Isolation of a Disulfide-stabilized, Three-Chain Polypeptide Fragment Unique to the Precursor of Human Collagen*

SUMMARY Cultured human diploid fibroblasts were labeled with [3H]-tryptophan, and the three-chain precursor of collagen in the medium was identified by immunoprecipitation and gel electrophoresis. [3H]Tryptophan was incorporated into the nonhelical, NH*-terminal extensions of the pro-al and pro-& chains of the precursor. Collagenase digestion of the helical segment of the precursor molecule left intact a disulfide-assembled protein fragment which contained all of the incorporated radioactivity. Reduction and alkylation of the fragment confirmed that it consisted of three chains of similar molecular weight. Large scale recovery of the labeled fragment was performed using gel filtration, collagenase digestion, and ion exchange chromatography. showed that the isolated fragment was still antigenically reactive. The molecular weights of the fragment and its constituent chains were determined from their electrophoretic mobilities in sodium dodecyl sulfate-polyacrylamide gels, from their elution positions in Sephadex by analytical

The molecular weights of the fragment and its constituent chains were determined from their electrophoretic mobilities in sodium dodecyl sulfatepolyacrylamide gels, from their elution positions in Sephadex gels, and by analytical ultracentrifugation. The latter two methods gave a molecular weight of 75,000 to 80,000 for the intact fragment and 25,000 for the component chains released by reduction.
The electrophoretic data gave somewhat higher molecular weights, which were judged to be anomalous.
The fragment had an amino acid composition quite different from that of collagen: it was poorer in glycine and proline, it contained half-cystine residues, and ~237~ of its residues were derived from aspartate and glutamate.
Collagen cx chains are synthesized as precursors (pro-al and pro-a2) with nonhelical, KHH2-terminal peptide extensions (propeptides) (1-9). Using cultured human diploid fibroblasts, we have presented biochemical (10, 11) and immunological (12) evidence that two pro-al chains and one pro-o!2 chain assemble intracellularly to form a molecule with a molecular weight of * approximately 360,000, which we have called pro-tropocollagen. The molecule is stabilized by disulfide bonds between the propeptides of all three chains and by noncovalent interactions between the helical, cx chain segments.
After secretion from the cell, native tropocollagen is generated by sequential, enzymatic excision of propeptide sequences from the precursor. Tryptophan is not present in tropocollagen but has been identified in pro-al chains (13,14). We report the use of [3H]tryptophan as a marker for pro-al and pro-a2 chains and describe a method for the isolation and purification of the disulfide-assembled, three-chain propeptide fragment. The method employs collagenase to digest the helical portion of pro-tropocollagen (14)) and the undigested propeptide fragment is then isolated by gel filtration and ion exchange chromatography. ;\lolecular weights are calculated for the fragment and the component chains released by reduction, and the amino acid composition for the intact fragment is given. Tris-IICl buffer, pH 7.5, containing 0.1 c/c sodium dodecyl sulfate.

MATERIALS
Columns were calibrated 17.ith blue dextran 2000, bovine serum albumin, (mol wt SS,OOO), horse heart cytochrome c (mol wt 12,000), and phenol red. Lyophilized medium or pooled fractions from columns were dissolved in the respective eluting buffers and applied to columns in volumes not exceeding 8 ml.
Ion Exchange ChromatograpkJ/-A DEAE-Sephadex column (A-50, 1.5 X 30 cm) was equilibrated with 0.025 M Tris-HCl buffer, pH 8.5. Samples were applied and eluted in the same buffer with a linear gradient of KaCl from 0 to 0.5 M over a total volume of 270 ml.
Acrylamide Gel Electrophoresis-Elcctrophorcsis on 5Gj0 acrylamide gels containing 0.176 sodium dodecyl sulfate and 0.5 M urea was performed as previously described (10). Authentic '*C-labeled a chains were added as internal markers.
In some experiments, parallel gels were calibrated lvith cytochrome c and '*%labelcd immunoglobulin A (mol wt 157,000) and immunoglobulin A heavy (mol wt 56,000) and light (mol wt 22,500) chains (a gilt of Dr. Michael Lamm, Sew York University School of l\ledicine).
Elcctrophoresis was also performed on 7c/1 acrylamidc gels polymerized in 0.05 RI 'I'ris-HCl buffer, p1-I 9.4, using the same buffer in the electrode chambers.
The latter gels were run for 2.5 hours at 3 ma per tube.
Some gels were fixed in 207; sulfosalicylic acid, stained in tlic same solution with 0.25(;, ('oomassie blue, ant1 destained with 7%;, acetic acid. Stailletl gels were sc-alnlcd at 600 nm using a Gilford densitonzeter and recorder, and areas under peaks were computed by planimetry. Stained and unstained gels were cut into l-mm slices and radioactivity was determined as previously described (10).
I,rtltlunoc/zel,lical I'rocedures-Preparatioii of rabbit antiserum against the serum-free culture medium has been described (12). This antiserun? forms three precipitin lines in double immunodiffusion plates when tested against the immunizing antigen. The antiserum contains antibodies specific for the SIIz-terminal propc~ptide determinants of pro-tropocollagen and does not recognize native tropocollagen. Titration of the ant,iserum against 31-i-labeled culture medium or against solutions of radiolabelcd propeptide were performed as described previously (12). Prior t,o gel elcctrophoresis, immune precipitates were dissolved in 0.01 M phosphate buffer, p1-I 7.0, containing 1% sodium clodecyl sulfate and 0.5 nr urea.
In some experiments, immune precipitates of radiolabeled medium wcrc suspended in 0.5 ml of 0.05 xz Tris-HCl buffer, pT1 7.5, containing 0.15 11 KaCl and 0.005 M CaCIP, and were digested with collagenasc as above. Digests were made 17; in sodium dodccyl sulfate and dialyzed against the running buffer for sodium dodccyl sulfate-acrylamide gels.
Collagenase Assay-Fractions from Scphadex G-100 columns were assayed for collagcnasc activity by a modification of the method of Grant and Alburn (15). Protein in O.l-ml aliquots from each column fraction was estimated by the method of Lowry et al. (16) using a bovine serum albumin standard. Other aliquots of column fractions were incubated at 37" for 60 min lvith 20 mg per ml of bovine Achilles tendon collagen. After incubation, insoluble collagen was removed by centrifugation, and the protein in the supernatant fluid Teas measured. The amount of protein solubilized in 60 min was taken as an index of collagenasc activity.
Reduction and AlkzJlation-Immune precipitates were dissolved in 0.5 RI Tris-HCl buffer, pH 8.5, containing 1% sodium dodecyl sulfate and were reduced with 0.1 JI dithiothreitol for 1 hour at 22". Iodoacctamide (recrystallized twice from petroleum ether) was added to give a final concentration of 0.24 31, and incubation was continued for 1 hour. The samples were then dialyzed against the running buffer for sodium dodecyl sulfateacrylamide gels. Desalted, lyophilized propeptide from DEAE-Sephades columns was dissolved in 0.5 M Tris-HCl buffer, pH 8.5, and reduced and alkylated as above. Samples then were dialyzed against the running buffer for sodium doclecyl sulfate-acrylamide gels or against the cluting buffer for &phadex G-100 gel filtration. gmino Acid Analyses-Amino acid analyses were performed with a Spinco model 120C amino acid analyzer equipped for high sensitivity.
Protein 'rras hydrolyzed iI, constant boiling HCl for I8 hours at 110" in vacua.
A partial specific volume of 0.72 cc' per g was cakulated from the amino acid composition (19), and data wcrc fittctl by computer using a least square:: methotl.

Experiments
Tvere performed to test wllcthrr ["II]tryptophan could be specifically incorporated into the propeptides of pro-otl and pro-ot2 chains.
Cultures were labelrtl for 24 hours with [31-I]tryptophan, and freeze-dried samples of the acidified medium were dissolvccl and subjcctcd to electrophorcsis on sodium dodecyl sulfate gels (Fig. 1A) Two major radioactive peaks were found. Peak A corresponds in electrophoretic mobility to intact protropocollagen and Peak B, to a denatured digestion intermediate derived from the latter (IO, 11). The other radioactive peaks are proteins of lower molecular weight than cr chains and prcsumably are not collagenous molecules.
To confirm that Peak A was pro-tropocollagen, [3H]tryptophan-labeled medium was precipitated with serum cont,aining antibodies to the NHz-terminal propeptide determinants of protropocollagen (12). Titration revealed that 457; of the radioactive material in the medium could be precipitated at antibody excess. When the dissolved immune precipitates mere subjected to electrophoresis on sodium dodecyl sulfate-polyacrylamide gels, all of the radioactivity was found in Peak A (Fig. IB). Gel electrophoresis of reduced and alkylated immune precipitates showed that both pro-al and pro-a2 chains were labeled (Fig.  IB).
To test if a three-chain, propeptide fragment could be derived from pro-tropocollagen by enzymatic digestion of the helical sequenccs, radiolabeled immune precipitates were digested with collagenasc, dissolved, and dialyzed for gel electrophoresis. with a mobility greater than immunoglobulin A (Fig. 2). When collagenase-treated immune precipitates were reduced and alkylated prior to electrophoresis, all of the radioactivity was recovered in a new peak with a molecular weight one-third that of the unreduced fragment (Fig. 2, and see below). These experiments show that a three-chain, propeptide fragment assembled through disulfide bonds can be obtained by collagenase digestion of pro-tropocollagen. Large Scale Isolation and Purifmation of the Propeptide Fragment--After dialysis and lyophilization of 1. tryptophan-labeled culture medium, the sample was dissolved and filtered through a Sephadex G-100 column (Fig. 3A), and the material eluting in the void volume (brackets) was pooled. Rechromatography demonstrated that all of the material was redelivered in the void volume. A sodium dodecyl sulfate-acrylamide gel of an aliquot of the excluded material confirmed that pro-tropocollagen was present in these fractions. When the pooled fractions were treated with collagenase and rerun on Sephadex G-100 columns, the pattern shown in Fig. 3B was obtained. Less radioactivity was delivered in the void volume, and a major radioactive peak (brackets) was now present which eluted just ahead of the bovine serum albumin marker. As the absorbance and radioactivity were not coincident in this region, it seemed likely that the radioactive species was contaminated with unlabeled protein(s). Assay of the column fractions showed that 35 to 50% of the collagenase activity applied to the column eluted with the major radioactive peak, and stained sodium dodecyl sulfate gels of an aliquot of the pooled fractions showed four protein bands, only one of which was labeled.
To purify further the presumptive propeptide fragment, the pooled fractions from the Sephadex G-100 column (Fig. 3B) were desalted, lyophilized, and chromatographed on DEAE-Sephadex with a linear salt gradient (Fig. 4). At least four major absorbing species were resolved, but only one major radioactive species was eluted between 150 and 180 ml. When the pooled fractions (brackets) were desalted and lyophilized, and the protein was subjected to electrophoresis on sodium dodecpl sulfate gels, a single symmetrical peak of radioactivity was obtained with a mobility identical with that of the propeptide fragment isolated by collagenase digestion of immune precipitates (see above). Reduction and alkylation of the sample prior 7036 to electrophoresis gave a single radioactive peak with a mobility identical with that obtained for the reduced samples of Fig. 2. When the pooled DEAE-Sephadex fraction was titrated with anti-pro-tropocollagn serum, 100% of the radioactivity was precipitated at antibody excess. A double immunodiffusion assay with the same antigen and antiserum gave a single line of precipitation (Fig. 5A). These data show that a disulfideassembled, three-chain propeptide fragment had been isolated _ __ and that it retained its antigenicity after purification. To estimate the purity of the preparation, 40 pg (estimated by the Lowry method (16)) of the propeptide fragment were run on a Tris-HCI-buffered gel (pH 9.4) without denaturants. After staining, only a single band was detected (Fig. 5B) which contained all of the radioactive material applied to the gel (Fig. 5C). When 100 pg of the same material was run on a sodium dodecyl sulfate gel (Fig. 50), staining revealed a major band in the predicted position, which contained all of the radioactive material, and a lighter stained band running just ahead of the major band. Densitometry showed that the minor band contained 6% of the total detectable protein in the gel. These results indicate that the propeptide fragment is contaminated by a small amount of unlabeled protein, possibly derived from the collagenase preparation.
The recoveries of radioactivity and absorbing species (280 nm) at each step in the purification are shown in Table I. After DEAE-Sephadex chromatography, 15.6 '% of the radioactivity and 5.5% of the absorbing species were recovered. Since immune precipitation showed that only 45% of the initial radioactivity in the medium was incorporated into the propeptide sequences of pro-tropocollagen, the final recovery of radioactive propeptide was approximately 35%.

Molecular
Weight Measurements-The molecular weights of the propeptide fragment and its constituent chains were derived from calibration of the acrylamide and Sephadex gels with globular proteins. Plots of log molecular weight versus mobility or elution volumes were employed. For acrylamide gels (e g.  Absorbance wovered was calculated to be 105,000, and chains of 33,000 molecular weight were generated by reduction and alkylation. By contrast, data from Sephadex G-100 columns gave a value of 75,000 for the intact fragment and 26,000 for the individual chains released by reduction.
Calibration of a Sephadex G-200 column in the presence of O.lc10 sodium dodecyl sulfate gave a value of 80,000 for the intact fragment, in agreement with the results obtained without denaturant.
Since different results were obtained by clectrophoresis and gel filtration, the molecular weight of the propeptide fragment also was determined by sedimentation equilibrium.
No heterogeneity was detected in the sample by interference and Schlieren optics, and :I value of 75,400 was obtained, in good agreement with the gel filtration data. We therefore believe that the mobilities of the intact and reduced propcptide fragments on sodium dodecyl sulfate gels are somewhat anomalous and consider the fragment to be composed of three chains, each of approximately 25,000 molecular weight. Amino Acid Composition-The amino acid composition of the purified fragmrnt is shown in Table II. We caution that errors might have been introduced by the presence of an unidentified contaminant at a lerel of approximately 6% (see above). Under the conditions used for t,he analyses, hydroxylysinc and hydroxyproline, if present, would not have been uniquely identified. The purified material is relatively rich in aspartatc and glutamate which comprise 237, of the total residues. This is consistent with the aciliic behavior of the fragment on the DEAE-Sephadex column and iu the Tris-HCl buffered gel. As expected, halfcystine residue+ not found in a: chains, are present.
Although tryptophan wns not measured, it must be present in both pro-al and pro-a2 propeptide components (see Fig. IL?). However, spectroscopic analysis (not' shown) shows a low absorbance at 2S0 nm relative to the absorbance at 235 nm, suggesting that the tryptophan content is low. These data clearly show that the composition of the propeptide fragment differs from that of tropocollagcn. DISCCSSIOS I'ro-tropoco!l:~gerl is the major soluble protein in the medium of human diploicl fibroblast cultures (10-12). As its enzymatic conversion to tropocollagen occurs slowly under standard culture conditions (ll), the collected and acidified culture fluid has proven to bc a good source for the isolation and characterization of the int,act procursor molecule.
Using Collagenase treatment of the immune precipitates digested the helical a! chain segments of the precursor and generated a protein fragment which, upon reduction, gave polypeptide chains with one-third the molecular weight of the assembled fragment.
Given these observations, tryptophan-labeled protein from a large volume of culture medium was subjected to gel filtration, collagenase digestion, and ion exchange chromatography.
After these procedures, a radioactive protein was isolated in reasonable yield which, by immunologic and biochemical criteria, was identified as the three-chain, disulfide-assembled propeptide fragment of the precursor molecule. It is generally agreed that tropocollagen has a molecular weight of 285,000, and we have calculated a molecular weight of 360,000 for pro-tropocollagen using sodium dodecyl sulfate gels calibrated with collagen components (10). The propeptidc fragment should account for the difference between these two molecular weights.
By the gel filtration and centrifugation analyses reported here, the molecular weight of the purified fragment was found to be 75,000 to 80,000, in good agreement with the expected value.
However, the molecular weights of the fragment and its component chains in sodium dodecyl sulfate gels were significantly greater than t'hose determined by the other two techniques. Others have reported anomalous electrophoretic mobilities in sodium dodecyl sulfate gels for propeptides derived from pro-arl chains, particularly in 5% gels, where the sieving effect is reduced (13, 14). Our results also suggest that the mobilities of the fragment and its component chains are anomalous. Glycoprotcins are known to behave anomalously iu sodium dodecyl sulfatr gels (20), and carbohydrate has been found in a propeptide derived from pro-al chains (14). We therefore suggest that the presence of carbohydrate could account for the observed anomalous mobilities.
Compared to tropocollagen, the propeptide fragment is rela-by guest on March 23, 2020 http://www.jbc.org/ tirely rich in aspartate and glutamate and poor in glycine and proline.
Moreover, it contains cysteine and tryptophan which are not present in tropocollagen.
In these respects, the data are in agreement with the results obtained for pro-al propeptides from embryonic cranial bone (13) and the skin of dermatosparaxic cattle (14).
It must be emphasized, however, that our analyses were performed on a fragment containing propeptide sequences of both pro-al and pro-a2 chains; therefore, critical comparison with the published data for pro-al propeptides is not possible. Additional problems arising in such comparisons are: tissue and species differences, the varying lengths of pro-a chains analgsed in different instances, and the demonstrated presence of a small quantity of contaminating protein in our analyses. Most of the collagen precursors isolated from the culture medium are disulfide-assembled molecules, but a smaller quantity of noncovalcntly assembled three-chain precursors is also present (10-12).
The latter arise through the action of an extracellular enzyme which sequentially excises the propeptides of protropocollagen.
The enzymatically shortened pro-cr chains have been shown to lack some or all of their cysteine residues and to migrate just behind the marker al and ot2 chains in sodium dodecyl sulfate gels (10, 11). The experiment of Fig. 1A shows that there is only background radioactivity in the region of the marker chains and so, shortened pro-a chains also lack tryptophan.
These data demonst,rate that pro-a! chains released by denaturation of digestion intermediates can be expected to differ in molecular weight and amino acid composition from pro-or chains released by reduction and denaturation of unprocessed pro-tropocollagen.
Studies of pro-a chains isolated from embryonic cranial bones initially give no evidence for disulfide assembly (I, 2, 7). However, Monson and Bornstein (21) have recently reported that a disulfide-assembled collagen precursor can be isolated from this system with a different method of extraction.
This suggests that the earlier analyses may have been performed on shortened pro-oc chains released by denaturation of a partially digested precursor molecule.
Given the accumulating evidence for disulfide assembly of pro-a chains in a variety of culture systems (see also Refs. 5,8,22), it appears that the manner of assembly and extracellular processing of the collagen precursor molecule is similar for most tissues.