Size of Human Lens β-Crystallin Aggregates Are Distinguished by N-terminal Truncation of βB1

The aggregates formed by the interactions of the human lens β-crystallins have been particularly difficult to characterize because the β-crystallins comprise several proteins of similar structure and molecular weight and because their sequences were not known until recently. Previously, it could not be ascertained whether the species of various acidities were different proteins or modifications of the same proteins. The recent determination of the sequences permits calculation of molecular weights and unambiguous identification of the various β-crystallins and their modified forms by mass spectrometry. In this investigation, the components of the three sizes of β-crystallin aggregates, β1 (∼150,000), β2 (∼92,000), and β3 (∼46,000), were determined. The principal differences among the different β-crystallin aggregates was the presence of βA4 in β1 and β2, but not β3, and the length of the N-terminal extension of βB1. The size of the β-crystallin aggregate correlated with the length of the N-terminal extension of βB1, indicating that the flexible N terminus of βB1 is critical to the formation of higher molecular weight aggregates of β-crystallins. Separation of the components by ion exchange under non-denaturing conditions showed that βB2 occurs as homo-dimers and homo-tetramers as well as contributing to hetero-oligomers. Other β-crystallins were present only as hetero-oligomers.

The lens is a transparent avascular tissue with a high concentration of proteins closely packed to give a refractive index that will focus light on the retina (1). The lens proteins, which have monomeric molecular weights of approximately 20,000 -30,000, form aggregates with molecular weights up to 1 million. It is believed that the clarity of the lens depends on the proper assembly of these aggregates (2). In the mammalian lens there are three primary groups of crystallins, ␣-, ␤-, and ␥-crystallins, each composed of proteins homologous within the group but differing from the proteins in other groups. The ␣-crystallins have aggregate molecular weights of 180,000 -1,000,000, the ␤-crystallins are aggregates of 40,000 -150,000, and the ␥-crystallins do not aggregate. Within the ␣and ␤-crystallin subgroups, the proteins have been named according to their acidities. For example, the ␣-crystallins include ␣A-crystallin, the more acidic, and ␣B-crystallin, the more basic. Within the ␤-crystallins of various mammalian species, seven ␤-crystallins, ␤A1, ␤A2, ␤A3, ␤A4, ␤B1, ␤B2, and ␤B3, have been identified; the ␥-crystallins include ␥S, ␥A, ␥B, ␥C, ␥D, and ␥E. Not all the genes are expressed in each species. In human lenses, expressed genes include those coding for ␣A (3), ␣B (4), ␤A1, ␤A3, ␤A4 (5), ␤B1 (6), ␤B2 (7), ␤B3 (5), ␥S (8), ␥C, and ␥D (9,10). The ␤and ␥-crystallins have considerable homology, both being composed of two globular domains forming "Greek key motifs" joined by a linker segment (11)(12)(13)(14). All of the ␣and ␤-crystallins as well as ␥S-crystallin are acetylated at the N terminus.
The ␤-crystallins of the mammalian lens form aggregates that fractionate by size exclusion chromatography into two or three subgroups of varying proportions depending on the chromatographic conditions (10,15,16). When they separate into two subgroups they are referred to as ␤ HIGH (␤ H ) and ␤ LOW (␤ L ); when they separate into three subgroups they are ␤ H , ␤ L1 , and ␤ L2 (17) or ␤ 1 , ␤ 2 , and ␤ 3 (15). The unique structural features of the ␤-crystallins that distinguish them from the ␥-crystallins, which do not aggregate, include N-and C-terminal extensions from the globular domains and a distinctive linker sequence. There is evidence supporting the importance of both of these features in the formation of aggregates. Investigations of mutant rat ␤-crystallins have demonstrated that replacing the linker sequence of ␤B2-crystallin with the ␥-crystallin linker prevents dimer formation, whereas deletion of the N-and C-terminal extensions does not affect folding of the domains or dimer formation (18,19). On the other hand, in studies of bovine ␤-crystallins, it has been noted that the largest of the ␤-crystallins, ␤B1, which has a long N-terminal extension, is a component only of the highest molecular weight aggregates (␤ 1 ), implying that the long N terminus is vital to formation of large aggregates (16,20,21).
Six human ␤-crystallin sequences are now known: ␤B1 (6), ␤B2 (7), and ␤B3, ␤A1, ␤A3, and ␤A4 (5). Mass spectrometric determinations of the molecular weights of the crystallins confirmed the cDNA-determined sequences and identified three of the crystallins, ␤B1, ␤A1, and ␤A3, in forms with truncated N termini. ␤B1 missing 15, 34, 39, and 40 residues of the N terminus were identified; ␤A3 was found in one truncated form, missing the first 22 residues. Because ␤A3 and ␤A1 have the same sequence except that the N terminus of ␤A1 begins with residue 19 of the ␤A3 sequence, truncated ␤A3-(23-215) is the same as ␤A1-(5-197) (5). The previous studies of the ␤-crystallins did not describe which ␤-crystallins were present in the various size aggregates. The goal of this study was to use molecular weights determined by mass spectrometry to identify the ␤-crystallins and their truncated forms in each of the subgroups and, from these data, derive an understanding of the crystallin features contributing to aggregate formation.

EXPERIMENTAL PROCEDURES
Lenses were obtained from the National Disease Research Interchange (Philadelphia, PA). Trypsin was purchased from Worthington, BDH Aristar grade urea was from Gallard-Schlesinger Industries (Carle Place, NY), and all other chemicals were from Sigma. All chemicals were of reagent grade and used without further purification.
The lenses examined in this study were from donors 16, 27, 37, 56, * This work was supported by National Institutes of Health Research Grant EY RO1 07609 and National Science Foundation Grant 9413023. 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. and 70 years old. All lenses were clear; the donors had no diseases known to affect lens opacity. Each lens, analyzed individually, was homogenized in 2.5 ml of a buffer (0.05 M NaHSO 3 , 0.05 M Tris, 0.02 M EDTA, 0.02% NaN 3 , pH 7.4). The water-soluble crystallins, removed as the supernatant after centrifugation at 27,000 ϫ g for 15 min, were fractionated into ␣-, ␤ 1 -, ␤ 2 -, ␤ 3 -, and ␥-crystallins by size exclusion chromatography (70 ϫ 2 cm Sephacryl S-300HR column) using the homogenizing buffer as the mobile phase at a flow rate of 15 ml/h. The column was calibrated using thyroglobulin (M r 669,000), alcohol dehydrogenase (M r 150,000), bovine serum albumin (M r 66,000), and ovalbumin (M r 45,000). The absorbance of the eluate was monitored at 280 nm. Because there was some overlap between the peaks for ␤ 3 and ␥-crystallins, the ␤ 3 -crystallins were separated from the ␥-crystallins by further size exclusion chromatography (Sephadex G-75) using the above buffer. The size exclusion buffer was replaced with the ion exchange buffer by passing each of the ␤-crystallin fractions through Sephadex G-25 (1 ϫ 5 cm column) at a flow rate of 1.5 ml/min.
Ion Exchange HPLC 1 -After isolation by size exclusion chromatography, each of the ␤-crystallin fractions was further fractionated by anion exchange chromatography using a Mono-Q HR column (Pharmacia Biotech Inc.) connected to an analytical HPLC system (Rainin Instrument Co. Inc., Woburn, MA) operating at flow rate of 1.0 ml/min. Elution of the crystallins was monitored by the absorbance at 280 nm. Buffer A was 0.01 M Tris at pH 7.63; buffer B was the same as buffer A plus 1.0 M NaCl. The sample was eluted following a gradient of 0% B for 3 min, 0 -11% B for 17 min, 11-50% B for 14 min, and 100% B in 1 min. The Mono-Q column was washed between runs with 2 ml of 2.0 M NaCl dissolved in 1:1 water and methanol.
Reversed Phase HPLC-The ␤-crystallin fractions from anion exchange chromatography were further fractionated by reversed phase HPLC (4.6 ϫ 150 mm Vydac C 4 column, 300 Å) using a gradient of acetonitrile and water, both solvents with 0.1% trifluoroacetic acid. The proteins were eluted using a gradient of 10 -30% acetonitrile for 5 min, 30 -50% for 30 min, and 50 -98% for 2 min. The absorbance was monitored at 280 nm; the fractions were collected manually, lyophilized to dryness, and stored at Ϫ80°C before analysis.
Electrospray Ionization Mass Spectrometry (ESIMS)-The molecular weights of the isolated proteins were determined using a quadrupole electrospray ionization mass spectrometer (Micromass Platform II, Manchester, United Kingdom) calibrated with horse heart myoglobin (M r 16,951). The sample was introduced in a solution of 1:1 acetonitrile: water with 2% formic acid at a flow rate of 5 l/min. Molecular weight determinations of proteins were Ϯ0.01%, yielding an uncertainty of Ϯ 2 Da for a protein with a molecular mass of 20 kDa.

RESULTS
The data presented were obtained with the 56-year-old lens. Not all molecular weight determinations were performed on all lenses, but the chromatograms for the size exclusion, ion exchange, and reversed phase fractionation for the different lenses were similar, suggesting that there were only minor age-related differences in the water-soluble ␤-crystallins from ages 16 to 70. For the size exclusion chromatographic conditions used in this study, ␤-crystallins gave three peaks with aggregates of approximate molecular weights of 150,000 (␤ 1 ), 92,000 (␤ 2 ), and 46,000 (␤ 3 ) (Fig. 1). By employing several chromatographic techniques, size exclusion, non-denaturing anion exchange HPLC, and reversed phase HPLC, along with mass spectrometric determination of molecular weights of the fractionated proteins, it was possible to unambiguously determine the components of the ␤-crystallin aggregates (Table I). Each ␤-crystallin was identified by matching its molecular weight as determined by mass spectrometry with a molecular weight calculated from the known sequences of the ␤-crystallins. For all identifications, the agreement between the experimentally determined and calculated molecular weights was within 2 atomic mass units.
Anion exchange chromatography was performed without urea in the buffer to observe which components were associated in the ␤-crystallin aggregates. Under these non-denaturing conditions, three peaks labeled A, B, and C ( Fig. 2) were evident. The intensities of A, B, and C varied considerably among the different groups of aggregates. For example, peak A from ␤ 1 was small (Fig. 2a), but it was approximately one-third of ␤ 2 (Fig. 2b) and ␤ 3 (Fig. 2c). Peak B was a relatively minor component of ␤ 1 but a major component of ␤ 2 and ␤ 3 . Peak C was a major component of all three size exclusion fractions. Although the chromatograms were generally similar for all lenses, the relative intensities of peaks B and C derived from ␤ 2 and ␤ 3 varied somewhat, from about 3:2 to 2:3, among the lenses.
The reversed phase chromatogram of proteins eluting in peak A had only one peak, with an elution time corresponding to ␤B2 (Fig. 3a) (7). ESIMS analysis of this peak showed the presence of one protein with a molecular weight (M r 23,291) ( Fig. 4) that confirmed its identity as ␤B2 (M r 23,291) (7). The peak labeled AЈ (Fig. 2c) also had a molecular weight matching ␤B2. Since deamidation of a protein causes it to elute later on anion exchange but changes the molecular weight by only 1 mass unit and therefore cannot be distinguished by ESIMS determination of its molecular weight, it seemed likely that AЈ is a deamidated form of ␤B2.
Reversed phase analysis of ion exchange peaks B and C gave similar chromatograms, each with two major peaks (Fig. 3, b and c). One peak had a retention time of ␤B2 (peak 1 of Fig. 3), and the other had a retention time corresponding to ␤B1 (peak 3 of Fig. 3) (6) and ␤A4 (5), which co-elute with these reversed phase conditions. Identification of ␤B2, ␤B1, and ␤A4 was confirmed by ESIMS analysis (Table I). Finding the same masses for the proteins in B and C suggested that B and C contain the same proteins but perhaps with the proteins in C modified by deamidation. The ␤B2-crystallins in peaks B and C, whether from ␤ 1 , ␤ 2 , or ␤ 3 , were all intact (M r 23,291) (Fig.  4). ␤A4-crystallin, which also was present only as the intact protein (M r 22,282), was found in size exclusion fractions ␤ 1 and ␤ 2 ; a mass corresponding to ␤A4 was not detected in ␤ 3 . In contrast, ␤B1-crystallin was identified by ESIMS in several forms (Fig. 5) with molecular weights corresponding to ␤B1 and its truncated products at the N terminus (6). The relative amounts of ␤B2:␤B1/␤A4, based on their intensities in the reversed phase chromatograms, were approximately 3:1 in peak B and 1:1.5 in peak C. A minor component, labeled 2 in Fig. 3, present in both ion exchange peaks B and C from ␤ 1 and ␤ 2 , was identified by its reversed phase elution time as ␤A3/ ␤A1 (5).
The electrospray mass spectra for the proteins with reversed phase elution times appropriate for ␤B1 were remarkably dif- The fractions labeled ␤ 1 , ␤ 2 , and ␤ 3 were further fractionated by anion exchange (Fig. 2) and reversed phase HPLC (Fig. 3). ferent for ␤ 1 , ␤ 2 , and ␤ 3 (Fig. 5, a, b, and c, respectively). The ␤B1-crystallins showed progressive N-terminal truncation from ␤ 1 to ␤ 2 to ␤ 3 . For all the lenses, intact ␤B1 (M r 27,933) was a major component only of ␤ 1 , which also contained ␤B1 missing the first 15 residues (M r 26,535) and sometimes minor amounts with as many as the first 39 residues (Fig. 5a). In contrast, all the ␤B1 in ␤ 2 was truncated at the N terminus with 15-41 residues missing (Fig. 5b), and most of ␤B1 in ␤ 3 was even further truncated at the N terminus, with at least 34 -41 residues missing (Fig. 5c).
Since non-denaturing conditions were used in the anion exchange chromatography, the proteins isolated in this separation could be expected to retain their native associations. Finding only ␤B2 in peak A of the ion exchange chromatograms of    , peak A (Fig. 2); b, peak B (Fig. 2); and c, peak C (Fig. 2)). The reversed phase retention times of the ion exchange components, peaks A, B, and C were the same for the three ␤-crystallin oligomers, ␤ 1 , ␤ 2 , and ␤ 3 . The relative amounts of each component differed, as discussed in the text. The dotted line indicates the %CH 3 CN in the eluting solvent. ␤ 2 and ␤ 3 indicated that some dimers of ␤ 3 and some tetramers of ␤ 2 are ␤B2 homo-oligomers. The small peak A of ␤ 1 indicated that ␤B2 homo-oligomers are a very minor component of ␤ 1 . Neither ␤B1 nor ␤A4 was found alone in the ion exchange fractions, implying that these ␤-crystallins do not form homo-oligomers. DISCUSSION In previous investigations the components of the various aggregates of human ␤-crystallins were identified primarily by SDS-polyacrylamide gel electrophoresis (15,16,22). Because the sequences of the human ␤-crystallins were not known and because there appeared to be several proteins with similar molecular weights, the exact composition of each of the aggregates could not be determined. Data from fetal lenses showed the presence of a 29-kDa protein, later identified as ␤B1 in both ␤ 1 and ␤ 2 (16), whereas data from lenses more than 5 years old indicated the presence of ␤B1 only in ␤ 1 (15,16). From the SDS-polyacrylamide gel electrophoresis data presented in those studies, it appeared that ␤B1 was not a component of the lowest molecular weight aggregates. A protein of 24 -26 kDa, first called ␤Bp then renamed ␤B2, was a major component of ␤ 1 , ␤ 2 , and ␤ 3 from lenses of all ages (15,16,22). The other ␤-crystallins were not identified.
Results from our mass spectrometric investigation of ␤-crystallins, separated by ion exchange and reversed phase HPLC, have led to unambiguous identification of the components of each of the subgroups of ␤-crystallins (Table I). The data show that ␤B1 in a variety of forms truncated at the N terminus is present in the aggregates of ␤ 1 , ␤ 2 , and ␤ 3 and that the size of the aggregates correlates with the length of the N terminus of ␤B1. The largest aggregates, ␤ 1 , are composed primarily of ␤B1 (both intact and with the first 15 residues missing), intact ␤B2, and intact ␤A4. The principal forms of ␤B1 present in ␤ 1 , ␤B1-(1-251) and ␤B1-(16 -251), correspond to bovine proteins previously identified as ␤B1a and ␤B1b (23). These two forms of ␤B1 are the major ␤B1-crystallins found in newborn human lenses (6). The ion exchange chromatography performed under non-denaturing conditions indicated that the oligomers of ␤ 1 are hetero-oligomers of ␤B1, ␤A4, and ␤B2. Because ␤B1 and ␤A4 co-elute on reversed phase HPLC, the relative amounts of each component were not ascertained. A very minor component in ␤ 1 had the correct retention time for ␤A3. Intact ␤A3 has previously been detected only in fetal lenses; in adult lenses it was found missing the first 22 residues of the N terminus. An insufficient amount of this minor component was isolated from the 56-year-old lens for molecular weight determination, but it was presumed to be ␤A3 truncated at the N terminus based on its reversed phase HPLC elution time. A molecular weight corresponding to ␤A3- , which is the same as ␤A1- , was determined for this protein isolated from other adult lenses.
The dimers of ␤-crystallin, ␤ 3 , included homo-oligomers of ␤B2 and hetero-oligomers of ␤B1 and ␤B2. The ␤B1 in ␤ 3 was degraded at the N terminus even further than in ␤ 2 , with most of ␤B1 missing 34 or more residues from its N terminus. Finding ␤B1, although only in forms truncated at the N terminus, in ␤ 3 is in opposition to previous reports that ␤B1 exists only in larger aggregates (15,17,20,21). These differing observations are easily explained by the fact that the ␤B1 products without the N-terminal 34 -41 residues have molecular weights of 24,834, similar to the molecular weight of ␤B2 (23,291) and may not have been recognized as ␤B1 by SDS-polyacrylamide gel electrophoresis analysis. Identification would have been further complicated by the fact that ␤B1 minus these residues has a pI (6.38) similar to the pI of ␤B2 (6.33). ␤B3, which has previously been isolated only from fetal or newborn lenses (5), was not detected in any of the subgroups of the ␤-crystallins from these adult lenses.
For bovine lens crystallins, it has been demonstrated that the various aggregates of ␤-crystallins are in a dynamic equilibrium with the size of aggregates affected by concentration, temperature, and ionic strength (24). Hydrophilic interactions appeared to be the main factor affecting the association-dissociation equilibrium (24). Even though these data may not be directly applicable to human lenses because the composition of bovine ␤-crystallins differs considerably from human ␤-crystallins, it is interesting to consider the effect hydrophilicity might have on the stability of the various human ␤-crystallin aggregates. The hydrophilicities of the ␤-crystallin or portions of them can be calculated from the Bull and Breese indices (25). The long N-terminal extension of ␤B1, which appears to play a unique role in the formation of large aggregates as demonstrated by the presence of intact ␤B1 only in ␤ 1 -crystallins, has a Bull and Breese index (BB) of ϩ365. The hydrophilicity indices for ␤B1 in the middle size aggregates of ␤ 2 and the dimers of ␤ 3 , where 15-41 residues are missing from the N terminus, range from BB Ϫ20 to Ϫ57. (The more positive numbers indicate a more hydrophilic sequence). Further evidence of the likelihood that hydrophilicity is important in aggregate formation is demonstrated by the fact that ␤B2 forms both homo-dimers and tetramers, but ␤B1 and ␤A4 do not form homo-oligomers. Overall, ␤B2 is a more hydrophilic protein (BB ϩ62) than either ␤A4 (BB ϩ30) or ␤B1(BB ϩ5).
Comparison of the sequence of the first 34 residues of human ␤B1 with the N terminus of bovine ␤B1 shows that this region is not as rich in alanine and proline as bovine ␤B1 (23), but it does include 8 alanines, 5 prolines, and 4 glycines and no bulky amino acids, giving this region considerable flexibility. These characteristics, along with its hydrophilicity, may allow the N terminus to be flexible for easy interaction with charged portions of the other ␤-crystallins, stabilizing high molecular weight aggregates. Previous investigations using mutant forms of ␤B2 have demonstrated that the linker portion (residues 80 -88, PIKVDSQEH) that joins the two domains of ␤B2 is critical to the formation of dimers (19), whereas the N-and C-terminal extensions of ␤B2 are not essential (18). If this linker region were the sole determinant of dimer formation, it might be expected that ␤B1, which contains a region highly homologous with the ␤B2 linker (residues 140 -148, PIKMDAQEH), would also form homo-oligomers. Failure to detect homo-oligomers of ␤B1 in any of the fractions suggests that the sequence of the linker portion is not the only factor that determines the stability of dimers.
The fact that ␤A4 is in ␤ 1 and ␤ 2 , but not in ␤ 3 , suggests that it too may contribute to the formation of higher aggregates. If so, its interactions are different from those proposed for other ␤-crystallins. The sequence features previously proposed to be favorable to aggregation, the long N terminus and the ␤B2 linker sequence, are missing in ␤A4. ␤A4 has a short N terminus and the sequence of the linker region, PAACANHRD, is very different from the sequence of ␤B2 that favors dimer formation.
In contrast to ␤B1, which was found in many truncated forms at the N terminus, ␤B2 and ␤A4 were present only as intact proteins. None of the fractions of ␤-crystallins contained proteins with molecular weights consistent with truncated products of either ␤B2 or ␤A4. These results do not support the presence of truncated ␤B2 at the N terminus in older lenses, a modification that has been proposed to explain a lack of reactivity of older human lenses with an antisera to residues 1-12 of ␤B2 (26).
Despite the similarity between the sequences of human and bovine ␤-crystallins, there are striking differences between the composition of the aggregates of the human and bovine ␤-crystallins (27). In both species, intact ␤B1 is found only in ␤ 1 (17). Slingsby and Bateman (27) also reported finding no ␤B1 in ␤ 2 and ␤ 3 of bovine lenses and identified the lower molecular weight proteins in ␤ 2 and ␤ 3 as ␤A3 and ␤B3. Because their techniques may not have recognized truncated ␤B1, it is possible that some truncated ␤B1-crystallins were present in ␤ 2 and ␤ 3 but incorrectly identified. One similarity between bovine and human lenses is that ␤B2 is present as homo-dimers as well as being a major component of all three sizes of aggregates in the lenses of both species. In their recombination studies, Slingsby and Bateman (27) showed that bovine ␤B2 could form homo-dimers as well as hetero-dimers with other acidic and basic ␤-crystallins. Our results for human lens ␤-crystallins showed that, in addition to forming homo-dimers and hetero-dimers, ␤B2 also formed homo-and hetero-tetramers.
In bovine lenses, both ␤B3 and ␤A3 are major components, whereas in humans ␤B3 is produced only in fetal and newborn lenses, and ␤A3 is a very minor component. Furthermore, the ␤A3 that is present in adult lenses is truncated at the N terminus (5). Slingsby and Bateman (27) concluded that, for bovine ␤-crystallins, oligomers larger than dimers required the presence of an acidic ␤-crystallin with a long N terminus, such as ␤A3. This conclusion was supported by studies of mutant rat ␤A3, showing that ␤A3 without the first 29 residues of the N terminus formed smaller aggregates than ␤A3 with only 6 residues missing (28). Such a role for ␤A3 in human lenses is improbable because very little ␤A3 is present in adult lenses and that which is present lacks the first 22 residues of the N terminus. It is much more likely that the long N terminus of ␤B1 is the major determinant of the size of the ␤-crystallin aggregates in human lenses.