Rates of carbamylation of specific lysyl residues in bovine alpha-crystallins.

Previous investigations indicate that some forms of cataract may be due to the reactions of isocyanate with lens proteins. The present investigation was directed toward identifying the products of these reactions and determining rate constants for their formation. Bovine alpha-crystallins were incubated with isocyanate and separated into alpha A- and alpha B-crystallins by reversed-phase HPLC (high-performance liquid chromatography). Products of the reaction of isocyanate with alpha-crystallins were analyzed by mass spectrometry and isoelectric focusing. Proteolytic digests of carbamylated alpha A were analyzed by HPLC and fast atom bombardment mass spectrometry to determine the extent of reaction of each of the 7 lysyl residues present in alpha A. These results demonstrate that incubation of alpha-crystallins in 0.1 M KNCO leads to partial carbamylation of all 7 lysines of alpha A-crystallin. The extent of modification after 24 h of incubation varied from 7% at Lys 88 to 61% at Lys 11. Rate constants for the reaction of specific lysyl residues with isocyanate ranged from 5 to 54 x 10(-2) M-1 h-1. The distribution of reaction products, as determined by isoelectric focusing, indicates that the physiologically relevant initial stages of carbamylation of the 7 lysyl residues of alpha A proceed in a noncooperative manner.

Several post-translational modifications of the lysyl residues of lens proteins have been implicated in cataractogenesis. The reaction of isocyanate with the lysyl residues is of interest because it may be responsible for the high incidence of cataract common to diseases that are accompanied by elevated levels of isocyanate. Isocyanate is endogenous to the lens because it is in equilibrium with urea. Under physiological conditions, the level of isocyanate is approximately 1% that of urea (Dirnhuber and Schutz, 1948). It has been suggested that the increased incidence of cataract accompanying severe diarrhea or chronic renal failure may be due to the reaction of isocyanate with the lens proteins since both diarrhea and renal failure are associated with elevated levels of blood urea (Harding and Rixon, 1980;van Heyningen and Harding, 1986). Because carbamylation of the eNH2 removes the positive charge on lysyl residues, conformational changes may occur which might disrupt the close packing required for transparency and cause increased aggregation and lens opacity (Beswick and Harding, 1984). This hypothesis is supported by a correlation between the severity of nuclear cataract and a decrease in the number of free amino groups in lens proteins (Garcia-Castiiieiras and Miranda-Rivera, 1983).
The principal objectives of the present investigation were to isolate and identify the products of the reaction of isocyanate and a-crystallins, and to determine rate constants for formation of these products. Of particular interest were the rate constants for carbamylation of each of the 7 lysyl residues in aA-crystallin, and whether these reactions may be accelerated through synergistic effects. These objectives were accomplished through a new type of application of fast atom bombardment mass spectrometry (FABMS),' which was used to determine the extent of carbamylation of each of the lysyl residues. The FABMS results were compared with isoelectric focusing (IEF) results to determine if rates of carbamylation of a-crystallins are affected by carbamylation of other lysyl residues on the same molecule. In addition to their relevance to the reactions of isocyanate and a-crystallins, these results suggest that this combination of mass spectrometry and isoelectric focusing may be the basis of a new approach for detecting conformational changes occurring as a result of modification of specific residues.

EXPERIMENTAL PROCEDURES
Isolation and Carbamylation of Lens Crystallins-Bovine aA-crystallins were isolated and identified as previously reported (Smith et al., 1991). Homogenates of whole bovine lenses were separated by gel filtration into the a-, p-, and y-crystallin fractions using Sephadex G-200. The a-crystallins (500 pg/ml) were carbamylated by incubation in 0.1 M KNCO, 0.2 M NHIOAc containing 0.02% NaN3 at pH The abbreviations used are: FABMS, fast atom bombardment mass spectrometry; HPLC, high-performance liquid chromatography; m/z, mass/charge; IEF, isoelectric focusing; TEMED, N,N,N',N" tetramethylethylenediamine. 7.2 and 37 "C for various times. Complete (or nearly complete) modification was effected by incubating the protein (300 pg/ml) in the same buffer, but with 6 M guanidine hydrochloride, at 37 "C for 24 h. A control was prepared by incubating a-crystallins in the same buffer with 0.1 M KC1 in place of the 0.1 M KNCO. After incubation, the samples were dialyzed against water to remove salts. The acidic (aA) and basic (aB) forms of the a-crystallins were separated by reversed-phase HPLC (Rainin Instrument Co., Woburn, MA) using a Vydac 0.46 X 25-cm CI8 column with UV detection at 280 nm. A linear gradient of 30-60% CH3CN in HzO, both with 0.1% of trifluoroacetic acid, was used to elute the proteins. The gene products, aA2 and aB2, coelute with their respective phosphorylated forms aAl and aB1 and aBO (Smith et aZ., 1991(Smith et aZ., , 1992. The fractions corresponding to aA-crystallin were collected, freeze-dried, and stored at -10 "C until analyzed. Isoelectric Focusing of the Carbamylated Proteins-The carbamylation of ah-crystallin was investigated by IEF in a vertical slab gel apparatus (Hoefer SE 400, San Francisco, CA). The 0.75-mm-thick gel contained 8 M urea, 2% Triton X-100, 3% pH 4-6.5 ampholyte, 4% acrylamide, 0.03% ammonium persulfate, and 0.1% TEMED. The aA-crystallin samples, incubated in isocyanate or KC1 (control), were dissolved in a buffer containing 8 M urea, 8% Triton X-100, and 1% ampholyte (pH 4-6.5) and placed in the wells. Isoelectric focusing was continued for 18 h with a voltage of 400 V. The gel was fixed with a solution of CH30H/HOAc/H20 (25:15:60%), and the ampholyte was removed with a solution containing CH30H/HOAc/Hz0 (501040%). The gel was stained with Coomassie Blue. The quantities of protein in each band were determined by scanning the gels at 590 nm.
Proteolysis of d-Crystallins-The freeze-dried aA-crystallins (150 pg) were dissolved in 500 pl of either 0.03 N HC1 for peptic (Sigma) digestion, 0.1 M Tris-HC1, 0.01 M CaClZ, pH 8.3, for chymotryptic (Worthington) digestion, or 0.2 M Tris, pH 8.2, for Asp-N (Boehringer-Mannheim) digestion with an enzyme to substrate ratio of 1:50. The solutions were incubated at 37 "C for 1 h, for the peptic and chymotryptic digestions, or for 7 h for the Asp-N digestion. The resulting peptides were separated by reversed-phase HPLC using a linear gradient of 5-60% CH3CN in HzO, both with 0.1% trifluoroacetic acid. The carbamylated and uncarbamylated peptides were collected in the same fraction, freeze-dried, and analyzed by FABMS.
Mass Spectrometric Methods-The FABMS analyses were performed with a Kratos MS-50 mass spectrometer (Kratos Analytical, Manchester, United Kingdom) equipped with an RF magnet and a DS-90 data acquisition system. The mass range (m/z 400-2000) was calibrated with CsI. Lyophilized HPLC fractions were dissolved in 30% formic acid and analyzed using a 1:l mixture of glycerol and thioglycerol as the matrix. A scan rate of 30 s per decade was used to record the FAB mass spectra. For the instrument conditions used in this investigation, 1 nmol of peptide usually gave an excellent spectrum from which the molecular weight of the peptide and the intensity could be determined. Mass spectrometric data were processed with Kratos MACH3 software on a SUN workstation. Peptide identifications were confirmed from the molecular weights of the residues present after incubation of the peptide in carboxypeptidase Y (Sigma) or B (Worthington) in 0.1 M NH,OAc, pH 7.0, at 37 "C for 10-70 min (Caprioli and Fan, 1986;Qin et al., 1992). Peak intensities for the modified and unmodified peptides, averaged over at least 10 scans, were used to determine the extent of carbamylation of each lysyl residue.
Determination of Isocyanate Concentration-To determine the extent of hydrolysis of the isocyanate during the incubation period, a solution of 0.1 M KNCO was incubated in the same buffer that was used for the carbamylation experiments. Samples (10 91) were removed at various time intervals, and the concentration of KNCO determined by reacting the remaining NCO-with a peptide, Ile-Serbradykinin (16 nmol in 20 pl of a buffer, 0.1 M pyrophosphate, pH 8.0, with 6 M guanidine hydrochloride) at 45 "C for 30 min. The carbamylated and uncarbamylated peptides were separated by reversed-phase HPLC and quantified by their UV absorbance. The HPLC response was calibrated with solutions of 0.02-0.10 M KNCO.

RESULTS
HPLC Analysis of Carbamylated a-Crystallins-After incubation, the isocyanate and buffers were removed from the proteins by dialysis and the a-crystallins were separated into a Aand aB-crystallins by reversed-phase HPLC (Fig. 1). Previous studies demonstrated that aB elutes at 39-43% acetonitrile and a A elutes at 44-50% (Smith et al., 1991). Similar results were obtained in the present study for analysis of a-crystallins incubated in isocyanate. However, it was noted that the peaks became broader with a larger portion of both a A and aB eluting later as the incubation time increased, as illustrated in Fig. 1A for unmodified a-crystallin and in This chromatographic behavior is consistent with formation of a mixture of a-crystallins that are carbamylated at different sites and to different extents. When the crystallins were incubated under denaturing conditions with 6 M guanidine hydrochloride, the original peak disappeared and a new sharp peak with a longer retention time appeared, suggesting that a uniform fully carbamylated product had been formed. aAcrystallin to be used for further analyses was collected in an elution volume that included uncarbamylated as well as all forms of the carbamylated protein.
Modifications to d-Crystallins-The products of the reaction of isocyanate with aA-crystallin were identified by analyzing proteolytic digests of the protein incubated for 24 h in isocyanate by a combination of HPLC and FABMS. Asp-N was used to fragment the protein, because it gave peptides that could be assigned to the entire 173 amino acids in the acrystallin. Nearly all of the peptides detected by FABMS could be assigned to unmodified segments of aA. The FAB mass spectra of the unidentified peaks could be assigned to pept.ides with lysyl residues with a +43 mass unit modification, indicating that all the lysyl residues had been partially carbamylated. No evidence was found for other modifications, such as the carhamylated cysteinyl residues which have been detected in 711-crystallin carbamylated under similar incubation conditions.2 Quantificution of Carham~vlution hy IRF-The extent of carhamylation was determined as a function of time by isoelectric focusing. Since carbamylation lowers the isoelectric point in a predictable manner, a-crystallins carbamylat.ed to different extent,s give a series of discrete IEF bands. Separation of carbamylated and uncarbamylated nA-crystallin by I E F is illust,rated by the bands in lane 1 of Fig. 2 for a sample prepared by combining unincubated nA with nA that had been incubated in 6 M guanidine hydrochloride and isocyanate. The four hands present in lune 1 are due to nA that is not carbamylated (nAZ(0) and nAl(O)), and nA that is fully carbamylated (nAZ(7) and nAl (7)). The numbers in parentheses indicate the number of carbamylated lysyl residues per molecule, while nA2 and nAl refer to nonphosphorylated and phosphorylated tuA-crystallin, respectively.
Lanes 2-5 in Fig. 2 show the extent of carhamylation of tuA-crystallins after 2, 4, 9, and 24 h of incubation in 100 mM KNCO. Although the phosphorylated and nonphosphorylated forms of uncarhamylated tuA-crystallin were easily separated by IEF (lane 1 ), some of the partially carbamylated forms were not separat.ed. T o identify species present in these bands, tuA2 and tvAl were isolated by preparative IEF, partially carbamylated, and analyzed by analytical IEF (results not shown). From these experimenh, it was evident that nA2 with one sit.e carbamylated had an isoelectric point between nAl a n d nA2. For nA with more than two sites of carbamylation, t h e isoelectric p0int.s of the carbamylated nA2 proteins overlapped with nAl with one less carbamylation site. The three hands in lane 6, obtained for nA-crystallin that had been extensively carhamylated in 6 M guanidine hydrochloride, are in order of decreasing pH nA2(6) + nA1(5), nA2(7) + nA1(6), and nAl (7). Densitometer measurements of lanes 2-5 indicated that an average of 1.2, 1.4, 2.3, and 2.7 lysines per molecule were carhamylated after 2, 4, 9, and 24 h of incubation, respectively.

Quantification of Carhamylation by FARMS-To determine
rate const.ants for carhamylation of each of t h e 7 lysyl residues in nA-crystallin, it was necessary to quantitatively assess the extent of carhamylation at each position. This was achieved by enzymatically fragmenting the protein and determining t h e relative ahundances of uncarbamylated and carbamylated peptides by FARMS. As indicated in Table I The m / t values of the carhamylated peptides are 4 3 mass units higher. enzymes were used to generate a set of peptides that included the 7 lysyl residues in nA. Selection of proteases was important, because large errors could be introduced if carhamylation of t h e lysyl residue affected production of the diagnostic peptide. For example, trypsin could not be used, because it cleaves COOH-terminal to Lys hut does not cleave at carbamylated Lys.
Not only must the carbamylated and noncarhamylated peptides he formed at the same rate, the FARMS response to the modified and unmodified forms of the lysine-containing peptides must also be known. Because modification of a sinEle sidechain may change the FARMS response, the effect of carbamylation was investigated. Oxidized €3-chain of insulin was used as a model peptide. Incubation of this peptide in isocyanate led to carhamylation of the NH2 terminus and L y P . T h e fully carbamylated peptide, as indicated by FARMS, was prepared by incubating the peptide with isocyanate in 6 M guanidine hydrochloride for 24 h. Equal amounts of uncarbamylated and carhamylated R-chain were combined, separated from reagents hv HPLC, and digested with V8-protease to give three peptides corresponding to segments 1-13, 14-21, and 22-30. T h e digest was fractionated by HPLC. The uncarbamylated and carbamylated forms of the 22-30 segment, which were expected to be equimolar, were collected in the same fraction and analyzed by FARMS. The spectrum (Fig. 3) has two peaks, corresponding to the molecular ions of the uncarhamylated ( m / z 1087) and the carbamylated ( m / z 1130) peptides. The ratio of the intensities of the ions averaged over 10 scans was 0.99, indicating that the FARMS response was not suhstantially affected by carbamylation of 1 lysyl residue. Results of similar investigations with several model peptides in which both the NH? terminus as well a s 1 lysyl residue were carbamylated indicated that carbamylation of 1 lysyl residue changes the FARMS response by less than 10%.
This comhination of peptide mapping and FARMS was used to determine the extent of carbamylation of all 7 lysyl residues in aA-crystallin as a function of the time the protein was incubated in isocyanate. Results of this investigation are presented in Fig. 4, which is a plot of percent modification uersus incubation time. These data show that all 7 lysyl residues are carbamylated, some more rapidly than others. T h e initial rapid increase of modification was followed by a plateau region, where the extent of modification varied from 7% for LysW to 61% for Lys" following 24-h incubation. Since the extent of modification did not increase significantly for incubation to 48 h, the possibility that the isocyanate was undergoing hydrolysis was considered. Analysis of a 100 mM isocyanate solution, incubated using the same conditions as were used to incubate the protein showed that the level of isocyanate decreased by 50% within the first 24 h of incubation (Fig. 4). Because the concentration of isocyanate was much greater than that of the protein, the decrease in isocyanate was attributed to hydrolysis (Kemp and Kohnstam, 1956). By combining the extent of modification of individual residues, it was determined that 1.2, 1.7, and 2.4 lysyl residues of aA-crystallin were carbamylated after 4, 9, and 24 h, respectively.
Since the combination of peptide mapping and FABMS is a new approach for determining the extent of modification at specific residues in proteins, the reliability of the method was investigated using different peptides that contain the same lysyl residue. suggest that the uncertainty in these measurements is usually less than 10%. Rate constants for the carbamylation a t each lysyl residue were calculated assuming that the reaction was first order with respect to the concentration of unmodified lysine, which was determined from the percent modification after adjustment for hydrolysis of isocyanate during the incubation (Fig.  4). The second-order rate constants for the reaction a t each lysine (Table 11) Tardieu et al., 1986). These subunits consists of two gene products, aA and aB, which have 0,1, or 2 serine residues phosphorylated (Spector et al., 1985;Voorter et al., 1986;Smith et al., 1991Smith et al., , 1992. The quaternary structure of acrystallin is of current interest (van den Oetelaar et d., Walsh et al., 1991) because of its probable relation to cataract, as well as the remarkable temperature stability of a-crystallins (Maiti et al., 1988). According to the three-layer micellar model (Tardieu et al., 1986;Walsh et al., 1991) there is a stable core of 18 aA subunits surrounded by an outer layer of approximately 24 aA or a B subunits. Rates of chemical modification of specific residues likely depend on quaternary structure, and may therefore be used as a probe of this structure. For example, the reactivity of the single cysteinyl residue in bovine aA suggests that there are three groups of LYA subunits, leading to the proposal of the three-layer model. If the subunits in the core are tightly bound by hydrophilic surfaces of the subunits, as proposed by Walsh et al. (1991), one would expect considerable variation in the exposure of lysyl residues in different positions in the sequence, as well as lysyl residues in the same position but on different molecules. In apparent contrast to the three-layer model, Schurtenberger and Augusteyn (1991) have presented results consistent with polymerization of monomeric units into linear chains.
One of the goals of this investigation was to determine whether some lysyl residues in aA-crystallin are particularly reactive nucleophiles or whether all lysyl residues have similar reactivities. As indicated in Table 11, the rate constants for carbamylation of the 7 lysyl residues in aA vary from 5 t o 54 X lo-' M" h-'. These results indicate that all of the lysyl residues in aA-crystallin can be carbamylated readily and must therefore have access to the solvent. As models for the quaternary structure of a-crystallin evolve, it will be important to include solvent accessibility for lysyl residues in all seven positions of aA-crystallin. The range of rate constants suggests that Lys", which is most reactive, is on the surface and has direct exposure to the solvent, while LysM, the least reactive, has less access to the solvent. This observation is consistent with the limited proteolysis results of Siezen and Hoenders (1979) which demonstrated that Arg" is especially prone to tryptic cleavage. The variation in reactivities of the lysyl residues may also reflect the different pK. values of the protonated +amines. Whether these different reactivities are due to tertiary or quaternary structure cannot be ascertained from the present results. It is significant however, that plots of ln[Lys] versus time of carbamylation are linear (Fig. 5), indicating that specific lysyl residues on different molecules have the same reactivity.
The present investigation has also used isoelectric focusing electrophoresis to examine the reaction of isocyanate with acrystallin. Since carbamylation of lysine decreases the positive charge on a protein by one unit, aA-crystallin subunits with different numbers of carbamylated lysyl residues can be separated, as illustrated in Fig. 2. The density of these bands is a direct measure of the distribution of the number of modified lysyl residues per molecule of aA, and may be used to examine the possibility that carbamylation of 1 residue alters the conformation causing an increased reactivity of other lysyl residues, a scenario that may be relevant to the role of isocyanate in some forms of cataractogenesis (Beswick and Harding, 1984;Crompton et al., 1985). For example, carbamylation of Lye?, the most reactive lysyl residue, might induce a conformational change that causes the remaining lysyl residues to be rapidly carbamylated. For limited reaction times, this model would lead to two populations of molecules, one unmodified and another highly modified. Analysis by IEF would give a bipolar distribution of bands. The relative abundances of the variously modified aA-crystallin was determined by scanning the IEF gel presented in Fig. 2. The product distribution obtained for the 24-h incubation, given in Fig. 6, shows that the experimental data are consistent with a uniform progression of carbamylation (noncooperative reaction) and that the distribution of products is not bipolar, as would be expected for a cooperative model.
Although IEF gives no information about carbamylation of specific lysyl residues, it does give information that complements and corroborates results obtained by mass spectrome-  Fig. 2. Calculated: distribution determined from the extent of modification at each lysyl residue (Fig. 4) with the assumption of a noncooperative model of reaction. try. For example, the number of carbamylated lysine residues per subunit after incubation in isocyanate for 4, 9, and 24 h determined by the two methods is in good agreement (see "Results"). A more stringent test of the internal consistency of the IEF and mass spectrometric results was conducted by assuming that a noncooperative reaction model dominates and using the mass spectrometric product distribution to calculate the IEF pattern. This calculated IEF pattern is given in Fig. 6 where it can be compared with the product distribution obtained by scanning the IEF gel. These product distributions are remarkably similar, further demonstrating the internal consistency of the IEF and mass spectrometric results and supporting the idea of noncooperative reactions of the lysyl residues of aA-crystallin.
The high reactivity of Lys" near the NH2 terminus suggests that this region is exposed to solvent, in spite of the fact that the region is hydrophobic. Studies by Ifeanyi and Takemoto (1991) indicate that the NH, terminus is involved in membrane recognition, a reaction that would be favored by both accessibility and hydrophobicity of the NH2 terminus. The high reactivity of Lys" also offers an explanation for the change in circular dichroism that occurs with carbamylation of a-crystallins. The circular dichroism studies of Beswick and Harding (1984) showed that carbamylation affected tryptophan and tyrosine chromophores. However, the effects of isocyanate incubation on tryptophan and tyrosine chromophores could not be determined independently. Since there are 6 tyrosines in aA-crystallin, none very close to the lysines, and only 1 tryptophan (Trpg), close to the highly reactive Lys", it seems likely that the altered circular dichroism spectrum following carbamylation is due to a change in the environment of Trp' caused by carbamylation of Lys".
The high reactivity of Lys" reported here also supports observations made in studies of aging bovine crystallins. Using antisera to nine regions of aA-crystallin, Takemoto et al. (1989) showed that changes occurred with aging which made the NH2-and COOH-terminal regions less reactive toward their antibodies. They suggested that the decreased recognition of the COOH terminus might be due to degradation, and decreased recognition at the NH2 terminus might be due to lower availability of the NH,-terminal epitope. Our data showing that lysines in both the COOH-terminal and NHp-terminal regions, Lys" and Lys'=, are highly reactive suggest the possibility that the decrease in antibody recognition at both termini might be due to age-related modification of the lysines in these regions. The low reactivity of Lys7* and Lysm supports their conclusion that this region i s less accessible to the solvent. However, the present results do not indicate that carbamylation leads to increased exposure (reactivity) of these internal lysyl residues. In a study of human lens tissue, Takemoto et al. (19901, again using antisera to the COOH-terminal region of aA-crystallin, found evidence suggesting that cataractous lenses have modifications in the region 164-173. Our results suggest the possibility that these modifications may be due to the high reactivity of LYS'~~. Previous investigations of carbamylation of lens proteins have been severely restricted by the experimental methods used. In some investigations, the carbamylated protein has been hydrolyzed by acid followed by amino acid analysis. Carbamylated lysine residues were detected as homocitrullin (Harding and Rixon, 1980). Unfortunately, acid hydrolysis also hydrolyzes the carbamylated lysine (Stark et al., 1960) and gives no information about the location of carbamylated residues. Other investigators (Garcia-Castiiieiras and Miranda-Rivera, 1983) have circumvented acid hydrolysis by using the reaction of 2,4,6-trinitrobenzenesulfonic acid with lysyl residues to quantify the free lysyl residues. However, addition of SDS to prevent precipitation of the proteintrinitrobenzenesulfonic acid derivative may cause errors, because SDS can also bind to the free amino groups, preventing the reaction of the amino group with trinitrobenzenesulfonic acid (Habeeb, 1966). Furthermore, this method only indicates whether the lysyl group is free; it gives no information about the nature or location of the modification. Radiolabeled isocyanate has also been used to investigate the reaction of isocyanate with a-crystallin (Harding and Rixon, 1980;Crompton et al., 1985). Although this approach indicated the uptake of radiolabel, neither the products nor their locations were determined.
The use of mass spectrometry in the present investigation of the reactions of isocyanate with a-crystallin gives a particularly detailed account of the reactivities of individual lysyl residues in aA-crystallin. All of the major products of this reaction were isolated, identified, and located with respect to the primary structure of the protein. Upon determining the relative response of FABMS to uncarbamylated and carbamylated peptides, the extent of carbamylation of specific lysyl residues was determined from the FAB mass spectra. This approach is of general significance because it should be applicable to different types of modifications, to modifications of different types of residues, and to different proteins.