture of ␣A-crystallin homo-oligomer and monomeric ␥-crystallin, 12.5 M each, was incubated at 37°C and the mixture passed through a 100-kDa membrane, all of the ␥-crystallin came through. But when the mixture was incubated in presence of 3 mM ATP, about 45% of the ␥-crystallin was retained over the membrane. We carefully considered the possibility of aggregation of the substrate proteins in the presence of ATP, but we found no evidence for it at the conditions used for our experiments. Thus, the results indicated that ATP triggered association between the chaperone ␣A-crystallin and its substrate ␥-crystallin. If the chaperone ␣A-crystallin is replaced by BSA in such an experiment, no binding between BSA and ␥-crystallin was observed in the presence of ATP, indicating that the chaperone is required for this interaction. Equilibrating 12.5 M of ␣A-crystallin with varying concentrations (2-25 M) of ␥-crystallin in the presence of 3 mM ATP, a binding isotherm (Fig. 1A) was constructed after determining the unbound and bound concentration of substrate (S) by membrane filtration. By assuming a single binding site per subunit of the chaperone (C), the dissociation constant (K d ) of the complex (CS) for the process CS ϭ C ϩ S is given by Equation 1, where C o and S o indicate the total concentration of the chaperone and substrate, respectively, and S indicates the equilib-rium concentration of the substrate in the mixture. Equation 1 can be rearranged to give Equation 2.
The dissociation constant obtained from the slope of the linear plot of C o /(S o Ϫ S) against 1/S ( Fig. 1B) was 8.2 M. The data were also analyzed by the Scatchard equation (Equation 3), assuming n identical noninteracting sites per subunit of the chaperone, where is the number of moles of substrate bound per mol of chaperone; n is the number of binding sites, and K d is the dissociation constant. The stoichiometry n and K d values obtained from the plot of /S against (  Table I ␣ L -crystallin as chaperone and ␥-crystallin and carbonic anhydrase as substrates. In all cases, binding between the chaperone and substrate was observed in the presence of ATP at 37°C. In absence of ATP at this temperature, the substrate proteins remained largely (85-98%) unbound (Fig. 2). Only a small amount of substrate (2-15%) was involved in an extremely low affinity binding having K d values in the hundreds of micromolar range (Table I). However, in the presence of ATP at 37°C, unbound substrates dropped in the range of 40 -60%, indicating a significant amount of substrates remained bound to ␣-crystallin. This enhanced binding in the presence of ATP was clearly indicated by a more than 1 order of magnitude decrease in K d values in the presence of ATP at 37°C (Table I).
It is of interest that the affinity of binding of a substrate to ␣B-crystallin is higher than that to ␣A-crystallin (Table I). When ATP is replaced by its nonhydrolyzable analogue ATP␥S at 37°C, a similar result was obtained (Fig. 2), indicating that ATP hydrolysis is not required for promoting the substrate-␣crystallin association.
We observed that the complex once formed in the presence of ATP at 37°C or higher temperatures remained stable even when the temperature was reduced to 25°C. However, removal of ATP from the complex by dialysis induced slow dissociation and extensive dialysis fully dissociated the complex indicating its reversible character with respect to ATP concentration.
Enhanced Substrate Binding Capacity of ␣-Crystallin at Various ATP Concentrations and Temperatures-Binding between various ␣-crystallins and its substrate proteins was found to be ATP concentration-dependent. A typical result of the association between recombinant human ␣A-crystallin and ␥-crystallin at 37°C as a function of the concentration of ATP is shown in Fig. 3A. The association constant is increased from 0.065 as the ATP concentration is increased from 0.2 to 3 mM above where it tends to level off. A very similar dependence on ATP concentration was also observed for the chaperone-substrate binding when either ␣B-crystallin or ␣ L -crystallin was used as chaperone (data not shown).
The effect of ATP on substrate binding was also dependent on temperature. In the absence of ATP there is negligible binding between ␣-crystallin and its substrate at 37°C and below. Binding is, however, enhanced in the presence of ATP at all temperatures. We define "binding enhancement" by ATP as the difference in percentage binding of the substrate in the presence and the absence of ATP at a given temperature. A typical result of the binding enhancement for the (␣A-␥) and (␣〉-␥)-crystallin system, at 3 mM ATP, as a function of temperature is shown in Fig. 3B. No enhancement was observed at 25°C for both systems. The enhancement rose to ϳ40% at  37°C, ϳ43% at 45°C, and ϳ45% at 55°C for ␣A-crystallin. A similar trend was also observed for binding between ␣B-crystallin and ␥-crystallin, but the enhancements were 12-18% higher in the 37-55°C range than the values observed for ␣A-crystallin. Fig. 3B also shows that although the binding enhancement increased somewhat beyond 37°C, 80 -90% of maximum enhancement was achieved at the physiological temperature. Binding between ␣-crystallin and CA was also enhanced as indicated at all the temperatures by a substantial decrease in K d values (Table I) in the presence of ATP. We noted that the influence of ATP was more pronounced, when various ␣-crystallins bind to its natural substrate, e.g. ␥-crystallin as against its non-natural substrate CA. This can be seen from the fact that the dissociation constants of ␣-␥ complexes are always lower than those of the respective ␣-CA systems ( Table I).
Assay of ATP-induced Binding by Gel Filtration HPLC-ATP-induced binding between the chaperone ␣-crystallin and ␥-crystallin was directly monitored through gel filtration HPLC. Because the presence of ATP interfered with the monitoring of the 280-nm peak of substrate protein, we used FITClabeled substrates for these experiments. We previously checked (33) that this modification of the substrate ␥-crystallin did not alter its interaction with the chaperone ␣-crystallin.
The gel filtration profile of a reference mixture of bovine ␣ Lcrystallin and FITC-labeled bovine ␥-crystallin incubated at 25°C (without ATP), monitored for FITC absorbance at 494 nm, showed only one major peak (ϳ95%) corresponding to the position of free ␥-crystallin and one minor peak (ϳ5%) corresponding to free FITC (Fig. 4, trace 1). Addition of 3 mM ATP to the sample mixture incubated at 25°C as well as in the eluting buffer resulted in a minor peak (ϳ2%) at void volume showing a negligible interaction of ␣ L -crystallin with ␥-crystallin (Fig. 4, trace 2) at this temperature. Incubation of ␣ L -and ␥-crystallin at 37°C in the absence of ATP showed a profile very similar to that of the reference with an additional minor peak (ϳ3.5%) at void volume indicating very little binding between them (Fig. 4,  trace 3). However, in the presence of 3 mM ATP, the 37°C preincubated sample showed a substantial increase (ϳ60%) in the area of the ␣ L -crystallin peak at void volume with concomitant loss in the area of the ␥-crystallin peak indicating complexation of ␥-crystallin with ␣ L -crystallin (Fig. 4, trace 4). Use of ATP␥S in place of ATP at 37°C gave an ϳ55% increase in the area of the void volume peak (Fig. 4, trace 5). These results, yielding a K d of 2.4 M at 37°C, are in reasonable agreement with the K d of 2.8 M obtained from the membrane filtration data (Table I).

␣A-, Human ␣B-, and Bovine ␥-Crystallins in the Presence of ATP
Bis-ANS, a hydrophobic molecule, has a low fluorescence quantum yield in aqueous solution. But when it binds to the hydrophobic sites of protein, its quantum yield dramatically increases (12). This dye has been widely used for probing the surface-exposed hydrophobicity of ␣-crystallin (12,15,36,37). We have done fluorescence titration of various ␣-crystallins by bis-ANS both in the presence and the absence of 3 mM ATP. The titration curves at 37°C are shown in Fig. 5. In presence of ATP, both ␣Aand ␣B-crystallins undergo a conformational transition associated with the exposure of additional hydrophobic sites (Fig. 5, A and B). Contrary to this, the substrate protein bovine ␥-crystallin shows much lower bis-ANS fluorescence intensity, and on addition of ATP, no significant changes could be observed (Fig. 5C). The binding data were analyzed by  Table II. It can be seen that the presence of 3 mM ATP increased n for binding to ␣A-crystallin from 0.40 to 0.50 per subunit and decreased the K d from 1.8 to 1.2 M. For ␣B-crystallin, n increased from 0.72 to 0.83, and the K d decreased from 1.1 to 0.69 M in 3 mM ATP. For the bis-ANS binding to ␥-crystallin in the absence of ATP, the binding stoichiometry remained lower (n ϭ 0.22) and the dissociation constant relatively higher (K d ϭ 11 M) as compared with those for ␣Aor ␣B-crystallin. The binding parameters changed little in the presence of ATP (Table II).
The increase of exposed hydrophobic sites of the chaperone is specific to ATP only, as ADP and AMP led to a marginal increase in the bis-ANS fluorescence as shown in the results of the bis-ANS binding to ␣ L -crystallin (Fig. 6A). The effects become clearly visible if the difference in bis-ANS fluorescence intensity in the presence and absence of the nucleotides (⌬F) was plotted against the bis-ANS concentration (Fig. 6A, inset). It is interesting to note that ATP␥S also produced effects similar to ATP (Fig. 6A, inset). The Scatchard plots for these data are shown in Fig. 6B. The parameters n and K d obtained from these plots are also shown in Table II. In the absence of any nucleotide, n and K d values for bis-ANS binding to ␣ L -crystallin were ϳ0.5 and 1.8 M, respectively. These values changed little when either AMP or ADP was added (Table II). In the presence of ATP, n increased to over 0.6 and K d decreased by more than a factor of 2. ATP␥S also produced similar effects. These results are consistent with the percentage of enhanced binding of ␥-crystallin to ␣ L -crystallin at 37°C as determined by the membrane filtration method that shows that whereas 3 mM ATP led to 46% binding enhancement, ADP and AMP at the same concentration resulted in only a negligible (Ͻ2%) binding enhancement (Fig. 6C).

Enhanced Structural Stability of ␣-Crystallin in the Presence of ATP
Limited tryptic digestion studies of ␣-crystallin reported by us (38) as well as by others (31) showed that cleavage sites located at the C-and N-terminal regions are relatively easily accessible by trypsin, but those on the ␣-crystallin domain are less accessible. We have checked whether the addition of ATP leads to any differences in the accessibility of these sites of ␣A-crystallin by trypsin. The SDS-PAGE profile of tryptic digestion products of ␣A-crystallin in the absence and presence of 3 mM ATP at different digestion times at a 1:50 (w/w) ratio of trypsin to chaperone is shown in Fig. 7A. After 1 h of digestion it became clear that the presence of ATP has slowed down the cleavage. After 1.5 h of tryptic digestion, the band for the full-length ␣A-crystallin almost disappeared in the absence of 3 mM ATP (Fig. 7A, lane 5); this band but is distinctly visible in the presence of ATP (Fig. 7A, lane 6). This suggests that ATP somewhat increases the structural stability of the ␣〈-crystallin. The digestion experiment was also carried out at a 1:100 ratio between trypsin and chaperone (w/w) (Fig. 7B). The qualitative features are similar to Fig. 7A, and ATP-induced structural stabilization is distinctly visible at 3 h of digestion (Fig.  7B, lanes 5 and 6).
We have also compared the equilibrium urea unfolding patterns of ␣A-crystallin in the presence and absence of 3 mM ATP by tryptophan fluorescence measurements at various urea concentrations. Because the wavelength of maximum emission ( max ) of native and unfolded ␣A-crystallin is 337 and 350 nm, respectively, the data were plotted as the ratio of intensities at 337 and 350 nm as a function of urea concentration (Fig. 7C). Both profiles in the presence and absence of ATP show sigmoidal shape. Analysis of the sigmoidal curves revealed that the transition midpoint (C1 ⁄2 ) of the curve in presence of ATP occurs at 2.82 M urea compared with 2.51 M urea in the absence of ATP, indicating somewhat higher stability in the presence of ATP. It can be seen that the profiles lacked strong cooperativity indicating multistate transitions. Sun et al. (39) analyzed the unfolding profile of recombinant ␣A-crystallin by a three-   Table  III. The standard free energy change of unfolding of ␣A-crystallin at zero urea concentration at 37°C in the absence of ATP is 19.9 kJ/mol. This value of ⌬G 0 compares well with a value of 26.7 kJ/mol reported by Sun et al. (39) for the unfolding of ␣A-crystallin at 25°C at infinite dilution of guanidine hydrochloride. In presence of 3 mM ATP at 37°C, ⌬G 0 increases to 24.4 kJ/mol (Table III). Thus, ATP induces an enhanced stability (⌬⌬G 0 ) of 4.5 kJ/mol of ␣A-crystallin. This difference, although small, is quite reproducible.

Effect of ATP-induced Association between ␣-Crystallin and Its Substrate on Its Chaperone Activity
To explore the possible implication of this effect of ATP on the chaperone-substrate interaction, we studied its effect on the chaperone-like properties of ␣-crystallin. Because the effect is significant around the physiological temperatures, we chose the chaperone assay systems at 37°C. Cleavage of the disulfide bond of insulin leads to the aggregation of the insulin B chain, and the presence of ␣-crystallin prevents this aggregation (12). We have compared the chaperone activity of recombinant ␣Bcrystallin incubated with insulin at 37°C in the presence and absence of 3 mM ATP (Fig. 8A). At a 0.5:1 (w/w) ratio of ␣Bcrystallin to insulin, 70% protection against aggregation is obtained at 1 h in the absence of ATP (Fig. 8A, trace 2). In the presence of 3 mM ATP, 94% protection is obtained (Fig. 8A,  trace 3). Therefore, ATP-induced association between insulin and ␣B-crystallin prior to the assay has enhanced the chaperone activity of ␣B-crystallin.
␣-Crystallin is also known to assist the in vitro refolding of many enzymes (16, 17, 25, 26). We studied the refolding of LDH at 37°C by recombinant ␣B-crystallin in the presence and absence of 3 mM ATP by 100-fold dilution of 6 M GdnHCl solution of LDH into refolding buffer (Fig. 8B). In the absence of both ␣B-crystallin and ATP in the refolding buffer only 5% activity could be recovered. In the presence of ␣B-crystallin without ATP ϳ40% activity could be regained. When the refolding buffer contained both ␣B-crystallin and 3 mM ATP, 48% activity could be regained. The error bar of each data point shown in Fig. 8B represents the standard error of LDH activity measurements in triplicate experiments. It reflects that al-though the enhancement of the LDH activity yield in the presence of ATP is small, it is real and clearly beyond the limits of experimental error. This result is in agreement with that reported by Muchowski and Clark (17) who observed the activity yield of ␣B-crystallin assisted refolding of citrate synthase to increase by 10% in the presence of ATP. This effect of ATP, leading to a marginal increase in refolding yield, is also seen with ␣Aand ␣ L -crystallin and also with other substrates such as malate dehydrogenase or carbonic anhydrase (data not shown). When ATP is replaced by ATP␥S, 46% activity is regained. This shows that ATP hydrolysis is not required for this activity. Both the insulin aggregation assay and the refolding activity assay reveal that presence of ATP has enhanced the chaperone activity of ␣-crystallin. DISCUSSION We have shown that ATP, instead of dissociating the chaperone-substrate complex, holds them gently together. The effect is not only seen with the highly purified recombinant preparation of ␣Aand ␣B-crystallin but also with bovine ␣ Lcrystallin (Table I). The weak binding can be inferred from the fact that the binding is reversible at least partly, and the binding affinity is low compared with that, for example, between ␣B-crystallin and the destabilized T4 lysozymes that are in the low submicromolar range (40). Here both chaperone and substrate, involved in the binding interaction, exist in states not far from the native states of the proteins. Native state fluctuations can give rise to "ladder states," having small energy differences centering around the ground state energy (41,42). In a series of papers (40,(43)(44)(45) from Mchaourab and co-workers, it has been demonstrated that ␣-crystallin can efficiently recognize and bind these degenerate native-like states of the substrates with varying affinity. From thermodynamic considerations, they have shown that the minor alteration in the stabilization energy of a substrate can trigger binding with the chaperone (44,45). In fact, for a given pair of proteins, initiation of binding interaction may reflect a delicate shift in the free energy states of any one or both of the interacting proteins. Our data clearly indicated that ATP induced distinct changes, in terms of the exposure of hydrophobic surfaces and structural stabilization by 4.5 kJ/mol, to the chaperone, but no similar changes were observed for the substrate. Thus, in this case, it is the ATP-induced shifting of the energy states of the chaperone that perhaps triggers the binding. ATP may also induce microscopic fluctuations to both chaperone and substrates in addition to these macroscopic changes adding to the affinity for binding.
We estimate that at 1:1 mole ratio between ␣-crystallin and ␥-crystallin, about half of the total substrate protein remains associated with the chaperone at 37°C. This estimate is significant in the context of the eye lens that contains a dense mass of proteins having a subunit concentration exceeding 20 mM in the cortical region (7,23). With 6 mM ATP physiologically present in the lens (27), much of its substrate proteins may remain in close contact with the ␣-crystallin. This ensures that if a substrate molecule is under stress and get partly unfolded, it has far less chance to form intermolecular aggregates, which could otherwise threaten the lens transparency. Moreover, such a high concentration of protein in the lens may only be possible with efficient packing. ATP may play an important role in the packing of these protein molecules in that dense environment by promoting association. By restricting freedom of the substrate proteins in the lens, it also reduces their chance to aggregate and improve refractive properties of the lens.
It is interesting to note that ATP not only promotes the binding of substrates with ␣-crystallin, it also enhances the binding of partially unfolding substrates also (Table I and Fig.  3). Our results also show that this induction of association between ␣-crystallin and its substrate by ATP occurs through additional exposure of hydrophobic sites on the surface of ␣-crystallin (Figs. 5 and 6). ␣-Crystallin has exposed hydrophobic pockets in the absence of ATP that binds to bis-ANS (12,15,36,37). The substrate protein ␥-crystallin or carbonic anhydrase binds much less bis-ANS. Binding between ␣-crystallin and its substrate at 37°C is negligible in the absence of ATP (Fig. 2) indicating that the already exposed sites are not sufficient to induce association. It is the additional exposure of bis-ANS-binding sites on ␣-crystallin by ATP that led to association. It may be pointed out here that the presence of ATP has already been reported to increase the binding of ␣-crystallin to lens membranes by 35% (46). This additional exposure is not because of unfolding or a molten globule-type swelling of ␣-crystallin but is due to reorganization of the hydrophobic residues. Partial unfolding or swelling would have reduced the stability of the structure of ␣-crystallin. Instead, the urea unfolding data (Fig. 7C) of ␣-crystallin reveal that in presence of ATP its structure become somewhat more compact and stabilized by 4.5 kJ/mol. The resistance to trypsin digestion in the presence of ATP (Fig. 7, A and B) also supports the compact structure of ␣-crystallin. Thus, the presence of ATP provides additional resistance against cleavage of ␣-crystallin by the endogenous proteases of the lens.
The mere presence of ATP is not enough to explain the results as dialyzing most (Ͼ95%) of the ATP out still showed the effect. The physical separation of free ATP from ␣-crystallin in the gel filtration column also showed the ATP effect, although somewhat reduced. An ATP-binding model seems to explain the results. Equilibrium studies of the ATP binding with ␣-crystallin reported a binding constant of 8 ϫ 10 3 M Ϫ1 at 37°C (24). Two charged clusters in the ␣-crystallin domain, namely Lys 82 , His 83 , Lys 90 , Lys 92 and Arg 116 , His 119 , Arg 120 , Lys 121 , Arg 123 , have been proposed as possible ATP-binding sites in ␣B-crystallin (31). Because the ␣-crystallin domain is highly conserved, human ␣A-chain also contains similar clusters, namely Arg 65 , Arg 68 , Lys 70 , Lys 78 , His 79 and Arg 112 , His 115 , Arg 116 , Arg 117 , Arg 119 . Numerous hydrophobic residues flank such clusters (22). Binding of ATP would neutralize the charge cluster pulling them inside the globular structure, which may reorient some of these hydrophobic residues to the surface of ␣-crystallin to trigger the association with the substrate.
Whether ATP hydrolysis is involved in its role is an important question. Our results show that ATP hydrolysis is not required to induce association between ␣-crystallin and its substrate, because nonhydrolyzable ATP␥S was able to produce the same effect. ATP␥S was also able to reproduce the resistance to trypsin digestion, hydrophobic exposure, as well as the stabilizing effect in urea unfolding studies (data not shown) observed with ATP. The effect is specific to adenine triphosphate moiety as the monophosphate and diphosphate variety were unable to produce a similar effect (Fig. 6). This again shows that binding of ATP or its analogue and not its hydrolysis is responsible for this effect.
The association between ␣-crystallin and its substrate seems to be relevant also to the chaperone-like property of the former. It has been reported that if ␣-crystallin could be associated with substrate by short term preincubation at elevated temperature, its ability to prevent chemical, ultraviolet ray, and oxidation-induced aggregation of substrate proteins is greatly enhanced (47). The authors were unaware as to how such an association might take place in vivo. Our results provide concrete evidence that the presence of ATP may lead to such an association. Our in vitro insulin aggregation assay of the chaperone activity at the physiological temperature of 37°C (Fig.  8A) clearly shows that ATP-induced association between insulin and ␣-crystallin has indeed improved the chaperone-like properties of ␣-crystallin. We believe that such an association may be of vital importance to preserve the ocular lens against various physiological stresses.  Like classical molecular chaperones, ␣-crystallin has also been implicated in the in vitro refolding of unfolded substrates despite its questionable relevance in vivo (16,17,25,26,48,49). Whereas ␣-crystallin alone in some cases was able to reactivate the enzymes (25,48,49), the presence of ATP enhanced the reactivation yield in some cases (17,26). The role of ATP in such experiments is not fully understood. On the contrary, the role of ATP is much better understood for the refolding of enzymes mediated by the chaperones GroEL, DnaK, etc. (28). ATP weakens the interactions between chaperone and substrate and thus helps the latter to refold. The dissociation by ATP is often activated by its hydrolysis carried out by the fairly well known ATPase activity of the above-mentioned chaperones. Taking clue from such a mechanism, a similar role of ATP has also been envisaged for the chaperone activity of ␣-crystallin by some workers (25,26). However, the ATPase activity of ␣-crystallin is highly controversial. Although Kantorow et al. (29,30) reported weak autophosphorylation activity of ␣-crystallin, Horwitz (23) by using highly purified recombinant ␣〈and ␣〉-crystallin and sensitive assay method failed to detect any sign of ATPase activity. The role of ATP hydrolysis to refold substrates by ␣-crystallin has also been debated in the literature. Muchowski and Clark (17) reported that the refolding yield of substrate protein reduced when ATP was replaced by nonhydrolyzable ATP␥S. However, Rawat and Rao (16) reported that ATP hydrolysis is not required for refolding. However, both of these studies reported that the presence of ATP increased the refolding yield. Our results on the reactivation of LDH show that without ATP a maximum of 40% reactivation was observed. If ATP triggers the release mechanism, how can someone account for such a significant reactivation without ATP? Many small heat shock proteins including ␣-crystallin are believed to be ATP-independent chaperones (22,49,50). ATP enhanced the reactivation of LDH to 48%. We also show that ATP hydrolysis is not required for this enhancement of the reactivation yield, as replacing ATP by ATP␥S produced no significant change (Fig. 8B). We believe that such an enhancement is due to the increased binding between the substrate LDH and ␣-crystallin in the presence of ATP, which reduced the nonproductive substrate aggregation on rapid dilution, and is not due to the ATP-mediated release mechanism. To date no convincing evidence for the active mechanism of substrate release from its complex with ␣-crystallin is known. Perhaps the folding of bound substrates take place within the ␣-crystallinsubstrate complex without dissociation. This model is consistent with the "marsupium" (kangaroo's bag) model (51) found to be valid for enzyme reactivation by other chaperones such as tubulin (52) and nucleolar protein B23 (53).