The chick alpha2(I) collagen gene contains two functional promoters, and its expression in chondrocytes is regulated at both transcriptional and post-transcriptional levels.

Embryonic chick cartilages contain transcripts derived from the α2(I) collagen gene, although type I collagen is not normally found in these tissues; most of these RNAs are alternative transcripts initiating within intron 2. Use of the internal start site results in replacement of exons 1 and 2 with a previously undescribed exon and a change in the translational reading frame; thus, the alternative transcript cannot encode α2(I) collagen. We have demonstrated that production of the alternative transcript is due to activation of an internal promoter in chondrocytes and have identified a 179-base pair domain that is required for its activity. Furthermore, we have shown that the alternative transcript resulting from activation of the internal promoter turns over relatively rapidly; thus, the steady-state level of this transcript is less than predicted based on the transcription rate. The upstream promoter is only partially repressed in chondrocytes, suggesting that the lack of authentic α2(I) collagen mRNA may also be due in part to decreased mRNA stability. Thus, repression of α2(I) collagen synthesis in cartilage involves both transcriptional and post-transcriptional mechanisms. In contrast, repression of α1(I) collagen synthesis appears to be mediated primarily at the level of transcription.

Embryonic chick cartilages contain transcripts derived from the ␣2(I) collagen gene, although type I collagen is not normally found in these tissues; most of these RNAs are alternative transcripts initiating within intron 2. Use of the internal start site results in replacement of exons 1 and 2 with a previously undescribed exon and a change in the translational reading frame; thus, the alternative transcript cannot encode ␣2(I) collagen. We have demonstrated that production of the alternative transcript is due to activation of an internal promoter in chondrocytes and have identified a 179base pair domain that is required for its activity. Furthermore, we have shown that the alternative transcript resulting from activation of the internal promoter turns over relatively rapidly; thus, the steady-state level of this transcript is less than predicted based on the transcription rate. The upstream promoter is only partially repressed in chondrocytes, suggesting that the lack of authentic ␣2(I) collagen mRNA may also be due in part to decreased mRNA stability. Thus, repression of ␣2(I) collagen synthesis in cartilage involves both transcriptional and post-transcriptional mechanisms. In contrast, repression of ␣1(I) collagen synthesis appears to be mediated primarily at the level of transcription.
Normal skeletal development requires precisely regulated expression of the genes encoding type I collagen, the major collagen produced by both prechondrogenic mesenchymal cells and osteoblasts. As mesenchymal cells differentiate into cartilage-producing chondrocytes, they stop producing type I collagen and initiate synthesis of several cartilage-specific collagens (reviewed in Refs. 1 and 2). Type I collagen is a heterotrimer containing two ␣1(I) and one ␣2(I) subunits. In cells and tissues that produce type I collagen, the genes encoding these subunits are coordinately regulated (3)(4)(5)(6). We previously identified an unusual molecular mechanism that mediates the cessation of ␣2(I) collagen production in cartilage. Embryonic chick chondrocytes contain transcripts derived from the ␣2(I) collagen gene (7), despite the fact that these cells do not synthesize type I collagen. These transcripts initiate at an internal start site within intron 2 (8,9), rather than at the previously identified site at the beginning of exon 1 (10) (Fig. 1A). Use of this internal start site results in replacement of exons 1 and 2 with a previously undescribed exon (exon A) and a change in the translational reading frame; this unusual RNA cannot encode ␣2(I) collagen, since the potential open reading frames are out of frame with the collagen coding sequence. Hereafter we will refer to the transcript initiating at the internal start site as the alternative transcript, in contrast to the authentic ␣2(I) collagen mRNA, which initiates at the beginning of exon 1 and encodes the ␣2 subunit of type I collagen.
We initially predicted (9) that a developmentally programmed change from the upstream promoter to the presumptive internal promoter for transcription of the ␣2(I) collagen gene may be responsible for preventing synthesis of ␣2(I) collagen in chondrocytes. However, in the experiments described below, we demonstrate that repression of ␣2(I) collagen synthesis in chondrocytes is a complex process involving both transcriptional and post-transcriptional mechanisms. Specifically, we have shown that the appearance of the alternative transcript in chondrocytes is due to activation of an internal promoter in these cells. In contrast, the absence of the authentic ␣2(I) collagen mRNA in chondrocytes appears to be due only in part to repression of the upstream promoter and may also be due in part to decreased stability of the authentic ␣2(I) collagen mRNA in these cells. Finally, we have demonstrated that transcription of the ␣1(I) collagen gene is repressed in chondrocytes; thus, the mechanisms that prevent synthesis of ␣1(I) and ␣2(I) collagen in these cells appear to be different.

EXPERIMENTAL PROCEDURES
Cell Culture-Chondrocytes isolated from lower (caudal) sternal cartilage of 18-day-old chick embryos or from vertebral cartilage of 12-dayold embryos were grown for 5-7 days in suspension culture as described previously (11,12); the small number of contaminating adherent fibroblasts were discarded. Fibroblasts from skin of 12-day-old embryos were isolated and cultured as described previously (7,13).
RNase Protection Assays-The probe used for RNase protection assays was prepared from a cloned cDNA, pG.␣2alt (14); it includes the last 14 nucleotides of exon A, exons 3-6, and the first 5 nucleotides of exon 7 cloned into the EcoRI and BamHI sites of pGEM2, as diagrammed in Fig. 1B. The probe was prepared by transcription of the EcoRI-linearized plasmid with T7 RNA polymerase in the presence of [␣-32 P]UTP and annealed to 1.5 g of total cellular RNA (isolated according to Ref. 15) from cultured sternal chondrocytes and skin fibroblasts; unhybridized RNAs were removed by digestion with RNases A and T1 according to standard procedures (16), and protected RNAs were fractionated on a 6% denaturing polyacrylamide gel. The probe protects a 216-nucleotide fragment of the alternative transcript, which contains the exon A sequences, and a 202-nucleotide fragment of the authentic ␣2(I) collagen mRNA, which does not contain the exon A sequences. * This work was supported in part by National Institutes of Health Grants GM28840 and AR41042 (to S. L. A.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  Promoter Constructs and Transfection Analyses-The reporter plasmid used for all transfection analyses was p0CATntLPA3Zf(ϩ), hereafter referred to as p0CAT, which was developed and generously provided by James C. Alwine (University of Pennsylvania, Philadelphia, PA). This vector contains the chloramphenicol acetyltransferase (CAT) 1 gene (17), preceded by the multiple cloning site from pGEM3Zf(ϩ) and followed by the SV40 late polyadenylation signal (LPA) (18). The SV40 small t splice site included in the original CAT vectors (17) was removed, since it was not necessary for high levels of expression, and the SV40 LPA was included, because it routinely provided higher levels of expression in chondrocytes than the SV40 early polyadenylation signal used in the original CAT vectors. 2 Promoter fragments derived from the chick ␣2(I) collagen gene were inserted into the multiple cloning site.
Most promoter fragments were derived from the plasmid p323-1, a genomic 5.7-kilobase pair EcoRI DNA fragment extending from Ϫ1644 (the first nucleotide of exon 1 is ϩ1) to intron 4 (19); this plasmid was generously provided by Benoit de Crombrugghe (M. D. Anderson Cancer Center, Houston, TX). We have made numerous corrections and additions to the published sequence (20), which have resulted in changes in the nucleotide numbers; the corrected sequence of nucleotides ϩ1 to ϩ2397 has been submitted to GenBank (accession number U62128). The ␣2(I) collagen gene sequences contained in the various promoter constructs are summarized in Table I. The plasmid pC1CAT (hereafter referred to as C1), which contains the full-length internal promoter, extends from an AccI site at ϩ130 in exon 1 to an MscI site at ϩ2397 in exon A; the fragment was blunt-ended with the Klenow fragment of DNA polymerase I and inserted into the SmaI site in the multiple cloning site of p0CAT. Convenient restriction sites (SmaI, ϩ846; SacI, ϩ1158; PvuII, ϩ1764; BglII, ϩ1904; DraI, ϩ2022; BsaAI, ϩ2201; and PstI, ϩ2288) were used to create C2 through C11, in which various domains were deleted. The full-length upstream promoter in plasmid pF1CAT (F1), which extends from Ϫ1064 to ϩ110 in exon 1, was recloned from pHO1000 (21) into p0CAT; pHO1000 was also provided by Benoit de Crombrugghe.
Primary floating chondrocytes were collected from the culture medium by centrifugation, trypsinized and replated at a density of 5 ϫ 10 6 cells/10-cm plate in the presence of 4 units/ml hyaluronidase, as described previously (22). Primary skin fibroblasts were trypsinized and replated at a density of 2.25 ϫ 10 6 cells/10-cm plate. After 24 h, cells were incubated with plasmid DNAs in the presence of LipofectAMINE (Life Technologies, Inc.) for 5 h according to the manufacturer's recommendations. Transfections included 20 g of the plasmid C1, containing the full-length internal promoter, or equal molar amounts of the deletion constructs C2-C11 or the full-length upstream promoter F1; 0.4 g of pCh110, expressing ␤-galactosidase under control of the SV40 early promoter (23), was included in each transfection as an internal control; pCh110 was generously provided by Tom Kadesch (University of Pennsylvania, Philadelphia, PA). Cells were harvested 68 h after addition of DNA, and lysates were prepared by standard techniques (17,24). ␤-Galactosidase assays were performed according to standard protocols (25); CAT activity was then assayed by incubation for 2 h at 37°C (17,24) using volumes of cell extracts containing equal amounts of ␤-galactosidase activity. Films were scanned using a Molecular Dynamics densitometer and quantified using ImageQuant 3.3 software. The activity of each construct is expressed as percent acetylation/0.1 A 420 unit of ␤-galactosidase activity and represents the average of three to five independent experiments with cells isolated from different batches of embryos; each construct was assayed in duplicate in each experiment. The activities of various constructs were compared pairwise using an unpaired t test.
In control transfections with pCh110, similar transfection efficiencies were observed in chondrocytes and fibroblasts, as determined by in situ staining (26). Furthermore, extracts from equal numbers of chondrocytes and fibroblasts provided similar amounts of ␤-galactosidase activity, indicating that the SV40 early promoter displays comparable activity in the two cell types.
RNA Half-life Determination-RNA polymerase II transcription was inhibited by treatment of primary chondrocytes and skin fibroblasts with 1 or 5 g/ml actinomycin D for varying lengths of time; treatments were initiated at different times, so all cultures were harvested simultaneously and all were in culture for the same length of time. Total cellular RNAs were prepared, and the relative amounts of ␣2(I) collagen RNA were determined by RNase protection as described above using the probe prepared from pG.␣2alt (14) or by Northern hybridization using nick-translated pCg45 (31), an ␣2(I) collagen cDNA generously provided by Helga Boedtker, formerly of Harvard University (Cambridge, MA).

RESULTS
We demonstrated previously (8,9) that embryonic chick chondrocytes, which do not synthesize type I collagen, contain an alternative transcript of the ␣2(I) collagen gene in which exons 1 and 2 are replaced by exon A, a previously undescribed exon located within intron 2 ( Fig. 1A). In contrast, cells of mesenchymal origin that synthesize type I collagen (including osteoblasts, tendon and skin fibroblasts, skeletal and smooth muscle myoblasts, and prechondrogenic mesenchymal cells) contain the authentic ␣2(I) collagen mRNA, which initiates at the beginning of exon 1 and encodes the ␣2(I) subunit of type I collagen. Exon A is spliced out of this mRNA. The cell typespecific nature of these transcripts is illustrated by the RNase protection assay in Fig. 1C, in which a single-stranded RNA probe complementary to a portion of the alternative transcript (diagrammed in Fig. 1B) (14) was used to discriminate between the alternative transcript and the authentic ␣2(I) collagen mRNA. RNA from cultured chondrocytes protects a 216-nucleotide RNA fragment, which includes the exon A sequences, indicating that these cells contain predominantly the alternative transcript. In contrast, RNA from skin fibroblasts protects a 202-nucleotide fragment, excluding the exon A sequences, indicating that these cells contain almost exclusively the authentic ␣2(I) collagen mRNA. Analysis of many RNA preparations has indicated that the amount of the alternative transcript in chondrocytes is 10 -25% of the amount of the authentic ␣2(I) collagen mRNA in skin fibroblasts (Refs. 7-9 and Fig. 1C).
The Presumptive Internal Promoter of the Chick ␣2(I) Collagen Gene Is Active in Chondrocytes-The presence in chondrocytes of an alternative transcript that appeared to initiate within intron 2 (9) suggested the presence of an internal promoter in the ␣2(I) collagen gene that is used preferentially in a Domain D has been divided into four subdomains, D1-D4.
chondrocytes. The major transcription initiation site in chondrocytes (at ϩ2353) is preceded by an imperfect TATA box (TGTAAA) and a CCAAT box, located 25 and 73 nucleotides, respectively, upstream from the major start site (9), which could comprise elements of such a chondrocyte-specific internal promoter (Fig. 2). To determine whether this is a functional promoter in chondrocytes, a 2266-bp DNA fragment extending from ϩ130 in exon 1 through ϩ2397 in exon A was introduced into the multiple cloning site of the reporter plasmid p0CAT to create the plasmid C1. This plasmid directed a significant level of CAT activity in cultured chondrocytes (Fig. 2), 5.2 times higher than a construct containing the same DNA fragment in the reverse orientation (C1R) (p Ͻ 0.001). These results indicate that this DNA fragment constitutes a functional promoter and appears to contain most or all of the sequences that are necessary for expression in these cells. We used convenient restriction sites to construct a series of 5Ј-end deletion mutants of the internal promoter (Fig. 2, constructs C2-C5), to begin to identify the regions that are important for transcriptional activity in chondrocytes. The promoter domains removed by the successive 5Ј-end deletions have been designated A-D; domain E, which contains only 65 bp of DNA preceding exon A, including the TATA-like element, is present in all constructs. Deletion of domain A in construct C2 resulted in a statistically significant 2.4-fold increase in CAT activity (p Ͻ 0.005), suggesting the presence of a negative element in this region. The additional removal of domain B did not have a significant effect on CAT activity (compare C3 with C2). However, the further removal of domain C (in construct C4) resulted in a 63% decrease in activity relative to C3 (p Ͻ 0.05), suggesting the presence of a weak positive element in this domain. When domain D was also removed, a dramatic 95% decrease in activity was observed (p Ͻ 0.05; compare C5 with C4), suggesting that this domain contains one or more strong positive elements. Thus, domain D appears to be very important for the function of the internal promoter.
We subsequently constructed a series of internal deletions, to examine more precisely the role of domain D in transcriptional activity of the internal promoter. Deletion of domain D alone from the full-length promoter decreased CAT activity by 93% (p Ͻ 0.001; compare C6 with C1). The importance of this domain was confirmed by analysis of additional constructs containing various portions of the full-length promoter (C7 and C8); the constructs missing domain D were inactive regardless of which other domains were present. Furthermore, construct C4, which contains only domains D and E, displayed a relatively high level of CAT activity in chondrocytes (comparable to the full-length promoter C1), indicating that domains D and E alone (587 bp) constitute a functional promoter in these cells; the sequence of these domains is shown in Fig. 3. Thus, domain D is not only essential for activity of the internal promoter in chondrocytes, in combination with domain E, it appears to be sufficient for activity in these cells.
A 179-bp Region of Domain D Is Required for Activity of the Internal Promoter in Chondrocytes-Since deletion of domain D abolished promoter function in all contexts we examined, and since a promoter containing only domains D and E retained a relatively high level of activity in chondrocytes, we used available restriction sites to construct additional 5Ј-end deletions of C4, sequentially deleting subdomains D1-D4 (Fig. 4, constructs C9 -C11). Deletion of D1 resulted in a decrease in CAT activity (compare construct C9 with C4), suggesting the presence of a positive element in this region; however, the decrease was not statistically significant. Deletion of D2 (compare construct C10 with C9) had no additional effect on CAT activity. However, removal of an additional 179 bp in D3 rendered the promotor inactive, resulting in a 91% decrease in activity (p Ͻ 0.02; compare C11 with C10), indicating that subdomain D3 contains one or more elements that are essential for promoter activity in chondrocytes. Construct C11, which contains the CCAAT box (in subdomain D4) and the TATA-like element (in domain E), is inactive, indicating that, while these elements may be important for transcriptional activity, they are clearly not sufficient without additional upstream elements.
The Internal Promoter Is More Active in Chondrocytes than In Fibroblasts-To determine whether the internal promoter is chondrocyte-specific, we compared the ability of construct C4 to direct CAT activity in chondrocytes with that in skin fibroblasts. The activity of C4 in skin fibroblasts was 4.5 Ϯ 2.6% (n ϭ 4), only 33% of its activity in chondrocytes (shown in Fig.  2), a statistically significant decrease (p Ͻ 0.05).
Thus, we have demonstrated that the presumptive internal promoter of the ␣2(I) collagen gene is not only functional, but is also more active in chondrocytes than in fibroblasts, suggesting that its function is cell type-specific. In addition, we have identified a region of 179 bp that is essential for function of the internal promoter in chondrocytes.

The ␣2(I) Collagen Gene Is Transcribed at a High Rate in Chondrocytes, but the Resulting Transcripts Are Unstable-
The RNase protection assay in Fig. 1C demonstrates that the amount of the alternative transcript in chondrocytes is significantly less than the amount of the authentic ␣2(I) collagen mRNA in skin fibroblasts. This reduced steady-state RNA level could be due to relatively low transcriptional activity of the ␣2(I) collagen gene in chondrocytes or to reduced stability of the alternative transcript. We used nuclear runoff transcription assays to determine whether the transcriptional activity of the gene was lower in chondrocytes than in skin fibroblasts. Chondrocyte and fibroblast nuclei were pulse-labeled, and equal amounts of radioactive RNA were hybridized to identical slot blots containing genomic clones representing different regions of the ␣2(I) collagen gene (Table II). The transcription rate of each region of the gene in chondrocytes is expressed relative to the rate in skin fibroblasts, in which the gene is actively transcribed from the upstream promoter to form the authentic ␣2(I) collagen mRNA. The clones containing exons 5-8, 9 -24, 25-31, and 52 provide an estimate of the relative transcription rate throughout the gene, since these exons are present in both the alternative transcript and the authentic ␣2(I) collagen mRNA. The relative transcription rate in chondrocytes in this experiment ranged from 82 to 112% of the rate in skin fibroblasts, indicating that the overall transcription rate of the ␣2(I) collagen gene in chondrocytes is similar to that in skin fibroblasts and is much higher than would be predicted based on the steady-state RNA levels shown in Fig. 1C and previously published experiments (7-9).
The relatively high transcription rate of the ␣2(I) collagen gene in chondrocytes observed in the nuclear runoff transcription assays (Table II) appeared to be inconsistent with the low steady-state level of the alternative transcript in these cells (Fig. 1C), suggesting that transcripts of the ␣2(I) collagen gene may be less stable in chondrocytes than in fibroblasts. To determine whether decreased RNA stability is in fact responsible for the low level of transcripts derived from the ␣2(I) collagen gene in chondrocytes, we inhibited RNA polymerase II transcription with actinomycin D and initially used Northern hybridization analysis to determine the amount of RNA remaining at various times after the initiation of drug treatment. In two independent experiments, one of which is shown in Intron and exon sequences are shown in lowercase and uppercase letters, respectively, and labeled above the DNA sequence in bold type. The boundaries of the subdomains D1-D4 and E are indicated by brackets, and the subdomains are labeled in bold italics above the DNA sequence. The polypyrimidine tract is shaded, the nucleotides that comprise the CCAATT and TGTAAA boxes are in bold type, and potential transcription factor binding sites are labeled above their nucleotide sequences, which are underlined.

FIG. 4. 5-End deletion analysis of construct C4 identifies a 179-bp domain that is required for transcriptional activity of the internal promoter in chondrocytes.
Experimental conditions were identical to those described in the legend for Fig. 2. D1-D4 are subdomains of domain D defined by the 5Ј-end deletion mutants. tendon fibroblasts (32). In contrast, the half-life of the RNA in chondrocytes was less than 3 h. The RNase protection assay shown in Fig. 5B indicates that the alternative transcript is the predominant RNA species derived from the ␣2(I) collagen gene in chondrocytes in the absence or presence of actinomycin D. These results suggest that the alternative transcript in chondrocytes is intrinsically less stable than the authentic ␣2(I) collagen mRNA in skin fibroblasts.
Transcription from the Upstream Promoter Is Repressed in Chondrocytes-The experiments described above demonstrate that the internal promoter of the ␣2(I) collagen gene is active in chondrocytes and is significantly less active in skin fibroblasts. Thus, activation of this promoter appears to be responsible for the appearance of the alternative transcript in chondrocytes, but not in other cells of mesenchymal origin, as we originally predicted (8,9).
The small amount of the authentic ␣2(I) collagen mRNA in chondrocytes (Figs. 1C and 5B) suggested that the activity of the upstream promoter may be repressed in these cells, concomitant with activation of the internal promoter. We used nuclear runoff transcription assays to test this prediction (Table II); if all transcription of the ␣2(I) collagen gene in chondrocytes initiates at the internal promoter, radioactive transcripts from chondrocyte nuclei should not hybridize to a clone containing only exon 1. Much less hybridization to the exon 1-containing clone was observed with radioactive transcripts from chondrocyte nuclei than from fibroblast nuclei, suggesting that the upstream promoter is significantly repressed in chondrocytes. However, the amount of transcription from the upstream promoter in chondrocytes remained unexpectedly high, 19% of the amount in skin fibroblasts, suggesting that, while the absence of the authentic ␣2(I) collagen mRNA in chondrocytes may be due primarily to decreased transcriptional activity of the upstream promoter, it is likely to be due in part to decreased stability of the transcripts initiating at the upstream promoter as well.
To ensure that the continued transcription of the ␣2(I) collagen gene from the upstream promoter was not due to loss of differentiated properties of the cultured chondrocytes, we analyzed the transcription rate of the gene encoding type II collagen, the major cartilage-specific collagen (Fig. 6). Transcription of this gene was extremely high in chondrocyte nuclei, as expected, indicating that these cells are well differentiated chondrocytes; transcription of the type II collagen gene was essentially undetectable in cultured skin fibroblasts. As an additional control, we demonstrated that transcription of the sequences encoding 27 S ribosomal RNA was essentially identical in chondrocyte and fibroblast nuclei (Fig. 6). Thus, the absence of the authentic ␣2(I) collagen mRNA in chondrocytes appears to be due in part to repressed transcription from the upstream promoter, but may also be due in part to decreased stability of the transcripts initiating at the upstream promoter, since little authentic ␣2(I) collagen mRNA can be detected in these cells.
To determine whether the reduced transcriptional activity of the upstream promoter in chondrocytes could be reproduced in transient transfection assays, we introduced a 1174-bp DNA fragment extending from -1064 to ϩ110 in exon 1 into p0CAT to create the plasmid F1, which was tested for promoter function in cultured chondrocytes (Fig. 2) and skin fibroblasts. This promoter has been demonstrated previously to direct CAT activity in stable transformants of NIH 3T3 cells (16). The activity of F1 in chondrocytes was much higher than expected, based on the small amount of the authentic ␣2(I) collagen mRNA in these cells (Refs. 8 and 9; Figs. 1C and 5B) and on the amount of hybridization to exon 1 sequences in the nuclear runoff transcription assay in Table II; its activity was similar to that of the full-length internal promoter C1. Furthermore, its activity in chondrocytes was comparable to that in skin fibroblasts   (27); the ␣2(I) collagen genomic clone p031-1, containing exons 9 -24 (19); pHrA, a genomic DNA fragment encoding human 28 S ribosomal RNA (30); and the type II collagen cDNA pCs2 (29). A, the amount of hybridization of fibroblast transcripts to the type I collagen cDNAs and the cloned 28 S ribosomal RNA gene was assigned a value of 1.0 and the amount of hybridization in chondrocytes was calculated relative to that value. B, the amount of hybridization of chondrocyte transcripts to the cloned type II collagen cDNA was assigned a value of 1.0, and the amount of hybridization in fibroblasts was calculated relative to that value. We have not attempted to directly compare the transcription rates in the ␣1(I) and ␣2(I) collagen genes because of the large differences in size and base composition of the respective clones.
(data not shown). These results indicated that construct F1 contains most or all of the DNA sequences that are necessary for transcriptional activity; however, either additional DNA sequences may be required to repress transcription in chondrocytes or the chondrocyte culture conditions used for the transfections do not permit appropriate regulation of this promoter.
Transcription of the ␣1(I) Collagen Gene Is Repressed in Chondrocytes-In tissues and cells that produce type I collagen, transcription of the ␣1(I) and ␣2(I) collagen genes is coordinately regulated; for example, the ␣1(I) collagen gene is transcribed at 2-4 times the rate of the ␣2(I) collagen gene in cultured human skin fibroblasts (3,4). Since the ␣2(I) collagen gene is transcribed at a high rate in chondrocytes, comparable to that in skin fibroblasts (Table II), we used nuclear runoff transcription assays to determine whether the ␣1(I) collagen gene is also transcriptionally active in these cells. The transcription rates of the ␣1(I) and ␣2(I) collagen genes were calculated relative to the rates in skin fibroblasts.
Transcription of the ␣1(I) collagen gene was quite low in chondrocytes, about 25% of the transcription rate in cultured skin fibroblasts (Fig. 6). This is comparable to the relative transcription rate of the ␣2(I) collagen gene from the upstream promoter in chondrocytes (Table II), suggesting coordinate regulation of these two promoters. However, in this experiment the ␣2(I) collagen gene was transcribed at a higher rate in chondrocytes than in skin fibroblasts. Since the majority of the transcription in chondrocytes appears to initiate at an internal start site (Ref. 9; Figs. 1C and 5B), these results indicate that transcription of the ␣2(I) collagen gene from the internal promoter is not coordinately regulated with transcription of the ␣1(I) collagen gene. DISCUSSION We have shown above and in our previous studies (7)(8)(9)13) that the repression of type I collagen synthesis in cartilage, an important event in chondrogenesis, involves a complex interplay of transcriptional and post-transcriptional mechanisms. Transcription of the gene encoding the ␣1(I) collagen subunit is significantly repressed in chondrocytes (Fig. 6); in contrast, transcription of the ␣2(I) collagen gene remains high in these cells (Fig. 6, Table II). Repression of ␣2(I) collagen synthesis involves repression of the upstream promoter of the ␣2(I) collagen gene and activation of an internal promoter, resulting in production of an alternative transcript that cannot encode a collagen protein due to an accompanying change in the translational reading frame (9). In previous studies we showed that the alternative transcript of the ␣2(I) collagen gene is likely to be translated in chondrocytes, since it is associated with a small number of puromycin-releasable ribosomes (13). However, it has not yet been possible to identify a protein product of this RNA, suggesting that it may be present in very small amounts. The relative instability of the alternative transcript, resulting in low steady-state RNA levels, may reflect a requirement for precise regulation of either the RNA itself or its protein product.
Characterization of the Internal Promoter of the ␣2(I) Collagen Gene-We have demonstrated that an internal promoter (construct C4) consisting of 587 bp of 5Ј flanking sequence (domains D and E) is sufficient for transcriptional activity in chondrocytes (Fig. 2); the sequence of this internal promoter is shown in Fig. 3. We have identified a 179-bp subdomain of this promoter (D3) that is required for activity in chondrocytes. This subdomain lies 152 bp upstream from the major internal transcription start site at the beginning of exon A. It contains potential binding sites for numerous transcription factors, which may play a role in the chondrocyte-specific activation of this internal promoter (Fig. 3). For example, there are three potential binding sites for the basic/helix-loop-helix class of transcription factors, which play important roles during myogenesis (33), and five sites for binding of GATA-1 (34) and NF-E1 (35), members of the GATA class of transcription factors, which regulate hematopoiesis (36), as well as other developmental programs. In addition, there is a binding site for NF-IL6 (37) (C/EBP␤; Ref. 38), which plays a role in adipogenesis (39) and transactivates the mouse ␣1(I) collagen gene (40), as well as two binding sites for TCF1 (41), a member of the Ets family of transcription factors (42). The clustering in this relatively small region of binding sites for several transcription factors known to be involved in other cell differentiation pathways suggests that some of these proteins may be involved in the developmental control of the internal promoter of the ␣2(I) collagen gene. Experiments designed to identify the transcription factors that regulate this promoter are in progress. Interestingly, while the CCAAT box and the TATA-like element may be important for transcriptional activity of the internal promoter, as they are in many other genes, they are clearly not sufficient, since construct C11, which contains both of these elements, is inactive (Fig. 4).
The internal promoter of the ␣2(I) collagen gene, which is preferentially utilized in chondrocytes, bears little resemblance to other cartilage-specific promoters that have been characterized. The type II collagen genes require enhancer sequences in the first intron to confer chondrocyte-specific activity (43)(44)(45)(46). There is little similarity between the type II collagen enhancers and the sequences required for activity of the internal promoter of the ␣2(I) collagen gene in chondrocytes. The link protein gene is regulated in part by a glucocorticoid response element and an AT-rich element (47,48), neither of which appear in the internal promoter of the ␣2(I) collagen gene. Thus, a variety of combinations of transcription factors and their cognate DNA sequences appear to confer chondrocyte-specific transcriptional activity.
During the course of these studies, constructs similar to our C1 and C3 were described by Wang and Lukens (49). The deletion mutant similar to C3 displayed reduced activity in chondrocytes relative to the full-length promoter, suggesting the existence of positive elements in the deleted region (approximately comparable to our domains A and B). In contrast, C3 displayed an increase in activity compared to the full-length promoter, suggesting the existence of negative elements in domains A and/or B. This discrepancy could be due either to differences between the constructs or to differences between the cells used for analysis of the promoters. We have used a relatively homogeneous population of immature chondrocytes from lower sternal cartilage of 18-day-old chick embryos, while Wang and Lukens (49) used cells from whole sternum of 14day-old embryos, a mixed population in terms of chondrocyte maturation. We have demonstrated recently that the abundance of the alternative transcript changes dramatically during development of the sternum (14).
The Alternative Transcript Is Less Stable than the Authentic ␣2(I) Collagen mRNA-The authentic ␣2(I) collagen mRNA in skin fibroblasts displayed a half-life of about 9 h (Fig. 5), comparable to that previously reported for this mRNA in human skin fibroblasts (3) and chick tendon fibroblasts (32). In contrast, the alternative transcript in chondrocytes displayed a half-life of less than 3 h, suggesting that it is intrinsically less stable than the authentic ␣2(I) collagen mRNA. This decreased stability could be due to the unique structure of the alternative transcript. For example, exon A may contain sequences that decrease RNA stability. Alternatively, the intrinsic instability of the alternative transcript may be due to the presence of the small open reading frame; the presence of an in-frame translation termination codon in exon 7 could be perceived as a premature stop codon, which may result in decreased RNA stability (50).
Regulation of the ␣1(I) and ␣2(I) Collagen Genes Differs in Chondrocytes-In normal tissues and cells that produce type I collagen, transcription of the ␣1(I) and ␣2(I) collagen genes is coordinated primarily by transcriptional mechanisms; the ␣1(I) collagen gene is transcribed at 2-4 times the rate of the ␣2(I) collagen gene in cultured human fibroblasts (3,4). The promoters of the ␣1(I) and ␣2(I) collagen genes appear quite different, and the DNA sequences and transcription factors responsible for this coordinate transcription have not yet been identified, although several transcription factors have been identified that appear to be involved in the regulation of both genes (51). In addition, the promoters of the human ␣1(I) and ␣2(I) collagen genes each inhibit the activity of the other promoter in competition cotransfection experiments (52), suggesting requirements for common transcription factors.
Chondrocytes are the only normal cell type described to date in which the ␣2(I) collagen gene is transcribed at a high rate in combination with a low rate of ␣1(I) collagen gene transcription, although a transformed Syrian hamster fibroblast cell line that produces only ␣2(I) collagen subunits has been described (53). The nuclear runoff transcription assay in Fig. 6 illustrates significantly repressed transcription of the ␣1(I) collagen gene in chondrocytes, while the ␣2(I) collagen gene continues to be transcribed at a high rate. Our results appear to differ from those of Askew et al. (54), who reported that the ␣1(I) collagen gene was transcribed at equal rates in control and bromodeoxyuridine-treated chondrocytes. This difference could be due to several factors. First, there could be differences in ␣1(I) collagen gene transcription at different stages of cartilage maturation, as we have observed for ␣2(I) collagen gene transcription (14). Furthermore, in our experiments the transcription rate of the ␣1(I) collagen gene has been compared with that in skin fibroblasts, rather than with that in bromodeoxyuridinetreated chondrocytes.
Despite the disparity in overall transcription rates of the genes encoding type I collagen in chondrocytes, it seems likely that the upstream promoter of the ␣2(I) collagen gene is repressed coordinately with the single promoter of the ␣1(I) collagen gene in these cells. In contrast, the internal promoter of the ␣2(I) collagen gene is utilized at a high rate in chondrocytes and is not coordinately regulated with the ␣1(I) collagen gene promoter.
In summary, during chondrogenesis, the authentic ␣2(I) collagen mRNA is replaced by an alternative transcript of the ␣2(I) collagen gene that does not encode a collagen protein. This change in the RNA is responsible for the cessation of ␣2(I) collagen synthesis during chondrogenesis. The appearance of the alternative transcript in chondrocytes is due to the activation of an internal promoter, and we have identified a 179-bp domain that is essential for the activity of this promoter. In contrast, the cessation of ␣1(I) collagen synthesis during chondrogenesis appears to be due simply to repressed transcription of the ␣1(I) collagen gene. Thus, the mechanisms preventing synthesis of ␣1(I) and ␣2(I) collagen in chondrocytes appear to differ.