The Use of 2-Dimensional CNBr Peptide Maps for the Analysis of Crosslinked Peptides in Bone Collagen

Collagen Rel. Res. Vol. 1/1981, p. 247-256

The Use of 2-Dimensional CNBr Peptide Maps for the Analysis of Crosslinked Peptides in Bone Collagen DAVID T. CHEUNG, PAUL D. BENYA, DANIEL P. HOFER and MARCEL E. NIMNI Departments of Orthopaedics, Medicine, and Biochemistry, Orthopaedic Hospital-University of Southern California, Bone and Connective T issue Research Programs, Orthopaedic Hospital, Los Angeles, California 90007, USA.

Abstract CNBr peptides from insoluble bovine cortical bone collagen were analyzed using a 2-D mapping technique. The major type I collagen CNBr peptides were detected by fluorography after general-Iabelling with 3H-NaBH4 in dimethylformamide. These maps were similar to those visualized by coomassie blue staining and demonstrated a proportional decrease of a 1CB6. New groups of peptides, different from those normally present in soluble type I collagen were detected. Some of these peptides were slightly larger and more acidic than a1CB6 and were highly labelled when the demineralized bone was specifically labelled for the presence of aldehydes and crosslinks with 3H-NaBH4 in a phosphate buffer, pH 7.4. Based on the size and charge characteristics of these specifically labelled peptides, they were tentatively identified as crosslinking peptides containing different combinations of a1CB6, a1CBO,1 and a1CB5. The specificity of the labelling method using 3H-NaBH 4 in phosphate buffer was demonstrated by the detection of other known crosslinked peptides and by the virtual absence of label in a1CB7, CB8, and CB3. We feel that this simple methodological approach developed in these experiments will prove to be very useful in the analysis of collagen crosslinks present in insoluble collagens derived from norm al tissues of various ages as well as from pathological states. Key words: Collagen CNBr Peptides, Crosslinks, 2-D Mapping Introduction Lysine- and hydroxylysine-derived crosslinks in hard tissue collagen, and their subsequent in vivo reduction or rearrangement to ketimines have been suggested by Mechanic et al. (1971, 1974) and Eyre and Glimcher (1973) to play important roles in determining the intractability of tissues such as bone and dentine. In order to understand how different kinds of crosslink affect the physi-

248

D. T. Cheung, P. D. Benya, P. Hafer and M. E. Nimni

cal properties of collagen fibers, it is desirable to locate these crosslinks along the collagen molecules. Crosslinks occur in bone and dentine collagen between the following pairs of peptides: a1CBO,1 and a1CB6 (Volpin and Veis, 1973); a1CB5 and a1CB6 (Eyre and Glimcher, 1973); a1CB6 and a2CB4 (Scott, Veis and Mechanic, 1976); a1CB7 and an unknown peptide in the a2 chain (Fujii, Corcoran and Tanzer, 1975); and a1CB6 with a lysine residue in the center of the helical region of a2 chain (Stimler and Tanzer, 1979). In contrast to these simple bifunctional crosslinks, the formation of "complex interfibrillar stable crosslinks" was proposed by Light (1979) based on the isolation of "poly a1CB6" Vlhich appeared to contain a1CB6, a1CBO,1 and a1CB5. The discovery and isolation of multifunctional crosslinks such as hydroxyaldolhistidine (Housley et al., 1975) and pyridinoline (Fujimoto et al., 1978) further demonstrate the complexity of the crosslinks that stabilize collagen matrices. Recently a 2-D mapping technique for the analysis of collagen CNBr peptides has been reported (Benya, 1981). The usual collagen CNBr peptides were weil resolved with this technique and uncleaved peptides appeared at predictable positions based on their size and charge. Since crosslinked peptides should behave in a fashion analogous to the uncleaved peptides, we applied this same technique to study the crosslinked CNBr peptides of bone collagen. This report presents the 2-D CNBr peptide mapping patterns of insoluble bone collagen labelled by a modification of the procedures of Tanzer and Mechanic (1968) to label peptides containing aldehydes and crosslinks (specific-Iabelling) and by a procedure developed in this laboratory to label all peptides to a high specific radioactivity (general-labelling). Striking differences were observed between the 2-D patterns of specifically labelled crosslinked peptides and gene rally labelled total peptides. These studies reveal the possible participation of a1CB6 in crosslink formation with a1CB5 and a1CBO,1 from neighboring collagen molecules in bone matrix. Materials and Methods Bovine femoral cortical bone from a 2-year-old calf was purchased fresh from the slaughter house. 3H-NaBH 4 , (9.8 Ci/mmole) was obtained from Amersham Corp., Ill., and stored at room temperature and kept unopened until used. Dimethylformamide (DMF, silylation grade) was obtained from Pier ce Chemical Co., Ill. and stored under nitrogen at all times. All other chemicals used were either reagent grade or described in a previous paper (Benya, 1981).

Demineralization 01 Bovine Cortical Bone Freshly cut bovine bone pieces (1 cm3) were washed in a chloroform/methanol (1: 1 v/v) mixture for two days at 4 oe. Sampies were washed with methanol for 6 hours and treated with 0.25 M EDTA (pH 7.4) for two weeks at 4 °C with frequent changes of fresh EDTA solution. These partially demineralized bone pieces were cut into slices (0.5 mm thick) and treated with EDTA solution for at least one additional week to compiete the demineralization. Slices were washed with 0.05 M phosphate buffer (pH 7.4) containing 1M NaCI to e1iminate any remaining EDTA and washed with distilled water just before use.

The Use of 2-Dimensional CNBr Peptide Maps

249

Preparation 01 3H-NaBH4 in DMF All containers and liquid handling tools such as Reacti-Vials (pierce Chemical Co., m.) and syringes were pre-washed sequentially with 0.1 M HCI for 30 min., 0.1 M NaOH for 30 min., and 50 mM NaBH 4 overnight, and then rinsed with distilled water. Immediately before use, they were purged with nitrogen until dry. 200 ,ul of DMF was added to a new ampule of 3H-NaBH4 (total of 100 mCi) and then immediately transferred into a nitrogen purged 1 ml Reacti-Vi al. Successive 200 ,ul portions of DMF were added into the ampule and transferred into the Reacti-Vial in the same manner until the final volume was 1 ml. The solutions were weil mixed by shaking and always used immediately. Speci(ic Labelling 01 Crosslinks and Aldehydes All specific labelling and washing procedures were carried out in phosphate buffer (pH 7.4) at room temperature. Demineralized bone slices (8 mg wet weight) were first equilibrated in 0.5 M phosphate buffer (pH 7.4). The reduction was carried out by immersing the bone slices in 0.5 ml of 0.5 M phosphate buffer (pH 7.4) preferably in a capped glass tube. An aliquot of 20 ,ul of freshly prepared 3H-NaBH4 /DMF solution (approximately 10 mole of 3H-NaBH 4/mole of collagen chain) was added directly to the phosphate buffer. The reduction mixture was then vortexed quickly and allowed to stand for one hour at room temperature. At the end of the ho ur, the highly radioactive phosphate buffer solution was decanted and properly disposed. The labelled bone slices were washed repeatedly with 5 mls of 0.5 M phosphate buffer to remove the majority of the soluble radioactivity. Finally, they were washed in a large volume of 0.5 M acetic acid overnight. It is recommended that these procedures be carried out in a chemical hood and care taken in the disposal of all radioactive wastes. General Labelling 01 Type land Bone Collagen Peptides All general labelling and washing procedures were carried out in anhydrous DMF at room temperature. 2 mg of purified bovine Type I collagen was lyophilyzed from acidic acid, and 0.5 ml of DMF added to it just before labelling. Bovine bone slices (8 mg wet weight) were washed in distilled water and thoroughly dehydrated in DMF. The final volume for labelling was 0.5 ml DMF in a capped glass tube. To both kinds of sampies, 20 ,ul of freshly prepared 3H-NaBH4 /DMF solution (approximately 1 mole of 3H-NaBH4 /3 moles of lysine in bone collagen) was added and mixed thoroughly. The reaction was allowed to continue for one ho ur. After the labelling reaction, the purified collagen sam pie was mixed with 0.5 mls of phosphate buffer (pH 7.4) and dialyzed against 1 liter of 0.5 M acetic acid until a minimal amount of radioactivity was found in the dialysates. For bone slices, 3H-NaBH4 /DMF was removed and properly disposed. The labelIed slices were washed sequentially for 30 min. with 5 ml of DMF, 0.5 M phosphate buffer and overnight with 0.5 M acetic acid. CNBr Cleavage Demineralized bovine bone slices or purified soluble bovine fetal skin Type I collagen were cleaved by CNBr (10 mg/mI in 70 Ofo formic acid, 2 to 4-fold weight excess) for 4 ho urs at 30°C (Miller, 1971).

250

D. T. Cheung, P. D. Benya, P. Hofer and M. E. Nimni

2-D Mapping of CNBr Peptides Soluble peptides from the CNBr cleavage of bovine bone and skin Type I collagen were subjected to 2-D mapping. This procedure consisted of nonequilibrium isoelectric focusing in the first dimension and NaDodS0 4 electrophoresis in the second dimension (Benya, 1981). Fluorography After electrophoresis, gels were fixed in 20 % trichloroacetic acid at 20°C for one hour prior to processing for fluorography (Bonner and Laskey, 1974; Laskey and Mills, 1975). Purification 01 Fetal Bovine Skin Type I Collagen Freshly frozen fetal bovine skin was first minced and extracted in neutral salt solution. The residue was extracted with 0.5 M acetic acid. Collagen from the acid extract was purified by aseries of neutral-salt and acid-salt precipitations (Benya, Padilla and Nimni, 1977).

Results 2-D CNBr Peptide Map from Generally LabelIed Bone Collagen The general labelling procedure was used to label insoluble demineralized cortical bone slices. After labelling, 97 Ofo of the radioactivity was solubilized by CNBr yielding 4 X 10 6 cpm/mg wet demineralized bone. More than 99 Ofo of the total solubilized radioactivity was also soluble in ampholine solution, and this material was subjected to 2-D mapping. After the focusing step, less than 3 Ofo of the total radioactivity was found above the focusing gel. Thus, the 2-D map shown in Figure 1 A represents approximately 94 Ofo of the labelIed peptides in the original bone slice. The majority of al and a2 CNBr peptides of Type I collagen were present in the 2-D map shown in Figure 1 A. Each peptide appeared as aseries of spots due to charge heterogeneity. This pattern was similar to that of the isolated soluble skin Type I collagen (data not shown). There was no evidence for the presence of collagen Types II, III (Benya, 1980) or V (Benya, in preparation). However, clear differences could be observed when this map was compared to the reported al(I) CNBr peptide pattern (Benya, 1981). The amount of alCB6 appeared to be considerably decreased. The decrease of alCB6 was accompanied by appearance of additional peptide spots found in regions not occupied by any known Type I collagen CNBr peptides. This suggested the involvement of CB6 in crossIinking. The pattern of the radioactive map was consistent with the coomassie blue stained map shown in Figure 1 B. This indicated that artifacts were not introduced by the in vitra generallabelling procedure. 2-D CNBr Peptide Map from Specifically LabelIed Bone Collagen Insoluble bovine demineralized bone was reduced with 3H-NaBH4 /P0 4 at pH 7.4, in order to label only the reducible crosslinks and aldehydes in collagen (Tan-

The Use of 2-Dimensional CNBr Peptide Maps

251

zer and Mechanic, 1968). After labelling, 97010 of the radioactivity was solubilized by CNBr cleavage yielding approximately 2 X 105 cpm/mg wet demineralized bone. More than 99 '010 of the CNBr solubilized radioactivity was also soluble in ampholine solution and this material was subjected to 2-D mapping. After the focusing step, less than 15 % of the total radioactivity was found above the focusing gel. Thus, 82 Ufo of the labelled peptides in the specifically labelled bone were present in the 2-D map shown in Figure 2. The 2-D map of specifically labelled peptides (Fig. 2) was quite different from that of the total peptides after general labelling. Specifically, only a trace amount of label was present in a1CB8, CB3, CB7, and aCB4. This indicated that these peptides are not effective substrates for lysyl oxidase or they are involved in crosslinking immediately after enzyme action. In addition to these differences, significant labelling was found in the areas not occupied by Type I collagen CNBr peptides. Most prominent were the peptides that mapped above and on the acid side of alCB6 (A, B, C, D, and E peptides in Fig. 2). These were considered to be crosslinked peptides because they were labelled during reduction with 3H-NaBH4 /P0 4 and because of their increased molecular weight relative to alCB6. This group

Fig. IA. 2-D map of "H-NaBH4 /DMF generally-labelled bone collagen CNBr peptides. Each peptide exists as a horizontal set of charge isomer spots. All a2 peptides have a superscript 2 for identification. The initials CB preceding each peptide have been omitted for clarity. The polarity and the direction of simple migration during the isoelectric focusing step are indicated by the symbols +, -, and --+.

252

D. T. Cheung, P. D. Benya, P. Hofer and M. E. Nimni

Fig. 1B. 2-D map ot coomassie blue stained bone collagen CNBr peptides. Note the similarity to the profile seen in Figure 1A. of peptides appeared to consist of five different composite peptides contammg a1CB6. Other peptides with molecular weights between 42,000 and 45,000 daltons (F) and slightly less than a1CB7 (G) were also specifically labelled. In the regions between a2CB3-5 and the uncleaved peptides CB8-3, 3-7, 7-6, there were at least two additional kinds of crosslinked peptides (H and I). Due to the size and charge characteristics of the most basic of these, we considered it to be a1CB6 X a2CB4 as reported by Scott, Veis and Mechanic (1976). It should be emphasized that although a2CB4 was involved in crosslinking, no peptide with the mobility of a2CB4 was labelled with 3H-NaBN4 /P0 4 In contrast, a1CB6 was specifically labelled indicating the presence of aldehydes. The crosslink in a1CB6 X a2CB4 may be derived from an aldehyde donor in a1CB6 and an amine donor in a2CB4. A plot of mobility versus log M r was derived from Figure 2 and was used to estimate the molecular weights of the specifically labelled crosslinked peptides A-E (Table 1). The charge separation in 2-D maps resolved the peptides of similar size so that they could be identified as different peptides. This size and charge information was utilized as described for uncleaved peptides (Benya, 1981) to estimate the identity of peptides A-E (see Discussion). o

The Use of 2-Dimensional CNBr Peptide Maps

253

Fig. 2. 2-D map of 3H-NaBH4 /P0 4 specifically-labelled bone collagen peptides. Some of the labelIed peptides are enclosed in boxes identified as A-G and 6 ald. Open circles represent peptides that are not specifically labelIed but appear in the generally-labelled map (Fig. 1).

Table 1. Estimated Molecular Weights of Specifically LabelIed Peptides A-E in Figure 2 Peptide

M r (Observed)"

Suggested Peptide Composition

M r (Calculated) ""

6 ald A B C D E

20.2K 21.7K 22.9K 23.8K 26.0K 25.3K

6 6 X 0.1 6 X (0.1). 6X5 6 X 5 X 0.1 6 X 5 X (O.lh

20.2K 22.1K 24.0K 23.7K 25.6K 27.5K

" The observed molecular weights expressed as 103 (K) daltons were estimated from a plot of mobility vs M r of the known CNBr peptides in the same 2-D map. "." The calculated molecular weights were based on the proposed identity of these composite peptides (see discussion). 16

Collagen 1/3

254

D. T. Cheung, P. D. Benya, P. Hofer and M. E. Nimni

Discussion We have combined eXlstmg and newly developed methodology for labelling collagen crosslinks and aldehydes in order to identify CNBr peptides containing such functional groups. Easily discernable differences have been demonstrated between 2-D CNBr peptide maps of bone slices labelled by procedures designed to label peptides indiscriminately (general-Iabelling) and by the standard method for labelling aldehydes and cross links (specific-Iabelling). Although quantitative data was not determined, a large decrease in the amount of a1CB6 was easily visualized when insoluble bovine cortical bone collagen was compared with the insoluble type I skin collagen, or the a1(I) CNBr peptide map (Benya, 1981). This is in agreement with the suggestion that the majority of the a1CB6 is involved in the intermolecular crosslinks in insoluble bone collagen (Light, 1979). The appearan ce of specifically labelIed peptides with slightly higher molecular weight on the acidic side of a1CB6 suggests that these may contain reducible crosslinks derived from a1CB6 crosslinked to sm aller peptides such as a1CBO,1 and a1CB5 (Volpin and Veis, 1973; Eyre and Glimcher, 1973). Preliminary information concerning the composition of the crosslinked peptides can be deduced by estimating their size and charge characteristics from the location of each peptide on the map. For ex am pie, the molecular weight of the peptides A and B, in Figure 2, immediately above a1CB6 were estimated to be 21,700 and 22,900 daltons, respectively. Peptide A was shifted in the acidic direction relative to a1CB6 and peptide B was shifted a similar distance from peptide A. This suggests sequential additions of a small acidic peptide onto a1CB6. These criteria are met by a1CBO,1 which has a molecular weight of 1,881 daltons and a net charge of - 2 if the lysine residue at 9N is oxidized to allysine. Crosslinking between a1CB6 and a1CBO,1 would result in a composite peptide more acidic than a1 CB6 with a molecular weight of 22,080 daltons and mapping characteristics similar to peptide A. Similarly, if another a1CBO,1 peptide were crosslinked to peptide A, a larger (Mr = 23,970 daltons) and more acidic composite peptide would result and map similar to peptide B. Although the measured molecular weights of peptides A and B were slightly sm aller than the calculated values for a1CB6 X a1CBO,1 and a1CB6 X (a1CBO,1)2 (Table 1), this may be explained by the non-helical nature of a1CBO,1. A similar assumption concerning helical and non-helical domains has been used to explain the electrophoretic mobility of prepro collagen (Sandell and Veis, 1980). The molecular weights of the peptides C, D and E in Figure 2 were 23,800, 25,300 and 26,000 daltons, respectively. The trend of increased molecular weight accompanied by increased negative charge was again observed. The molecular weight of peptide C was consistent with the combined molecular weight of a1CB6 and a1CB5. Also, since a1CB5 has similar charge characteristics to a1CB6, crosslinking between the two should yield a composite peptide with the same charge characteristics of a1CB6. Therefore, we suggest that peptide C may be a1CB6 X a1CB5. Based on the rationale for increased negative charge mentioned above, we also suggest that peptides D and E are a1CB6 X a1CB5 X a1CBO,1 and a1CB6 X a1CB5 X (a1CBO,1)2 respectively. It is likely that the peptides that are labelIed F in Figure 2 are crosslinked peptides, similar to peptides A-E, with the exception that these composite peptides may each contain two a1CB6 peptides. This is consistent with their charge characteristic and apparent molecular weight ranging

The Use of 2-Dimensional CNBr Peptide Maps

255

between 42,000 and 45,000 daltons. The presence of peptide G could not be explained by any combination of presently known bovine Type I collagen CNBr peptides. At present, direct chemical techniques are being used to verify the identity of each set of spots and evaluate the possible existence of attached noncollagenous materials. In the course of developing 2-D mapping for crosslinked peptide analysis, we employed DMF as a solvent for storing 3H-NaBH4 of high specific radioactivity. We also attempted to carry out the reduction of crosslinks and aldehydes in anhydrous DMF to avoid rapid decomposition of 3H-NaBH4 • As shown in Figure lA, the resulting 2-D map indicated the occurrence of a generallabelling phenomenon. Since labelling to high specific activity was possible with this method, it was used to circumvent the limited sampIe capacity of the 2-D mapping technique and the insensitivity of staining techniques. The mechanism of this general labelling is under current investigation. Additional data suggest that labelling occurs via the methylation of the c-amino groups of lysine re si du es (Cheung, manuscript in preparation). Specific reduction with high specific activity 3H-NaBH4 can still be obtained if the reduction is carried out in phosphate buffer at neutral pH. However we have observed that storage of 3H-NaBH4 in DMF at -20°C or -70°C, or changing the reaction conditions slightly can cause a loss in the specificity of the labelling. Our best criteria for the specificity of labelling techniques is to compare generally and specifically labelled 2-D maps of the same material. While a2CB4 was found to be involved in crosslinking, a2CB4 itself was not observed to contain aldehydes. It is possible that a2CB4 contributed only its unoxidized lysine or hydroxylysine residues during the formation of crosslinks, or that all lysine or hydroxylysine residues were immediately crosslinked after oxidation. Nevertheless, alCB7, CB8, CB3, and the majority of a2CB4 which are located in the interior or the helical region of the collagen molecule, do not appear to be substrates for lysyl oxidase. On the contrary, the presence of radioactivity in alCB6 in the specifically-labelled 2-D map indicates the presence of aldehydes. This reflects the fact that terminal peptides may be better substrates for lysyl oxidase. The presence of aldehydes in alCB6 also suggests that these aldehydes are not immediately incorporated into cross links or that they may never be involved in crosslink formation. We have applied the technique of 2-D CNBr peptide mapping to the analysis of insoluble bone collagen. The development of the general labelling technique has allowed very small tissue sampIes to be analyzed. Coupled with the specific labelling by reduction of crosslinks and aldehydes, a rapid survey of total collagen CNBr peptides and crosslinked peptides was accomplished. This methodological approach will facilitate the analysis of crosslinked collagen peptides from many normal and pathological tissues. Abbreviations: CNBr, cyanogen bromide; 2-D, two-dimensional 3H-NaBH4 , tritiated sodium borohydride; DMF, dimethylformamide; M r, molecular weight; 3H-NaBHiP04 ; 3H-NaBH4 in 0.5 M phosphate buffer; 3H-NaBH/DMF, 3H-NaBH4 in anhydrous DMF; EDTA, ethylenediaminetetraacetic acid.

256

D. T. Cheung, P. D. Benya, P. Hofer and M. E. Nimni

Acknowledgements The authors wish to thank Silvia R. Padilla for her excellent technical assistance and Evelyn Calbes for helping in the preparation of this manuscript. The research was supported by grants AM10358 and AM16404 from the National Institutes of Health, Bethesda, Maryland 20014 and funds from the Orthopaedic Hospital, Los Angeles, California 90007. References Benya P. D., S. R.Padilla, and M. E. Nimni: The progeny of rabbit articular chondrocytes synthesize collagen types land III and type I trimer, but not type 11 .. Verifications by cyanogen bromide peptide analysis. Biochemistry, 16: 865, 1977. Benya, P. D.: Two-dimensional CNBr peptide patterns of collagen types I, II and IIL Coll. Res. 1: 17-20, 1981. Bonner, W. M. and R. A. Laskey: A film detection method for tritium labelled proteins and nucleic acids in poly acryl amide gels. Eur. J. Biochem. 46: 83-88, 1974. Eyre, D. R. and M. J. Glimcher: Analysis of a crosslinked peptide from calf bone collagen: Evidence that hydroxylysyl glycoside participates in the cross link. Biochem. Biophys. Res. Commun. 52: 663-671, 1973. Fujii, K., Corcoran D. and M. L. Tanzer: Isolation and structure of a cross-linked tripeptide from calf bone. Biochemistry 14: 4409-4413, 1975. Fujimoto, D., Moriguchi, T., Ishida T., and H. Hayashi: The structure of pyridinoline, a collagen crosslink. Biochem. Biophys. Res. Commun. 84: 52-57, 1978. Housley, T. J., Tanzer, M. L., Henson, E., and P. M. Gallop,: Collagen crosslinking: Isolation of hydroxyaldolhistidine, a naturally occurring crosslink. Biochem. Biophys. Res. Commun. 67: 824-829, 1975. Laskey, R. A. and A. D. Mills: Quantitative film detection of 3H and HC in poly acryl amide gels by fluorography. Eur.}. Biochem. 56: 335-341, 1975. Light N. D.: Bovine type I collagen, a study of cross-linking in various mature tissues. Biochim. Biophys. Acta 581: 96-105, 1979. Mechanic, G. L., Gallop, P. M. and M. L. Tanzer: The nature of crosslinking in collagens from mineralized tissues. Biochem. Biophys. Res. Commun. 45: 644-653, 1971. Mechanic, G. L., Kubaki, Y., Shimokawa, H., Nakamoto, K., Sasaki, S. and Y. Kawanishi: Collagen crosslinks: Direct quantitative determination of stable structure crosslinks in bone and dentin collagens. Biochem. Biophys. Res. Commun. 60: 756-763, 1974. Miller, E. J.: Isolation and characterization of cyanogen bromide peptides from the a1(II) chain of chick cartilage collagen. Biochemistry 10: 3030-3035, 1971. Sandell, L. and A. Veis: The molecular weight of the cell-free translation product of a1(I) pro collagen mRNA. Biochem. Biophys. Res. Commun. 92: 554-562, 1980. Scott, P. G., Veis, A. and G. Mechanic: The Identity of a cyanogen bromide fragment of bovine dentin collagen containing the site of an intermolecular cross-link. Biochemistry 15: 3191-3198,1976. Stimler, N. P. and M. L. Tanzer: Isolation and characterization of a double chain intermolecular crosslinked peptide from insoluble calf bone collagen. J. Bio/. Chem. 254: 666-671, 1979. Tanzer, M. L. and G. Mechanic: Collagen reduction by sodium borohydride: Effects of reconstitution, maturation and lathyrism. Biochem. Biophys. Res. Commun. 32: 885-892, 1968. Volpin, D. and A. Veis: Cyanogen bromide peptides from insoluble skin and dentin bovine collagens. Biochemistry 12: 1452-1464, 1973. David T. Cheung, Ph. D., Bone and Connective Tissue Research Laboratory, Orthopaedic Hospital, D & T 5th Floor, 2400 S. Flower Street, Los Angeles, California 90007, USA.