Different levels of regulation accomplish the switch from type II to type I collagen gene expression in 5-bromo-2'-deoxyuridine-treated chondrocytes.

The shift of chick embryo chondrocytes to a fibroblastic phenotype by 5-bromo-2'-deoxyuridine (BrdUrd) has been used to examine the molecular basis of the switch from type II to type I collagen gene expression. Transcription rates of each of these three collagen genes before and after this shift were measured in nuclear run-on transcription assays with double-stranded 3'-cDNA probes specific for each of these three mRNAs. Degradation rates of each of these RNAs were calculated from the rate of decrease in the concentration of each RNA after the inhibition of synthesis with actinomycin D. The shut-off of the expression of the type II collagen gene during this shift was shown to occur at the transcriptional level, since the transcription rate of this gene decreased dramatically. The decay rate of the type II mRNA (half-life of approximately 15 h) is not significantly faster in BrdUrd-treated cells. The alpha 1(I) gene is transcribed at similar rates in untreated and shifted chondrocytes, but the steady state level of alpha 1(I) RNA in chondrocytes is only 1.5% of that in shifted cells. Although the measured degradation rate of the total alpha 1(I) RNA from untreated chondrocyte cultures is approximately the same as in shifted cells (half-life of approximately 12 h), indirect evidence suggests that this alpha 1(I) RNA is derived from a low level of fibroblast contamination of these chondrocyte cultures. The alpha 1(I) RNA synthesized by untreated chondrocytes is assumed therefore to be broken down very rapidly in the nucleus. The alpha 2(I) gene is also transcribed in untreated chondrocytes at rates similar to shifted cells but, unlike alpha 1(I) RNA, its steady state level in untreated chondrocytes is approximately 30% of its level in shifted chondrocytes. The increased level of alpha 2(I) RNA in shifted cells may be regulated in part by an increase in stability of the alpha 2(I) mRNA, which has half-lives of 5.2 and 10.4 h, respectively, in untreated and shifted chondrocytes. The alpha 2(I) RNA in the untreated chondrocytes was found to have a different 5' end from that present in the BrdUrd-shifted chondrocytes or in chick embryo fibroblasts. The presence of this altered RNA in untreated chondrocytes explains the absence of synthesis of the fibroblastic alpha 2(I) collagen polypeptide chains in these chondrocytes, despite the presence of the alpha 2(I) RNA as measured with 3' probes.


Different Levels of Regulation Accomplish the Switch from Type I1 to Type I Collagen Gene Expression in 5-Bromo-2"deoxyuridine-treated
Chondrocytes" (Received for publication, November 26, 1990) G . Roger Askew$, Sandia Wang, and Lewis N. LukensO

From the Department of Molecular Biology and Biochemistry, Wesleyan University, Middletown, Connecticut 06459
The shift of chick embryo chondrocytes to a fibroblastic phenotype by 5-bromo-2'-deoxyuridine (BrdUrd) has been used to examine the molecular basis of the switch from type I1 to type I collagen gene expression. Transcription rates of each of these three collagen genes before and after this shift were measured in nuclear run-on transcription assays with double-stranded 3'-cDNA probes specific for each of these three mRNAs. Degradation rates of each of these RNAs were calculated from the rate of decrease in the concentration of each RNA after the inhibition of synthesis with actinomycin D. The shut-off of the expression of the type I1 collagen gene during this shift was shown to occur at the transcriptional level, since the transcription rate of this gene decreased dramatically. The decay rate of the type I1 mRNA (half-life of approximately 15 h) is not significantly faster in BrdUrdtreated cells. The al(1) gene is transcribed at similar rates in untreated and shifted chondrocytes, but the steady state level of al(1) RNA in chondrocytes is only 1.5% of that in shifted cells. Although the measured degradation rate of the total al(1) RNA from untreated chondrocyte cultures is approximately the same as in shifted cells (half-life of approximately 12 h), indirect evidence suggests that this al(1) RNA is derived from a low level of fibroblast contamination of these chondrocyte cultures. The al(1) RNA synthesized by untreated chondrocytes is assumed therefore to be broken down very rapidly in the nucleus. The a2(I) gene is also transcribed in untreated chondrocytes at rates similar to shifted cells but, unlike al(1) RNA, its steady state level in untreated chondrocytes is approximately 30% of its level in shifted chondrocytes. The increased level of a2(1) RNA in shifted cells may be regulated in part by an increase in stability of the a2(I) mRNA, which has half-lives of 5.2 and 10.4 h, respectively, in untreated and shifted chondrocytes. The a2(I) RNA in the untreated chondrocytes was found to have a different 5' end from that present in the BrdUrd-shifted chondrocytes or in chick embryo fibroblasts. The presence of this altered RNA in untreated chondrocytes explains the absence of synthesis of the fibroblastic a2(I) collagen polypeptide chains in these chondro-cytes, despite the presence of the a2(I) RNA as measured with 3' probes.
The expression of the collagen type I and type I1 genes is tightly regulated during embryonic development, with each type of collagen being made in specific tissues at specific times (1, 2). Type I collagen is relatively widely distributed, being present in bone, skin, tendons, and ligaments, whereas type I1 collagen is located almost exclusively in hyaline cartilage. A useful system for investigating the molecular mechanisms that regulate the expression of these genes is provided by the 5-bromo-2'-deoxyuridine (BrdUrd)'-induced shift from type I1 to type I collagen synthesis in primary cultures of chondrocytes (3,4). Primary cultures of embryonic chick sternal chondrocytes synthesize the specialized products characteristic of differentiated chondrocytes, including type I1 collagen. However, exposure of these cells to low levels of BrdUrd for 8 to 10 days causes these cells to lose their chondrocyte phenotype, so that they no longer synthesize type I1 collagen or other cartilage-specific products. The previously floating chondrocytes attach to the bottom of the culture dish, and they now synthesize a number of proteins, notably type I collagen and fibronectin, that are characteristic of fibroblasts. Since the mesenchymal cells that give rise to chondrocytes synthesize type I collagen (5), it is possible that this shift to a fibroblastic phenotype represents a dedifferentiation back to the mesenchymal phenotype. A similar shift in phenotype can be induced by the tumor promoter, phorbol-12-myristate-13-acetate (6, 7), or by repeated subculture of the chondrocytes in monolayer (reviewed in Ref. 5) or by their transformation with Rous sarcoma virus (8-10).
Our previous studies of the changes in steady state levels of the type I and type I1 collagen mRNAs in nucleus and cytoplasm during the BrdUrd shift suggested that the primary control of the type I1 gene was at the transcriptional level (11). The level of type I1 RNA fell sharply in BrdUrd-treated cultures, to undetectable levels in the nucleus by day 6 and in the cytoplasm by day 8. The type I mRNAs, however, appeared to be present at appreciable levels in the nucleus of untreated chondrocytes, an observation that suggested that the type I genes were transcribed in untreated chondrocytes, although these cells do not synthesize type I collagen. An additional unexpected finding was that the level of a2(I) mRNA in the cytoplasm of untreated chondrocytes, as measured with cDNAs from the 3' end of the mRNA, was relatively high, being one-third of its level in shifted cells, although The abbreviations used are: BrdUrd, 5-bromo-:!'-deoxyuridine; kb, kilobase pair(s); bp, base pair(s); nt, nucleotides; PCR, polymerase chain reaction. untreated chondrocytes do not synthesize a2(1) collagen chains. The level of cytoplasmic al(1) mRNA was found, as expected, to be very low in chondrocytes relative to shifted cells (less than 3%).
To assess the relative contributions of changes in transcription and degradation rates to the control of the levels of these collagen mRNAs during the BrdUrd-induced shift, measurements of transcription rates and degradation rates for each mRNA were undertaken. The results reported here confirm the conclusion that the expression of the type I1 collagen gene is controlled primarily at the transcriptional level. Both type I collagen genes, however, appear to be transcribed in chondrocytes at rates comparable to the BrdUrd-shifted cells. The control of these genes is, therefore, not primarily due to changes in transcription rates. In the case of the al(1) gene, indirect evidence suggests that the control is primarily at the level of nuclear stability. The a2(I) RNA shows relatively minor changes in the transcription rate but a significant increase in stability during the BrdUrd-induced shift. The chondrocyte form of the a2(I) RNA was found to have a different 5' than the 5' end present in fibroblasts or in the BrdUrd-shifted cells, in agreement with the discovery of Adams et al. (12,13). Splicing of the chondrocyte-specific first exon, "exon A," to exon 3 of the usual fibroblastic a2(I) mRNA disrupts the collagenous reading frame (13), so this altered a2(I) RNA does not code for collagen sequences. Therefore, expression of the a2(I) gene in chondrocytes and BrdUrd-shifted cells is regulated by choice of the transcription start site.

MATERIALS AND METHODS
Cell Culture-Chondrocytes were isolated from the sterna of 14day-old chick embryos (a gift from Arbor Acres, Glastonbury, CT or purchased from Spafas, Norwich, CT) by the floater selection method of Schiltz et al. (14), except that Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Gibco Laboratories) was used. The chondrocytes were induced to shift to a fibroblastic phenotype by culture in the presence of 6.5 X M bromodeoxyuridine (Boehringer Mannheim).
Cloned cDNAs-Collagen cDNA subclones were constructed in the plasmid vector pBS (Stratagene) for use in the nuclear run-on experiments and as templates for the synthesis of the antisense [32P]RNA probes used in the RNase protection assays. To avoid cross-hybridization caused by the sequence homology between these genes, the cloned cDNA probes were prepared from regions of minimal homology, and the homopolymer stretches in the 3' ends of these mRNAs were also avoided. The a2(I) probe and type I1 probe are both from the 3"nontranslated region ( The clone pBSCg54-344 contains a 344-bp AluI-KpnI fragment from the chick al(1) cDNA clone pCg54 (15), ligated into the HincII and KpnI sites of pBS. The clone pBSCg54-169 contains the 5'-AluI to 3"HpaII al(1) sequence from pBSCg54-344, which was transferred to pBS by ligating the HindIII-HpaII fragment from pCg54-344 into the HindIII and AccI sites of pBS. The clone pBSCg45-236 contains a 236-bp HaeIII-Hind111 fragment from the 3"nontranslated region of the a2(I) cDNA clone pCg45 (15), ligated into the HincII and HindIII sites of pBS. The clone pBSCg12-215 contains a 215-bp HinfI fragment from the 3"nontranslated region of the type I1 collagen cDNA clone pCgII-12 (16), filled in and blunt-end-ligated into the HincII site of pBS. The chicken ribosomal 27 S cDNA clone, pllD2, used in the run-on experiments, contains a 0.9-kb insert in the PstI site of pBR322 and was a gift from Drs. Y. Capetanaki and E. Lazarides.
Evidence That Each cDNA Probe Hybridizes Only with Its Cognate Collagen mRNA-To test the specificity of these cDNAs (Fig. lA, solid boxes) synthetic type I and I1 collagen mRNA sequences (Fig. lA, lines above solid boxes) that overlapped as much as possible, these cDNA probes were synthesized in vitro by phage T3 or T7 polymerases from the appropriate cDNA sequences cloned into the vector pBS. The template for the al(1) RNA was the al(1) cDNA clone pBSCAll (17), designated pBSal(I)-1.4 in Fig. 1, a gift from Dr. William Upholt that contains 1.4 kb of al(1) sequence from the nucleotide -541 to approximately +870, numbered as in Ref. (15). This sequence was constructed in Dr. Helga Boedtker's laboratory by joining sequences from pCg54 and pCg26 (15) and by removing inverted sequences and was cloned into pBS by Upholt's group (17). The a2(I) cDNA template, pBSa2(1)-1.8, contains the KpnI HindIII fragment, nucleotides -738 to +1042, from pCg45 (15), ligated into the KpnI and HindIII sites of pBS. The type I1 cDNA clone pBSal(II)-1.4, a gift from Dr. William Upholt, contains a type I1 sequence extending from -85 to the poly(A) tail a t +1330, ligated into the PstI and SmaI sites of pBS. This clone was constructed by splicing together the inserts of the type I1 collagen cDNA clones p2-34 and p2-20 a t their common BamHI sites.2 The synthetic type I mRNA sequences were made by linearizing the templates with HindIII and transcribing with T7 polymerase, and the type I1 mRNA sequence was synthesized with T 3 polymerase after linearizing the template with EcoRI. The reactions were performed in the presence of [a-32P]UTP according to the transcription kit instructions (Stratagene) except that unlabeled UTP (50 pM final concentration) was added to promote full length transcription. Full length RNA probes were gel-purified on a 6% polyacrylamide sequencing-type gel.
When each of these three 32P-labeled synthetic collagen RNAs was hybridized under the conditions used in the run-on experiments to a Southern blot containing the three types of cDNA probes (Fig. lA, solid boxes; the longer al(1) probe of 344 bp was used), the 'lP-labeled RNA hybridized only with the cognate cDNA probe and did not crosshybridize to the other two cDNA probes (Fig. 1B). These cDNA probes can therefore be used to measure each of these three collagen mRNAs without interference from the other two. Since only the type I1 synthetic mRNA extends to the poly(A) addition site (18), it is formally possible that sequences in the type I mRNAs that are 3' to the synthetic mRNA sequences could cross-hybridize to the cDNA probes, but this appears very unlikely. The al(1) synthetic RNA ends 79 nt before the first poly(A) addition site (15) and approximately 1,580 to 2,180 nt before the second set of poly(A) addition sites (19,20); the a2(I) synthetic RNA ends 44 nt before the first poly(A) addition site (15) and approximately 510 nt before the second set of poly(A) addition sites (21). The 1,500 to 2,100 nt beyond the first poly(A) addition site in the al(1) mRNA have not been sequenced, but there are, to our knowledge, no examples of such 3'-nontranslated sequences in collagen mRNAs being homologous to upstream mRNA sequences. A computer-assisted comparison of the remaining 3' sequences with the known upstream type 1 and I1 collagen mRNA sequences showed no significant homologies.
RNA Isolation and Actinomycin D Chase-The preparation of total cellular RNA was as described previously (11). To measure the degradation rates of collagen mRNAs, actinomycin D was added to cultures at a concentration of 5 pg/ml, a level which was found to inhibit RNA synthesis completely, as judged by the absence of ['HI uridine incorporation. Subsequently, at time intervals that successively doubled, five 100-mm tissue culture dishes were chilled on ice, and total cellular RNA was isolated.
RNase Protection-Total RNA from untreated and BrdUrd-treated chondrocytes was assayed for the levels of al(I), a2(I), and type I1 collagen RNAs by RNase protection (22). Antisense RNA probes were labeled by incorporation of [o~-'~P]UTP during synthesis by either T 3 polymerase (pBSCg45-236 and pBSCg12-215) or T7 polymerase (pBSCg54-169). Probes were purified using a 6% sequencing type gel and resuspended in hybridization buffer (22). Samples of total cellular RNA, 0.2 to 20.0 pg, depending on the relative abundance of the specific RNA to be measured, were lyophilized and resuspended in 30 p1 of hybridization buffer containing 50,000 cpm of antisense RNA probe and incubated overnight at 55 "C. Following hybridization, samples were digested for 30 min at 30 "C with 350 p1 of RNase digestion buffer (22) containing 4 pg/ml RNase A (Sigma) and 70 units/ml RNase T1 (Boehringer Mannheim). Reducing the amounts of both RNases to one-tenth the recommended levels (22) reduced degradation of the protected fragment and gave much cleaner autoradiograms. The termination of the RNase digestion, recovery of protected RNA hybrids, and electrophoresis on 6% polyacrylamide M. Barembaum, unpublished results. sequencing gels were as described (22). The autoradiograms from this procedure and from the run-on assays (see below) were quantitated with a laser beam scanning densitometer. In all cases, multiple exposures were scanned to ensure that the absorbance fell within the linear range of absorbance uersus radioactivity.
Run-on Transcription Assays-After chilling the cells on ice, nuclei were isolated by the procedure of Konieczny and Emerson (23), performed a t 4 "C. Nuclei, after storage at -80 "C, were assayed in run-on transcription reactions as described (23). except that heparin, 1 mg/ml, was included to inhibit initiation of transcription (24), and incubation was for 20 min a t 27 "C. The '"P-labeled run-on transcripts (23) were adjusted to 2-5 X lo6 cpm/ml in hybridization buffer: 50% formamide, 50 mM sodium phosphate, pH 7.4, 5 X SSC (1 X SSC is 0.15 M NaCI, 0.015 M sodium citrate), 1 mM EDTA, 0.2% sodium dodecyl sulfate, and 1 X Denhardt's solution (25). Hybridization was performed at 42 "C for 48 h in the presence of 2 pg of each cDNA, excised and purified free from vector sequences. The DNA was dotor slot-blotted to Genescreen (Du Pont-New England Nuclear). This level of DNA was found to be in excess relative to the RNA present in run-on reactions (data not shown). Washing and RNase treatment of the filters was essentially as described by Greenberg (26).
Polymerase Chain Reaction-The procedure was basically that of Conboy et al. (27). First strand cDNA synthesis was carried out on total cellular RNA with Moloney murine leukemia virus reverse transcriptase (Bethesda Research Laboratories) in the presence of an oligodeoxynucleotide primer containing 18 n t identical with the antisense strand of n2(I) exon 3. Duplicate first strand cDNA samples were then amplified with AmpliTaq (Perkin-Elmer) with the same 3' primer as above but with either a 5' primer identical with the first 21 n t of exon 1 or a 5' primer identical with n t +9 to +28 of exon A. Twenty-five PCR cycles were performed with a DNA Thermal Cycler (Perkin-Elmer Cetus) at 94 "C for 1 min, a t 40 "C for 1 min, and at 72 "C for 3 min. The primers were supplied by Oligos Etc., Inc. PCR products were analyzed on agarose gels consisting of 3% NuSieve (FMC Corp.) and 1% ultrapure agarose (Bethesda Research Laboratories), and the gels were stained with ethidium bromide.

Transcription Rates of Type I and Type II Collagen Genes in Untreated and BrdUrd-treated Chondrocytes-The amount
of radioactive type I and I1 collagen transcripts synthesized in nuclear run-on assays by nuclei from untreated or BrdUrdtreated chondrocytes was measured by hybridization of the transcripts to immobilized cDNAs and by densitometry of autoradiograms (Fig. 2). The hybridization to nonspecific DNA (pUC) was subtracted, and the net hybridization was corrected for the relative extent of 27 S rRNA synthesis by the two types of nuclei prior to calculating the relative transcription rates for the collagen genes in untreated chondrocytes as compared to BrdUrd-treated chondrocytes. The radioactive collagen transcripts were shown to be produced by RNA polymerase 11, since a-amanitin at 2 pg/ml completely abolished their synthesis (data not shown).
These measurements showed that the al(1) gene is transcribed a t approximately the same rate in chondrocyte nuclei as in the nuclei from BrdUrd-treated cells. The al(1) transcription rate in untreated chondrocytes was found to be 0.86 times the rate in BrdUrd-treated chondrocytes in the experiment shown in Fig. 2, after normalizing the radioactivity incorporated into the al(1) transcripts to the radioactivity incorporated into 27 S ribosomal RNA by the nuclei from the two types of cells. The al(1) cDNA probe used in this experiment, excised from the plasmid pCg54 (15), contains a long collagenous sequence, 602 nucleotides coding for the triple helical region, and 572 nucleotides coding for the COOHterminal propeptide. Similar results were obtained in subsequent experiments with the 344-bp probe that was shown to be specific for al(1) sequences (Fig. 1); a representative example is shown in the bottom horizontal line of Fig. 2. In this experiment, the relative rate of al(1) transcription by chondrocyte nuclei was found to be 1.06 times the transcription B, the cDNA probes were excised from the parent plasmids and electrophoresed on agarose gels prior to transfer to nitrocellulose by blotting. Lanes 1-3 of each gel contained, respectively, the nl(I), n2(I), and type I1 cDNA probes, which have sizes, respectively, of 344,236, and 215 bp (the longer nl(1) probe was used). Each blot was hybridized with the '*P-labeled in uitro-synthesized mRNA sequences diagrammed in A, as indicated below each panel. The conditions for hybridizing and washing the blots were exactly the same as those used in the nuclear run-on transcription assay. The autoradiograms were overexposed to allow for detection of weak cross-hybridization.  (1) gene is approximately the same in untreated and BrdUrd-treated chondrocytes. The variability in the values obtained in this assay, however, prevents us from determining a precise ratio; we can only conclude that the difference in transcription rates is relatively modest, not differing by more than a factor of 2 in either direction.
The relative transcription rate of the a2(I) gene in untreated chondrocytes was found to be 0.98 times the transcription rate in BrdUrd-treated chondrocytes in the experiment shown in Fig. 2, after correcting for the relative rates of ribosomal RNA synthesis. In this experiment, the 236-bp a2(I) cDNA of proven specificity (Fig. 1) was used. Additional assays with this cDNA, as well as earlier assays with the large 2,500-bp cu2(1) cDNA insert from the plasmid pCg45, also demonstrated that untreated chondrocyte nuclei synthesized d ( 1 ) transcripts at approximately the same rate as the nuclei of BrdUrd-shifted cells. Again, the variability in this assay does not allow us to exclude a %fold difference in these rates.
The nuclear run-on measurements of the transcription rate of the type I1 collagen gene gave an experimental result markedly different from that obtained for the type I genes. The type I1 gene showed a marked and consistent decrease in transcription rate in the nuclei from BrdUrd-treated chondrocytes. In Fig. 2, the transcription rate of the type I1 gene in the BrdUrd-treated chondrocytes is below the detectable level, but in other experiments the rate varied from 0 to 0.23 times the rate in untreated chondrocytes. Since this variability probably arises in this case not only from the variability of the assay, but also from variability in the completeness of the BrdUrd-induced shift at the time of preparing the nuclei, the exact ratio is probably not very significant. In the case of this gene, however, the nuclear run-on measurements lead to the important conclusion that the turn-off of expression of the type I1 collagen gene in BrdUrd-treated cells is achieved primarily by control at the level of transcription.
Degradation Rates of Type I and 1 1 Collagen mRNAs in Untreated and BrdUrd-treated Chondrocytes-To assess the contribution of mRNA stability in the BrdUrd-induced changes in the levels of the type I and I1 collagen RNAs ( l l ) , the degradation rates were measured for each of these RNAs in untreated chondrocytes and in BrdUrd-treated chondro-cyte~. The level of each RNA was measured by RNase protection of RNA probes synthesized from the gene-specific cDNA clones shown in Fig. 1 (the shorter al(1) cDNA was used). The degradation rates were obtained by observing the decrease in the concentration of each RNA at various time intervals after the addition of sufficient actinomycin D to inhibit further mRNA synthesis.
The results of these assays are shown as representative autoradiograms in Fig. 3 and as decay curves in Fig. 4, where the results from scanning the autoradiograms from two or more experiments are plotted. Since the d ( 1 ) gene is transcribed as rapidly in chondrocytes as in BrdUrd-shifted cells, but its steady state level in chondrocytes is only 1.5% that of shifted cells (see below), the expectation was that its degradation rate must be much faster in chondrocytes than in shifted cells. The experimental results, however (Figs. 3, A and C, and 4B), show no significant change in the degradation rate of al(1) RNA in chondrocytes, with an approximate halflife of 12 h in both types of cells. The fibroblastic decay rate, along with the low levels of this RNA, suggests that this RNA is derived from the perichondrial fibroblasts that contaminate these chondrocyte cultures. The variable levels of d ( 1 ) mRNA present in different chondrocyte RNA preparations3 support the conclusions that this RNA is not an authentic chondrocyte product. From observations in other systems (see "Discussion") it is likely that the turnover of al(1) transcripts is extremely rapid in chondrocytes and not measurable by the method employed here.
The a2(I) RNA appears to be degraded more rapidly in untreated chondrocytes (Figs. 3, B and C, and 4C), t1I2 = 5.2 h, than in BrdUrd-treated chondrocytes, tlIz = 10.4 h. While this is a relatively modest change in stability, it may account for the higher steady state level of a2(I) RNA in BrdUrdtreated chondrocytes. Since this RNA has a different 5' end in chondrocytes than in BrdUrd-treated chondrocytes (see below), it is not surprising that the two different RNA species might have different decay rates. Because of the relatively high level of the a2(I) RNA in untreated chondrocytes, the low level of contamination of these cultures by fibroblasts (3% maximum) would introduce a maximum contamination by fibroblast a2(I) RNA of 10%.
The half-life of type I1 RNA after BrdUrd treatment is not significantly different from its half-life in untreated chondrocytes (Figs. 3, D and E and 4A), with an average value of approximately 15 h. The decrease in expression of this gene in BrdUrd-treated chondrocytes is achieved, therefore, primarily by decreasing its transcription rate, rather than by changes in the stability of the mRNA.
Relative Levels of the Type I and Type 1 1 Collagen mRNAs in Chondrocytes and BrdUrd-treated Chondrocytes-The RNase protection assays used to measure decay rates provided, as a by-product, a very sensitive and specific measurement of the levels of these RNAs in the time zero samples, taken before the addition of actinomycin D. These measurements confirmed our earlier measurements that used RNA dot-blots and Northern blots (11). The level of al(1) RNA in total cellular RNA is very low in chondrocytes relative to BrdUrd-shifted cells (1.5 & 1.3%). The level of a2(I) RNA in chondrocytes, relative to shifted cells, is again found to be surprisingly high (28 & 7.5%). The marked decrease seen in the level of type I1 RNA in BrdUrd-treated cells compared with untreated cells (13.4 * 6%) is also consistent with the earlier measurements, although in the earlier measurements the type I1 mRNA was undetectable in shifted cells. It is possible that the RNA dot-blot and Northern blot procedures used previously, unlike the more sensitive RNase protection assay, would not have detected the low level of type I1 RNA in shifted cells, but it is also likely that, due to variation in the time course of the BrdUrd-induced shift in different experiments, the shift may not have gone to completion in these cultures.
The a2(I) RNA in These Chndrocytes Differs in Its 5'-End from the Fibroblast-type a2(I) mRNA-It was important to know whether the a2(I) in these chondrocytes contained the altered 5'-end previously reported by 13) to be present in embryonic chick chondrocytes, since such an RNA would not be expected to code for collagen d ( 1 ) chains (13). The presence of this altered a2(I) RNA in these chondrocytes would indicate that a change in the location of the transcription start site was a major regulatory mechanism in the turn-on of the synthesis of the a2(I) collagen chains in BrdUrd-treated cells. Since the presence of this altered RNA is influenced by the culture conditions (12) and might also be influenced by the age of the sternum from which the chondrocytes were obtained (see, for example, Ref unincubated '"P-labeled probes alone; and lane C contains "'P-labeled probes treated exactly as the samples in lanes 0-8, except that tRNA was substituted for the cellular RNA during the hybridization. Due to transcribed vector sequences, the probes are larger than their protected sequences, with sizes of 220, 262, and 249 bp, respectively, for the al(I), a2(I) and type I1 probes. In A-C, both type I probes were used to assay cul(1) and a2(I) RNAs simultaneously. The autoradiogram in B was overexposed in A to allow measurements of cul(1) RNA. The apparently lower level of RNA in the 0-time samples in C and D is due to experimental error, as is the anomalously high levels of both type I RNAs in the 2-h sample in A and B, since these results were not seen in other experiments (compare Fig. 4). In D, lane P, the band visible near 215 bp is the 220-bp cul(1) probe which was added, along with the 249-bp type I1 probe, to this lane. this a2(I) RNA species. For this purpose, the PCR technique was used to detect specifically both kinds of a2(I) RNA, the fibroblast form containing the fibroblast exons 1 and 2 spliced to exon 3, and the chondrocyte form containing the chondrocyte exon A within intron 2 (13), spliced to exon 3 (see diagram in Fig. 5). Total RNA was used to synthesize first strand cDNA from a 3' antisense primer complementary to the exon 3 sequence of a2(I) mRNA. Duplicate portions of this cDNA were then amplified with the same 3' primer, but with either a 5' primer identical with the first 21 nucleotides of exon 1 or a 5' primer identical with nucleotides +9 to +28 of exon A. If fibroblast-type a2(I) mRNA is present, an amplified fragment of 232 bp should be made in the presence of the exon 1 primer, and if chondrocyte a2(I) RNA is present, an amplified fragment of 105 bp should be made in the presence of the exon A primer, as diagrammed in Fig. 5.
In the presence of calvaria RNA and the exon 1 primer, the 232-bp PCR product expected from fibroblast-type a2(I) mRNA was made, as expected, since calvaria RNA is a rich source of fibroblast-type a2(1) mRNA (Fig. 5, lane 4 ) . In an identical reaction, except with the exon 1 primer replaced by the exon A primer, very little 105-bp PCR product was made from the same calvaria RNA (Fig. 5, lane I). In 1 primer (lane 6 ) . The BrdUrd-treated chondrocytes, unlike the untreated chondrocytes, contain predominantly the fibroblast-type a2(I) mRNA, since this RNA gives a relatively good yield of the 232-bp product in the presence of the exon I primer (lane 5 ) , but gives very little 105-bp product in the presence of the exon A primer (lane 2). These results show, therefore, that these chondrocytes contain the previously reported altered 5' end that begins at exon A in intron 2 (13). BrdUrd treatment, however, shifts the transcription start site utilized by these cells to the fibroblastic site, at the start of exon 1. Similar conclusions have been reached by means of RNase protection experiments:'

DISCUSSION
The studies reported here support the conclusion that the switch that turns off the expression of the type I1 collagen gene during the BrdUrd-induced shift is acting primarily at the transcriptional level. The nuclear run-on transcription assays show that the type I1 gene is transcribed to a much lesser extent in BrdUrd-shifted cells than in untreated cells, while the degradation rate of type I1 mRNA shows little or no change. The conclusion that the type I1 gene is controlled primarily at the transcriptional level in this system is consistent with our earlier observation that the level of type I1 mRNA decreases to very low levels in both the nucleus and cytoplasm of chondrocytes exposed to BrdUrd for 6 to 8 days chondrocytes. After the indicated time of exposure to actinomycid D, the amount of each of the three collagen mRNAs was measured by densitometry of autoradiograms (see Fig. 3). The values with the average deviation from the mean are plotted on a logarithmic scale as the percent of the time zero value, and the curve is drawn by the method of least squares. The values for al(1) and type I1 RNAs are from two experiments and for a2(I) from three experiments. Regression analysis shows no significant differences between the decay rates of the type I1 and al(1) mRNAs in the two-cell cell types ( t test for difference in regression slopes: p = 0.29 and 0.12, respectively). The tr2(I) RNA decays, however, significantly faster in untreated chondrocytes ( p = 0.037).
(11). The control of type I1 collagen gene expression has also been observed to occur primarily at the transcriptional level in a different experimental system. In this system, cultures of adherent dedifferentiated chondrocytes, derived from chick embryonic tibia, are induced to undergo a phenotypic shift in the reverse direction from that studied here, by transfer to suspension culture. As these cultures shift from the synthesis of type I to type I1 collagen, the increase in type I1 mRNA concentration is accompanied by a corresponding increase in the transcription rate of this gene (29, 30). The further analysis of the mechanism that regulates the transcription rates of the type I1 collagen gene, such as the identification of the DNA sequences and the trans-acting transcription factors that are involved, is likely to require the isolation of presently unavailable clones of the 5' region and 5"flanking sequences of the chick type I1 collagen gene.
Although the untreated chondrocytes do not produce type I collagen polypeptides, they appear to transcribe the type I genes as rapidly as BrdUrd-treated chondrocytes, which do produce type I collagen. This conclusion is based on the similar transcription rates observed in nuclear run-on assays RNA. Combinatioca of chondrocyte or fibroblast specific 5' primers and a common 3' primer give a 105-bp PCR product from the chondrocyte form of (u2(I) (bottom arrow to right of gel) and a 232bp piece from the fibroblast form (middle arrow), as shown in the diagram below the gel. The PCR products were separated by electrophoresis on an agarose gel and were visualized with ethidium bromide. T o measure the levels of the two types of a2(I) RNAs, first strand cDNA synthesis was performed with 1, 2, and 4 pg of total cellular RNA isolated from, respectively, calvaria, BrdUrd-shifted chondrocytes, and untreated chondrocytes. Lane @ X is a size marker of HaeIIIdigested 6x174. Lanes I and 4 contain the PCR products from calvaria RNA with, respectively, the chondrocyte-type 5' primer and the fibroblast-type 5' primer. Lanes 2 and 5 resemble lanes 1 and 4, except the RNA was from BrdUrd-shifted chondrocytes; lane 3 and 6 likewise resemble lanes 1 and 4, but the RNA was from untreated chondrocytes. The band of approximately 384 bp in lanes I to 3 (top arrow to right of gel) is the size expected if unspliced n2(I) RNA of either the fibroblast or chondrocyte type was amplified from the 5' exon A primer and the 3' exon 3 primer. This band is not useful, therefore, for distinguishing the two types of (u2(I) RNA. The corresponding amplified product from unspliced fibroblast-type a2(I) RNA from the 5' exon 1 primer to the same 3' primer in exon 3 would be much larger (2.7 kb) and was not seen in this experiment. with nuclei from untreated or BrdUrd-shifted chondrocytes. Two observations, in agreement with the experience of others (23), suggest that this nuclear run-on system is accurately reflecting the activity of the nuclei in intact cells rather than transcribing DNA indiscriminately. First, the marked decrease in the rate of transcription of the collagen type I1 gene in the nuclei from BrdUrd-shifted cells, relative to the rate in untreated cells, while little change is occurring in the relative rate of transcription of the type I genes, indicates that a specific gene can be selectively transcribed by this system. Second, the nuclei from chicken intestinal epithelial cells which would be expected to synthesize little or no type I or I1 collagen mRNAs, transcribed these three genes at rates that were 4 to 8% of the rates observed in the nuclei of untreated chondrocytes (data not shown).
The turn-on of type I collagen gene expression in these