Cell, Vol . 4, 4 5 -50, January 1975, Copyright©1975 by MIT
Procollagen Peptidase : Its Mode of Action on the Native Substrate Burton Goldberg, Mark B . Taubman, and Allen Radin The Department of Pathology New York University Medical Center New York, New York 10016
Summary Procollagen peptidase was recovered from the medium of human and mouse fibroblast cultures by precipitation with ammonium sulfate . The test substrate for the in vitro enzymatic reaction was radioactively-labeled, disulfide-linked procollagen prepared from the medium of human fibroblast cultures . The enzymatic digests were analyzed by electrophoresis in polyacrylamide gels containing sodium dodecyl sulfate and urea . The human and mouse enzymes reacted with the substrate to generate the same intermediates and final products . Procollagen peptidase acts as an endopeptidase which cleaves each of the three procollagen chains in turn . The final products of the reaction are collagen and a three-chain, disulfide-linked fragment derived from the nonhelical aminoterminal residues of procollagen . Introduction The procollagen molecule contains the triple helical structure of collagen, but the constituent chains of the precursor ("pro a chains") have additional nonhelical residues ("propeptides") at their aminotermini . The propeptides of the three chains together weigh about 75,000 daltons, and their amino acid composition differs from that of the helical portion of the procollagen molecule . In particular, the propeptides have much less proline and glycine and contain tryptophan and half-cystine residues, the latter forming disulfide bonds between the three pro a chains (Bornstein, 1974, for review) . Presumably collagen is generated from procollagen in vivo by the enzymatic removal of the propeptides . Bornstein, Ehrlich, and Wyke (1972), Lapiere, Lenaers, and Kohn (1971), and Kohn et al . (1974) have reported the extraction or purification of a neutral protease from animal connective tissues which converts pro a chains to collagen a chains in vitro . However, in these reports the test substrate was not the native disulfide-assembled procollagen molecule with a molecular weight of about 360,000, but rather pro a chains that had already undergone some enzymatic shortening and that were no longer disulfide-assembled . We have used a substrate containing radioactively-labeled, native, disulfide-assem bled procolla-
gen in an in vitro assay for procollagen peptidase activity . The substrate was isolated from the medium of human fibroblast cultures, and the enzymes from the medium of human and mouse fibroblast cultures . We present evidence that procollagen peptidase functions as an endopeptidase to generate collagen and a three-chain, disulfide-assembled propeptide fragment . Results Pulse-chase experiments with human skin fibroblast cultures demonstrated that conversion of procollagen to collagen proceeded very slowly, presumably because procollagen peptidase activity was rate limiting (Goldberg, Epstein, and Sherr, 1972 ; Goldberg and Sherr, 1973) . In contrast, cultures of the mouse fibroblast line 3T6 converted the precursor in the medium much more rapidly, and hence we anticipated that this medium would be a better source of the enzyme . Accordingly, enzyme preparations from the media of the two culture systems were compared for activity in an in vitro assay . The radioactively-labeled test substrate was prepared only in human cultures, as the more rapid conversion in the mouse system precluded the isolation of significant amounts of native undigested procollagen . Figure la is the gel pattern of the substrate as isolated from the medium of human cultures incubated for 24 hr with the isotopes . The radioactive peak I is the native disulfide-assembled procollagen molecule with the composition (pro al )2 • pro a2 . 3 H-tryptophan labels only the propeptides of procollagen, while 14 C-proline and glycine preferentially label the helical portion of the pro a chains (Sherr, Taubman, and Goldberg, 1973) . As some of the radioactively-labeled procollagen was converted to collagen during the labeling interval, the substrate also contained intermediates and products of this reaction . Radioactive peaks II and III will be shown to be intermediates, and peak IV a product of the enzymatic conversion of procollagen to collagen . Collagen is also represented in the test substrate, as shown by the 14 C-labeled a chains in the characteristic 2/1 ratio . As these chains lack tryptophan, they do not contain the 3 H label . Human and mouse enzyme preparations were incubated with the substrate at 37°C, and equal aliquots of the digests were removed at intervals and analyzed in SDS-polyacrylamide gels . Both enzymes generated the same intermediates and final products from the substrate (see below) . However, for identical concentrations of enzyme protein and radioactively-labeled substrate, the reaction went to completion in 8-10 hr with mouse enzyme, but was
Cell 46
0
10
50 20 30 40 mm from ORIGIN
60
70
1z
30
j 8 HOURS IT
I
(c) al N i
3 HOURS
0
c X G
a-
0 0
0 2 ro 0 0
I
10
I
i
20 30 40 50 mm from ORIGIN
I
60
70
0
10 20
30
40
50
60
70
mm from ORIGIN
Figure 1 . SDS-Polyacrylamide Gels of Test Substrate and Substrate-Enzyme Digests (a) substrate before addition of enzyme ; (b-d) enzyme (1 .3 mg/ml) added to substrate (0 .3 mg/ml) and digests sampled at intervals . Gels contain equal amounts of radioactivity . 3H cpm ( •- •) ; 14C cpm (0 0) .
incomplete after 24 hr with human enzyme . Accordingly, we report only the data for the more active mouse preparation . Figure 1 b shows the effect of enzyme on substrate after 1 hr of incubation . As compared to the original substrate, there is less native procollagen and more of the other molecular forms . In particular, radioactivity in peaks III and IV and in collagen a chains has increased 2-3 fold . Additionally, a 14C_ labeled chain with a molecular weight of about 110,000 runs just behind the al chain . Whereas the
ratio of 3 H cpm/ 14 C cpm is about 1 for the native procollagen, it is clear from this gel that this ratio progressively increases in molecules II, III, and IV, respectively. After 3 hr of enzymatic digestion (Figure 1c) collagen a chains and peak IV are the most prominent molecular species, and the amount of radioactivity in peak III now exceeds that in either peak I or peak II . A new peak of 74 C radioactivity (V) is now detected between peaks I and II . In gels of 8 hr digests (Figure 1 d) peak V is clearly resolved, as native procollagen has largely disappeared . As
Procollagen Peptidase 47
compared to the 3 hr sample, peaks II and III are less prominent, but radioactivity in collagen a chains and in peak IV has increased . Gels of digests incubated for 10-24 hr (not shown) contained only peaks IV, V, and 14 C-labeled collagen a chains . The amount of radioactivity resolved in peak V did not increase with the longer incubations . Identification of Products of the Reaction The 14 C-labeled chains generated in the reaction were designated as al and a2 collagen chains, as the reference collagen chains moved to the same positions in the gels . Since these two chain classes have the same content of proline and glycine, it is evident that a major product of the reaction is radioactive collagen with the composition (al)2 • a2 . The much smaller amount of 14 C radioactivity recovered in peak V requires further explanation . Calibration of the gels with collagen components assigned a weight of about 300,000 to this molecule, and we considered that it might contain covalently linked a chains . Addition of /3-aminopropionitrile (50 µg/ ml) at the start of the in vitro reaction did not prevent the appearance of peak V in the digests, indicating that the molecule did not require a lysyl oxidase function for its covalent assembly (Pinnell and Martin, 1968) . Limited digestion with pepsin did not change the molecule's position in the gels, but reduction with 2-mercaptoethanol caused its radioactivity to move to the position of a chains . These results indicate that peak V is type Ill collagen, a molecule composed of three identical and genetically distinct a chains linked by disulfide bonds between pepsin-resistant carboxy-terminal sequences (Miller, Epstein, and Piez, 1971 ; Chung and Miller, 1974) . As 3 H-tryptophan was a unique marker for the propeptides, gel peaks II, III, and IV must have contained at least some of these sequences . Calibration of the gels with globular proteins assigned peak IV a molecular weight of about 75,000, indicating that it contained most, if not all, of the nonhelical residues of native procollagen . When appropriate digests were reduced with 2-mercaptoethanol, peak IV no longer appeared in the gels, and all of its radioactivity was recovered in a peak with a molecular weight of about 25,000 (not shown) . Peak IV thus contained three subunits of equal size linked by disulfide bonds . Immunologic studies confirmed the above data . Substrate was digested with enzyme for 18 hr, so that peak IV and collagen were the only radioactive products . Peak IV was isolated from the digest by a combination of gel filtration and ion exchange chromatography, and tested as a competing antigen in a radioimmunoassay for human procollagen . In this assay the antibodies are specific for propep-
tides, and the standard antigen is a three-chain, disulfide-linked propeptide fragment of 75,000 daltons obtained by collagenase digestion of procollagen (Taubman, Goldberg, and Sherr, 1974) . Peak IV gave a reaction of antigenic identity in the assay, and so we conclude that it is a threechain, disulfide-linked molecule which is generated by the en bloc excision of most of the propeptide residues of native procollagen . Kinetics and Specificity of the Reaction The gels shown in Figure 1 are from an experiment in which the digest was sampled over an incubation interval of 24 hr . The radioactivities in the various gel peaks of each sample were calculated, and the data showed that 3 H radioactivity lost from native procollagen was successively recovered in gel peaks II, III, and IV as the reaction went to completion . Figure 2 shows the recoveries of radioactivity in the propeptide fragment (peak IV) and in collagen as the reaction progressed . Reaction velocities for the initial 3 hr are approximately linear and similar for the two curves, supporting the view that these two molecules are products of the same enzymatic reaction . After 10 hr, when the reaction was virtually complete, all the radioactivity of the native procollagen was recovered in these final products . Figure 3 shows that the yield of the propeptide fragment in the reaction is dependent upon the enzyme protein concentration . In these experiments the substrate concentration was that of Figure 1, and the digests were incubated for 3 hr . The observed reaction cannot be ascribed to proteases isolated with the substrate, for when the latter was incubated alone for 24 hr at 37°C, the gel pattern of Figure 1 a was still obtained . Addition of ethylenediaminetetraacetate (0 .13 M) to the reaction mixture caused an 85% decrease in the yields
PROPEPT/LLE FRAGMENT
COLLAGEN ,~,- •o--- -. io
6-
0'1 0
2
4
6
8
r 10
24
HOURS Figure 2 . Progress Curves for the Enzymatic Generation of the Propeptide Fragment and Collagen Data from experiment of Figure 1 .
Cell 48
of collagen and the propeptide fragment, indicating that divalent cations were required for enzymatic activity . The reaction was not inhibited by the trypsin inhibitor, N-a-p-tosyl-L-lysine-chloromethyl ketone HCI (10 -4 M), nor by soybean trypsin inhibitor (0 .5 mg/ml) . Nor does the reaction resemble the effect of a pepsin-like protease ; limited digestion of the substrate with pepsin (not shown) releases collagen a chains, but gel peak IV is not generated . Instead, the propeptide residues are incompletely recovered in gels in several peptides with molecular weights of less than 75,000 . Discussion We believe our experiments demonstrate the intermediates and products of the specific enzymatic conversion of procollagen to collagen . We cite the following to support this view : (a) the reaction requires the addition of enzyme protein, and the enzymatic activity is then proportional to the protein concentration over a narrow range ; (b) all the intermediates and products generated in the digests are present in the medium of intact fibroblast cultures (Goldberg et al ., 1972 ; Goldberg and Sherr, 1973 ; Sherr et al ., 1973) ; (c) nonspecific proteases do not generate the same products when reacted with the test substrate ; and (d) the kinetics of the reaction and its limited susceptibility to inhibition indicate that a specific enzyme activity is being measured .
8-
We emphasize that the enzyme was recovered from the culture medium . Attempts to detect activity in homogenates or extracts of the cells have been unsuccessful to date . This suggests that the enzyme may be in precursor form or inhibited prior to secretion from the fibroblast . Figure 4 represents a model for the action of procollagen peptidase in accord with the experimental data . The enzyme is shown to act as an endopeptidase, cleaving each of the pro a chains in turn near their helical regions . The procollagen molecule remains assembled throughout the enzymatic reaction by virtue of covalent disulfide bonds between the nonhelical aminoterminal propeptides, and noncovalent interactions between the helical regions of the chains . However, noncovalently assembled chains generated in the reaction are separated in SDS-urea gels according to their molecular weights . We thus propose that enzymatic cleavage of only one pro a chain separates peak II, an intermediate with an estimated molecular weight of 265,000, from the noncovalently associated helical chain . When two pro a chains are cleaved, gel peak III (molecular weight = 170,000 daltons) is resolved, and when all the chains are clipped, triple helical collagen and the covalently-assembled globular propeptide fragment (IV) are generated . The model clarifies the different ratios of 3 H/ 14 C radioactivity in the various gel peaks and the observation that the 3 H radioactivity in native procollagen is sequentially recovered in gel peaks II to IV as the reaction proceeds . Moreover, the model is consistent with the gel patterns given by digests before and after reduction with 2-mercaptoethanol . The molecular weights given in Figure 4 are compatible with calibrations of the gels, but should be considered as approximate values . The model is drawn for type I procollagen, but our data indicate that it applies to the conversion
I
IZ
(360,000)
Ill
(265,000)
(70,000
II 1'
(75,000)
COLLAGEN (285,000) 112
112
112
112
Figure 4 . Model for Endopeptidase Action of Procollagen Peptidase
0
4 2 PROTEIN CONC . (Mg/Ml)
6
Figure 3 . Amount of Propeptide Generated as a Function of Enzyme Protein Concentration 0 .3 mg/ml of substrate was incubated for 3 hr with increasing concentrations of enzyme protein . Samples were applied to gels, and the radioactivity in gel peak IV plotted .
The native procollagen molecule (I) has the composition (pro a1)2 - pro a2 ; its pro a chains are covalently linked by disulfide •) between the nonhelical aminoterminal propepbonds ( • tides . Roman numerals correspond to the radioactive gel peaks of Figure 1 and apply only to disulfide-linked chains . Molecular weights derived from calibration of gels are given in parentheses . Successive enzymatic cleavage of the three chains of procollagen generates disulfide-assembled molecules (solid lines) and noncovalently-associated chains (dotted lines) which are separated in SDS-urea gels .
Procollagen Peptidase 49
of type III procollagen as well . At the end of the in vitro reaction 14% of the 14 C-labeled chains were recovered in a covalently-assembled molecule (gel peak V) consistent with type III Collagen . We therefore presume that a small amount of radioactively-labeled type III procollagen was contained in peak I of the test substrate and was converted by the enzyme . However, the reaction intermediates and the final helical molecule generated from type III procollagen would not dissociate in gels, as the chains would remain linked by disulfide bonds at their carboxytermini . In the model of Figure 4 all the propeptide residues of a given pro a chain are shown to be excised in one enzymatic step, that is, native collagen a chains are generated directly . It is possible that the initial excision might not remove all the propeptide residues, and a few may remain with the helical chain . For example, in our 1 hr digests (Figure 1 b) we resolved 14 C-labeled chains slightly heavier than the native al chains, and these were still detected as a shoulder on the al peak in the 3 hr digests (Figure 1 c) . After 8 hr, only native al and a2 chains were identified in the gels . These data suggest that the propeptides of each pro a chain may be excised in two steps, and perhaps two enzyme activities are required . This view receives some support from the studies of dermatosparaxis, a recessively inherited disease of cattle in which the conversion of procollagen to collagen is defective (Lapiere et al ., 1971) . The procollagen accumulating in the tissues of dermatosparactic animals differs from the native procollagen in our test substrate by having about 60% fewer propeptide residues and by lacking interchain disulfide bonds . The dermatosparactic pro a chains thus resemble the heavier helical chains noted in our digests after the short incubations (Figures 1 b and 1c) . This suggests that in dermatosparaxis primary excision of most of the propeptides from native procollagen occurs normally, but the enzymatic removal of the remaining residues is the step that is blocked . It is of interest that Kohn et al . (1974) have presented evidence that procollagen peptidase purified from normal calf tendon functions as an endopeptidase in excising the abbreviated propeptides from dermatosparactic procollagen . The collagen molecules generated from the precursor aggregate to form native collagen fibers in vivo, but it is not known if the propeptide fragment plays any role in the organization of the intercellular matrix . Human serum contains a molecule antigenically identical to the propeptide fragment (Taubman et al ., 1974), suggesting that the fragment is not subject to rapid proteolytic turnover in vivo . Detailed structural analyses of the fragment generated in our in vitro system should help to clarify the role
of the propeptide residues in the assembly, secretion, and extracellular metabolism of procollagen . Experimental Procedures Cells, Culture Conditions, Preparation of Substrate Experiments were performed with confluent cultures of human skin fibroblasts (CRL 1121, American Type Culture Collection) or the mouse fibroblast line 3T6 (Todaro and Green, 1963) . Culture conditions were as previously described (Goldberg et al ., 1972) . To prepare the radioactively-labeled substrate, 3 x 108 human fibroblasts were placed on serum-free, ascorbate-supplemented medium containing 100 µCi of 3 H-tryptophan and 20 µCi each of 14 Cproline and glycine (all New England Nuclear) . After 24 hr the medium was collected, centrifuged at 5000 x g for 15 min to remove any cellular debris, dialyzed successively against 0 .5 M acetic acid and water at 4°C, and then lyophilized . The freeze-dried material was stored at -20°C and dissolved in standard buffer (50 mM Tris-HCI, pH 7 .5 ; 150 mM NaCl ; 5 mM CaCl 2) for use in the in vitro assay . Enzyme Preparation Washed, confluent cell layers were incubated with serum-free medium for 24 hr at 37°C ; thereafter all manipulations were at 4°C . The medium was centrifuged to remove any cellular debris and then dialyzed against the standard buffer for 48 hr . Ammonium sulfate was added to 39% (w/v) ; the mixture was stirred for 18 hr, and then centrifuged for 15 min at 15,000 x g . The pellet thus obtained from 1 I of culture medium was suspended in 10 ml of standard buffer, dialyzed against the buffer for 18 hr, and insoluble material removed by centrifugation . The supernatant was used as a source of enzyme . In Vitro Assay Substrate and enzyme solutions were mixed at 4°C and incubated with shaking at 37°C for appropriate intervals . Final reaction volumes generally did not exceed 1 .0 ml . At intervals, samples of equal volume were withdrawn from the digests, chilled, made 2% in SDS and 1 M in urea, and immersed in a 60°C water bath for 1 hr . Samples were then applied directly to polyacrylamide gels . Electrophoresis in SDS-Urea Polyacrylamide Gels Electrophoresis in 5% gels and measurement of radioactivities in gel slices were performed as previously described (Goldberg et al ., 1972) . The gel patterns as plotted contain equal amounts of 3H and 14 C radioactivity, respectively . For reduction, samples were heat denatured at 60°C in the presence of 1 % 2-mercaptoethanol . Calibration of the gels for assignments of molecular weights was performed with purified collagen components and noncollagen proteins as previously described (Goldberg et al ., 1972 ; Sherr et al ., 1973) . Gel Filtration and Ion Exchange Chromatography A digest was applied to a Sephadex G-100 column (5 x 100 cm) and eluted with the standard buffer . Radioactive fractions in the included volume were pooled, dialyzed against H20, lyophilized, and dissolved in 0 .025 M Tris-HCI buffer (pH 8 .5) . The sample was applied to a DEAE-Sephadex column (A-50, 0 .9 x 50 cm) and eluted in the above buffer with a linear gradient of NaCl from 0-0 .7 M over a total volume of 250 ml . The 3H-tryptophan-labeled peak was pooled, dialyzed against H 2 0, and lyophilized. This is a modification of a published procedure for the isolation of the propeptides of human procollagen (Sherr et al ., 1973) . Materials N-a-p-tosyl-L-lysine-chloromethyl ketone HCI was obtained from the Sigma Chemical Company ; soybean trypsin inhibitor and pepsin from Worthington Biochemical Corp . ; and (3-aminopropionitrile HCI from Calbiochem .
Cell 50
Acknowledgments The expert assistance of Sheila Heitner and technical aid of Josa Williamson are gratefully acknowledged . This work was supported by an NIH grant . M . T. and A . R . are recipients of NIH Medical Scientist Fellowships . Received October 24, 1974 References Bornstein, P . (1974) . Ann . Rev. Biochem . 143, 567 . Bornstein, P ., Ehrlich, H . P ., and Wyke, A. W . (1972) . Science 175, 544 . Chung, E ., and Miller, E . J . (1974). Science 183, 1200 . Goldberg, B ., and Sherr, C . J . (1973) . Proc . Nat . Acad . Sci . USA 70, 361 . Goldberg, B ., Epstein, E . H ., Jr ., and Sherr, C . J . (1972) . Proc . Nat . Acad. Sci. USA 69, 3655 . Kohn, L . D ., Isersky, C ., Zupnik, J ., Lenaers, A ., Lee, G ., and Lapiere, C. M . (1974) . Proc . Nat . Acad . Sci . USA 71, 40 . Lapiere, C . M ., Lenaers, A ., and Kohn, L . D . (1971) . Proc . Nat . Acad . Sci . USA 68, 3054 . Miller, E . J ., Epstein, E . H ., Jr ., and Piez, K . A . (1971) . Biochem . Biophys . Res . Commun . 42, 1024 . Pinnell, S . R ., and Martin, G . R ., (1968) . Proc . Nat . Acad . Sci . USA 61, 708 . Sherr, C. J ., and Goldberg, B . (1973) . Science 180, 1190 . Sherr, C . J ., Taubman, M . B ., and Goldberg, B . (1973) . J . Biol . Chem . 248, 7033 . Taubman, M . B ., Goldberg, B ., and Sherr, C . J . (1974) . Science, in press . Todaro, G . J., and Green, H . (1963) . J . Cell Biol . 17, 299 .
Procollagen Peptidase : Its Mode of Action on the Native Substrate Burton Goldberg, Mark B . Taubman, and Allen Radin The Department of Pathology New York University Medical Center New York, New York 10016
Summary Procollagen peptidase was recovered from the medium of human and mouse fibroblast cultures by precipitation with ammonium sulfate . The test substrate for the in vitro enzymatic reaction was radioactively-labeled, disulfide-linked procollagen prepared from the medium of human fibroblast cultures . The enzymatic digests were analyzed by electrophoresis in polyacrylamide gels containing sodium dodecyl sulfate and urea . The human and mouse enzymes reacted with the substrate to generate the same intermediates and final products . Procollagen peptidase acts as an endopeptidase which cleaves each of the three procollagen chains in turn . The final products of the reaction are collagen and a three-chain, disulfide-linked fragment derived from the nonhelical aminoterminal residues of procollagen . Introduction The procollagen molecule contains the triple helical structure of collagen, but the constituent chains of the precursor ("pro a chains") have additional nonhelical residues ("propeptides") at their aminotermini . The propeptides of the three chains together weigh about 75,000 daltons, and their amino acid composition differs from that of the helical portion of the procollagen molecule . In particular, the propeptides have much less proline and glycine and contain tryptophan and half-cystine residues, the latter forming disulfide bonds between the three pro a chains (Bornstein, 1974, for review) . Presumably collagen is generated from procollagen in vivo by the enzymatic removal of the propeptides . Bornstein, Ehrlich, and Wyke (1972), Lapiere, Lenaers, and Kohn (1971), and Kohn et al . (1974) have reported the extraction or purification of a neutral protease from animal connective tissues which converts pro a chains to collagen a chains in vitro . However, in these reports the test substrate was not the native disulfide-assembled procollagen molecule with a molecular weight of about 360,000, but rather pro a chains that had already undergone some enzymatic shortening and that were no longer disulfide-assembled . We have used a substrate containing radioactively-labeled, native, disulfide-assem bled procolla-
gen in an in vitro assay for procollagen peptidase activity . The substrate was isolated from the medium of human fibroblast cultures, and the enzymes from the medium of human and mouse fibroblast cultures . We present evidence that procollagen peptidase functions as an endopeptidase to generate collagen and a three-chain, disulfide-assembled propeptide fragment . Results Pulse-chase experiments with human skin fibroblast cultures demonstrated that conversion of procollagen to collagen proceeded very slowly, presumably because procollagen peptidase activity was rate limiting (Goldberg, Epstein, and Sherr, 1972 ; Goldberg and Sherr, 1973) . In contrast, cultures of the mouse fibroblast line 3T6 converted the precursor in the medium much more rapidly, and hence we anticipated that this medium would be a better source of the enzyme . Accordingly, enzyme preparations from the media of the two culture systems were compared for activity in an in vitro assay . The radioactively-labeled test substrate was prepared only in human cultures, as the more rapid conversion in the mouse system precluded the isolation of significant amounts of native undigested procollagen . Figure la is the gel pattern of the substrate as isolated from the medium of human cultures incubated for 24 hr with the isotopes . The radioactive peak I is the native disulfide-assembled procollagen molecule with the composition (pro al )2 • pro a2 . 3 H-tryptophan labels only the propeptides of procollagen, while 14 C-proline and glycine preferentially label the helical portion of the pro a chains (Sherr, Taubman, and Goldberg, 1973) . As some of the radioactively-labeled procollagen was converted to collagen during the labeling interval, the substrate also contained intermediates and products of this reaction . Radioactive peaks II and III will be shown to be intermediates, and peak IV a product of the enzymatic conversion of procollagen to collagen . Collagen is also represented in the test substrate, as shown by the 14 C-labeled a chains in the characteristic 2/1 ratio . As these chains lack tryptophan, they do not contain the 3 H label . Human and mouse enzyme preparations were incubated with the substrate at 37°C, and equal aliquots of the digests were removed at intervals and analyzed in SDS-polyacrylamide gels . Both enzymes generated the same intermediates and final products from the substrate (see below) . However, for identical concentrations of enzyme protein and radioactively-labeled substrate, the reaction went to completion in 8-10 hr with mouse enzyme, but was
Cell 46
0
10
50 20 30 40 mm from ORIGIN
60
70
1z
30
j 8 HOURS IT
I
(c) al N i
3 HOURS
0
c X G
a-
0 0
0 2 ro 0 0
I
10
I
i
20 30 40 50 mm from ORIGIN
I
60
70
0
10 20
30
40
50
60
70
mm from ORIGIN
Figure 1 . SDS-Polyacrylamide Gels of Test Substrate and Substrate-Enzyme Digests (a) substrate before addition of enzyme ; (b-d) enzyme (1 .3 mg/ml) added to substrate (0 .3 mg/ml) and digests sampled at intervals . Gels contain equal amounts of radioactivity . 3H cpm ( •- •) ; 14C cpm (0 0) .
incomplete after 24 hr with human enzyme . Accordingly, we report only the data for the more active mouse preparation . Figure 1 b shows the effect of enzyme on substrate after 1 hr of incubation . As compared to the original substrate, there is less native procollagen and more of the other molecular forms . In particular, radioactivity in peaks III and IV and in collagen a chains has increased 2-3 fold . Additionally, a 14C_ labeled chain with a molecular weight of about 110,000 runs just behind the al chain . Whereas the
ratio of 3 H cpm/ 14 C cpm is about 1 for the native procollagen, it is clear from this gel that this ratio progressively increases in molecules II, III, and IV, respectively. After 3 hr of enzymatic digestion (Figure 1c) collagen a chains and peak IV are the most prominent molecular species, and the amount of radioactivity in peak III now exceeds that in either peak I or peak II . A new peak of 74 C radioactivity (V) is now detected between peaks I and II . In gels of 8 hr digests (Figure 1 d) peak V is clearly resolved, as native procollagen has largely disappeared . As
Procollagen Peptidase 47
compared to the 3 hr sample, peaks II and III are less prominent, but radioactivity in collagen a chains and in peak IV has increased . Gels of digests incubated for 10-24 hr (not shown) contained only peaks IV, V, and 14 C-labeled collagen a chains . The amount of radioactivity resolved in peak V did not increase with the longer incubations . Identification of Products of the Reaction The 14 C-labeled chains generated in the reaction were designated as al and a2 collagen chains, as the reference collagen chains moved to the same positions in the gels . Since these two chain classes have the same content of proline and glycine, it is evident that a major product of the reaction is radioactive collagen with the composition (al)2 • a2 . The much smaller amount of 14 C radioactivity recovered in peak V requires further explanation . Calibration of the gels with collagen components assigned a weight of about 300,000 to this molecule, and we considered that it might contain covalently linked a chains . Addition of /3-aminopropionitrile (50 µg/ ml) at the start of the in vitro reaction did not prevent the appearance of peak V in the digests, indicating that the molecule did not require a lysyl oxidase function for its covalent assembly (Pinnell and Martin, 1968) . Limited digestion with pepsin did not change the molecule's position in the gels, but reduction with 2-mercaptoethanol caused its radioactivity to move to the position of a chains . These results indicate that peak V is type Ill collagen, a molecule composed of three identical and genetically distinct a chains linked by disulfide bonds between pepsin-resistant carboxy-terminal sequences (Miller, Epstein, and Piez, 1971 ; Chung and Miller, 1974) . As 3 H-tryptophan was a unique marker for the propeptides, gel peaks II, III, and IV must have contained at least some of these sequences . Calibration of the gels with globular proteins assigned peak IV a molecular weight of about 75,000, indicating that it contained most, if not all, of the nonhelical residues of native procollagen . When appropriate digests were reduced with 2-mercaptoethanol, peak IV no longer appeared in the gels, and all of its radioactivity was recovered in a peak with a molecular weight of about 25,000 (not shown) . Peak IV thus contained three subunits of equal size linked by disulfide bonds . Immunologic studies confirmed the above data . Substrate was digested with enzyme for 18 hr, so that peak IV and collagen were the only radioactive products . Peak IV was isolated from the digest by a combination of gel filtration and ion exchange chromatography, and tested as a competing antigen in a radioimmunoassay for human procollagen . In this assay the antibodies are specific for propep-
tides, and the standard antigen is a three-chain, disulfide-linked propeptide fragment of 75,000 daltons obtained by collagenase digestion of procollagen (Taubman, Goldberg, and Sherr, 1974) . Peak IV gave a reaction of antigenic identity in the assay, and so we conclude that it is a threechain, disulfide-linked molecule which is generated by the en bloc excision of most of the propeptide residues of native procollagen . Kinetics and Specificity of the Reaction The gels shown in Figure 1 are from an experiment in which the digest was sampled over an incubation interval of 24 hr . The radioactivities in the various gel peaks of each sample were calculated, and the data showed that 3 H radioactivity lost from native procollagen was successively recovered in gel peaks II, III, and IV as the reaction went to completion . Figure 2 shows the recoveries of radioactivity in the propeptide fragment (peak IV) and in collagen as the reaction progressed . Reaction velocities for the initial 3 hr are approximately linear and similar for the two curves, supporting the view that these two molecules are products of the same enzymatic reaction . After 10 hr, when the reaction was virtually complete, all the radioactivity of the native procollagen was recovered in these final products . Figure 3 shows that the yield of the propeptide fragment in the reaction is dependent upon the enzyme protein concentration . In these experiments the substrate concentration was that of Figure 1, and the digests were incubated for 3 hr . The observed reaction cannot be ascribed to proteases isolated with the substrate, for when the latter was incubated alone for 24 hr at 37°C, the gel pattern of Figure 1 a was still obtained . Addition of ethylenediaminetetraacetate (0 .13 M) to the reaction mixture caused an 85% decrease in the yields
PROPEPT/LLE FRAGMENT
COLLAGEN ,~,- •o--- -. io
6-
0'1 0
2
4
6
8
r 10
24
HOURS Figure 2 . Progress Curves for the Enzymatic Generation of the Propeptide Fragment and Collagen Data from experiment of Figure 1 .
Cell 48
of collagen and the propeptide fragment, indicating that divalent cations were required for enzymatic activity . The reaction was not inhibited by the trypsin inhibitor, N-a-p-tosyl-L-lysine-chloromethyl ketone HCI (10 -4 M), nor by soybean trypsin inhibitor (0 .5 mg/ml) . Nor does the reaction resemble the effect of a pepsin-like protease ; limited digestion of the substrate with pepsin (not shown) releases collagen a chains, but gel peak IV is not generated . Instead, the propeptide residues are incompletely recovered in gels in several peptides with molecular weights of less than 75,000 . Discussion We believe our experiments demonstrate the intermediates and products of the specific enzymatic conversion of procollagen to collagen . We cite the following to support this view : (a) the reaction requires the addition of enzyme protein, and the enzymatic activity is then proportional to the protein concentration over a narrow range ; (b) all the intermediates and products generated in the digests are present in the medium of intact fibroblast cultures (Goldberg et al ., 1972 ; Goldberg and Sherr, 1973 ; Sherr et al ., 1973) ; (c) nonspecific proteases do not generate the same products when reacted with the test substrate ; and (d) the kinetics of the reaction and its limited susceptibility to inhibition indicate that a specific enzyme activity is being measured .
8-
We emphasize that the enzyme was recovered from the culture medium . Attempts to detect activity in homogenates or extracts of the cells have been unsuccessful to date . This suggests that the enzyme may be in precursor form or inhibited prior to secretion from the fibroblast . Figure 4 represents a model for the action of procollagen peptidase in accord with the experimental data . The enzyme is shown to act as an endopeptidase, cleaving each of the pro a chains in turn near their helical regions . The procollagen molecule remains assembled throughout the enzymatic reaction by virtue of covalent disulfide bonds between the nonhelical aminoterminal propeptides, and noncovalent interactions between the helical regions of the chains . However, noncovalently assembled chains generated in the reaction are separated in SDS-urea gels according to their molecular weights . We thus propose that enzymatic cleavage of only one pro a chain separates peak II, an intermediate with an estimated molecular weight of 265,000, from the noncovalently associated helical chain . When two pro a chains are cleaved, gel peak III (molecular weight = 170,000 daltons) is resolved, and when all the chains are clipped, triple helical collagen and the covalently-assembled globular propeptide fragment (IV) are generated . The model clarifies the different ratios of 3 H/ 14 C radioactivity in the various gel peaks and the observation that the 3 H radioactivity in native procollagen is sequentially recovered in gel peaks II to IV as the reaction proceeds . Moreover, the model is consistent with the gel patterns given by digests before and after reduction with 2-mercaptoethanol . The molecular weights given in Figure 4 are compatible with calibrations of the gels, but should be considered as approximate values . The model is drawn for type I procollagen, but our data indicate that it applies to the conversion
I
IZ
(360,000)
Ill
(265,000)
(70,000
II 1'
(75,000)
COLLAGEN (285,000) 112
112
112
112
Figure 4 . Model for Endopeptidase Action of Procollagen Peptidase
0
4 2 PROTEIN CONC . (Mg/Ml)
6
Figure 3 . Amount of Propeptide Generated as a Function of Enzyme Protein Concentration 0 .3 mg/ml of substrate was incubated for 3 hr with increasing concentrations of enzyme protein . Samples were applied to gels, and the radioactivity in gel peak IV plotted .
The native procollagen molecule (I) has the composition (pro a1)2 - pro a2 ; its pro a chains are covalently linked by disulfide •) between the nonhelical aminoterminal propepbonds ( • tides . Roman numerals correspond to the radioactive gel peaks of Figure 1 and apply only to disulfide-linked chains . Molecular weights derived from calibration of gels are given in parentheses . Successive enzymatic cleavage of the three chains of procollagen generates disulfide-assembled molecules (solid lines) and noncovalently-associated chains (dotted lines) which are separated in SDS-urea gels .
Procollagen Peptidase 49
of type III procollagen as well . At the end of the in vitro reaction 14% of the 14 C-labeled chains were recovered in a covalently-assembled molecule (gel peak V) consistent with type III Collagen . We therefore presume that a small amount of radioactively-labeled type III procollagen was contained in peak I of the test substrate and was converted by the enzyme . However, the reaction intermediates and the final helical molecule generated from type III procollagen would not dissociate in gels, as the chains would remain linked by disulfide bonds at their carboxytermini . In the model of Figure 4 all the propeptide residues of a given pro a chain are shown to be excised in one enzymatic step, that is, native collagen a chains are generated directly . It is possible that the initial excision might not remove all the propeptide residues, and a few may remain with the helical chain . For example, in our 1 hr digests (Figure 1 b) we resolved 14 C-labeled chains slightly heavier than the native al chains, and these were still detected as a shoulder on the al peak in the 3 hr digests (Figure 1 c) . After 8 hr, only native al and a2 chains were identified in the gels . These data suggest that the propeptides of each pro a chain may be excised in two steps, and perhaps two enzyme activities are required . This view receives some support from the studies of dermatosparaxis, a recessively inherited disease of cattle in which the conversion of procollagen to collagen is defective (Lapiere et al ., 1971) . The procollagen accumulating in the tissues of dermatosparactic animals differs from the native procollagen in our test substrate by having about 60% fewer propeptide residues and by lacking interchain disulfide bonds . The dermatosparactic pro a chains thus resemble the heavier helical chains noted in our digests after the short incubations (Figures 1 b and 1c) . This suggests that in dermatosparaxis primary excision of most of the propeptides from native procollagen occurs normally, but the enzymatic removal of the remaining residues is the step that is blocked . It is of interest that Kohn et al . (1974) have presented evidence that procollagen peptidase purified from normal calf tendon functions as an endopeptidase in excising the abbreviated propeptides from dermatosparactic procollagen . The collagen molecules generated from the precursor aggregate to form native collagen fibers in vivo, but it is not known if the propeptide fragment plays any role in the organization of the intercellular matrix . Human serum contains a molecule antigenically identical to the propeptide fragment (Taubman et al ., 1974), suggesting that the fragment is not subject to rapid proteolytic turnover in vivo . Detailed structural analyses of the fragment generated in our in vitro system should help to clarify the role
of the propeptide residues in the assembly, secretion, and extracellular metabolism of procollagen . Experimental Procedures Cells, Culture Conditions, Preparation of Substrate Experiments were performed with confluent cultures of human skin fibroblasts (CRL 1121, American Type Culture Collection) or the mouse fibroblast line 3T6 (Todaro and Green, 1963) . Culture conditions were as previously described (Goldberg et al ., 1972) . To prepare the radioactively-labeled substrate, 3 x 108 human fibroblasts were placed on serum-free, ascorbate-supplemented medium containing 100 µCi of 3 H-tryptophan and 20 µCi each of 14 Cproline and glycine (all New England Nuclear) . After 24 hr the medium was collected, centrifuged at 5000 x g for 15 min to remove any cellular debris, dialyzed successively against 0 .5 M acetic acid and water at 4°C, and then lyophilized . The freeze-dried material was stored at -20°C and dissolved in standard buffer (50 mM Tris-HCI, pH 7 .5 ; 150 mM NaCl ; 5 mM CaCl 2) for use in the in vitro assay . Enzyme Preparation Washed, confluent cell layers were incubated with serum-free medium for 24 hr at 37°C ; thereafter all manipulations were at 4°C . The medium was centrifuged to remove any cellular debris and then dialyzed against the standard buffer for 48 hr . Ammonium sulfate was added to 39% (w/v) ; the mixture was stirred for 18 hr, and then centrifuged for 15 min at 15,000 x g . The pellet thus obtained from 1 I of culture medium was suspended in 10 ml of standard buffer, dialyzed against the buffer for 18 hr, and insoluble material removed by centrifugation . The supernatant was used as a source of enzyme . In Vitro Assay Substrate and enzyme solutions were mixed at 4°C and incubated with shaking at 37°C for appropriate intervals . Final reaction volumes generally did not exceed 1 .0 ml . At intervals, samples of equal volume were withdrawn from the digests, chilled, made 2% in SDS and 1 M in urea, and immersed in a 60°C water bath for 1 hr . Samples were then applied directly to polyacrylamide gels . Electrophoresis in SDS-Urea Polyacrylamide Gels Electrophoresis in 5% gels and measurement of radioactivities in gel slices were performed as previously described (Goldberg et al ., 1972) . The gel patterns as plotted contain equal amounts of 3H and 14 C radioactivity, respectively . For reduction, samples were heat denatured at 60°C in the presence of 1 % 2-mercaptoethanol . Calibration of the gels for assignments of molecular weights was performed with purified collagen components and noncollagen proteins as previously described (Goldberg et al ., 1972 ; Sherr et al ., 1973) . Gel Filtration and Ion Exchange Chromatography A digest was applied to a Sephadex G-100 column (5 x 100 cm) and eluted with the standard buffer . Radioactive fractions in the included volume were pooled, dialyzed against H20, lyophilized, and dissolved in 0 .025 M Tris-HCI buffer (pH 8 .5) . The sample was applied to a DEAE-Sephadex column (A-50, 0 .9 x 50 cm) and eluted in the above buffer with a linear gradient of NaCl from 0-0 .7 M over a total volume of 250 ml . The 3H-tryptophan-labeled peak was pooled, dialyzed against H 2 0, and lyophilized. This is a modification of a published procedure for the isolation of the propeptides of human procollagen (Sherr et al ., 1973) . Materials N-a-p-tosyl-L-lysine-chloromethyl ketone HCI was obtained from the Sigma Chemical Company ; soybean trypsin inhibitor and pepsin from Worthington Biochemical Corp . ; and (3-aminopropionitrile HCI from Calbiochem .
Cell 50
Acknowledgments The expert assistance of Sheila Heitner and technical aid of Josa Williamson are gratefully acknowledged . This work was supported by an NIH grant . M . T. and A . R . are recipients of NIH Medical Scientist Fellowships . Received October 24, 1974 References Bornstein, P . (1974) . Ann . Rev. Biochem . 143, 567 . Bornstein, P ., Ehrlich, H . P ., and Wyke, A. W . (1972) . Science 175, 544 . Chung, E ., and Miller, E . J . (1974). Science 183, 1200 . Goldberg, B ., and Sherr, C . J . (1973) . Proc . Nat . Acad . Sci . USA 70, 361 . Goldberg, B ., Epstein, E . H ., Jr ., and Sherr, C . J . (1972) . Proc . Nat . Acad. Sci. USA 69, 3655 . Kohn, L . D ., Isersky, C ., Zupnik, J ., Lenaers, A ., Lee, G ., and Lapiere, C. M . (1974) . Proc . Nat . Acad . Sci . USA 71, 40 . Lapiere, C . M ., Lenaers, A ., and Kohn, L . D . (1971) . Proc . Nat . Acad . Sci . USA 68, 3054 . Miller, E . J ., Epstein, E . H ., Jr ., and Piez, K . A . (1971) . Biochem . Biophys . Res . Commun . 42, 1024 . Pinnell, S . R ., and Martin, G . R ., (1968) . Proc . Nat . Acad . Sci . USA 61, 708 . Sherr, C. J ., and Goldberg, B . (1973) . Science 180, 1190 . Sherr, C . J ., Taubman, M . B ., and Goldberg, B . (1973) . J . Biol . Chem . 248, 7033 . Taubman, M . B ., Goldberg, B ., and Sherr, C . J . (1974) . Science, in press . Todaro, G . J., and Green, H . (1963) . J . Cell Biol . 17, 299 .
