Quantitative analysis of collagen expression in embryonic chick chondrocytes having different developmental fates.

A quantitative determination of collagen expression was carried out in cultured chondrocytes obtained from a tissue that undergoes endochondral bone replacement (ventral vertebra) and one that does not (caudal sterna). The "short chain" collagen, type X is only expressed in the former while the other "short chain" collagen type IX, was primarily expressed in the latter. These two tissues also differ in that vertebral chondrocytes express moderate levels of both type I procollagen mRNAs which were translated into full length procollagen chains both in vivo and in vitro, while caudal sternal chondrocytes did not. The percent of collagen synthesis was about 50% in both cell types, but sternal cells expressed twice as much collagen as vertebral cells even though type II procollagen was more efficiently processed to alpha-chains in vertebral chondrocytes than in sternal chondrocytes. The number of type II procollagen mRNA molecules/cell was found to be about 2300 in vertebral chondrocytes and about 8000 in sternal cells, in good agreement with the results reported by Kravis and Upholt (Kravis, D., and Upholt, W. B. (1985) Dev. Biol. 108, 164-172). There were about 630 copies of type I procollagen mRNAs with an alpha 1/alpha 2 ratio of 1.6 in vertebral chondrocytes compared with 5100 copies and an alpha 1/alpha 2 ratio of 2.2 in osteoblasts, and less than 40 copies in sternal cells. Since the rate of type I collagen chain synthesis was 50 times greater in osteoblasts than in vertebral cells, type I procollagen mRNAs were about six times less efficiently translated in vertebral cells than in osteoblasts. The type I mRNAs in vertebral chondrocytes were polyadenylated and had 5' ends that were identical in osteoblasts, fibroblasts, and myoblasts. Moreover, type I mRNAs isolated from vertebral chondrocytes were translated into full length preprocollagen chains in vitro in rabbit reticulocyte lysates. Thus, chondrocytes isolated from cartilage tissues with different developmental fates differed quantitatively and qualitatively in total collagen synthesis, procollagen processing, and distribution of collagen types.

A quantitative determination of collagen expression was carried out in cultured chondrocytes obtained from a tissue that undergoes endochondral bone replacement (ventral vertebra) and one that does not (caudal sterna). The "short chain" collagen, type X is only expressed in the former while the other "short chain" collagen type IX, was primarily expressed in the latter. These two tissues also differ in that vertebral chondrocytes express moderate levels of both type I procollagen mRNAs which were translated into full length procollagen chains both in viuo and in vitro, while caudal sternal chondrocytes did not. The percent of collagen synthesis was about 60% in both cell types, but sternal cells expressed twice as much collagen as vertebral cells even though type I1 procollagen was more efficiently processed to a-chains in vertebral chondrocytes than in sternal chondrocytes.
The number of type I1 procollagen mRNA molecules/ cell was found to be about 2300 in vertebral chondrocytes and about 8000 in sternal cells, in good agreement with the results reported by Kravis and Upholt (Kravis, D., and Upholt, W. B. (1985) Deu. Biol. 108, 164-172). There were about 630 copies of type I procollagen mRNAs with an alia2 ratio of 1.6 in vertebral chondrocytes compared with 5100 copies and an al/a2 ratio of 2.2 in osteoblasts, and less than 40 copies in sternal cells. Since the rate of type I collagen chain synthesis was 50 times greater in osteoblasts than in vertebral cells, type I procollagen mRNAs were about six times less efficiently translated in vertebral cells than in osteoblasts. The type 1 mRNAs in vertebral chondrocytes were polyadenylated and had 5' ends that were identical in osteoblasts, fibroblasts, and myoblasts. Moreover, type I mRNAs isolated from vertebral chondrocytes were translated into full length preprocollagen chains in vitro in rabbit reticulocyte lysates. Thus, chondrocytes isolated from cartilage tissues with different developmental fates differed quantitatively and qualitatively in total collagen synthesis, procollagen processing, and distribution of collagen types.
Chondrocytes which undergo endochondral bone replacement and chondrocytes which remain as hyaline cartilage have been shown to differ qualitatively and quantitatively in the expression of two "short chain" collagens, type X and type IX. These collagens are smaller than the fibrillar collagens (less than 100 kDa) and have been shown to be unique to cartilage (Mayne and Irwin, 1986). Type X has been immunologically localized to cartilage undergoing hypertrophy and endochondral replacement (Schmid and Linsenmayer, 1985;Capasso et al., 19% Gibson et ai., 1982). Organ and cell culture studies have demonstrated that type X collagen was only expressed by hypertrophic cells Ninomiya et al., 1986;Jimenez et al., 1986;Gibson and Flint, 1985;Habuchi et al., 1985;Schmid and Linsenmayer, 1983), while type IX collagen was expressed primarily, but not exclusively, in tissues remaining as hyaline cartilage Reginato et al., 1986;Wu and Eyre, 1984).
In order to further characterize the phenotype of chondrocytes isolated from two different cartilage tissues which have different developmental fates, a quantitative study was made of total collagen synthesis, type If and type I collagen synthesis, type I1 procollagen processing, and the steady state levels of type I and 11 procollagen mRNAs in caudal sternal chondrocytes, a tissue which remains as hyaline cartilage, and in ventral vertebral chondrocytes, a tissue which undergoes endochondral replacement. The most striking difference between these two cell types is that vertebral but not sternal cells have moderate levels of type I procollagen mRNAs which are translated into type I procollagen chains both in vivo and in vitro.

MATERIALS AND METHODS
Sternal and Vertebral ~h o n d r~y~ Cell C u~t~r e s -C h o n~o c~s were prepared from the caudal half of 16-day-old chicken embryo sterna and from the ventral half of 12-day-old chicken embryo vertebra as described previously (Finer et a t , 1985). Identical culture conditions and periods of time of cell growth in culture were used for both chondrocyte cultures. Cells were initially grown in minimum essential medium for a 7-day period and only floating, nonattached cells were used for subcultivation. Cells were subcultivated into Dulbecco's modified essential medium, and greater than 95% of the sternal cells remained as floating cells after 72 h while the vertebral cells all attached. All experiments were performed 7 days after the subcultivation. Myoblasts were prepared as described by Gerstenfeld et al. (1984) and cultured in the presence of 25 pg/ml ovotransferrin and 10% fetal bovine serum for 7 days. Osteoblasts were prepared as described by Gerstenfeld et al. (1987) and after subcultivation, day 3 cultures were used for RNA extraction and pulse labeling. Tendon fibroblasts were prepared as described by Schwarz et al. (1976) and cultured in the presence of 25 pg/ml ascorbate and 10% bovine serum for 12 days.
Protein Pulse-labeling, Cell-free Translation, and Protein Analysis-Cultures were labeled for 1 h using 5 ml of Dulbecco's modified 5112 essential media supplemented with 250 pCi of [3H]proline (110 mCi/ mmol) as described previously (Gerstenfeld et al., 1983). After the 1b pulse, media was removed, cell layers were washed with phosphatebuffered saline, and 5 ml of serum-free Dulbecco's modified essential media was added. Chase time points were 0.5,1,2.5,6, and 24 h after the label had been removed. The total cell layer proteins were extracted using 6 M guanidine HCI (Gerstenfeld et al., 29841, and radiolabeled type I procollagen protein was prepared from osteoblast cell cultures as described by Gerstenfeld et al. (1987). Pepsin treatment of chondrocyte media was carried out by resuspending 500,000 cpm of nondialyzable labeled protein in 5 ml of 5% (v/v) acetic acid and digesting the protein with 10 pg/ml pepsin at 4 "C for 24 h.
Samples were dialyzed against distilled water until a neutral pH was obtained, and then lyophilized. 75,000 cpm were used/lane for each gel.
For indirect immunoprecipitation, labeled proteins (100,000 cpm total radiolabeled cell extracts for osteoblast cultures and 750,000 cpm for vertebral cultures) were adjusted to 0.5 ml of 150 mM NaCI, 50 mM Tris-HC1, pH 7.6. Insoluble material was removed by centrifugation at 12,000 rpm for 1 min. The supernatants were reacted at 22 "C for 90 min with rabbit antisera against type I procollagen . Indirect immunoprecipitation resulted following addition of goat anti-rabbit antisera during the subsequent 2-h incubation. The precipitates were collected through a 1 M sucrose cushion and washed as described previously (Sonenshein et aL, 1978). Collagen peptides were identified by digesting proline-labeled protein (50,000 cpm) with collagenase, and the amount of collagenase-sensitive material was determined by the method of Peterkofsky and Diegelmann (1971). CNBr peptide mapping of collagen was carried out as described by Gerstenfeld et al. (1983) with comparison to the previously published data for fibrillar collagens (Bornstein and Sage, 1980) and for type X collagen Schmid and Conrad, 1982).
RNA was translated in rabbit reticulocyte lysates prepared by the method of Lodish and Nathan (1972) and treated with micrococcal nuclease (Pelham and Jackson, 1976). Lysates were optimized for the translation of procollagen mRNA as previously described (Gerstenfeld et ai., 1983) using f3H]leucine as the label. Proteins were analyzed by electrophoresis on 5-10% continuous gradient sodium dodecyl sulfate (SDS)'-poIyac~lamide gels (Laemmli, 1970). Fluoro~aphy was carried out by the method of Bonner and Laskey (1974) and Laskey and Mills (1975). The gels were dried and exposed to preflashed x-ray film from 3 to 7 days at -50 "C. Quantitation of the fluorographs was performed on a LBK Ultrascan I1 densitometer (LKB Broma, Sweden).
Analysis of Procollagen mRNA Expression-Total RNA was extracted from embryonic calvaria and chondrocytes using a modification of the phenol-proteinase K method (Gerstenfeld et al., 1983). Polyadenylated mRNA was prepared as described by Tate et al. (1983). RNA was electrophoresed on 25-cm agarose gels containing 2.2 M formaldehyde by method of Lehrach et al. (1977), and blotting was carried out as described by Thomas (1980). Specific subfragments used as probes were a 550-base pair Bum-Pst fragment for al(I1) prepared from pCs2 (Young et al., 1984); a 380-base pair Bum-Sal fragment for al(1) prepared from pMFlA8 (Finer et al.,198%) and a 364-base pair Aua-Sau3A fragment for a2(I) prepared from pMF2l (Tate et al., 1983). Fragments were either labeled by random primer labeling or for the 5' end type I clones by SP6 runoff reactions (Meiton et at., 1984). Hybridization conditions were as described by or Finer et al., 1985. 5 "C higher temperatures were used for hybridizations with labeled SP6 runoff products. al(1X) mRNA levels were analyzed using a -600 base pair cDNA cloned and sequenced by Vesna Maksimovic, corresponding to nucleotides 406-969 (Ninomiya and Olsen, 1984), and type X mRNA levels were determined by hybridization to a 33-base primer synthesized and high pressure liquid chromatography purified by Operon Technologies (San Pablo, CA). The 33-base primer was the anti-sense sequence from base pair 290 to 323 from the reported sequence by Ninomiya et al. (1986). Primer was labeled by polynucleotide kinase with ["PI ATP (Maniatis et ul., 1982). Relative mRNA levels were determined by scanning densitometry of underexposed autoradiograms of slot blots and from dot blot analysis.
In order to convert the slot blot or dot blot data to number of mRNA molecules/cell, the dots were excised and counted or an underexposed slot blot autoradiogram was scanned on the LKB The abbreviations used are: SDS, sodium dodecyl sulfate; Pipes, 1,4-piperazinediethanesulfonic acid.
Ultrascan I1 densitometer. For each sample measurement, duplicates of three concentrations of RNA at 1, 2, and 4 pg were used. An average value of cpms or relative densitometric areas were extrapolated from the slope of the line generated from the densitometric or cpm values. The RNA to DNA ratio was determined by measuring these nucleic acid components separately in the total nucleic acid extracted from the cells so that final copy number could be expressed per DNA or per cell. Relative densitometric areas and cpm are used interchangeably. In the equations presented below, cprn was used. Probe back hybridization was determined by measuring the cpm or relative densitometric areas of six DNA concentrations from 20 to 500 pg. For each probe and each hybridization, the back hybridization was determined, and a straight line was derived from which the slope determined, as cpm/pg of probe. For each probe, it is assumed that the RNA to DNA or RNA to RNA hybridization is the same as the observed DNA to DNA or RNA to DNA hybridization of the probe.
Half-values were used when double-strand^ DNA was used as a probe. Accordingly, 1) cpm/pg RNA X pg RNA/pg DNA = cpm/pg DNA; 2) cpm/pg DNA X (ng RNAlcpm)' = ng RNA/pg DNA 3) ng RNA/pg DNA X ( p g DNA/cell)b = ng RNA/cell; 4) ng RNA/cell X IO-' X (Avagadro's number)"/(Mr of probe in g/rnolld = molecules RNA/cell; in which (a) is determined from back hybridization of the probe to itself; ( b ) the DNA content/diploid chicken cell is 2.5 X lo-' pg DNA/cell, derived from Mirsky and Ris (1956); (c) Avagadro's number is 6.0225 X loz3 molecules/mol; ( d ) the molecular weight is 1.33 x IO5 for the al(I), 1.30 X 105 for the a2(I), and 1.68 x IO6 for the al(I1) collagen probe determined by probe length in bases X (350 or 365) mean molecular weightbase for either DNA or RNA probes.
In the experiments presented here, the following data were empirically determined for each probe: a(l) back hybridization was 6.36 X 10' cpm/ng probe and cpm/pg RNA was 4.9 x lo3; a d ) back hybridization was 8.7 X lo' cpm/ng probe and cpm/pg RNA was 3.4 X lo3.
The RNA/DNA ratio in 7-day myoblasts was 1.5.
Analysis of 5' mENA Ends: RNase Protection of ~~A /~N A Hybridization and Primer Extension Analysis-The plasmid SP64/ AH500 and SP65/E$X320 were constructed to map correct initiation of mRNAs transcribed from the chicken proa2(I) and proal(1) cob lagen genes, respectively. In order to construct SP64/AH500, pCg 5.7, a genomic clone containing the first four exons of the chicken proa2(1) collagen gene (Tate et al., 1983) was opened at its unique ApaI site. The 3' overhang was removed by incubation with T4 DNA polymerase. EcoRI linkers were added to the blunt end, and the DNA was digested with HaeIII. These reactions were carried out under standard conditions. The 500 nucleotide EcoRI ( A~I ) / H~I I I fragment was isolated from 0.8% low melt agarose gel, ligated to an EcoRI/SmaI digest of the vector SP64 according to Frischauf et al. (1980), and transformed into Escherichia coli DH5 (Hanahan, 1983). This plasmid encodes 100 base pairs of 5"flanking sequence, the 203base pair first exon, and 310 base pairs of the first intron. SP65/ BX310 was constructed by ligation of the 313-base pair BarnHI/AccI fragment from XSA/S51, a genomic clone encoding the first five exons of the chicken proal(1) collagen gene (Finer et aL, 1987a) to a BarnHI/ SmaI digest of the vector SP65. The ligation and transformation were carried out as described above. This plasmid contains 220 nucleotides of 5"flanking sequence and 93 nucleotides of exon 1.
Labeled antisense RNA probes were prepared by digestion of SP64/ AH500 with EcoRI or digestion of SP65/BX310 with BamHI, followed by in uitro run off transcription with SP6 RNA polymerase as described by Melton et a1 (1984). Labeled RNA was resuspended in 300 pl of 2 mM EDTA, pH 7.5. One to 20 pg of total RNA to be assayed were ethanol precipitated and resuspended in 3 pl of probe. Each sample was adjusted to 80% formamide, 400 mM NaCl, 40 mM Pipes, pH 6.7,l mM EDTA. The samples were incubated for 10 min at 85 "C, followed by 12-18 h at 55 "C. Following hybridization, the samples were diluted 10-fold and adjusted to 300 mM NaCI, 5 mM EDTA, 10 mM Tris-HC1, pH 7.5. RNase A and RNase T1 (PL biochemicals) were added to final concentrations of 250 and 1 pg/ml, respectively, and incubated for 30 min at 30 "C. The samples were adjusted to 0.5% SDS, 250 pg/ml proteinase K (Merck) and digested for 15 min at 37 "C. After digestion, the samples were phenol extracted, ethanol precipitated, resuspended in Maxam and Gilbert sequencing loading buffer and run on 6% polyacrylamide, 8.3 M urea DNA-sequencinggels (Maxam and Gilbert, 1980). After hybridization to normal prod(1) and proaZ(1) mRNAs, the 313 nucleotide runoff transcript from SP65/BX310 should be trimmed to 93 nucleotides, and the 500 nucleotide runoff transcript from SP64/AH500 should be trimmed to 203 nucleotides, respectively.
Primer extension of proal(1) mRNA was carried out on poly(A)+ mRNA using a 14-mer corresponding to the sequence starting at the position of the AccI restriction site, depicted in Fig. 5A. The expected extension product of the al(I) primer is 93 base pairs. The sequence used for primer extension of proa2(1) mRNA was a 15-mer corresponding to the first 15 bases of exon 3 of this gene. The expected size of a2(I) extension is 229 base pairs. Primer hybridization and cDNA synthesis were carried out as described by Tate et al. (1983).

RESULTS
Collagen Synthesis and Processing-Total collagen synthesis was examined in vertebral and sternal chondrocyte cultures. A summary of these results presented in Table I indicates that sternal chondrocytes synthesized approximately two times more collagen/cell than vertebral chondrocytes. Both types of chondrocytes, however, devoted a very high percentage of their protein synthesis to collagen production, with sternal cells devoting a slightly higher percentage, 56%, compared with 50% for vertebral cells. The variation in distribution of the newly synthesized collagen between the cell layer and media compartments, shown in Table I, reflects the physical differences in the growth patterns between the two culture systems. Sternal cells grew as nonadherent floating cells while vertebral cells grew as adherent polygonal cells. Thus, most of the newly synthesized collagen of the vertebral cultures was secreted and incorporated into the matrix of the cell layers. In contrast, the floating sternal cells secreted their collagen into a gelatinous matrix which forms in the culture media, but upon separation of the cells from the media by centrifugation, the newly synthesized collagen was left in the supernatant while the cells were pelleted.
The analysis of the newly synthesized collagen proteins in sternal and vertebral cells is shown in Fig. U. This analysis compared the proteins isolated from the cell layers, the media, and the pepsin-treated media of sternal and vertebral chondrocyte cell cultures pulsed for 1 h followed by a 2.5-h chase.
A prominent proal(I1) collagen band is present in the profiles of both vertebral and sternal cell extracts, with only minor differences detected in the profiles of newly synthesized intracellular proteins (Fig. U, left panel). Proal(I1) collagen appeared to be more rapidly processed in vertebral cells than in sternal cells. The identification of the processing intermediates was identical to that made by Schmid and Coilrad (1982) and was verified by cyanogen bromide peptide mapping of each of the identified bands. Two other prominent proteins were found in the vertebral but not in the sternal chondrocyte profiles. One was a noncollagenous protein of approximately 220 kDa and the other was a collagenase sensitive protein of about 60 kDa. These proteins were identified as fibronectin and type X collagen based on their size and previous identification (Schmid and Conrad, 1982;Adams et al., 1982;Gerstenfeld et al., 1985). Scanning densitometry of the media protein profiles indicated that the 60-kDa collagen comprised -40% of the total collagenous media proteins.
In comparison, two collagenase-sensitive bands of 85 and 70 kDa were seen in profiles of the sternal media proteins but not in the vertebral media profiles. These correspond to the a3-and al-chains of type IX collagen. This identification was based on these proteins' collagenous nature, their reported molecular weight von der Mark et al., 1984;van der Rest et al., 1985)., and their chromatographic behavior on DEAE-cellulose: which was consistent with that reported by von der Mark et (11. (1984). Scanning densitometry of the media profiles indicated that the 85 and 70 kDa collagens constituted about 12% of the sternal culture media collagen. * L. Gerstenfeld, unpublished data. 4.0 X 1 0 4 93 a Protein synthesis and pg of DNA/lOO-mm dishes were determined from an average of three separate preparations from cells labeled for 24 h. Each determination was from at least triplicate samples. Total cpm and percent of collagen were calculated using the formula of Peterkofsky and Diegelmann (1970) as modified by Schwarz et al. (1976), and the corrected values for cpm of collagen are reported in the table. All measurements are expressed as the mean result and no measurements had a range greater than 18%. A, vertebral (V) and sternal ( S ) cell cultures were pulse labeled for 1 h followed by a 2.5-h chase. Cell layer, media and pepsin-digested media proteins were extracted as described under "Materials and Methods." 75,000 cpm of each sample were electrophoresed on a 5-10% SDS-polyacrylamide gradient gel. The gel was exposed to preflashed film for 3 days. FN, fibronectin; pro, unprocessed procollagen; p N and PC, partially processed procollagen retaining the amino-terminal or carboxy-terminal propeptide, respectively; 60K, type X collagen, formerly called short chain collagen; arrows identify collagenase-sensitive polypeptides of 85 and 70 kDa believed to be the a3 and a1 chains of type IX collagen; 0, osteoblast control. B, immunoprecipitation of [3H]proline pulse-labeled protein with type I antisera. V, vertebral and 0, osteoblast media proteins were used for the immunoprecipitation experiments.
Pepsin digestion of the media proteins from 24-h labeled samples was carried out to remove noncollagenous proteins from the media proteins and to verify the identification of the collagen bands. The sternal sample contained a single major pepsin-resistant protein corresponding to the al(I1) chain, while the vertebral sample contained two prominent pepsinresistant species the al(I1) collagen band and one corresponding to a 45-kDa pepsin-resistant protein (Fig. IA, rightpanel).
The reduction of type X chain from 60 to 45 kDa following pepsin treatment has been reported previously (Schmid and Linsenmayer, 1983). Upon longer fluorographic exposure of the sternal lanes, additional pepsin fragments of 50, 35, 30, and t 2 0 kDa were detected, presumably derived from type IX collagen chains.
It was not possible to positively identify either proal(1) and proaf(1) or the a(1) collagen chains in the cell layer and media profiles in Fig. lA, indicating that if type I collagen was present, it was being synthesized at very low levels. A minor band with the mobility of a2(I) chains, however, could be seen in the pepsin-digested media from vertebral cells. The ratio of the a2(I) to the al(I+II) was 1:24. Assuming that the ratio of al(1) to a2(I) is 2:1, the ratio of type I1 to type I collagen chains in vertebral cells is 22:3, that is, the rate of type I1 chain synthesis is about seven times greater than that of type I. However, since pepsin may preferentially digest a2(I) collagen, this may be an underestimation of the amount of type I collagen.
To confirm the pepsin results, indirect immunoprecipitation of the radiolabeled type I collagen was carried out. Proteins synthesized by either vertebral chondrocytes or day 3 osteoblasts (used as control) were precipitated with antisera to type I collagen  and then examined by polyacrylamide gel electrophoresis. These results are shown in Fig. 1B. The vertebral and osteoblast samples both contain the a1 and a 2 chains of type I collagen, but the osteoblast sample also had procrl(1) and proa2(1) chains. It was necessary to use eight times more radiolabeled protein to obtain the barely detectable immunoprecipitated products from the vertebral samples. The gel profiles were scanned, and the osteoblast sample had almost seven times more labeled type I collagen chains than the vertebral sample. Since the cpm could be normalized to DNA and the total cpm/ immunoprecipitation was known, it was calculated that there was roughly 50 times more synthesis of type I chains/osteoblast cell than per vertebral cell.
Either short term pulse labeling, shown in Fig. 1, or steady state 24-h labeling indicated that more processed al(I1) molecules had accumulated in the vertebral cultures than in sternal cultures. These results would suggest that vertebral cultures processed their collagen more efficiently than sternal cultures. In order to examine this possibility, a pulse chase experiment was carried out. These results, shown in Fig. 2A, clearly demonstrate that the vertebral cultures converted their procollagen to a-chains, while sternal cultures did not. Moreover, sternal cells secrete collagen at a slower rate than vertebral cells as indicated by the lower amounts of labeled protein seen in the culture media at earlier time points. Analysis of the kinetics by scanning densitometry (Fig. 2B) demonstrated that the amino-terminal propeptide was processed more efficiently than the carboxyl-terminal propeptide with an overall tm of >6 h for vertebral cultures. However, in sternal cultures less than 20% of the proa-chains were converted to the processing intermediates after 24 h. As expected, neither the 60-kDa collagen species in the vertebral cultures nor the 85-and 70-kDa collagen species in the sternal cultures were converted to lower molecular weight polypeptides.
Analysis of Procollagen mRNA Expression-To compare the observed differences in the synthesis of types I and I1 procollagen and types IX and X collagen in sternal and vertebral chondrocytes ( Fig. 1) with differences in the levels of the mRNAs encoding these proteins, three types of analyses were carried out. A qualitative assessment of type I and I1 procollagen mRNAs and of type IX and X collagen mRNAs was carried out by Northern blot analysis of polyadenylated and total RNA, respectively, isolated from either vertebral or from sternal cells. The quantitative determination of each of these RNAs was then obtained from slot blot analyses of total RNA,

FIG. 2. Pulse chase analysis of type I1 procollagen processing in vertebral and sternal chondrocytes.
Each pulse-labeled media sample was resuspended in 500 pl and 50 pl was loaded/lane. A, cultures were labeled with [3H]proline for 1 h. Chase was initiated at this time by replacing the pulse media with fresh Dulbecco's modified essential media minus serum. The time of chase and the source of media proteins is denoted in the figure. The gel was exposed to preflashed film for 5 days. Proteins are labeled as in Fig. 1. B, gel profiles depicted in A were scanned with an LKB Ultrascan I1 densitometer. The processing kinetics are expressed as percent of the total densitometric area of the total type I1 collagen pro and processing species. while the number of copies of type I procollagen mRNAs/cell was obtained both from slot blot analysis and mRNA protection experiments.
Both sternal and vertebral chondrocytes expressed high levels of type I1 procollagen mRNAs but only vertebral cells expressed type I mRNAs detectable by Northern blots (Fig.  3). The al(1X) mRNA was readily detectable on blots of sternal RNA but only visible in blots of vertebral RNA if longer autoradiographic exposures of blots containing two or three times more RNA were used, while type X collagen mRNA was detectable in vertebral but not in sternal chondrocyte RNA (Fig. 4) and remained undetectable even a t higher RNA concentrations. Quantitative slot blot analyses of comparable mRNA quantities demonstrated that the type I1 procollagen mRNA level was 3.5 times greater in sternal cells than in vertebral cells. In contrast, type I mRNAs were undetectable in sternal cells while the level of type I1 mRNA was almost four times that of type I mRNAs in vertebral cells.
Because mRNA protection assays are about 10-50 times more sensitive than Northern blot analyses, the former were carried out to better assess the presence of type I procollagen mRNAs in sternal chondrocytes. The results are shown in Fig. 5, A-C and are summarized in Table 11. Although only visible on the longer exposure of the autoradiograph (Fig. 5C), low levels of both proal(1) and proa2(1) mRNAs could be detected in sternal cells. Since the mRNA copy number/cell

FIG. 3. Northern blot analysis of steady state levels of type I and I1 procollagen mRNAs in vertebral and sternal cells.
Micrograms of poly(A)+ mRNA applied to the 0.8% agarose, 2.2 M formaldehyde gel, procollagen mRNA species hybridized to pCs2 (Young et al., 1984), pCg45 (Lehrach et al., 1978), or Cg54 (Lehrach et al., 1979) to identify proal(II), proa2, and proal collagen mRNAs, respectively, and the cellular source of the RNA are denoted in the future. Autoradiographs were exposed for 8 h. had been determined for proal(1) and proa2(1) mRNAs in myoblasts, as described under "Materials and Methods," this analysis allowed an accurate and very sensitive determination of type I mRNA levels in both vertebral and sternal chondrocytes as well as in tendon fibroblasts and osteoblasts (Table  11). Type I mRNA levels in sternal cells were 15-fold lower than in vertebral cells, over 100-fold lower than in osteoblasts. In contrast, the 8,000 copies/cell of type I1 mRNA in sternal cells was almost 3.5-fold higher than the 2,300 copies/cell in vertebral cells and 1.5 times greater than that of type I (proal and proa2) mRNAs in osteoblasts. While the difference between the type I1 procollagen mRNA levels in these two cell types might seem surprising, it is consistent with the finding that sternal cells synthesize twice as much collagen as vertebral cells, as well as the earlier results of Kravis and Upholt (1985) who reported that there were about 10,000 copies of type I1 mRNA/cell in RNA isolated from 17-day embryonic sterna, and about 2,000 copies/cell in RNA isolated from 7day cultures of stage 24 limb mesenchyme.
The difference between the 4 1 ratio of type I1 to type I procollagen mRNAs levels in vertebral cells, and the 7:l ratio of type I1 to type I a chain synthesis in these cells, determined from the data in Fig. lA, suggests that type I1 procollagen mRNAs were more efficiently translated in vertebral cells than type I mRNAs, but only by a factor of less than two. If pepsin digestion had preferentially digested the a2(I) collagen, the levels of type I collagen synthesis may have been underestimated, and there may be no difference in the translation efficiency of these mRNAs. However, if one compares the 8fold higher levels of type I procollagen mRNA in osteoblasts to that in vertebral chondrocyte with the 50-fold higher a chain synthesis in these two cell types, obtained from the data in Fig. lB, it is apparent that type I procollagen mRNAs were about six times more efficiently translated in osteoblasts than in vertebral chondrocytes.
Analysis of 5' Ends of Type I Procollagen mRNAs-The apparent inefficient translation of type I mRNAs in vertebral chondrocytes could have resulted from a chondrocyte-specific factor which is limiting for collagen translation in vertebral cells, or it could result from an altered primary structure of the mRNA, such as the loss of the poly(A) tail or altered 5' ends of the type of procollagen mRNAs. However, Northern blot analyses of oligo(dT)-bound vertebral RNA (Fig. 3) showed that both proarl(1) and proa2(1) collagen mRNAs were polyadenylated, and mRNA protection experiments using in uitro synthesized transcripts containing the 5' ends of both type I mRNAs (Fig. 5, A and B ) clearly demonstrated that the 5' ends of both of these mRNAs were the same in vertebral chondrocytes as in osteoblasts, tendon fibroblasts, and myoblasts in which they were efficiently translated. In order to independently verify the 5' sequence of al(1) and a2(I) mRNAs found in vertebral cells, primer extension analysis was carried out. Because an alternate splicing pattern might occur for exon two of the proa2(1) gene (Aho et al., 1984;Tate et al., 1983), a primer was constructed corresponding to the first 15 base pairs of exon three (Tate et al., 1983). This should have resulted in a fully extended product 229 bases long. Since no similar alternate splicing could be seen in the DNA sequence of the first four exons of the proal(1) gene (Finer et al., 1987), a 14-base pair primer ending in the AccI site at +93, shown in Fig. 5A, was used. As can be seen from these results (Fig. 6), both primers resulted in extensions having the predicted size for calvaria and vertebral mRNAs. These results confirm the RNA protection data and show that the 5' ends of both type I mRNAs are identical in RNA isolated from embryonic calvaria and vertebral chondrocytes.
In Vitro Translation of RNA Isolated from Vertebral C e h -Since other secondary sequence modification may affect mRNA translation ability but not be detected by primer extension or mRNA protection assays, the in uitro translatability of type I and I1 mRNAs was tested in a reticulocyte lysate translation system using increasing concentrations of polyadenylated RNA from vertebral cells. The fluorograph of the gel showing the [3H]leucine labeled proteins obtained is shown in Fig. 7. At the lowest RNA concentration, type I1 procollagen mRNA was more efficiently translated than either type I procollagen mRNA. Even though the proal(1) and proa2(1) bands were expected to be less intense than the proal(I1) band since there were four times more type I1 than type I mRNAs in the vertebral RNA sample, these bands are barely visible. At the three higher RNA concentrations, however, the type I procollagen mRNAs were much more efficiently translated than type I1 mRNAs, with no proal(I1) band visible at the two highest RNA concentrations. This unexpected effect of RNA concentration on the relative translation efficiencies of type I and I1 procollagen mRNA appears to be unique to vertebral RNA. When mRNA from sternal cell was translated together with increasing amounts of mRNA from calvaria, there was no decrease in the intensity of the proal(I1) in comparison to the procrl(1) band at the higher RNA concentrations.* This suggests that there may have been an inhibitor of type I1 mRNA translation in the vertebral RNA preparation. Such an inhibitory factor has been identified in mRNA isolated from a chondrosarcoma tumor (Paglia et aZ., 1981). Whatever the cause of the odd dependence of the translation efficiency on RNA concentration, the type I procollagen mRNAs are clearly capable of being translated into full length type I prepropeptides in uitm at moderate to high RNA concentrations.

DISCUSSION
Chondrocytes which undergo endochondral bone replacement and chondrocytes which remain as hyaline cartilage have previously been found to differ in that short chain type  Fig. 5 to myoblast values using the following equation: (densitometric area X RNAPNA ratio) for cell type (densitometric area X RNAIDNA ratio) for myoblasts X copy number of a myoblast These values were independently confirmed by slot or dot blot analysis. Myoblast copy number was determined using calculations presented under "Materials and Methods." Number of type I1 mRNA molecules/cell were determined from slot blot analysis and back hybridization of the probe to itself as described under "Materials and Methods." All values were determined from averages of at least duplicate experimental determinations, and variations in calculated copy number were +12%.  X collagen is only expressed by the former while another short chain collagen, type IX, is primarily expressed by the latter (Reginato et aZ., 1986;Jimenez et d., 1986;Gibson et al., 1984;Schmid and Linsenmayer, 1985;Habuchi et aZ., 1985;Gibson and Flint, 1985;Capasso et al., 1984;Linsenmayer, 1983, 1985;Gibson et aZ., 1982). While confirming this differential expression of the two short chain collagens, we have also shown that these two cell types differ in that vertebral chondrocytes, which undergo endochondral replacement, express moderate levels of both type I procollagen mRNAs which are translated into type I procollagen chains both in uiuo and in vitro, while caudal sternal chondrocytes, which remain as hyaline cartilage, do not. Moreover, sternal cells expressed twice as much collagen as vertebral cells, but type I1 procollagen was more efficiently processed to a-chains in vertebral chondrocytes than in sternal cells.
By making a quantitative determination of the relative rates of type I1 and type I procollagen synthesis in embryonic chick vertebral and sternal chondrocytes, the rates of type I chain synthesis in vertebral cells and osteoblasts, and of the number of mRNA molecules/cell in each of these cells as well as in myoblasts and tendon fibroblasts, we were able to show that type I procollagen mRNAs are about as efficiently translated in vertebral chondrocytes as type I1 mRNA, but only about a sixth as efficiently translated in vertebral chondrocytes as in osteoblasts. This lower translational efficiency was shown not to result from an altered structure of the type I mRNAs in vertebral cells. Both proal(1) and procr2(1) collagen mRNA were polyadenylated since oligo(dT)-bound RNA was used for Northern blot analysis. Moreover, mRNA protection experiments and primer extension analysis demonstrated that the 5 ' -untranslated region of both type I procollagen mRNAs is identical in vertebral chondrocytes, which do not translate these mRNAs efficiently, in uiuo, and in fibroblasts, myoblasts, and osteoblasts, which do translate type I procollagen mRNAs efficiently. These data are also consistent with that of Marini et al. (1988) who showed that the 5' ends of both chicken and human proa2(1) collagen mRNA were identical in RNA isolated from fibroblasts and osteoblasts. While our results cannot rule out very small changes in nucleotide sequence, they demonstrate that there have been no major changes due to alternate slicing at the 5' ends of these genes.
Finally, additional evidence that the lower translational efficiency of type I mRNA in vertebral cells was not a result of an altered primary structure of these mRNAs was obtained by showing that both mRNA species were translated in vitro in a heterologous rabbit reticulocyte lysate into full length prepropeptide chains. However, while type I procollagen mRNAs were much more efficiently translated in vitro than type I1 mRNAs, at moderate to high concentrations of RNA, the opposite prevailed at low RNA concentrations. This unexpected mRNA titration profile suggests there must be at least two factors affecting the in uitro translation of these mRNAs. One, acting at low RNA concentrations in vitro,