INTRODUCTION

The composition of the vertebrate eye lens is dominated by a group of structural proteins known as crystallins. The largest of these, -crystallin, is a dynamic multimeric complex composed of two types of homologous subunits, A- and B-crystallin (1). A few years ago, it was discovered that these subunits are also constitutively expressed in various non-lenticular tissues, suggesting that their function is more than merely structural (2, 3, 4, 5). In addition, increased expression of B-crystallin has been observed in a variety of neurodegenerative disorders such as multiple sclerosis (6), Alexander's disease (7, 8), and Alzheimer's disease (9, 10). Although the physiological significance of this extralenticular expression is unknown, it certainly relates to the fact that -crystallins belong to the family of small heat shock proteins (hsp)¹ (11, 12). Among the shared features of -crystallins and other small hsp are the conferring of thermotolerance (13, 14), interaction with actin (15, 16), phosphorylation (17, 18), and intracellular relocalization upon stress (19, 20). Furthermore, in vitro experiments have shown that A- and B-crystallin as well as hsp25 can act as molecular chaperones by suppressing aggregation of denaturing proteins (21, 22, 23). Unfortunately, the mechanism of this functional activity is unclear because the three-dimensional structure of -crystallin is unknown.

Native -crystallins occur as polydisperse particles with an average molecular mass of about 800 kDa. Understanding the multimeric arrangement of these subunits is essential for defining structure-function relationships. Recently, solvent-accessible cylindrical (24) and open micellar arrangements (25) were proposed as alternatives for previous models such as the solid three-layered (26, 27) and rhombododecahedric structures (28). However, none of these quaternary structure models is based on substantial experimental evidence. The subunits mainly comprise -sheets (29, 30), and several biophysical studies endorse a tertiary structure composed of two compact domains and an unstructured C-terminal region (31, 32, 33, 34). Indeed, as shown by NMR spectroscopy, -crystallin (35) and hsp25 (36) have unstructured, flexible, and solvent-exposed C-terminal extensions. Finally, like other chaperones, such as GroEL, -crystallin seems to suppress aggregation by providing appropriately placed hydrophobic surfaces for denaturing protein substrates (37, 38, 39).

The purpose of the present work is to study the influence of the C-terminal extension on chaperone-like activity. It is known that the extensions are very sensitive to various kinds of enzymatic and non-enzymatic modifications (for a review see Ref. 1). For example, enzymatic truncation of the C terminus with calpain II or trypsin results in a decreased chaperone-like activity (40, 41). Although sequence alignment shows that the extensions in -crystallins and small hsp are quite variable, their composition is remarkably dominated by polar and charged amino acid residues (12, 42). This suggests that the extensions may have a role in solubilizing the hydrophobic complexes formed between -crystallin particles and denatured proteins. To explore such a hypothesis, we have studied the structure and functional behavior of a number of A-crystallin mutants in which charged and hydrophobic residues were inserted into the C-terminal extension.

EXPERIMENTAL PROCEDURES

Ampicillin, chloramphenicol, isopropyl-1-thio--D-galactopyranoside, bovine pancreas insulin, dithiothreitol, and protease inhibitors were obtained from Sigma. 8-Anilino-1-naphthalenesulfonic acid was purchased from Molecular Probes. The Escherichia coli strain used for expression experiments was B BL21(DE3)pLysS (43).

cDNAs encoding wild-type and mutants of bovine A-crystallin were transformed in the host E. coli B BL21(DE3)pLysS. Induction, cell lysis, and fractionation were essentially performed as described by Merck et al. (33). Proteins were purified via anion exchange chromatography with a Fast Flow DEAE-Sepharose column (Pharmacia Biotech Inc.) and a DE52 column (Whatman) (44). For structural and functional studies, wild-type and mutants were refolded in urea as described previously (44).

Chaperone-like activity was determined essentially as described by Farahbakhsh et al. (45). Briefly, various amounts of wild-type and mutant A-crystallins (61, 122, or 245 µg) were preincubated with 245 µg of bovine pancreas insulin for 3 min at 40 °C. Denaturation of insulin was initiated by addition of 20 µl of 1 M dithiothreitol and scattering was monitored for 15 min at 360 nm using a Perkin-Elmer Lambda 2 UV-visible spectrophotometer equipped with a thermostated circulating water bath and a thermocouple to register the sample temperature. Total volume was 1.0 ml, and all solutions were in phosphate buffer (0.1 M Na₂SO₄, 20 mM NaP_i, pH 6.9).

¹H NMR spectra were recorded at 400 and 600 MHz on Varian-Unity 400 and Bruker DMX-600 spectrometers, respectively. Proteins were at a concentration of 28 mg/ml. Measurements were performed in phosphate buffer (0.1 M Na₂SO₄, 20 mM NaP_i, 99.9% D₂O), and, depending on the protein sample, the pH was varied from 6.9 to 7.5 to optimize solubility. All two-dimensional spectra were acquired with 512 t₁ increments over 2048 t₂ points using the time-proportional phase incrementation method to obtain pure-phase spectra (46) with 32 or 64 scans/t₁ increment. Clean-TOCSY spectra (47) using a MLEV-17 spin lock (48) were acquired with various mixing times in the range from 30 to 66 ms. The NOESY spectrum (49) of the ALDKG mutant had a mixing time of 120 ms. For spectra acquired in D₂O, the residual solvent resonance was removed by presaturation for 1.5 s with a low powered, continuous pulse using the transmitter. For the two-dimensional experiments in H₂O, solvent suppression was accomplished by application of the WATERGATE scheme (50) for the read pulse. In this scheme, solvent suppression is achieved by a combination of gradient-tailored excitation and pulsed-field gradients. Spectra were processed with either Varian or Bruker software or, in the main, with FELIX 95.0 software (Biosym). For processing, the data were zero filled to a matrix size of 2048 × 2048 data points prior to multiplication in both dimensions with shifted sine-bell or Gaussian window functions, followed by Fourier transformation. For spectra acquired at 25 °C, chemical shifts were referenced to the residual water resonance at 4.76 ppm. The temperature inside the probe at 25 °C and higher temperatures was calibrated with ethylene glycol.

Samples containing reconstituted A-crystallin were diluted with phosphate buffer (0.1 M Na₂SO₄, 20 mM NaP_i, pH 6.9) to a final concentration of 200 µg/ml and centrifuged for 15 min at 15,000 × g to remove possible light scattering particles. Fluorescence spectra were measured at 25 °C on a Hitachi F-3000 spectrofluorometer equipped with a four-cuvette holder accessory and a thermostated circulating water bath. The excitation wavelength was set to 295 nm with a 5-nm band pass. Fluorescence emission was detected over a range of 300-420 nm with a 3-nm band pass at right angles to the incident beam after passing through a 10-mm quartz cuvette.

Samples containing reconstituted A-crystallin were diluted with phosphate buffer (0.1 M Na₂SO₄, 20 mM NaP_i, pH 6.9) to a final concentration of 100 µg/ml. Scattering at 360 nm was measured as a function of temperature using a Perkin-Elmer Lambda 2 UV-visible spectrophotometer equipped with a thermostated circulation water bath and a thermocouple to register the sample temperature.

Circular dichroism (CD) spectra were obtained on a Jasco 720 CD spectropolarimeter. Solutions of A-crystallin at concentrations of 0.25-0.30 mg/ml were prepared in phosphate buffer (0.1 M Na₂SO₄, 20 mM NaP_i, pH 6.9). Experiments were performed over a temperature range from 21 °C to 70 °C using a quartz cell of path length 1 mm. Samples were heated at a rate of 1 °C/min and were left to equilibrate for an empirically determined period of 220 s at each temperature prior to accumulation of data. Spectra acquired at a particular temperature represent an average of four scans.

Gel permeation analysis was performed on a Superose 6 HR 10/30 prepacked column as described before (44). High molecular mass standards (Pharmacia Biotech Inc.) were used for calibration. Protein concentrations were determined in triplicate using the Bradford method (51). Binding studies with the hydrophobic probe 8-anilino-1-naphthalenesulfonic acid were performed as described before (52).

RESULTS

To explore the influence of the C-terminal extension of bovine A-crystallin on chaperone-like activity, we have characterized the structural and functional properties of a number of mutants, which were originally designed to assess the substrate requirements for transglutaminases (53). As shown in Table I, these mutants contain modified C termini in which charged and hydrophobic residues are introduced. Wild-type and mutants were induced in B BL21(DE3)pLysS host cells. Except for the ALWKG mutant, all proteins were predominantly present in the water-soluble fraction of the bacterial lysates. Apparently, the presence of a hydrophobic tryptophan residue in the C-terminal region is unfavorable for the solubility of A-crystallin in the E. coli host cell. The expression products were purified by native and denaturing anion exchange chromatography, resulting in 95-98% pure proteins as estimated by SDS-PAGE (Fig. 1). Additionally, their identity was confirmed by amino acid analysis (data not shown). To avoid isolation and purification artifacts during further analysis, all purified proteins were simultaneously subjected to reconstitution. The refolded proteins were examined for their ability to form multimeric complexes on a Superose 6 gel permeation column. Table I shows that most of the altered C termini do not affect multimerization significantly. The only exception was the ALWKG mutant, where a slightly smaller multimeric mass was observed.

Table I. The influence of C-terminal mutations in bovine A-crystallin on multimeric mass Samples of 40 µg of protein in 200 µl of phosphate buffer, pH 6.9, were analyzed on a Superose 6 gel-permeation column. Different preparations of wild-type and mutant A-crystallins were analyzed at least in duplicate. Molecular masses were estimated by calibrating the column with thyroglobulin (669 kDa), ferritin (440 kDa), and catalase (232 kDa). C-terminal residues of mutant A-crystallinsa Molecular mass MDa  170 171 172 173 -Ala-Pro-Ser-Ser-OH (wild-type) 0.64  ± 0.01 -Ala-Pro-Ser-Lys-OH 0.64  ± 0.01 -Ala-Leu-Gly-Lys-Gly-OH 0.65  ± 0.01 -Ala-Leu-Arg-Lys-Gly-OH 0.65  ± 0.01 -Ala-Leu-Asp-Lys-Gly-OH 0.64  ± 0.01 -Ala-Leu-Trp-Lys-Gly-OH 0.61  ± 0.02 a  Wild-type and mutants are referred to in text using amino acid one-letter codes.

SDS-PAGE of wild-type and mutant A-crystallins.

The chaperone-like activity of wild-type and mutants was assessed by determining their ability to prevent the chemically induced aggregation of insulin B-chains at 40 °C. Reduction of the disulfide bonds between the A- and B-chains of insulin with dithiothreitol, rapidly leads to a specific precipitation of aggregated B-chains. As shown in Fig. 2, this event can easily be monitored by measuring the absorbance at 360 nm. Insulin was reduced at different A-crystallin to insulin mass ratios (indicated as percentages). Fig. 2A shows that aggregation of insulin B chains is almost completely prevented in the presence of 100% of wild-type A-crystallin. The APSK, ALGKG, and ALRKG mutants display a small decrease of functional activity (Fig. 2, compare panels B-D with panel A). This decrease is probably caused by the positively charged lysine and arginine residues rather than the uncharged leucine and glycine residues, because there is no substantial difference in activity between the ALGKG and the APSK mutant (Fig. 2, compare panel C with panel B). The ALDKG mutant has the same net charge as the wild-type protein, but the polarity of its extension is obviously increased. However, this has no detectable effect on the ability to bind denaturing substrates (Fig. 2, compare panel E with panel A). On the contrary, a dramatic effect on chaperone-like activity is observed upon introducing a hydrophobic tryptophan residue. As Fig. 2F shows, a considerable level of insulin B aggregation occurs even at an ALWKG mutant to insulin mass ratio of 100%, whereas the wild-type and the other mutants provide almost complete protection at this ratio (Fig. 2, A-E). Thus, unlike charged residues, a strongly hydrophobic residue in the C-terminal extension can seriously disturb functional activity.

Chaperone-like activity of wild-type and mutant A-crystallins.

The ALDKG and ALWKG mutants were subjected to two-dimensional ¹H NMR spectroscopy to determine whether the diminished functional activity of the latter is related to an altered conformation of the C-terminal extension. Fig. 3 shows the cross-peaks arising from -CH resonances in TOCSY spectra at 25 °C of the ALDKG and ALWKG mutants in D₂O. The majority of the cross-peaks in the ALDKG spectrum (Fig. 3A) were assigned using the sequential-assignment procedure of Wuthrich (54), which involved the combined use of TOCSY and NOESY spectra in H₂O (data not shown). The chemical shifts of the assigned resonances in the ALDKG spectra (TOCSY and NOESY) are very similar to those it shares with wild-type A-crystallin (35) and random coil peptides (55). Therefore, it is clear that this mutant has a flexible C-terminal extension consisting of the last 10 residues (from Glu-165 to Gly-174) and that this extension has little preferred conformation. Comparison of panels A and B (Fig. 3) shows that the ALWKG mutant displays fewer cross-peaks than the ALDKG mutant. For the ALWKG mutant, cross-peaks from the -CH protons were present for the first part of the C-terminal extension (Glu-165 to Ala-170) but not for the remainder (Leu-171 to Lys-173). In the ALDKG mutant, Gly-174 -CH₂ does not give rise to cross-peaks in this region of the TOCSY spectrum as these protons are equivalent. It was apparent from the one-dimensional ¹H spectra of both mutants, however, that the strong singlet resonance from the -CH₂ protons of Gly-174 in the ALDKG mutant (at 3.73 ppm) was absent in the ALWKG mutant (data not shown). No cross-peaks were observed from Lys-166 -CH in the TOCSY spectra of the ALWKG mutant, whereas weak signals were observed in the spectra of wild-type A-crystallin (35) and the ALDKG mutant. This absence could be explained by additional weakness arising from a reduced flexibility due to Lys-166 being at the start of the C-terminal extension and/or the proximity of the Lys-166 -CH resonance to the water resonance resulting in cross-saturation effects. Despite the absence of cross-peaks from the -CH protons of residues in the terminal part of the extension in the ALWKG mutant, resonances were observed from their side chains, e.g. weak cross-peaks were present in the TOCSY spectra between the ,-CH and -CH₃ protons of Leu-171 and from the - to -CH₂ protons of Lys-173 and Lys-166. In the one-dimensional ¹H NMR spectrum of the ALWKG mutant, broad resonances were also observed at chemical shifts indicative of a tryptophan residue. Presumably, these resonances arise from the solvent-accessible aromatic protons of Trp-172. In the TOCSY spectrum, however, these resonances did not give rise to any cross-peaks, which implies that they have reduced flexibility. In conclusion, the NMR data indicate that the terminal sequence of the C-terminal extension in the ALWKG mutant (Leu-171 to Gly-174) has reduced flexibility compared with the first part of the extension (Glu-165 to Ala-170) and compared with the entire C-terminal extension in the ALDKG mutant.

The ALWKG mutant was subjected to fluorescence spectroscopy to probe the polarity of the micro-environment in which the introduced tryptophan is located. In general, the fluorescence emission maximum of tryptophan residues can vary from about 330 nm (completely buried) to 350 nm (completely exposed) (56). The ALWKG mutant has two tryptophan residues, one at position 9 and the other at position 172 in the C-terminal extension. Both residues contribute to the measured emission spectrum shown in Fig. 4. Since the ALDKG and ALWKG mutants share the Trp-9 in the N-terminal domain, the contribution of the additional Trp-172 in the latter could be determined by calculating the difference between the measured emission spectra. As shown in Fig. 4, the extrapolated spectrum of Trp-172 has a _max of 348 nm, indicating that this residue is almost fully exposed to the solvent. Furthermore, the emission intensity of Trp-172 is much lower than that of Trp-9, which also indicates a polar environment because exposed residues display a relatively low quantum yield.

To determine whether the thermostability of A-crystallin is conserved upon C-terminal alterations, wild-type and mutants were heated and the absorption at 360 nm was monitored as a function of temperature. As shown in Fig. 5A, the ALWKG mutant, but not the wild-type and the other mutants, rapidly starts to precipitate at 59 °C. To probe further the reduced temperature stability of the ALWKG mutant, the conformation of the ALWKG and ALDKG mutants was investigated as a function of temperature using CD and NMR spectroscopy. Both mutants gave at room temperature a CD spectrum that is typical for -crystallin, i.e. the profiles were characteristic for a high proportion of -sheet structure, showing an ellipticity minimum at 218 nm (data not shown). With increasing temperatures above 44 °C, an increase in ellipticity at 205 nm was observed for the ALDKG mutant, such that when the ellipticity was plotted against temperature, a sigmoidal-shaped curve was obtained with a half-point at around 50 °C (Fig. 5B). The increase in ellipticity at 205 nm is indicative of an increase in random coil structure of the protein at these higher temperatures. Similar behavior was observed for native -crystallin, except that the midpoint of this transition was at around 60 °C (57). The CD spectra of the ALWKG mutant were similar to those for the ALDKG mutant over a range from 20 °C to 52 °C (data not shown). However, between 52 °C and 55 °C, the ALWKG mutant precipitated, which was reflected in the uncharacteristic change of ellipticity at 205 nm (Fig. 5B). Nevertheless, the similarities between the CD spectra for both mutants below 52 °C suggest that the ALWKG mutant does not undergo any gross unfolding or unusual conformational changes prior to precipitation.

Thermostability of wild-type and mutant A-crystallins.

Temperature-dependent NMR analysis of the ALWKG mutant was consistent with this result. One-dimensional ¹H NMR spectra were acquired in 5 °C increments from 25 °C to 55 °C with TOCSY spectra being acquired at 35, 45, and 55 °C. Compared with the spectrum at 25 °C (Fig. 3), there was little change in the NMR spectrum with increasing temperature up to 50 °C. Thus, the same cross-peaks were present and no new cross-peaks were observed, indicating that there was no major unfolding or change in flexibility of the protein at higher temperature. At 55 °C, however, the spectrum became broad and sensitivity was lost due to the sample precipitating out of solution. It would seem, therefore, that increasing temperature up to approximately 55 °C had little effect on the flexibility of the C-terminal extension of the ALWKG mutant or other regions in the molecule.

DISCUSSION

Since it became clear a few years ago that -crystallin is a small heat shock protein with a role in various non-lenticular tissues, there has been great interest in determining the mechanism by which -crystallin interacts in vitro with aggregation-prone proteins. Up to now, it is not known which domains or structural elements of -crystallin subunits are involved in this interaction. In this paper we have focused on structural and functional effects of amino acid replacements in the C-terminal extension of bovine A-crystallin. As shown in Fig. 2, chaperone-like activity is only slightly influenced upon introduction of additional charged residues in this region. In contrast, Fig. 2F demonstrates that insertion of a hydrophobic tryptophan residue decreases the aggregation-suppressing behavior dramatically. To understand this functional behavior at a structural level, the ALWKG mutant was subjected to biophysical examination to explore the structural aspects of its extension. Two-dimensional ¹H NMR analysis of the ALDKG and ALWKG mutants showed that substitution of a hydrophilic aspartic acid residue by a hydrophobic tryptophan residue at position 172 leads to a pronounced reduction in flexibility specifically in the vicinity of this residue (Fig. 3). The introduction of tryptophan at position 172 makes the terminal region of the extension relatively hydrophobic in character, i.e. the apolar amino acids Ala-170, Leu-171, Trp-172, and Gly-174 are only offset by the charged Lys-173. Considering this in light of the fact that -crystallin is known to expose a significant hydrophobic surface to solution (37, 39), the reduced flexibility of Ala-170 to Gly-174 is likely to be the result of an extension-domain interaction. This interaction might explain why the multimeric mass of the ALWKG mutant appears to be somewhat lower than that of the wild-type and the other mutants (Table I), i.e. immobilization of the C-terminal extension slightly diminishes the average Stokes' radius of the multimeric particles. Furthermore, the observation that the immobilized tryptophan is still solvent-accessible (Fig. 4) indicates that its binding site is not buried in a domain structure. This result supports the possibility that the hydrophobic extension interacts with substrate binding sites, as these sites are probably located on the surface of the multimeric particles to allow for an efficient aspecific interaction with unfolding proteins.

The immobilization of the terminal part of the ALWKG extension could explain the decreased chaperone-like activity of the mutant in several ways. First of all, as mentioned above it might be possible that binding of the extension reduces the amount of hydrophobic surface available for unfolding substrates. However, this concept is apparently not supported by binding studies with the fluorescent probe 8-anilino-1-naphthalenesulfonic acid, because at 40 °C no significant differences in absolute hydrophobicity were found between wild-type, the ALDKG mutant, and the ALWKG mutant (data not shown). Another possibility is that the immobilized extension diminishes the accessibility of the substrate binding sites by steric hindrance. On the basis of the present data, this hypothesis cannot be excluded, although it does not provide a rationale as to why -crystallin subunits need a flexible C-terminal extension at all. Sequence comparisons show that the extension is quite variable, but so far no mammalian -crystallin subunits have been found lacking such an extension (12, 58). -Crystallin multimers seem to have a defined number of substrate binding sites, which may become occupied by denaturing proteins until a level of saturation is reached (44, 59). This occupation is accompanied by an increase of hydrophobicity because the substrate proteins are bound in a state that is at least partially unfolded (60). Considering this increasing hydrophobicity together with the observation that the extension of wild-type A-crystallin is not affected by substrate binding (60), it might be possible that the extension functions as a sort of solubilizer in the complex. That is to say, the strong hydration of the polar extensions compensates for the relatively high surface hydrophobicity of the complexes formed between A-crystallin and unfolded substrates. Consequently, the poor functional activity of the ALWKG mutant can be explained by the comparatively high tendency of ALWKG-substrate complexes to aggregate because their extensions are immobilized and far less hydrated than those in wild-type A-crystallin and the ALDKG mutant.

The heat-induced precipitation of the ALWKG mutant (Fig. 5A) is a remarkable observation because the wild-type and the other mutants are very thermosoluble. In addition, native -crystallin and mutants of A-crystallin that were analyzed previously, including the elongated A^ins-polypeptide, do not precipitate up to at least 75 °C (30, 44, 52). The CD and NMR experiments do not provide any evidence that precipitation of ALWKG mutant at around 55 °C is preceded by precocious unfolding of domain structures. Therefore, this insolubilization probably has a more indirect cause. Although -crystallin does not denature upon heating, at least some of the -sheet structure is irreversibly lost in the temperature range between approximately 50 °C and 70 °C (57). In addition, differential scanning calorimetry has revealed a broad endothermic transition for -crystallin at about 60 °C (57, 61). Interestingly, hydrophobic probe binding studies suggest that the apparent thermotropic transition of -crystallin is accompanied by a significant increase in surface hydrophobicity (37, 39). In general, proteins become more hydrophobic with increasing temperature because the water ordering around their apolar surfaces tends to melt out (56). Although this thermodynamic phenomenon may play a role to some extent, it does not explain why the surface hydrophobicity of -crystallin is still increased after a cooling down period (39). Therefore, it is conceivable that the increased surface hydrophobicity is directly correlated to an irreversible structural transition. Fig. 5B shows that the ALDKG and ALWKG mutants display a loss of -sheet structure with increasing temperature similar to that for native -crystallin. Furthermore, the precipitation of the ALWKG mutant occurs between 52 and 59 °C, which is in the same range as the thermotropic transition temperature of -crystallin. Because of this, precipitation of the ALWKG mutant may also be related to a structural transition and a concomitant increase in surface hydrophobicity. Basically then, the temperature-induced insolubilization of the ALWKG mutant may have the same cause as its diminished chaperone-like activity, namely that the solubilizing capacity of the immobilized extensions is insufficient to compensate for the increasing hydrophobicity of the domain core of the protein.

Whereas introduction of a tryptophan causes structural changes, i.e. immobilization of the C-terminal extension and a concomitant functional deterioration, insertion of additional charged residues in the C-terminal extension of A-crystallin has only minor effects on the protein's functionality. Perhaps the very presence of a hydrated and flexible extension is more important for solubility and chaperone-like activity than its precise amino acid sequence. The importance of the extension has been noticed before because the C-terminal domain of rat A-crystallin becomes completely insoluble upon removal of the 23 C-terminal residues.² Furthermore, the observation that the ALWKG mutant is soluble under normal conditions may well be due to the fact that part of its extension, i.e. Glu-165 to Ala-170, is still flexible.

We did not study the effect of hydrophobic residues other than tryptophan. However, it is conceivable that any hydrophobic stretch which restricts the conformational freedom of the C-terminal extension will also affect functional activity. This implies that the polar extension of wild-type A-crystallin has an important physiological function. In fact, the evolutionary conservation of such a polar extension may be dictated by the fact that a non-hydrophobic unstructured C-terminal region is required for increasing the capacity of A-crystallin to form soluble complexes with aggregation-prone polypeptides.