Domain Interaction Sites of Human Lens βB2-Crystallin*

βB2-crystallin, the major component of β-crystallin, is a dimer at low concentrations but can form oligomers under physiological conditions. The interaction domains have been speculated to be the β-sheets, each of which is formed by two or more β-strands. βB2-crystallin consists of 16 β-strands, 8 in the N-terminal domain and 8 in the C-terminal domain. Domain interaction sites may be removed by destroying the β-strands, which can be done by site-specific mutations, substituting the β-formers (Val, Phe, Leu) with Glu or Asn, strong β-breakers. We have cloned the following β-strand-deleted mutants, I20E, L34E, V54E, V60E, V73E, L97E, I109E, I124E, V144E, V152E, L162E, L165E, and V187E and their corresponding X → Asn mutants. We also made two mutants, V46E and V129E, that were not on the β-strand as controls. Disruption of protein-protein interactions was screened by a mammalian two-hybrid system assay. Protein-protein interactions decreased for all β-strand-deleted mutants except I20E, L34E, and L162E mutants; this effect was not seen in the two mutant controls, V46E and V129E. The sequences around Val-54, Val-60, Val-73, and Leu-97 in the N-terminal region and Ile-109, Ile-124, Val-144, Val-152, Leu-165, and Val-187 in the C-terminal region that formed β-strands appear to be important in dimerization. Some selected mutant proteins that showed strong (V46E and V129E) and reduced (V60E, V144E, V60N, and V144N) interactions were expressed in bacterial culture and were studied with spectroscopy and chromatography. The V60E and V144E mutants were found to be partially unfolded and incapable of forming a complete dimer.

␤B2-crystallin is the major component of ␤-crystallin and is a dimer at low concentrations (1)(2)(3). At high concentrations or in the lens, ␤B2crystallin forms hetero-oligomers with other ␤-crystallins. These oligomeric ␤-crystallins further participate in the formation of a supramolecular assembly that is important in lens function-lens transparency. A posttranslational or mutational modification in the interaction domain will disrupt the complex formation and the supramolecular assembly. Therefore, the determination of the interaction domains is essential for understanding crystallin functions. For a dimer or oligomeric proteins, the interaction sites or domains are usually ␤-sheets formed by two or more ␤-strands. ␤B2-crystallin was reported to have a high ␤-sheet content (2). Based on the structure from x-ray refraction studies for homodimer ␤B2-crystallin (4,5), each subunit includes 16 ␤-strands. Table 1 gives the distribution of those 16 ␤-strands along with that predicted from the PHD method (6). The distribution for ␥B-crystallin is also included for comparison (7). Each subunit has four Greek key motifs, and each motif contains four ␤-strands. Using site-specific mutation and a two-hybrid system assay, one can determine which ␤-strands are involved in the dimerization. It has been reported, based on x-ray structures of 29 proteins (8,9), that amino acids such as Val and Ile are strong ␤-formers and Leu and Phe are ␤-formers but that Glu and Pro are strong ␤-breakers and Asn, Lys, and Asp are ␤-breakers. Using site-specific mutation, one can eliminate the ␤-strands that are responsible for dimerization. We have prepared many X (Val or Ile or Leu) 3 Glu or Asn mutants for the two-hybrid system screening assays ( Table  2) and found that most ␤-strand-deleted mutants, especially those in the C-terminal domain, have reduced protein interactions. The results are consistent with our recent study of the Q155* ␤B2-crystallin mutant (10); the Q155* mutation decreased protein-protein interactions (dimerization), and the expressed mutant protein became partially unfolded. The truncation of 51 amino acid residues in the C-terminal region removes three ␤-strands that participate in intermolecular domain interactions. To study the effects of the ␤-strand-deleted mutations on biophysical properties, we cloned two such mutants (V60E or V60N and V144E or V144N) and two controls (V46E and V129E) and observed that ␤-strand-deleted mutant genes resulted in conformational change (partial unfolding) and inability to form complete dimers.

MATERIALS AND METHODS
Subcloning Mutant Genes into the Two-hybrid System Vectors-We used the Clontech Mammalian Two-Hybrid System Assay Kit 2 (Clontech, Palo Alto, CA). The test protein (bait) was fused into the GAL4 DNA-BD in the pM vector, and the second test protein (prey) was fused into the VP16-AD in the pVP16 vector. The third vector, pG5SEAP, contains a reporter construct and is encoded with secreted alkaline phosphatase (SEAP), 2 an enzyme that enables sampling of culture medium without cell lysis.
We used the previously constructed plasmids pM-␤B2 and pVP16-␤B2 (11,12) to prepare 14 mutants of X 3 Glu substitution using the QuikChange mutagenesis kit (Stratagene, La Jolla, CA). The primers used were custom synthesized (Invitrogen) ( Table 2). The corresponding mutants of X 3 Asn substitution were also prepared.
Cotransfection and SEAP Assay-HeLa cells were used in the cell culture, and Lipofectamine 2000 (Invitrogen) was used in cotransfection as previously reported (10 -12). HeLa cells were grown at 37°C with 5% CO 2 with 10% serum and were seeded at 2 ϫ 10 5 cells in 500 l of medium/well in a 24-well plate. Plasmids pM-X (0.3 g) and pVP16-Y (0.3 g) and pG5SEAP reporter vector (0.3 g) were added to the wells containing 2 l of Lipofectamine 2000 reagent.
SEAP activity was detected 48 h after transfection using the BD Great EscAPe SEAP fluorescence detection kit (Clontech). The detailed protocol for SEAP detection is provided in the kit and was used in our recently reported study (10). The fluorescence substrate MUP * This work was supported by National Institutes of Health Grant EY13968 and by the Massachusetts Lions Eye Research Fund. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. 1  (4-methylumbelliferyl phosphate) provides an easy assay of SEAP by reading fluorescence at 360/449 nm. A standard linear curve was obtained using the positive placental alkaline phosphatase. The expression of GAL4 DNA-BD/protein fusion was verified by Western blot with the GAL4 DNA-BD monoclonal antibody (11,12). To see whether expression levels were a factor for different SEAP activities, protein concentrations in the soluble protein extracts were determined using the Pierce BCA assay. The results indicated that protein expression levels varied very little among the various culture cells.
Expression of ␤B2-Crystallin Mutants-The use of the QIAexpression Type IV kit (Qiagen, Valencia, CA) in the cloning, expression, and purification of His 6 -tagged ␤B2-crystallin has been described elsewhere (10). Briefly, the ␤B2-crystallin mutant genes in pM plasmids (e.g. pM-␤B2WT) were amplified by PCR using Pfu DNA polymerase (Stratagene) with the forward/reverse primers: CGGGGTACCCCGGCCT-CAGATCACCAG/CCCAAGCTTGGGGTTGGAGGGGTGGAA. Two restriction sites, KpnI and HindIII, were included in the primers. The PCR products and pQE-30 vector were doubly digested by KpnI and HindIII. The digested genes and vector were then ligated by DNA ligase under standard conditions. The ␤B2 cDNA inserts were verified by sequencing analysis.
The expression constructs containing various ␤B2-crystallin genes were transformed into Escherichia coli strain M15 (pREP4). Cell culture was performed to induce expression of various proteins by a standard protocol. The His 6 -tagged ␤B2-crystallins, either wild-type or mutant, were purified by nickel-nitrilotriacetic acid affinity chromatography.
SDS-PAGE and Immunoblotting-SDS-PAGE was performed in a slab gel (15% acrylamide) under reducing conditions according to the method of Laemmli (13). Western blotting was performed with polyclonal anti-␤-crystallin antibodies (a gift from Sam Zigler of NEI/National Institutes of Health). Protein concentrations were determined by The prediction from PHD program by Rost and Sander (6) for the wild-type human lens ␤B2-crystallin. The sequence numbering is based on the sequence of the wild-type ␤B2-crystallin. The sequences of the ␤-strands are shown in bold. b,c The distribution of ␤-strand is taken from Bax et al. (5) for bovine lens ␤B2-crystallin and Kumaraswamy et al. (7) for bovine lens ␥B-crystallin.

TABLE 2
The forward and reverse primers for subcloning ␤ B2-crystallin mutants of X 3 Glu substitution The primers for the corresponding mutants of X 3 Asn substitution were prepared by substituting GAG or GAA with AAC in the forward primers and CTC or TTC with GTT.
Spectroscopic Studies-CD spectra were obtained with an Aviv circular dichroism spectrometer (model 60 DS; Aviv Associates, Lakewood, NJ). Five scans were recorded and averaged and followed by a polynomial-fitting program. The CD was expressed as deg-cm 2 -dmol Ϫ1 .
Fluorescence was measured with a Shimadzu spectrofluorometer (model RF-5301PC; Shimadzu Instruments, Columbia, MD). Trp emission was scanned with an excitation wavelength at 295 nm. Bis-ANS fluorescence emission spectra were scanned between 460 and 560 nm with an excitation wavelength of 395 nm.
The majority of the mutations in the ␤-strands, including V54E, V60E, V73E, L97E, L109E, I124E, V144E, V152E, L165E, and V187E, decreased SEAP activities relative to the wild-type ␤B2-crystallin. The mutations that did not decrease SEAP activities included I20E, L34E, V46E, V129E, and L162E. In general, mutations in the C-terminal region reduced SEAP activity more than did those in the N-terminal region. The two control mutations, V46E and V129E, which are not in the ␤-strand, did not decrease the SEAP activity. These results indicated that disruption of particular ␤-strands that might participate in dimerization was responsible for the reduction of SEAP activity. Those mutations that resulted in low SEAP activities between themselves (e.g. V54E-V54E) also showed low SEAP activities with WT (e.g. V54E-WT) (Fig. 2).
The corresponding SEAP activities for the mutants with an X 3 Asn substitution are shown in Fig. 3. The results are basically the same as those of mutants of X 3 Glu substitution, indicating that substitution of either the charged glutamic acid or the neutral asparagines yielded the same result.
Biophysical Studies of ␤B2-Crystallin Mutants-Four ␤B2-crystallin mutants with a Val 3 Glu substitution (V46E, V54E, V129E, and V144E), two in the N-and two in the C-terminal region, as well as WT were prepared as His 6 -tagged proteins. The two mutants V46E and V129E did not involve destruction of ␤-strands, and the two mutants V54E and 144E did involve destruction of ␤-strands. The His 6 tagging did not change protein conformation, as the His 6 -tagged WT ␤B2-crystallin did not show any difference in Trp fluorescence and CD from non-His-tagged WT ␤B2-crystallin (10). For mutation of Val 3 Asn substitution, V60N and V144N mutants were prepared.
SDS-PAGE, Western Blot, and FPLC Size-exclusion Chromatography-SDS-PAGE and Western blot are shown in Fig. 4. The His 6 -tagged ␤B2-

FIGURE 1. Protein-protein interactions between ␤B2-crystallin mutants of X 3 Glu substitution as expressed by the fold increase of SEAP activity over that in the controls (plasmids without DNA inserts): N-terminal region (A) and C-terminal region (B).
The ␤B2-crystallin genes were fused into the pM containing the DNA binding domain (BD) and a pVP16 vector containing the transcript activity domain (AD) and cotransfected to HeLa cells with pG5SEAP reporter vector. The culture medium was assayed for SEAP activity. Statistical significance was calculated by the paired t-test. Significant decreases were observed for the mutants from WT (*, p Ͻ0.05; **, p Ͻ0.005). crystallin is located at a higher molecular weight than the non-tagged ␤B2-crystallin.
The elution profiles of the V46E and V129E mutants on FPLC are the same as that of the WT, but those of V60E and V144E mutants appeared to contain a mixture of dimers and monomers (Fig. 5). A rerun of collected fractions between 18 and 21 ml for V144E mutant showed the same profile. The corresponding mutants V60N and V144N did not show the effect.
CD and Fluorescence-Both far-UV and near-UV CD changed greatly for the V60E and V144E mutants (Figs. 6 and 7), indicating an alteration occurs not only in the secondary structure but also the tertiary structure. The change is similar for the two mutants, but the other two control mutants, V46E and V129E, did not show these changes. The CD data indicate that the mutations (Val 3 Glu) involving the addition of charge change CD more than those mutations (Val 3 Asn) involving no change of charge.
Trp and Bis-ANS Fluorescence-Trp fluorescence showed no shift in emission maximal wavelength and only a small change in intensity for V46E and V129E mutants but a large red shift (from 331 to 345 nm) for V60E and V144E or V60N and V144N mutants (Fig. 8), suggesting that Trp residues in these mutants are in a more hydrophilic environment than are those in V46E, V129E, and WT.
In comparison with ␣〈-crystallin, WT ␤B2-crystallin gave a relatively low bis-ANS fluorescence intensity (2), but the intensity increased greatly for the V60E and V144E or V609N and V144N mutants (Fig. 9), indicating that the buried hydrophobic surfaces became accessible to the bis-ANS probe in these mutants.

FIGURE 3. The corresponding protein-protein interactions between ␤B2-crystallin mutants of X 3 Asn substitution as expressed by the fold increase of SEAP activity over that in the controls (plasmids without DNA inserts): N-terminal region (A) and C-terminal region (B).
Significant decreases were observed for the mutants from WT (*, p Ͻ0.05; **, p Ͻ0.005).

DISCUSSION
The crystallographic structure of ␤B2-crystallin has been determined for a homodimer of a fragment of a 181-amino acid sequence without the N-and C-terminal extensions (5,15) and a tetrameric WT (16). The distribution of the ␤-strand in each subunit is different in dimers and tetramers. For the purpose of our study, the use of the crystallographic structure of homodimer ␤B2-crystallin should be adequate. Both the dimer and tetramer structures were determined with bovine lens ␤B2crystallin. Because we used human lens ␤B2-crystallin, we used a PHD program to predict the secondary structure (6). The predicted secondary structure is in good agreement with that from crystallography (Table  1). It is reasonable to assume that the structure of human ␤B2-crystallin is similar to that of bovine ␤B2-crystallin, because they share a high sequence homology (91%). We also included the distribution of ␤-strands for ␥B-crystallin for comparison because the structures of ␤B2and ␥B-crystallin are similar. Both ␤〉2and ␥〉-crystallin structures are characterized by having four Greek key motifs: two (1 and 2) in the N-terminal domain and two (3 and 4) in the C-terminal domain (5,7). Each motif consists of four ␤-strands, and four motifs form four ␤-sheets. Each ␤-sheet contains four ␤-strands that do not correspond exactly to those in the particular motif. The ␤ 1 -sheet consists of three ␤-strands in motif-1 (␤ 1 -, ␤ 2 -, and ␤ 4 -) and one ␤-strand in motif-2 (␤ 7 -). The corresponding ␤-strands in the other three ␤-sheets are: ␤ 2 -sheet (␤ 3 -, ␤ 5 -, ␤ 6 -, and ␤ 8 -), ␤ 3 -sheet (␤ 9 -, ␤ 10 -, ␤ 12 -, and ␤ 15 -), and ␤ 4 -sheet (␤ 11 -, ␤ 13 -, ␤ 14 -, and ␤ 16 -). In ␥B-crystallin, the two domains are connected by a flexible linker peptide and organized so that two ␤-sheets (sheet-1 and sheet-3) lie on the outside of the molecule and the other two (sheet-2 and sheet-4) are in partial contact (intramolecular domain association). ␤B2-crystallin has a structure similar to that of ␥-crystallin except that the linker peptide is extended in a way favoring an intermolecular association; the two subunits associate in a dimer in a swap form such that the N-terminal domain of the first subunit is adjacent to the C-terminal domain of the second subunit. The other major difference is that ␤B2-crystallin contains an N-extension and a C-extension, reported to be responsible for preventing oligomerization (association of dimers) (17,18). Domain interactions in the ␤B2-crystallin may be similar to those in the ␥B-crystallin, i.e. between sheet-2 of the first subunit and sheet-4 of the second subunit and between sheet-4 of the first subunit and sheet-2 of the second subunit. The sheet-2 consists of the ␤ 3 -, ␤ 5 -, ␤ 6 -, and ␤ 8 -strand in the N-terminal region of the first subunit, and the sheet-4 consists of the ␤ 11 -, ␤ 13 -, ␤ 14 -, and ␤ 16strand in the C-terminal region of the second subunit. The mutations in Val-60, Val-73, Leu-97 destroyed ␤ 5 -, ␤ 6 -, and ␤ 8 -strands located in motif-2, and the mutations in Val-152, Leu-165, and Val-187 destroyed ␤ 13 -, ␤ 14 -, and ␤ 16 -strands located in the motif-4. The destruction of ␤-strand was reflected in the large reduction in SEAP activity. The results of mutations at ␤ 2 -and ␤ 4 -sheets confirm that they play an important role in dimerization. However, the results of some mutations in ␤ 1 -and ␤ 3 -sheets were mixed: mutation in Val-54, Ile-109, Ile-124, and Val-144 decreased SEAP activities and in Ile-20 and Leu-34 did not decrease SEAP activities. The mutation sites of Val-46 and Val-129 are not in the ␤-strand, and the SEAP activities were not affected. Apparently, the association of the ␤ 2 -sheet of the first subunit and the ␤ 4 -sheet of the second subunit was not entirely responsible for the dimerization. The Val 3 Glu substitution introduced a negative charge; we made some Val 3 Asn substitutions to see whether the decrease of SEAP activity might be due to the introduction of the charge and found that the changes in SEAP activities were basically the same as that of mutants with the X 3 Glu substitution. The disruption of key ␤-strands is the critical mechanism for the decrease of SEAP activity.
The observation that V60E (V60N) and V144E (V144N) mutations greatly reduced SEAP activity indicated that the ␤ 5 -and ␤ 12 -strands were two of the many dimerization sites, which was further confirmed by their inability to form complete dimers. The control mutants V46E and V129E, in which mutations were in the non-␤-strand, did not show these effects. The partial unfolding in the V144E mutant was also seen in the V144N mutant, indicating that disruption of the ␤-strand by either a Val 3 Glu or a Val 3 Asn substitution gave the same result. As dimerization is the initial step in the formation of ␤-crystallin complexes seen in the human lens, to maintain the integrity of the subunit interaction domains is thus important in sustaining their stability. Any disruption of the domain structures by either posttranslational modification or mutation will disturb the complex formation that is critical for lens transparency. This was demonstrated in the congenital cataract CRYBB2 gene product Q155*, in which 51 amino acid residues were truncated, removing all ␤-strands in the motif-4 critical in dimerization and resulting in a partially unfolded and unstable protein (10). We believe that these events are primary effects and that the secondary effect is an inability of the protein to participate in complex formation. The determination of the dimerization domain is thus important not only for ␤〉2-crystallin but also for ␣〈and ␣B-crystallin, whose threedimensional structures have not been established. It is possible to obtain data about dimerization domains similar to those of ␤B2-crystallin by extending the present study to ␣-crystallin, with its ␤-strand distribution predicted by the PHD program. The sites of both R116C ␣A-crystallin and R120G ␣B-crystallin mutations, associated with various autosomal dominant congenital cataracts, are located in the ␤7-strand and were found to change protein-protein interactions profoundly (11),  Human Lens ␤B2-Crystallin FEBRUARY 3, 2006 • VOLUME 281 • NUMBER 5 although the mechanism might not be related to disruption of the ␤-strand. Two recent studies reported the interaction domains for ␣B-crystallin, with their results not in complete agreement; one used a peptide scan (19), and the other used a protein pin array (20). The latter report indicates that interaction domains are accessible to solvent and consist primarily of ␤-strands in the core region with a minor helical structure in the N-terminal region and a nonstructural sequence in the C-terminal region. An earlier study using site-directed spin labeling identified an interface consisting of two ␤-sheets of seven ␤-strands at ␣-crystallin domain that mediates the formation of dimeric ␣A-crystallin (21). The other relevant study was reported by J. King and coworkers (22) on human ␥D-crystallin. Using site-specific mutagenesis on ␥D-crystallin and an unfolding-refolding study, they found that some hydrophobic residues (Phe-56, Ile-81, Val-132, and Leu-145) are important in the domain interactions and they are in the ␤-strands that contribute to intra-domain interactions.
In summary, we have demonstrated that the two-hybrid system could be used to determine the subunit interaction domains for ␤B2-crystallin. Our studies confirm the key ␤-strands that play an important role in the dimerization.