Negative Charges in the C-terminal Domain Stabilize the αB-Crystallin Complex*

αB-Crystallin is one of the six known mammalian small heat-shock proteins (sHsps). These are characterized by the presence of a conserved sequence of 80–100 residues, which constitutes the putative C-terminal domain. Like other sHsps, αB-crystallin forms multimeric globular complexes, often in combination with related sHsps. Here we show that in a yeast two-hybrid system, αB-crystallin can specifically interact with itself as well as with αA-crystallin and Hsp27. Analyses of the separate domains show that the conserved C-terminal domain (CαB) is essential for this interaction between subunits. To try and detect residues that are important in subunit interaction, the CαB domain was used in a two-hybrid screen as bait to select randomly mutated CαB mutants. In this way we obtained nine mutants that were still able to interact with wild-type CαB despite the presence of up to 15 replacements. Similarly, we obtained 16 mutants that were unable to bind, because of the presence of just three to nine replacements. In binding CαB mutants, lysine residues were most often replaced by glutamic acid residues, and in non-binding CαB mutants, acidic residues were often found to be replaced by non-charged residues. This indicates that negative charges are important for subunit interaction and we propose a model to explain this role of acidic residues. Furthermore, we observed that two homologs of αB-crystallin, αA-crystallin and Hsp27, generally interact similarly with the binding and non-binding CαB mutants as does αB-crystallin. This suggests that interactions involved in the complex formation of these three sHsps are largely comparable.

␣B-Crystallin is a protein with dual function. It is an abundant structural protein in the vertebrate eye lens (1). Outside the lens it is especially expressed in tissues with high oxidative activity such as heart, striated muscle, and kidney, but also in degenerative disorders of the nervous system (2)(3)(4). ␣B-Crystallin belongs to the family of small heat shock proteins (sHsps) 1 (5,6). The sHsps vary in molecular mass between 12 and 43 kDa and are characterized by the presence of an 80 -100-amino acid long conserved C-terminal region. Most members of this family form large globular particles, up to 700 kDa in size, which can consist of a variable number of subunits.
They are capable of preventing the aggregation of unfolding proteins (7)(8)(9). This chaperone-like activity may contribute to the most important in situ property of the sHsps, namely their ability to increase the stress tolerance of cells (10). In many cell types ␣B-crystallin is constitutively expressed, while in others the expression can be induced by different kinds of stress (11,12). In normal cells, ␣B-crystallin is mainly present as a soluble cytoplasmic protein. It can form mixed complexes with ␣Acrystallin, as well as with Hsp27 and Hsp20 (9,(13)(14)(15)(16). Under certain stress conditions, ␣B-crystallin interacts with intermediate filaments, and in this way it is probably able to modulate the intermediate network (17)(18)(19).
The tertiary structure of ␣B-crystallin is unknown. This is due to the polydispersity and flexibility of the complex, which hampers crystallization. From CD measurements and secondary structure predictions it appears that ␣B-crystallin has mainly ␤-pleated sheet and only little ␣-helix structure (20). It probably has a two-domain structure, an N-terminal hydrophobic domain and the conserved C-terminal domain extending into an exposed flexible C-terminal arm (21,22). Most recently, data about the quaternary structure of human ␣B-crystallin has been obtained by cryo-electron microscopy, showing that ␣B-crystallin forms variable globular structures containing a large central cavity (23).
As an indirect means to obtain information about the quaternary structure formation, we have employed a yeast twohybrid system to study subunit-subunit interactions of ␣Bcrystallin. We present the data that leads us to conclude that negatively charged residues in the C-terminal domain are important for the complex formation of the sHsps.

EXPERIMENTAL PROCEDURES
Cloning-The vectors pJK103 (LacZ reporter plasmid containing two LexA operators), pJG4 -5 (prey plasmid), and pEG202 (bait plasmid) were kindly provided by Roger Brent (24). The complete coding region of the rat ␣B-crystallin gene was amplified by PCR using the 5Ј primer ATCTGCCATGGACATAGCCATCCAC (NcoI site in bold) and the 3Ј primer GTCCTCGAGCTACTTCTTAGGGGCTG (Xho site in bold). For the N-terminal gene fragment (N␣B) corresponding to amino acid residues 1-66 plus the C-terminal sequence Ala-Met-Gly, the 3Ј primer ATCTGCTCGAGCTAACCCATGGCTGAGAGCCCAGTGTCAA (NcoI and XhoI site in bold) was used and for the C-terminal gene fragment (C␣B) corresponding to amino acid residues 68 -175, containing an Arg 69 -Gly mutation, the 5Ј primer ATCTGCCATGGGTATGGAGAAG-GACAGGTTC (NcoI site in bold). The resulting PCR products were, after digestion, cloned in-frame into the pJG4 -5 and pEG202 vectors. The complete coding sequence of the bovine ␣A-crystallin gene (25) and human Hsp27 (kindly provided by Matthias Gaestel) were cloned inframe into the pEG202 vector. The gene coding for desmin (26) and rat vimentin (27), obtained by PCR using the primers ATCTGCCATGGC-CACCAGGTCCGTG (NcoI in bold) and ATCTGCCATGGTTAT-TCAAGGTCATCGTG (NcoI in bold) were cloned in-frame into the pJG4-5 and pEG202 vectors. PCR was done with Pwo DNA polymerase (Boehringer Mannheim) and the obtained clones were checked for errors by sequence analysis. Wild type C␣B and mutants thereof were cloned in the NcoI,XhoI sites of the pET16b vector.
Random Mutagenesis of the ␣B-Crystallin Gene-The template pJG-C␣B was amplified by using two vector primers.  (29) with plasmids expressing a LexA fusion protein, an activator fusion protein, and the LacZ reporter plasmid pJK103. Transformants were selected on plates lacking histidine, uracil, and tryptophan, and were replicated on plates containing 5-bromo-4-chloro-3indolyl-␤-D-galactose to select for ␤-galactosidase producing cells or on plates lacking, beside histidine, uracil, and tryptophan, also leucine to select for growth. For determining the activation of the LacZ reporter, six independent colonies were grown in media containing galactosidase to OD 660 ϭ 0.5. Cells were harvested by centrifugation and lysed with ether in 100 mM KH 2 PO 4 /K 2 HPO 4 , pH 7.2, 10 mM KCl, 1 mM MgSO 4 , and 0.4% ␤-mercaptoethanol. The protein concentration of the cell extract was measured with the Bradford assay (30). The cell extract was incubated at 30°C with 2 mM o-nitrophenyl-␤-D-galactopyranoside in a CERES BioTek UV900c multireader for 1 to 6 h during which the OD 405 was measured. The ␤-galactosidase activity was calculated in arbitrary units from the slope of the linear part of the reaction divided by the total protein concentration in the extract.
Expression and Purification of Recombinant Proteins-Proteins were expressed from pET16b vectors in BL21(DE3) cells (Invitrogen). A culture inoculated from a single colony was incubated at 37°C for about 4 h. At OD 600 ϭ 0.5 the protein production was induced by the addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 0.5 mM. After 4 h, cells were harvested by centrifugation and lysed by sonication in TNE (50 mM Tris, pH 8.0, 10 mM NaCl, 1 mM EDTA). Insoluble proteins were removed by centrifugation. The proteins were purified over a DEAE-Sepharose Fast Flow column (Pharmacia) followed by a Superdex 75 PG 16/60 size exclusion column (Pharmacia). Circular dichroism (CD) spectra of wild type and mutant C␣B at concentrations of 0.25 mg/ml in phosphate buffer, pH 7.5, were obtained on a Jasco 720 CD spectropolarimeter.

Specific Interactions between sHsp Subunits Can be Detected
with the Yeast Two-hybrid System-The yeast two-hybrid system allows detection of specific interactions between proteins (24,32). However, the suitability of this system to study the interactions of ␣B-crystallin might be hampered by the latter's chaperone-like activity. We therefore first tested the specificity of interactions of ␣B-crystallin with itself and other sHsps. To this end, we made constructs which code for full-length ␣Acrystallin, ␣B-crystallin, or Hsp27 fused to the LexA DNAbinding domain (bait fusion proteins), as well as a construct which codes for ␣B-crystallin fused to the B42 activation domain (prey fusion protein). Bait and prey plasmids, in different combinations, were introduced into the yeast reporter strain EGY48, together with the LacZ reporter pJK103 (24). Expression of both bait and prey ␣B-fusion proteins in the yeast strain resulted in a very high expression of ␤-galactosidase activity (Table I). Yeast carrying the ␣B-bait or ␣B-prey plasmids along with control vectors expressed about 100 times less ␤-galactosidase, which is near background level (Table I). This confirms the specificity of the interaction between ␣B-crystallin fusion proteins in the two-hybrid system. Similar activations of the reporter gene were obtained by coexpression of ␣B-prey fusion protein with ␣A or Hsp27 fused to LexA. These interactions were again very specific since the ␣A and Hsp27 fusion proteins did not interact with control proteins. It can be concluded that ␣A-, ␣B-crystallin, and Hsp27, despite their chaperone-like properties, do not have a tendency to bind nonspecifically to proteins.
No Interaction of ␣B-Crystallin with Intermediate Filaments in the Yeast System-␣B-Crystallin is known to associate in vitro with the intermediate filament proteins desmin and vimentin and colocalizes under stress conditions in situ with the intermediate filament network (18,19). To assess whether such interactions can be detected in the two-hybrid assay, we constructed hybrid genes encoding full-length vimentin and desmin fused to the LexA DNA-binding domain and the B42 activation domain. The vimentin and desmin hybrid genes were introduced into the yeast reporter strain along with the LacZ reporter and the appropriate ␣B-crystallin hybrid gene. No interaction, as assayed by ␤-galactosidase expression, could be detected between ␣B-crystallin and vimentin or desmin, in either combination (Table I). The lack of detectable interaction is not due to the inability of the intermediate filament fusion proteins to interact, since both vimentin and desmin readily interact with themselves. The absence of detectable interactions between ␣B-crystallin and the intermediate filament proteins might be due to interference of the LexA or B42 domain fused to the N terminus of the proteins. However, it is also possible that specific conditions, or specific helper proteins, which are absent in the yeast system, are required for proper interactions. It thus seems that the two-hybrid system is not a suitable system to study interactions between ␣B-crystallin and intermediate filaments.
The Conserved Domain of ␣B-Crystallin Is Essential for Complex Formation-The interaction between the ␣B fusion proteins as detected by the yeast two-hybrid assay, prompted us to try and localize the regions that are essential for this interaction. Considering that ␣B-crystallin is assumed to comprise two structural domains, we fused the N-terminal domain (N␣B, residues 1-66) and the C-terminal domain (C␣B, residues 68 -175) separately to the DNA-binding domain and the activation domain. The interactions between total ␣B-crystallin, N␣B and C␣B, in all possible combinations, were assayed in the two-hybrid assay (Table II). The C␣B-fusion protein clearly is able to interact very efficiently with itself and with the full-length ␣B-fusion protein. In contrast, the N␣B-fusion protein does not interact with itself, nor with ␣Bor C␣B-fusion proteins. This lack of interaction is not due to aberrant expression of the protein, since N␣B is detectably present in the yeast  cell extract (Western blot, data not shown). The two-hybrid system thus provides a sensitive and specific assay for the detection of C␣B-C␣B interaction, and should be suitable for selecting C␣B mutants that are either or not affected in their interaction with wild-type C␣B.
Random Mutagenesis of C␣B-To determine which amino acid residues are involved in the interaction between the Cterminal domains, and thus in the complex formation of ␣Bcrystallin, we generated two libraries of randomly mutated C␣B proteins. To introduce random point mutations, we performed a PCR under non-optimal conditions, reducing the fidelity of DNA synthesis (28). By varying the number of cycles and the amount of added template DNA in the PCR reaction, the number of mutations could be optimized. The two libraries were made in the prey vector. One contained clones with an average of 18 substitutions per C␣B gene, and was used for the selection of binding mutants with a larger number of amino acid changes. The other library contained clones with about 6 substitutions per C␣B gene, and was used to select non-binding mutants with a smaller number of amino acid changes. From sequence analysis of arbitrarily selected clones it appeared that the distribution of substitutions was random and that both transitions and transversions were generated. A 3 G and T 3 C transitions were most frequently observed (40 and 30% of the mutations, respectively). The preponderance of these transitions results in a biased array of amino acid replacements, favoring 26 of the 97 possible types of amino acid replacements that can be caused by a single base substitution. The frequencies of other transitions or transversions were all lower than 10% (data not shown).
C␣B Mutants Able to Interact with Wild Type C␣B-To select for mutants able to bind the C␣Bwt fusion protein, the library containing the highly mutated C␣B clones was transformed into the EGY48 yeast strain together with the C␣Bwt bait. Only about 3% of the transformants could grow on minimal medium lacking leucine, indicating that they contain a mutant able to interact with the C␣Bwt bait. Consequently, 97% of the mutants were too heavily mutated to be expressed and interact with C␣Bwt. We took 11 binding mutants and assayed them for expression of C␣B fusion protein by Western blot, using an antiserum directed against the C-terminal peptide of ␣B-crystallin. From this analysis it appeared that two clones did not detectably express the C␣B fusion protein, which might be due to instability of these mutants. These clones were not used any further. The yeast DNA preparations of the remaining nine C␣B mutants were introduced into E. coli to isolate the C␣B mutant prey vectors. The plasmids were then reintroduced into the yeast strain EGY48, along with the LacZ reporter plasmid and the C␣Bwt bait plasmid, to obtain several independent transformants of each mutant. The efficiency of the interaction of each mutant with the C␣Bwt was measured by determining the average ␤-galactosidase activity of at least six independent transformants. All mutants showed high enzyme activity (Table III, upper row), indicating that they are able to interact efficiently with C␣Bwt.
The deduced amino acid sequences of the nine mutants are shown in Fig. 1A. The mutants contained from 3-15 replace-ments, with an average of 9.6. Remarkably, 24% of the 86 detected replacements involved a change from lysine or arginine to another residue (not counting K 3 R changes). The inverse change only rarely occurred. Most often, in 14% of all cases, a lysine was replaced by glutamic acid (black in Fig. 1A). Glutamic acid or aspartic acid residues were less often mutated; only in 10% of the cases. As a consequence, most mutants have an increased ratio between acidic and basic amino acids, resulting in a decreased pI (Fig. 1A). Remarkably, too, is the frequent mutation of the highly conserved proline at position 86, being replaced in seven of the nine mutants. This proline has been replaced by very distinct types of residues (serine, glutamine, arginine, and leucine), suggesting that the replacement of this proline per se is somehow more important than the type of residue introduced.
C␣B Mutants Unable to Interact with Wild Type C␣B-To select for mutants that are unable to interact with the C␣Bwt fusion protein, the library containing C␣B mutants with about six-base substitutions per clone was transformed into the EGY48 yeast strain, along with C␣Bwt bait and reporter plasmid. The screening for mutants not able to interact with the C␣Bwt bait was done on plates containing 5-bromo-4-chloro-3indolyl-␤-D-galactose. Transformants were selected which were white, in contrast to the dark blue color observed with the C␣Bwt prey in conjunction with the C␣Bwt bait. Fifty-five percent of the obtained transformants were white. From these white colonies, 146 were assayed for activator-C␣B mutant fusion protein expression by Western blot, using the antiserum directed to the C-terminal peptide of ␣B-crystallin. It appeared that only 16 clones expressed the full-length activator-C␣B fusion protein. The yeast clones not expressing the fusion protein probably contain an early termination signal or an out-offrame mutation, and were discarded.
The plasmids coding for the 16 C␣B mutants were isolated and reintroduced into the EGY48 yeast strain. Independent transformants were assayed for ␤-galactosidase expression to quantify the interaction between C␣B mutant and C␣Bwt. No ␤-galactosidase expression could be detected, confirming that the mutants were unable to interact with C␣Bwt (Table III, upper row, and data not shown). By DNA sequence analysis the deduced amino acid sequences of the non-binding mutants were determined (Fig. 1B). All C␣B mutants were full-length and contained three to nine replacements, with an average of 5.6 per mutant. It thus is, unfortunately, not possible to identify individual amino acid replacements that frustrate interaction with C␣Bwt. Remarkably, frequently, in 20% of the 89 detected replacements, glutamic acid or aspartic acid was replaced by a non-charged residue (gray in Fig. 1B), mostly glycine. The basic residues arginine or lysine were less often (12%) mutated and only rarely replaced by an acidic residue. By comparing the replaced residues in the binding and non-binding mutants (Fig. 1C), it is obvious that in the non-binding mutants also some hydrophobic residues are preferentially replaced, in particular the residues Phe 84 , Ile 98 , Phe 118 , and Ile 161 . Nine of the 16 mutants have either specific hydrophobic residue(s) mutated or acid residue(s) replaced (Fig. 1B), suggesting that both types of changes might annihilate the interaction with C␣Bwt.
In an attempt to investigate the non-binding mutants in vitro, we expressed several of them in E. coli. The NB1 to NB6 mutants were expressed without fusion protein, while NB1 to NB3 were also expressed fused to the N-terminal domain of ␣B-crystallin. Unfortunately, the tested mutants could not be obtained properly folded, most likely caused by the limitations of the prokaryotic expression system. The C␣B mutants had a disturbed structure, as determined by far-UV circular dichro- ism spectroscopy, and the full-length mutant proteins formed inclusion bodies. None of the purified mutants could be correctly folded by denaturation/renaturation in urea.
Comparison of the Interaction of ␣B-Crystallin, ␣A-Crystallin, and Hsp27 with the C␣B Mutants-A specific interaction between ␣B-crystallin and ␣A-crystallin or Hsp27 subunits can be detected with the yeast two-hybrid system (see Table I). Since these proteins are homologs, it is likely that their interactions are very similar. By comparing the efficiency of interaction of the C␣B mutants with ␣B-crystallin, ␣A-crystallin, and Hsp27 fusion proteins, we might confirm this assumption. We first tested whether the C␣Bwt domain alone can interact with ␣A-crystallin and Hsp27. Coexpression of ␣A-crystallin or Hsp27 bait with the C␣B prey resulted in a very high induction of ␤-galactosidase. The ␤-galactosidase expression is for unknown reasons even higher than with the ␣B bait (Table III). Thus the C␣B domain can efficiently interact with ␣A and Hsp27. We then tested whether any conspicuous differences could be detected in the interactions of the C␣B mutants with the ␣B-crystallin, ␣A-crystallin, or Hsp27 bait proteins. These interactions, however, as measured by the ␤-galactosidase expression, were mostly very similar (Table III). Notably, the mutants B4 and B8, which can interact with the C␣B bait, do not bind to ␣B-crystallin, nor to ␣A-crystallin or Hsp27. It is possible that B4 and B8 cannot form a complex with the fulllength proteins due to their reduced affinity for the C-terminal The efficiency of interaction of prey C␣Bwt, ␣B-crystallin, ␣A-crystallin, and Hsp27 with the different bait C␣B mutants B1-9; binding mutants; NB1-6; non-binding mutants Unit definition ␤-galactosidase expression as in Table I 1. Comparison of binding and non-binding C␣B mutants. A, sequence alignment of C␣B mutants able to interact with wild type C␣B. The mutants were selected as interactors of LexA/C␣B fusion protein with the yeast two-hybrid system from a library containing randomly mutated C␣B clones. B, sequence alignment of C␣B mutants unable to interact with wild type C␣B. The mutants were selected as non-interactors of LexA/C␣B fusion protein from a library containing randomly mutated C␣B clones. Only those residues that differ from the wild type sequence of rat C␣B are indicated. Residues marked by a black box are K 3 E replacements. Residues indicated by a shaded box are replacements of acidic residues by non-acid residues, mostly glycine. Asterisks indicate non-binding mutants that have one or more of the hydrophobic residues Phe 84 , Ile 98 , Phe 118 , and Ile 161 mutated or one or more acidic residues replaced. C, diagram comparing the amino acid substitution frequency (y axis) at each position (x axis) in binding mutants (open bars) and non-binding mutants (closed bars).
domain (Table III). Only the mutants NB3 and NB6 behave differently toward the three sHsps, being able to interact with ␣A-crystallin, but not with ␣B-crystallin or Hsp27. This might be due to the ability of ␣A-crystallin to form a tighter complex with ␣B-crystallin than does ␣B-crystallin with itself (22). In summary, these results suggest that ␣B-crystallin, ␣A-crystallin, and Hsp27 interact with the C␣B domain in a very similar manner, although some remarkable differences in interaction do exist.

DISCUSSION
Specific Interactions of ␣B-Crystallin and Its Domains in the Yeast Two-hybrid System-The interactions of ␣B-crystallin as measured in the yeast two-hybrid system appear highly specific. This is shown by the ability of the LexA/␣B-crystallin fusion protein to interact with ␣B-crystallin, ␣A-crystallin, and Hsp27, fused to the activator protein B42, and the inability to interact with various control activator fusion proteins. Also by screening different cDNA libraries with ␣B-crystallin as bait only a very limited number of proteins were selected that are able to interact with ␣B-crystallin (33). The apparent inability of ␣B-crystallin to interact nonspecifically with proteins in the two-hybrid assay might indicate that the chaperone-like activity of ␣B-crystallin is annihilated by the fused LexA domain, although it has been shown that a protein fused to the Nterminal domain of ␣B-crystallin does not necessarily affect the chaperone-like activity (34). Another, and probably more likely explanation, is that the LexA/␣B-crystallin fusion protein is still able to interact with unfolding or aggregating proteins, but that such proteins are not present or available in the yeast two-hybrid system. Also an other member of the sHsp family, yeast Hsp42, is able to interact specifically with itself in the yeast two-hybrid system and does not seem to interact nonspecifically in this system (35). Remarkably, this protein did not detectably interact with its yeast homolog Hsp26.
We found that the N-terminal domain of ␣B-crystallin does not detectably bind to ␣B-crystallin or with one of its two domains. This result agrees very well with the finding that a degradation product of ␣B-crystallin is present as a monomer in calf lens extract, consisting of the N-terminal 80 residues (36). On the other hand, when the N-terminal domain is clipped off from Caenorhabditis elegans Hsp16-2 (37), Saccharomyces cerevisiae Hsp42 (35), mouse Hsp25, rat ␣A-crystallin, or bovine ␣B-crystallin (38), the remaining domain, containing mainly the conserved sHsp sequence, forms small complexes. The N-terminal domain is thus involved in the formation of larger complexes, but can only do so in the presence of the C-terminal domain. It would seem that the conserved C-terminal domain forms the building blocks, which associate into large complexes through interactions between the N-terminal domains (39).
C␣B Mutants-Not many mutagenesis studies have been performed that are relevant for the complex formation of sHsps; most studies were mainly concentrated on the chaperone-like activity. The amino acid replacements tested so far had only minor effects on complex size (40 -43). A dramatic effect, however, is observed upon phosphorylation of Hsp27 (44 -46). This protein contains three phosphorylatable serine residues, which are key elements in its structural organization. Phosphorylation of these residues correlates with a reduction of the in vivo complex size to small oligomers. Also substitution of these serine residues by glycine or aspartic acid reduces the complex size (47). ␣B-Crystallin has phosphorylatable serine residues too, but does not reduce its complex size upon phosphorylation (16). To get a better understanding of the complex formation of ␣B-crystallin, we have performed a mutagenesis study. We chose to study the interaction between C-terminal domains, because this interaction is probably less complicated than the interaction between complete subunits. By screening a library of randomly mutated C␣B domains for mutants that are able to interact with wild type C␣B, we have selected 9 binding mutants with an average of 9.6 mutations. The 86 replacements present in the binding mutants are tolerated, in that they do not inhibit folding and/or prevent binding of the C␣B mutants. It is possible that some of these mutations actually have a negative effect on the affinity for C␣Bwt, but were compensated by other mutations. Such compensating mutations might be the replacements of proline 87, because this residue is preferentially removed in the binding mutants, but not in the non-binding mutants (Fig. 1C). It is difficult to imagine how replacement of this highly conserved proline could be advantageous, but it might be possible that it allows protein flexibility to better cope with the other introduced mutations. Interestingly, this particular proline is highly conserved in animal sHsps, but is not conserved in plant sHsps (6).
Other compensating mutations involve charge changes. In the binding mutants, but not in the non-binding mutants, the residues Lys 72 , Lys 82 , Lys 90 , Lys 121 , and Lys 150 were frequently replaced by glutamic acidic. It has been shown, by comparing the pI of native and denatured ␣B-crystallin, that more basic residues than acidic ones are present inside the complex (13). The excess of basic residues inside the complex may generate repulsive forces to allow the rapid exchange of subunits between complexes (13,48,49). Replacing basic by acidic residues might thus stabilize the interaction between mutant and wild type C␣B, due to an increased electrostatic interaction and decreased electrostatic repulsion (see model in Fig. 2). Remarkably, although both ␣A-crystallin and Hsp27 have a lower pI, both interact very efficiently with the binding mutants, which indicates that the charged residues involved in the interaction must have been conserved between the three proteins.
To enhance the interaction between the mutant and wild type domains, the basic residues Lys 72 , Lys 82 , Lys 90 , Lys 121 , and Lys 150 should be located at the subunit interface. It has recently been determined by site-directed spin labeling that residue Arg 117 in ␣A-crystallin and Lys 141 in Hsp27 are located in a ␤-strand near the subunit interface (50,51). These residues occupy the same position as Lys 121 in ␣B-crystallin, strongly suggesting that Lys 121 is indeed at the interface of the subunits. It remains to be determined whether the other lysine FIG. 2. Model of the electrostatic interactions between wild type and mutant C␣B. Inside the ␣B-crystallin complex relatively more basic than acidic residues are present (13). The ionic attractions between basic and acidic residues are indicated with an ellipse, and the repulsion between two basic residues with a two-headed arrow. Replacement of a basic residue by an acidic one might enhance the attraction between subunits due to an increased electrostatic interaction and decreased electrostatic repulsion. Removal of an acidic residue might reduce the attraction between subunits due to a decreased electrostatic interaction and increased electrostatic repulsion.
residues are located at subunit interfaces too.
In the mutants unable to interact with wild type C␣B, acidic residues were relatively more frequently mutated than in the binding mutants (Fig. 1C). These residues, which were often replaced by a glycine residue, might interact with basic residues to stabilize subunit interactions or conformation. Their removal can thus decrease the affinity for neighboring subunits due to decreased electrostatic interaction and increased electrostatic repulsion (Fig. 2). Four of the 16 non-binding mutants contain a Glu 117 Gly substitution, which is not present in the binding mutants. Like Lys 121 , also Glu 117 can be expected to be located at the subunit interface (50,51). The preferential replacement of Glu 117 in the non-binding mutants therefore suggests that this residue has an important role in stabilizing the interactions between wild type subunits.
It can, however, not be excluded that some of the mutations in the non-binding mutants additionally affect the interaction by interfering with the formation of the correct tertiary structure of the protein. The inability of a series of these mutants to be properly folded in E. coli might even suggest this. However, it must be noted that the prokaryotic expression system is very different from the yeast system, and it is thus quite possible that the mutants are folded correctly in the yeast system. The ability of mutants NB3 and NB6 to interact with ␣A-crystallin (Table III) strongly indicates this possibility. Not having available correctly folded non-binding mutant proteins, it was unfortunately not possible to study the consequences of the various mutations on properties such as complex formation, complex stability, subunit exchange, and the chaperone-like activity. Reducing the number of replacements per mutant, such that the protein may be properly expressed in E. coli, might overcome this problem and will allow further studies.