Identification of S-Hydroxylysyl-methionine as the Covalent Cross-link of the Noncollagenous (NC1) Hexamer of the α1α1α2 Collagen IV Network

Collagen IV networks are present in all metazoans as components of basement membranes that underlie epithelia. They are assembled by the oligomerization of triple-helical protomers, composed of three α-chains. The trimeric noncollagenous domains (NC1) of each protomer interact forming a hexamer structure. Upon exposure to acidic pH or denaturants, the hexamer dissociates into monomer and dimer subunits, the latter reflect distinct interactions that reinforce/cross-link the quaternary structure of hexamer. Recently, the cross-link site of the α1α1α2 network was identified, on the basis of x-ray crystal structures at 1.9-Å resolution, in which the side chains of Met93 and Lys211 were proposed to be connected by a novel thioether bond (Than, M. E., Henrich, S., Huber, R., Ries, A., Mann, K., Kuhn, K., Timpl, R., Bourenkov, G. P., Bartunik, H. D., and Bode, W. (2002) Proc. Natl. Acad. Sci. U. S. A. 99, 6607-6612); however, at the higher resolution of 1.5 Å, we found no evidence for this cross-link (Vanacore, R. M., Shanmugasundararaj, S., Friedman, D. B., Bondar, O., Hudson, B. G., and Sundaramoorthy, M. (2004) J. Biol. Chem. 279, 44723-44730). Given this discrepancy in crystallographic findings, we sought chemical evidence for the location and nature of the reinforcement/cross-link site. Trypsin digestion of monomer and dimer subunits excised a ∼5,000-Da complex that distinguished dimers from monomers; the complex was characterized by mass spectrometry, Edman degradation, and amino acid composition analyses. The tryptic complex, composed of two peptides of 44 residues derived from two α1 NC1 monomers, contained Met93 and Lys211 post-translationally modified to hydroxylysine (Hyl211). Truncation of the tryptic complex with post-proline endopeptidase reduced its size to 14 residues to facilitate characterization by tandem mass spectrometry, which revealed a covalent linkage between Met93 and Hyl211. The novel cross-link, termed S-hydroxylysyl-methionine, reflects at least two post-translational events in its formation: the hydroxylation of Lys211 to Hyl211 within the NC1 domain during the biosynthesis of α-chains and the connection of Hyl211 to Met93 between the trimeric NC1 domains of two adjoining triple-helical protomers, reinforcing the stability of collagen IV networks.

duced its size to 14 residues to facilitate characterization by tandem mass spectrometry, which revealed a covalent linkage between Met 93 and Hyl 211 . The novel cross-link, termed S-hydroxylysyl-methionine, reflects at least two post-translational events in its formation: the hydroxylation of Lys 211 to Hyl 211 within the NC1 domain during the biosynthesis of ␣-chains and the connection of Hyl 211 to Met 93 between the trimeric NC1 domains of two adjoining triple-helical protomers, reinforcing the stability of collagen IV networks.
Collagen IV networks are components of basement membranes that underlie epithelia, compartmentalize tissue, and influence cell behavior. The networks are assembled from a family of six polypeptide chains (␣1-␣6) that associate forming three subtypes of triple-helical protomers with distinct chain compositions: ␣1␣1␣2, ␣3␣4␣5, and ␣5␣5␣6 (1, 2). The protomers self-assemble by end-to-end associations in which the amino termini of four protomers associate tail-to-tail forming the 7 S domain, and the carboxyl termini of two protomers associate head-to-head through the noncollagenous (NC1) 1 domains, forming dimers. At the interface of the head-to-head connection, the trimeric NC1 domains exist as a hexamer, a stable complex that can be excised by cleavage with collagenase for in vitro studies.
The NC1 domain plays a pivotal role in the assembly of the distinct collagen IV networks. In protomer assembly, the NC1 domains (monomers) of three chains interact, forming a NC1 trimer, to select and register chains for triple-helix formation. In the network assembly, the NC1 trimers of two protomers interact, forming a NC1 hexamer structure, to select and connect protomers. Upon exposure to acidic pH or denaturants, isolated NC1 hexamer dissociates into monomers and dimers, the latter reflecting the presence of cross-links that stabilize the trimer-trimer interface. The cross-links connect ␣1-like monomers (␣1-␣1, ␣1-␣5, and ␣3-␣5) and ␣2-like monomers (␣2-␣2, ␣2-␣6, and ␣4 -␣4) (3,4). For two decades, the reduc-ible dimers were thought to consist of monomers bound by disulfide cross-links (5,6). However, the recent x-ray crystal structures of the NC1 hexamers of bovine lens capsule basement membrane and human placenta basement membrane, determined independently by us (7) and Than et al. (8), respectively, have disproved this hypothesis. An alternative explanation was proposed by Than et al. (8) in which the cross-link is a thioether bond between Met 93 and Lys 211 that bridges the trimer-trimer interface; the evidence was based on electron density maps at 1.9-Å resolution, suggesting the existence of both cross-linked and noncross-linked residues at this site (8). However, in a subsequent study at the higher resolution of 1.5 Å, we found no evidence for this cross-link (9).
Given this discrepancy in crystallographic findings, we sought chemical evidence in the present study for the location and nature of the reinforcement/cross-link site. We used tryptic digestion in combination with mass spectrometry as a strategy to search for post-translational modifications that may go undetected by crystallography. The results revealed that the site is located at the trimer-trimer interface of the NC1 hexamer, characterized by a novel covalent cross-link: S-hydroxylysylmethionine. The cross-link is uniquely labile to conditions typically used for characterization, rendering it a challenge for detection. The findings are the first report of chemical evidence for the location and nature of the reinforcement/cross-link site, and the presence of Hyl within the NC1 domain.
Separation of Monomers and Dimers-Monomers and dimers of the NC1 domain were isolated as described elsewhere with minor modifications (10). Briefly, PBM hexamers (5 mg) were denatured in 0.2 M Tris-HCl, pH 8.5, buffer containing 4 M GdnHCl and 25 mM DTT and incubated in a boiling water bath for 20 min. Subsequently, the reduced proteins were alkylated with 50 mM iodoacetamide in the dark for 30 min at room temperature. To fractionate dimer and monomer subunits, the denatured hexamer sample was run through a Sephacryl S-300 column (120 ϫ 2.5 cm) that had been equilibrated in 50 mM Tris-HCl, pH 7.4, buffer containing 4 M GdnHCl. The fractions with higher absorbance at 280 nm were analyzed by 4 -20% linear gradient SDS-PAGE (Bio-Rad). The fractions containing monomer or dimers were pooled, concentrated, and washed with 50 mM ammonium bicarbonate, pH 7.8, buffer in Amicon Ultra filters 10,000 MWCO (Millipore Corp. Bedford, MA). In addition, to investigate the effect of DTT and denaturant as a function of temperature and time of incubation on hexamer dissociation, the chromatography analyses were performed on a TSK SW xl 3000 ToSo-Hass HPLC column equilibrated with 50 mM Tris-HCl, pH 7.5, containing 4 M GdnHCl, and connected to an ÄKTA Purifier HPLC chromatography system (Amersham Biosciences). Gels were stained with Bio-Safe colloidal Coomassie Brilliant Blue G-250 (Bio-Rad) for MS compatibility.
Separation of Tryptic Peptides by Size-exclusion Chromatography-NC1 dimers and monomers (reduced and alkylated) were separately incubated with sequencing grade-modified trypsin (Promega, Madison, WI) at a ϳ1:25 enzyme to protein ratio at room temperature for 16 h. Tryptic peptides were then separated in a Superdex TM peptide column (Amersham Biosciences) equilibrated with 50 mM ammonium bicarbonate, pH 7.8, and calibrated with small peptides. The elution of peptides was monitored by absorbance at 230 nm. The approximate molecular weight of the T 5k -complex was estimated by comparing its elution time to elution time of peptide standards.
Truncation of the T 5k -complex with Post-proline Endopeptidase-The T 5k -complex, isolated by gel filtration peptide column, was further truncated with post-proline endopeptidase from Flavobacterium meningsepticum (Seikagaku America, East Falmouth, MA) in 0.1 M ammonium bicarbonate, pH ϳ 7.8, for 3 h at 37°C. The products of the digestion were immediately analyzed by LC-ESI/MS (see below).
MALDI-TOF MS-Matrix-assisted laser desorption/ionization timeof-flight mass spectrometry (MALDI-TOF MS) analysis was performed on a Voyager 4700 mass spectrometer or on a Perceptive Biosystems Voyager Elite (Applied Biosystems, Foster City, CA). Tryptic peptides derived from the T 5k -complex were prepared by the dried-droplet method using ␣-cyano-4-hydroxycinnamic acid as a matrix. ␣-Cyano-4hydroxycinnamic acid was dissolved in water/acetonitrile/trifluoroacetic acid (50:49.9:0.1) at a concentration of 10 mg ml Ϫ1 . The peptides were initially identified by comparing the experimental masses of each peak with computer-predicted masses of tryptic peptides from the bovine NC1 domain (IV) sequences (7). Identity of some peptide sequences was confirmed by inducing ion fragmentation using the instrument in the tandem mode (MS/MS).

LC-ESI/MS/MS and LC-ESI/MS 3 -
The LC-ESI/MS/MS and LC-ESI/MS 3 analyses were performed on a ThermoFinnigan LTQ linear ion trap mass spectrometer equipped with a ThermoFinnigan Surveyor LC pump, microelectrospray source, and Xcalibur 1.4 instrument control and data analysis software. HPLC separation of the NC1 tryptic peptides was achieved with a C 18 capillary column at 0.7 ml min Ϫ1 flow rate. Solvent A was H 2 O with 0.1% formic acid and solvent B was acetonitrile containing 0.1% formic acid. The gradient program was: 0 -3 min, linear gradient 0 -5% B; 3-5 min, 5% B, 5-50 min, linear gradient to 50% B; 50 -52 min, linear gradient to 80% B; 52-55 min, linear gradient to 90% B; 55-56 min, 90% B in solvent A. A blank sample (0.1% formic acid) was run between the two analyses. Each sample was subjected to two LC-ESI/MS/MS analyses. In the first analysis, MS/MS spectra of the peptides were obtained using data-dependent scanning in which one full MS spectrum (mass range 400 -2000 atomic mass units) was followed by three MS/MS spectra. In the second run, several specific precursor masses were selected for MS/MS analysis in a targeted fashion. For the LC-ESI/MS 3 analyses, the MS 3 spectra were obtained using data-dependent scanning in which one full MS spectrum is followed by one MS/MS spectrum. The three most intense ions in the MS/MS spectrum were selected for a third fragmentation (MS 3 ).
Software for Sequence and Post-translational Modification Analysis-Samples of trypsin-digested monomers and dimers were analyzed using LC-MS/MS. The data base search algorithm SEQUEST was used to identify peptides from the fragment ions recorded in the tandem mass spectrum. P-Mod (11), a statistics based algorithm, allowed for the successful identification of hydroxylation sites.
Analytical RP-HPLC of the T 5k -complex-The T 5k -complex peak was analyzed on an ÄKTA Purifier liquid chromatography system run by UNICORN 4.11 software (Amersham Biosciences). Runs were performed on a Supelcosil LC-318 reversed-phase analytical HPLC column FIG. 1. Trypsin digestion reveals a structural difference between dimer and monomer subunits ␣1 NC1 domains. NC1 dimers and monomers were dissociated from bovine PBM NC1 hexamer in 4 M GdnHCl, reduced, and alkylated, and then fractionated by gelfiltration chromatography (Fig. 6, panel b). The dimer and monomer subunits were separately digested with trypsin, and the tryptic products were fractionated by size-exclusion chromatography on a Superdex peptide column as described under "Experimental Procedures." The fractionation was monitored by measurement of absorbance at 220 nm for monomers (dashed line) and dimers (solid line). The asterisk indicates the retention time of a tryptic (T) complex(s) of ϳ5000 molecular weight that distinguishes dimers from monomers.
(Sigma). The samples were loaded into the column that had been equilibrated with buffer A (95% water with 0.1% trifluoroacetic acid and 5% acetonitrile) with a flow rate of 1 ml/min. The peptides were eluted with a linear gradient up to 40% buffer B (0.085% trifluoroacetic acid, 95% acetonitrile and water) over 60 min. Peptide elution was monitored by absorbance at 215 nm.
Amino Acid Composition Analyses-Amino acid analysis was carried

Identification of a Structural Difference between Dimers and Monomers of the NC1 Domain: Isolation of the Reinforcement
Site by Trypsin Digestion-Trypsin digestion was used to identify structural differences between monomer and dimer subunits of the NC1 hexamer domain of collagen IV from bovine placenta basement membrane. To isolate monomers and dimers, the NC1 hexamers were reduced and alkylated in 4 M GdnHCl, then fractionated on a gel filtration column equilibrated with 4 M GdnHCl (Fig. 6, Ref. 9). The dimers and monomers were separately digested with trypsin under conditions for maximal cleavage. The peptide products were then fractionated over a size-exclusion chromatography column ( Fig.  1). A comparison of the profiles for monomer (dotted) and dimer (solid) reveals that dimers yield an extra peak (M r ϭ ϳ5000), and therefore are designated as T 5k -complex, indicative of a structural distinction between dimer and monomer subunits.
The chemical nature of the T 5k -complex was characterized by mass spectrometry. Because of the large size of the T 5kcomplex, the MALDI-TOF MS instrument was optimized to detect ions in the m/z range 500 -8000. Surprisingly, the most intense ion in the spectrum was m/z 1413.8 and the highest mass observed was m/z 3601.6 ( Fig. 2). The mass spectra were identical in linear or reflectron mode, which rules out a possible in-source fragmentation. Although the T 5k -complex was not observed, the combined mass of T-1414 plus T-3600 peptides is ϳ5014, a value close to that observed by size-exclusion chromatography.
To identify the sequence of T-1414 and T-3600 peptides, they were analyzed by MALDI-TOF-TOF tandem MS. Fig. 2B shows the fragmentation spectrum for the T-3600 peptide, which is consistent with the peptide sequence, 77 NDYSYWLST-PEPMPMSMAPITGENIRPFISR 107 , derived from the ␣1 NC1 domain. The mass of the T-1414 peptide was not predicted from the ␣1 NC1 domain sequence, therefore, it was considered to be a modified peptide. As shown in Fig. 2C, the fragmentation pattern of the T-1414 peptide is consistent with the sequence, 204 KPTPSTLKAGELR 216 , but with 16 extra mass units attached to Lys 211 . It is well known that lysine residues within the collagenous domain of collagen IV are post-translationally modified by a lysyl hydroxylase catalyzed addition of a hydroxyl group (12), converting lysine to 5-hydroxylysine (Hyl). Thus, the extra 16 mass units observed in the T-1414 peptide in comparison to the known primary sequence, suggests the presence of Hyl 211 (confirmed below). Therefore, the results are consistent with a stable complex (T-5014), composed of T-1414 and T-3600 peptides, which dissociate before or during sample preparation for MALDI-TOF MS analyses.
To test the hypothesis that T-1414 and T-3600 peptides form a stable complex, we analyzed the T 5k -complex, isolated by gel-filtration (Fig. 1), using LC-ESI/MS/MS. As shown in Fig. 3, the sample displayed two major peaks (Fig. 3A) each giving rise to a charge envelope containing ϩ3, ϩ4, ϩ5, ϩ6, and ϩ7 ions in the mass analyzer (Fig. 3B). The ions had an experimental The T-5014 and T-5030 complexes were further characterized by MALDI-TOF MS analyses. The T 5k -complex ( Fig. 1) was fractionated by reversed-phase HPLC yielding two major components that corresponded to T-5014 and T-5030 complexes (Fig. 4A). MALDI-TOF MS analyses (Fig. 4B) showed that the T-5014 complex is composed of T-1414 and T-3600 peptides, and that the T-5030 complex is composed of T-1414 and T-3616 peptides. The only difference between T-5014 and T-5030 complexes is the oxidation of methionine in the latter, which may have occurred during chromatography. Thus, the apparent masses of the complexes equal the sum of the masses of the constituent peptides, indicating that interaction of the peptides occurs without a change in mass (non-covalent), or a change in mass (covalent linkage) that is within the experimental error of the measurement for the 5,000 mass range. Moreover, the results establish that the T-5014 and T-5030 complexes dissociate under the conditions of MALDI-TOF MS analyses, a finding that explains why they were not observed in the earlier report (9) that used this method. The presence of fragments T-1460 and T-3554, which are the result of an alternative fragmentation of the tryptic complexes, are indicated in each spectrum (Figs. 1 and 4) and will be addressed below.
The T-5014 complex, isolated by HPLC (Fig. 4A), was also characterized by conventional amino acid analysis and Edman degradation. The amino acid composition (Table I) revealed a very close correlation between the experimental and theoretical values based on the sequences of the composite T-1414 and T-3600 peptides, identified by mass spectrometry, including one residue of Hyl in the T-1414 peptide. Amino acid analyses of monomer and dimer subunits also revealed the presence of one Hyl residue in monomer and two residues in the dimer (data not shown). Twenty cycles of Edman degradation revealed the presence of two peptides with sequences of 77 NDYSYWLSTPEPMPMSMAPITGENIR-PFISR 107 and 204 KPTPSTLK OH AGELR 216 , which correspond to T-3600 and T-1414 peptides (Fig. 5). The latter peptide contained Hyl at the eighth position, corresponding to Hyl 211 , as suggested by the above mass spectrometry studies. Furthermore, mass spectrometry and Edman degradation analyses indicate that this Hyl is not glycosylated. Moreover, that the T-5014 complex sequenced through 20 cycles without interruption at Met 93 or Hyl 211 indicates that the interaction between T-1414 and T-3600 peptides is labile under the conditions of Edman degradation. Evidence for a Covalent Cross-link in the Tryptic Complex (T-5014)-The chemical nature of the interactions between T-1414 and T-3600 peptides in the T-5014 complex was explored by LC-ESI/MS 3 mass spectrometry, using collision-induced dissociation for fragmentation. The fragmentation generated two modified peptides: T-1460 and T-3554 that differed in mass from the T-1414 and T-3600 peptides observed by MALDI-TOF (see above). Although the modified peptides (T-1460 and T-3554) were much less intense than T-1414 and T-3600 peptides, they were also observed in the MALDI-TOF MS spectra (Figs. 1 and 4). These results suggested that a chemical group of ϳ46 was transferred from the T-3600 to T-1414 peptide, possibly indicating a covalent cross-link between the two peptides. However, the location of mass changes within the two peptides could not be interpreted with confidence because of the complexity of MS 3 spectra because of the large masses of the peptides. To circumvent this problem, the T 5k -complex isolated in Fig. 1 was digested with a second protease to reduce its size for characterization by LC-ESI/MS 3 mass spectrometry.
The large number of prolyl residues in the T 5K -complex suggested the use of post-proline endopeptidase for truncation. This enzyme cleaves the peptide bond on the carboxyl side of prolyl residues (13), and it would reduce the T-5014 complex of 44 residues down to 14 residues, as depicted in Fig. 5. The truncated product would be composed of MSMAP (535.2) and STLK OH AGELR (989.6) peptides, which together exist as a complex with a theoretical mass of 1524.8 (designated as the P-1525 complex). The post-proline endopeptidase digestion product was analyzed by LC-ESI/MS 3 . Fig. 6a (left panel) presents a full MS spectrum showing two major ions at m/z 509.1 and 762.7 corresponding to the triply and doubly charged forms of the P-1525 complex, respectively. The experimental average mass of the P-1525 complex is 1523.9, approximately 1 mass unit less than the theoretical mass of 1524.8 (Fig. 5), indicating that interaction of the two constituent peptides of the P-1525 complex results in the loss of a single hydrogen.
Further MS/MS analyses of the doubly charged ion of m/z 762.6 revealed ions of m/z 488.3 and 1036.6 ( Fig. 6a, right  panel). These ions differ from the theoretical ones of 536.2 (535.2 ϩ H ϩ ) and 990.6 (989.6 ϩ H ϩ ), calculated for the two constituent peptides of the P-1525 complex (Fig. 5), by values of Ϫ48 and ϩ46 mass units, respectively. To obtain structural information about each of these ions and the nature of the 48 and 46 masses, each ion was submitted to a third collisioninduced dissociation fragmentation (MS 3 ) in the instrument. In Fig. 6b, the MS 3 spectrum at the m/z 488.3 fragment is shown; the fragmentation profile is consistent with the MSMAP peptide sequence, except that the b and y ion series demonstrate the loss of 48 atomic mass units of Met 93 . In contrast, in Fig. 6c, the fragmentation profile of the ion at m/z 1036.6 is consistent with the SKLK OH AGELR, but the y and b ion series demonstrate that Hyl 211 gained 46 atomic mass units of fragmentation of the P-1525 complex. These fragmentation profiles indicate a loss of a CH 3 S-group along with a proton (totaling 48 atomic mass units) from Met 93 and a gain of a CH 3 S-group onto Hyl 211 and the loss of a proton (totaling 46 atomic mass units).
Overall, the MS results provide compelling evidence that the two peptides of the P-1525 complex are connected by a covalent bond between the side chains of Met 93 and Hyl 211 , and that its formation is concomitant with the loss of 1 mass unit. A structure for the cross-link is proposed in Fig. 6d, in which the S atom of Met 93 is covalently linked to the C ⑀ atom of Hyl 211 , forming a sulfonium linkage. Alternatively, the S atom could be attached directly to the N atom on C ⑀ , to C ␦ , or the O atom on   The dissociation of PBM hexamer was also studied by SDS-PAGE with the samples treated with DTT and SDS. Fig. 7B shows a decrease in mobility of the monomer and dimer bands as expected by the reduction of intramolecular disulfide bonds (lanes 1 and 2). In addition, a gradual conversion of dimers into monomers (lanes 2-8) as a function of incubation time is clearly demonstrated. Importantly, incubation of hexamers for 2 h at 80°C in the absence of DTT (lane 1) did not change the 80:20 dimer:monomer ratio. Thus, DTT appears to exert two effects: the reduction of disulfide bonds and the breakage of the Met-Hyl cross-link.
Mass spectrometry analysis of the breakage of dimers into monomers can also provide information about the location and nature of the cross-link. In the dimer, the Hyl 211 -Ala 212 bond is resistant to trypsin cleavage, resulting in the excision of the T-1414 peptide with an intact Hyl 211 -Ala 212 bond (Fig. 5); such bonds are known to be susceptible to trypsin cleavage except when Hyl is modified by a carbohydrate unit (14). This poses the questions of whether the Hyl 211 -Ala 212 bond is resistant or susceptible to cleavage when present in the monomer, and whether it becomes susceptible after breakage of dimer into monomer by treatment with DTT. Thus, the status of the Hyl 211 -Ala 212 bond was explored in two kinds of monomers: monomers that were derived from hexamers and fractionated by gel-filtration chromatography (Fig. 7A, panel b), and monomers that were derived from the dimers (Fig. 7A, panel b) followed by treatment with 2% SDS plus 100 mM DTT and incubated for 2 h at 80°C. This objective was accomplished using SDS-PAGE to separate dimers and monomers, in-gel digestion of components with trypsin to release peptides, and LC-ESI/MS/MS analyses of tryptic peptides. The data sets were interrogated for the presence of the T-1414 peptide, which contain the intact Hyl 211 -Ala 212 bond, and T-887 peptide, which corresponds in mass to a peptide with the sequence of KPTPSTLK OH 211 and that indicates cleavage of the Hyl 211 -Ala 212 bond (Fig. 8). The T-1414 peptide was present in the dimer (lane 1), consistent with previous MS analyses, but not in the monomer (lane 2). Conversely, the T-887 peptide was present in the monomer but not in the dimer. The T-887 peptide was also present in monomer (lane 3) that had been treated at elevated temperatures, but with DTT. Moreover, the monomer that was derived by DTT treatment of dimer (lane 4) also revealed the presence of the T-887 peptide, but the complete absence of the T-1414 peptide. These results indicate that the Hyl 211 -Ala 212 bond is resistant to trypsin cleavage in the dimer subunit, but susceptible to cleavage in the monomer subunit, and that the resistance is broken by treatment with DTT. The findings are consistent with a covalent cross-link involving Hyl 211 rendering the Hyl 211 -Ala 212 bond resistant to trypsin cleavage. Moreover, the presence of Hyl 211 in both M-and D-hexamers suggests that the post-translational modification of Lys 211 in monomers is necessary but not sufficient for cross-linking of monomers to form dimers. DISCUSSION In a recent study (9), we presented evidence for the existence of two distinct kinds of hexamers, M ␣1␣2 -hexamers composed exclusively of monomers, and D ␣1␣2 -hexamers composed exclusively of dimers. The two extremes reflect a process that reinforces/cross-links the interaction of monomer subunits forming D-hexamers. The proportion of M-and reinforced D-hexamers indicate that the collagen IV network of PBM is a more stable structure than that of the lens capsule basement membrane, a feature that may be an important determinant of biological function. The reinforcement is not unique to the ␣1␣1␣2(IV) network, but also occurs in the ␣3␣4␣5(IV)and ␣1␣1␣2-␣5␣5␣6(IV) networks (1,3,4).
In the present study, the location and nature of the reinforcement/cross-link site of dimers of D ␣1␣2 -hexamers were investigated using trypsin digestion as a strategy to excise the site and using mass spectrometry for characterization. Fortuitously, a structural feature distinguishing dimers from monomer subunits could be excised as a low molecular weight complex that was easily purified for chemical and physical characterization. The tryptic complex is composed of two short peptides, comprising residues 77-107 and 204 -216 of an ␣1-NC1 monomer. The two peptides, respectively, contain Met 93 and Lys 211 post-translationally modified to Hyl 211 , and they correspond to one of the regions that are in close proximity at the trimer-trimer interface of the NC1 hexamer, as defined previously from the crystal structure (7)(8)(9). That the complex is derived exclusively from dimer subunits of D ␣1␣2 -hexamers, and its primary structure corresponds to a region that connects the trimer-trimer interface, provides the first chemical evidence for the location of the reinforcement/cross-link site (Fig.  9). Its location is in agreement with that proposed by Than et al. (8) based on connectivity observed by x-ray crystallography, but differs with respect to the presence of Hyl 211 instead of Lys 211 .
The chemical nature of the reinforcement/cross-link was elucidated by mass spectrometry of a truncated form of the tryptic complex. A second digestion with post-proline endopeptidase truncated the 44 amino acid residues down to 14, rendering the complex amenable to LC-ESI/MS 3 mass spectrometry. The fragmentation patterns, the loss the CH 3 S-group from Met 93 and its transfer to Hyl 211 , provided compelling evidence that the two peptides of the smaller complex are connected by a covalent bond between the side chains of Met 93 and Hyl 211 , and that its formation is concomitant with the loss of 1 mass unit, relative to the masses of unmodified Met 93 and Hyl 211 . The proposed structure, shown in Figs. 6d and 9, is a sulfonium ion in which the S atom of Met 93 is covalently linked to the C ⑀ atom of Hyl 211 . Alternatively, the S atom could be attached directly to the N atom on C ⑀ , to C ␦ , or the O atom on C ␦ of Hyl, either of which are consistent with all the LC-ESI/MS 3 results. The sulfonium ion linkage is consistent with the susceptibility of the cross-link to cleavage by DTT (Figs. 7-9), because methionine sulfonium ions undergo cleavage by sulfhydryl agents (15). Of particular note, the linkage is also susceptible to cleavage by the conditions of Edman degradation and MALDI-TOF analysis; the latter explains why the cross-link was not observed by MALDI-TOF analysis of dimer subunits in our previous study. Several amino acid residues, including methionine, are susceptible to radiation damage by synchrotron x-rays (16). The breakage in methionine is known to occur at the CH 3 S-group. This may be the reason why we did not detect the cross-link involving the Met 93 residue in the crystal structure (9) and why it is difficult to model accurately even at 1.5-Å resolution.
The novel cross-link is termed S-hydroxylysyl-methionine to denote a sulfur atom connection between Hyl and Met. At least two events occur in the formation of the cross-link: the posttranslational hydroxylation of Lys 211 to Hyl 211 within the NC1 domain during the biosynthesis of ␣-chains; and the connection of Hyl 211 to Met 93 between the trimeric NC1 domains. The cross-link connects two adjoining triple-helical protomers, reinforcing the stability of collagen IV networks.
The presence of hydroxylysine 211 in the NC1 domain is a novel feature. Hyl is a post-translational modification of lysine residues that typically occurs in X-Lys-Gly triplets in the collagenous domain of various types of collagens and collagen-like proteins (12). Hyl was previously noted in the preparation of NC1 domains (17), but it was thought to be a contaminant from the collagenase digestion of the collagenous domain or a residue in the two Gly-X-Y triplets at the NH 2 terminus of the isolated NC1 domains (Fig. 5, Ref. 18). In certain cases, hydroxylation occurs in nonhelical peptides with sequences of X-Lys-Ala(Ser) at the end of collagen I chains that are involved in aldehyde-derived cross-links (19). This consensus sequence is identical to the X-Lys 211 -Ala sequence in the T-1414 peptide.
Hyl residues have two important functions. They are essential for the stability of the intermolecular collagen cross-links, and their hydroxyl groups serve as attachment sites for the monosaccharide galactose or the disaccharide glucosyl-galactose (20 -22). The carbohydrate units influence the lateral packing of fibril-forming collagen molecules into fibrils and may facilitate the assembly of the 7 S domain of collagen IV (23). In a recent study, lysyl hydrolysase-3 was shown to be essential for the assembly of collagen IV networks, presumably because of the absence of the hydroxylysine-linked carbohydrates (24). Conceivably, such carbohydrate units attached to Hyl 211 might prevent the self-assembly of collagen IV protomers into networks within the intracellular environment, but upon deglycosylation in the extracellular space, network assembly could take place. In a related example, the NH 2 -and COOH-terminal propeptides of fibril-forming collagens prevent assembly of fibrils until removed by proteases in the extracellular environment.
Finally, the S-hydroxylysyl-methionine cross-link, likely occurs in all three collagen IV networks (␣1␣1␣2, ␣3␣4␣5, and ␣1␣1␣2-␣5␣5␣6 networks). Each NC1 domain of the six human ␣-chains exist as both monomers and dimers: the ␣1-like mono- FIG. 9. Summary of findings regarding the location and nature of the reinforcement/cross-link site. The top panel shows the space-filling model for the reinforced/cross-linked D-hexamer (left) comprised of two trimeric caps, each composed of two ␣1 monomers and one ␣2 monomer. The juxtaposition of two opposing ␣1 NC1 monomers at the hexamer interface generates two sites of reinforcement: site 1 and site 2, which are defined by the tryptic 5014 complex composed of T-3600 (red) and T-1414 (green) peptides. The side chain of Met 93 is colored in gold, and the side chain of Hyl 211 is colored in cyan. In site 2, the peptide (not the side chains) colors are reversed. Under denaturing conditions, the D-hexamer dissociates into dimer subunits (middle). The dimer upon trypsin digestion releases T-5014 complex (right), composed of the T-3600 peptide (31 residues, red) and T-1414 peptide (13 residues, green), which constitutes the reinforcement site. The T-1414 peptide contains a post-translational modification in which Lys 211 is converted to Hyl 211 . For mass spectrometry analyses to determine the chemical nature of the cross-link, the T-5014 complex was truncated by digestion with post-proline endopeptidase (PPE), which generated the P-1525 complex composed of the MSMAP peptide covalently cross-linked to the STLK OH AGELR peptide. The side chains of Met 93 and Hyl 211 are covalently connected by a sulfonium ion linkage between S ␦ of Met 93 and C ⑀ of Hyl 211 , as presented in Fig. 6d. A magnification of the proposed S-hydroxylysyl-methionine cross-link is shown. The bottom panel illustrates the non-reinforced/non-cross-linked M-hexamers (left), which dissociates into NC1 monomers (middle) upon denaturation. In this case trypsin digestion generates T-3600 and T-887 peptides that are not cross-linked in the hexamer. T-887 peptide, derived from the monomers, is a truncated version of T-1414 peptide derived from dimers. The structural relationship of these peptides indicates that in the dimer the Hyl 211 -Ala 212 bond is blocked to trypsin cleavage, but in the monomer this bond is susceptible to cleavage. This difference in susceptibility to trypsin cleavage provides independent support for the location of a cross-link at Hyl 211 , and the difference is consistent with a covalent linkage between Met 93 and Hyl 211 . The presence of Hyl 211 in both M-and D-hexamers suggests that the posttranslational modification of Lyl 211 in monomers is necessary but not sufficient for cross-linking of monomers to form dimers. mers form ␣1-␣1, ␣1-␣5, and ␣3-␣5 dimers, and the ␣2-like monomers form ␣2-␣2, ␣2-␣6, and ␣4-␣4 dimers (3,4), indicating the presence of a cross-link. As noted previously by Than et al. (8), the sequences encompassing Met 93 and Lys 211 are invariant among all six human ␣-chains, suggesting that the same post-translational modifications could occur in each chain. Of particular importance, the reinforcement/cross-link of the ␣3␣4␣5 network was recently found to sequester B-cell epitopes within the NC1 hexamer, rendering them inaccessible to pathogenic autoantibodies in patients with Goodpasture syndrome. Thus, the cross-link represents a novel molecular mechanism for establishing immune privilege (25).