Use of a new rat chondrosarcoma cell line to delineate a 119-base pair chondrocyte-specific enhancer element and to define active promoter segments in the mouse pro-alpha 1(II) collagen gene.

We show that a new rat chondrosarcoma (RCS) cell line established in long-term culture from the Swarm tumor displayed a stable differentiated chondrocyte-like phenotype. Indeed, these cells produced the collagen types II, IX, and XI and alcian blue-stainable cartilage-specific proteoglycans, but no type I or type III collagen. To functionally characterize their chondrocytic nature, the cells were stably transfected with a type II collagen/beta geo chimeric gene which confers essentially perfect chondrocyte-specific expression in transgenic mice. RCS cells expressed both beta-galactosidase and G418 resistance, in comparison with similarly transfected 10T1/2 and NIH/3T3 fibroblasts which did not. These cells were then used to perform a systematic deletion analysis of the first intron of the mouse type II collagen gene (Col2a1) using transient expression experiments to determine which segments stimulated expression of a luciferase reporter gene in RCS cells but not in 10T1/2 fibroblasts. Cloning of two tandem copies of a 156-base pair (bp) intron 1 fragment (+2188 to +2343) in a construction containing a 314-bp Col2a1 promoter caused an almost 200-fold increase in promoter activity in RCS cells but no increase in 10T1/2 cells. DNase I footprint analysis over this 156-bp fragment revealed two adjacent protected regions, FP1 and FP2, located in the 3'-half of this segment, but no differences were seen with nuclear extracts of RCS cells and 10T1/2 fibroblasts. Deletion of FP2 to leave a 119-bp segment decreased enhancer activity by severalfold, but RCS cell specificity was maintained. Further deletions indicated that sequences both in the 5' part of the 119-bp fragment and in FP1 were needed simultaneously for RCS cell-specific enhancer activity. A series of deletions in the promoter region of the mouse Col2a1 gene progressively reduced activity when these promoters were tested by themselves in transient expression experiments. However, these promoter deletions were all activated to a similar level in RCS cells by a 231-bp intron 1 fragment that included the 156-bp enhancer. The RCS cell-specific activity persisted even if the Col2a1 promoter was replaced by a minimal adenovirus major late promoter. This 231-bp intron 1 fragment also had strong enhancing activity in transiently transfected mouse primary chondrocytes. Our experiments establish the usefulness of RCS cells as an experimental system for studies of the control of chondrocyte-specific genes, provide an extensive delineation of segments in the Col2a1 first intron involved in chondrocyte-specific activity, and show that promoter sequences are dispensable for chondrocyte specificity.

ever, these promoter deletions were all activated to a similar level in RCS cells by a 231-bp intron 1 fragment that included the 156-bp enhancer. The RCS cell-specific activity persisted even if the Col2a1 promoter was replaced by a minimal adenovirus major late promoter. This 231-bp intron 1 fragment also had strong enhancing activity in transiently transfected mouse primary chondrocytes. Our experiments establish the usefulness of RCS cells as an experimental system for studies of the control of chondrocyte-specific genes, provide an extensive delineation of segments in the Col2a1 first intron involved in chondrocyte-specific activity, and show that promoter sequences are dispensable for chondrocyte specificity.
The differentiation of mesenchymal cells into chondrocytes results in the synthesis and secretion of a series of proteins characteristic of the extracellular matrix of cartilages. These include types II, IX, and XI collagens, the proteoglycan aggrecan, link protein, and cartilage matrix protein, the expression of which is part of a genetic program of differentiation specific for chondrocytes (1,2). Earlier studies on the biosynthesis of these cartilage components have used mainly primary chondrocytes, but the phenotypic instability of these cells, which varies with culture conditions, has made experiments on chondrocyte differentiation difficult (3)(4)(5)(6). A small number of chondrocyte cell lines that maintain at least part of the chondrocyte phenotype during culture have been isolated recently (7)(8)(9)(10)(11)(12). Such stable chondrocytic cell lines are essential for achieving further understanding of the genetic program of chondrocyte differentiation.
Because the gene for type II collagen is expressed at very high levels in chondrocytes (2,13,14), it is an excellent candidate for study of chondrocyte differentiation. Humans and mice with mutations in type II collagen often show severe cartilage defects and skeletal malformations (15)(16)(17). During embryonic development, the gene becomes active in cartilage anlages as early as the time of mesenchymal condensation preceding cartilage formation. Hence, type II collagen can be considered an early and abundantly expressed marker of chondrocyte differentiation. Although the gene for type II collagen is also expressed transiently in some extrachondrogenic sites during embryonic development (14), this occurs at much lower levels, and the significance of this extrachondrocytic expression is not understood, in contrast to the richly documented role played by this molecule in cartilage function.
Our long-term goal is to identify and characterize the DNAbinding proteins involved in the chondrocyte-specific activity of the type II collagen gene. In this study we sought to better delineate sequences within the mouse type II collagen gene (Col2a1) 1 that can confer chondrocyte-specific expression to reporter genes. Previous transient expression experiments using primary chick chondrocytes had identified a 620-bp chondrocyte-specific enhancer in the first intron of the rat Col2a1 gene (3). Subsequently, similar experiments showed that a 380-bp segment within this 620-bp fragment activated expression of a reporter gene 11-fold in chondrocytes and that a 260-bp subfragment activated expression 6-fold (4). A point mutation in this 260-bp fragment decreased this activity by about two-thirds. In other experiments, two silencer elements identified in the rat Col2a1 promoter were proposed to participate in inactivating the gene in nonchondrocytic cells (18).
In the present study we first examined the expression of a series of molecular markers to characterize the chondrocytic phenotype of a new rat chondrosarcoma (RCS) cell line derived from the Swarm chondrosarcoma tumor after long-term culture. The cells were further characterized by comparing them with 10T1/2 and NIH/3T3 fibroblasts after stable transfection with a mouse Col2a1 chimeric DNA construction and by establishing that the Col2a1-derived transgene was selectively expressed in chondrosarcoma cells. We then used these cells in transient expression experiments to delineate minimal sequences in the first intron of the mouse Col2a1 gene that were needed for chondrocyte-specific enhancer activity and to determine active sequences in the promoter of this gene.

MATERIALS AND METHODS
Obtention of the Rat Chondrosarcoma Cell Line-RCS cells were obtained from nodules of the transplantable rat Swarm chondrosarcoma (19) that were formed after long-term suspension culture. These nodules which initially contained primarily chondrocyte-like cells but also few fibroblast-like cells were dispersed with trypsin and collagenase. The fibroblastic components were removed by repetitive monolayer culture from which a population of less adherent chondrocyte-like cells were isolated. These cells were continuously propagated in serial monolayer culture for several years and constitute the RCS cell line used in the present study. The cells show a doubling time of less than 24 h and display a completely stable phenotype under standard culture conditions.
Other Cells-10T1/2, NIH/3T3, and C 2 C 12 cell lines were purchased from the American Type Culture Collection. Primary chondrocytes were prepared from rib cartilages of 1-to 3-day-old mice or from articular cartilages of 3-week-old rats as described previously (20). Rat osteosarcoma cells (ROS 17/2.8) were kindly provided by Dr. W. T. Butler (Dept. of Biological Chemistry, University of Texas Health Science Center, Houston, TX).
Northern Blot Analysis-Isolation of total RNAs, Northern blot analysis, and preparation of labeled DNA probes were performed as described previously (12,20).
Staining with Alcian Blue-Staining of cartilage-specific proteoglycans in cell culture monolayers was done with alcian blue at pH 1 as described (21).
Analysis of Collagens-Labeling of collagens synthesized by cells in culture was performed at confluence for 40 h in fresh culture medium supplemented with 15 Ci/ml of L-[2,3,4,5-3 H]proline (Amersham), 50 g/ml ascorbic acid, 100 g/ml 4-aminopropionitrile fumarate, 1 mM cysteine, and 1 mM pyruvate. Culture media were collected, centrifuged, and acidified with 0.5 M acetic acid. Monolayers containing the cells and their extracellular matrix were harvested in 0.5 M acetic acid. Pepsin at 0.25 mg/ml was added to these culture samples, and incubation was carried out for 4 h at 15°C. Pepsin was inhibited by raising the pH to 8 with NaOH and 50 mM Tris, and by addition of 10 g/ml pepstatin. Aliquots of the samples were digested with 25 g/ml purified bacterial collagenase (type VII; Sigma) in the presence of 2 mM CaCl 2 , 1 mM phenylmethylsulfonyl fluoride, and 10 g/ml leupeptin for 2 h at 37°C. Collagens in all samples were precipitated with 30% ammonium sulfate, washed with 80% ethanol, and subjected to sodium dodecyl sulfatepolyacrylamide gel (6%) electrophoresis after boiling for 2 min in loading buffer supplemented or not with 100 mM dithiothreitol. Prestained protein molecular weight standards were from Life Technologies, Inc. Gels were treated with Amplify (Amersham), and x-ray films were exposed for 1-3 days.
Permanent Transfections with a Col2a1-␤geo Chimeric Transgene-Two different plasmids were used for permanent transfections. The plasmid pPGK-HYG contains a hygromycin (HYG) resistance gene driven by a phosphoglycerate kinase-1 (PGK) gene promoter (22). The plasmid pCol2a1-␤geo was constructed in pBluescript II KS(ϩ/Ϫ) (Stratagene) by cloning a 7.2-kb fragment of the mouse Col2a1 gene upstream of a SA-␤geo-bpA cassette. 2 The Col2a1 fragment contained 3 kb of promoter sequences, exon 1, and 3034 bp of intron 1. It was released from a cosmid vector (23) in two pieces by digestion with NotI/ScaI and ScaI/XbaI and reconstituted in Bluescript by cloning at the NotI and XbaI sites. The Col2a1 translation initiation codon (ATG) was changed (to CTG) by PCR mutagenesis. The SA-␤geo-bpA cassette (24) encodes a protein (␤geo) with both ␤-galactosidase and neomycin transferase activities. A splice acceptor (SA) is present at its 5Ј end and the bovine growth hormone gene polyadenylation signal (bpA) at its 3Ј end. The cassette was released from the pSA-␤geo plasmid (24) by digestion with SpeI and Asp-718 and cloned at corresponding sites in pBluescript.
Permanent transfections were performed by the modified DNA-calcium phosphate coprecipitation method described by Chen and Okayama (25). 5 ϫ 10 5 cells were plated in 50-cm 2 tissue culture dishes and transfected with 1 g of pPGK-HYG alone or together with 20 g of pCol2a1-␤geo. Hygromycin-resistant cells were selected by treatment for 10 days with 200 g/ml hygromycin B. Southern analysis was performed on the pools of hygromycin-resistant cells to verify integration of the Col2a1 transgene. This was done by using the 3-kb Escherichia coli ␤-galactosidase gene (lacZ) as a probe. Selection of cells expressing the neomycin resistance gene was performed by treatment with G418 at 300 or 500 g/ml as indicated. ␤-Galactosidase activities were measured with a chemiluminescent assay kit (Tropix, Bedford, MA), and protein was assayed with the Bradford reagent (Bio-Rad).
DNA Constructions Used in Transient Transfections-Col2a1-luciferase chimeric genes were constructed in the pA 3 LUC vector (26). For some constructions, a modified version of this plasmid (pLuc4) was used in which a multiple cloning site was introduced immediately upstream of the luciferase gene (27). P1, P2, and P3 promoter fragments and I1 to I10 intron 1 fragments of the mouse Col2a1 gene were cloned by bluntend ligation. P4 and P5 promoters were generated by PCR with oligonucleotide primers containing a HindIII recognition site at their 5Ј end. I11 to I17 intron 1 fragments were generated by PCR using two oligonucleotide primers, one containing a BamHI recognition site at its 5Ј end and the other a BglII site at its 5Ј end. Some of these intron fragments were cloned as two or four tandem repeats as indicated in the figure legends. In order to form these repeats, the PCR products were digested with BamHI and BglII, ligated, and treated again with BamHI and BglII to keep only multimers ligated head to tail. Dimers or tetramers were isolated by electrophoresis in agarose gel and ligated to the appropriate vectors as indicated in the figures. Intron fragments were cloned in their natural 5Ј to 3Ј orientation. To construct the padMLPL, a 56-bp fragment of the adenovirus type 2 major late promoter (Ϫ46 to ϩ10) was isolated by EcoRI and SmaI digestion from the pUC19 plasmid, a HindIII linker was added to the SmaI site, and the EcoRI site was blunt-ended. This promoter fragment was cloned in the pA 3 LUC vector between the blunt-ended KpnI and HindIII sites. The sequences of all constructs were verified by dideoxy DNA sequencing.
Transient Transfections-DNA transfection experiments of RCS, 10T1/2, and C 2 C 12 cells were carried out by electroporation of cells harvested near midlog growth phase. RCS cells were detached from plastic and freed of their abundant extracellular matrix by treatment for 1 h at 37°C with 1.5 mg/ml bacterial collagenase (collagenase D, Boehringer Mannheim) in Dulbecco's modified Eagle's medium, followed by two washes in phosphate-buffered saline. 10T1/2 and C 2 C 12 1 The abbreviations used are: Col2a1, pro-␣1(II) collagen gene; RCS, rat chondrosarcoma cells; ␤geo, a fusion protein of Escherichia coli ␤-galactosidase and neomycin transferase; bp, base pair(s); kb, kilobase(s); PCR, polymerase chain reaction; FP1 and FP2, 5Ј-and 3Ј-DNase I footprinted regions in the 231-bp enhancer fragment, respectively. cells were harvested with trypsin/EDTA. 2 ϫ 10 6 cells were distributed in 0.4-ml electroporation cuvettes in 200 l of phosphate-buffered saline with 10 g of Col2a1-luciferase plasmids, and 2 g of pSV2␤gal plasmid (28) was used as an internal control of transfection efficiency. RCS cells were electroporated at 250 V, 500 F at 4°C, and 10T1/2 and C 2 C 12 cells at 300 V, 500 F at room temperature, using a Gene Pulser apparatus (Bio-Rad). Cells were replated in 20-cm 2 plastic dishes, medium was changed within 24 h, and the cultures continued for another 24 -48 h. Primary chondrocytes were transfected by DNA-calcium phosphate coprecipitation (25) within 24 h after release from cartilage. Five million cells were plated per 50-cm 2 dishes and transfected using 18 g of Col2a1-luciferase plasmids, and 6 g of pSV2␤gal plasmid. Twentyfour h before harvesting the primary chondrocytes, 50 g/ml ascorbic acid was added to the culture media in order to stimulate the expression of the differentiated phenotype (20), together with 1 mM cysteine and 1 mM pyruvate. Luciferase assays were performed according to Wood et al. (26) using a Monolight 2010 luminometer (Analytical Luminescence Laboratory, San Diego, CA) and D-luciferin (Sigma) as substrate. ␤-Galactosidase activity was measured by a colorimetric assay with resorufin-␤-galactopyranoside as substrate (Boehringer Mannheim) as described (29) or by chemiluminescence (see above). Luciferase activities were normalized for transfection efficiency by dividing the luciferase relative luminescent units by the ␤-galactosidase activities expressed either in absorbance milliunits (colorimetric assay) or in relative chemiluminescent units ϫ 10 Ϫ3 .
DNase I Protection Assay-Nuclear extracts from RCS and 10T1/2 cells were prepared according to Dignam et al. (30) except that leupeptin (10 g/ml) and pepstatin (10 g/ml) were added to all buffers. For DNase I protection assays on the coding DNA strand, plasmid pI9BS, which consisted of a 231-bp Col2a1 enhancer fragment cloned in the BamHI site of pBluescript II KS(ϩ/Ϫ), was cut with NotI, and its 3Ј ends were radiolabeled using [␣-32 P]dCTP and the Klenow fragment. The labeled fragment containing the enhancer was released from the vector by HindIII digestion, isolated by electrophoresis in an 8% polyacrylamide gel, and eluted in 10 mM Tris-HCl buffer, pH 8, containing 1 mM EDTA and 100 mM NaCl. Labeling of the noncoding strand was performed with [␥-32 P]ATP and T4 polynucleotide kinase after cleavage of the same Bluescript recombinant plasmid by NotI. About 10 fmol of the labeled fragments were mixed without or with 10 -15 g of nuclear extract proteins in the presence of 0.2 g of poly(dA-dT) and 50 mM NaCl or KCl in a buffer containing 20 mM HEPES, pH 7.9, 0.1 mM EDTA, 0.5 mM dithiothreitol, and 5% glycerol. The 50-l assay mixtures were incubated for 15 min at room temperature prior to addition of DNase I (Worthington Biochemical Corp.) for 60 s at room temperature. 0.1 ng of DNase I was used for free DNA and 2 ng for DNA preincubated with protein. The reactions were terminated by addition of 10 l of a stop buffer containing 5% sodium dodecyl sulfate, 125 mM EDTA, and 13.2 g of tRNA. Protein was extracted with phenol/chloroform, and DNA was precipitated by ethanol, resuspended in 3 l of loading buffer containing 80% deionized formamide and 10 mM NaOH, and heated at 80°C for 3 min. The same number of radioactive counts was loaded in each lane of a sequencing gel. G ϩ A sequencing was done according to Maxam and Gilbert (31) and run in the same gel as standard. Fig. 1 shows that RCS cells exhibited a polygonal or round shape in both nonconfluent (A) and confluent (B) cultures. They secreted a very abundant extracellular matrix which stained intensely with alcian blue (Fig. 1C), indicating the presence of large amounts of cartilage-specific proteoglycans. Fig. 2A shows that the cells contained large amounts of type II procollagen RNA (lane 2), although the hybridization signal was clearly less intense than in primary mouse rib chondrocytes (lane 1) and primary rat articular chondrocytes (not shown). The cells also contained RNA for ␣2(IX) procollagen, another characteristic chondrocyte-specific marker, showing a signal that was stronger in intensity than that found for primary mouse chondrocytes (Fig. 2B). No RNA for type X collagen, a characteristic marker of hypertrophic chondrocytes, could be detected in the RCS cells, whereas it was found in mouse rib primary chondrocytes, which presumably contained a substantial proportion of hypertrophic cells (Fig. 2C). Only a weak signal for type X collagen transcript was detected in the rat chondrocytes, most probably due to their articular cartilage origin. Importantly, no RNA for ␣1(I) procollagen was observed in RCS cells unlike what is seen in primary chondrocytes after a short time in culture (Fig. 2C). In contrast, the mouse pro-␣1(I) probe hybridized strongly with RNA isolated from rat primary chondrocytes and rat osteosarcoma cells. No RNA for either osteocalcin or osteopontin was detected, but a probe for matrix Gla protein gave a strong signal (Fig. 2D). Fig. 3 shows that RCS cells, like mouse primary chondrocytes, synthesized large amounts of type II collagen whose ␣1 chain characteristically migrated slightly slower than the ␣1(I) collagen chain, as well as type XI collagen whose ␣1 and ␣2 chains migrated more slowly than the ␣1(II) collagen chain, whereas its ␣3 chain comigrated with the ␣1(II) collagen FIG. 1. Morphology of the RCS cells and staining with alcian blue. A, nonconfluent culture of RCS cells 24 h after plating. B, confluent culture 3 days after plating. C, alcian blue staining of a confluent culture. The extracellular matrix synthesized by the RCS cells is produced as a thick viscous layer of material that easily detached during the staining procedure. This might account for the nonuniform alcian blue staining. A and B are dark-field photographs. C is a bright-field photograph. Bars in A, B, and C correspond to 100 m. chain. Both collagen types accumulated more abundantly in the cell and matrix layers than in the culture media. RCS cells also produced a ladder of pepsin-resistant polypeptides with an apparent molecular mass comprised between 150 and 250 kDa under nonreducing conditions. These polypeptides which were sensitive to bacterial collagenase most likely correspond to type IX collagen chain trimers undigested or partially digested by pepsin. After reduction, they gave rise to smaller collagenasesensitive peptides with apparent molecular mass of 86, 70, and 60 kDa which likely correspond to intact or partially digested type IX collagen chains (32). Trimers of ␣1(III) chains and ␣2(I) chains, which were the major pepsin-resistant polypeptides in 10T1/2 cell cultures, were undetectable in the RCS cell samples.

Phenotype of the Rat Chondrosarcoma Cell Line-
Altogether, these results indicated that the RCS cells express a number of characteristic chondrocyte differentiation markers, including type II, IX, and XI collagens, and cartilagespecific proteoglycans. The absence of type X collagen RNA suggests that they were frozen in a state before transition into hypertrophic chondrocytes. Importantly, the complete absence of type I collagen indicates that these cells, unlike other chondrocytic cell lines previously described (7-11), do not express a partial mesenchymal or fibroblastic phenotype.
Stable Transfection of RCS, 10T1/2, and NIH/3T3 Cells with a Chimeric Col2a1 Construct-To further characterize RCS cells, we compared these cells with 10T1/2 and NIH/3T3 fibroblasts after stable transfection with a Col2a1-␤geo chimeric construction containing 3 kb of 5Ј-flanking sequences of the mouse pro-␣1(II) collagen gene, exon 1, and about 3 kb of intron 1, linked to the ␤geo reporter gene (Fig. 4A). The ␤geo gene encodes a fusion polypeptide that includes both E. coli ␤-galactosidase and neomycin resistance activities. In other experiments, we showed that this construction directed essentially perfect chondrocyte-specific ␤-galactosidase activity in transgenic mice. 2 RCS cells, 10T1/2, and NIH/3T3 fibroblasts were stably transfected with this construction together with an expression vector containing a hygromycin resistance gene driven by a phosphoglycerate kinase promoter (pPGK-HYG).
After a first selection of the cells with hygromycin, a Southern hybridization with the DNA of the pooled colonies showed a signal of about equal intensity for the three cell types, indicating that the Col2a1 transgene had been stably integrated (data not shown). To determine whether the Col2a1 transgene A pool of cell clones that had stably integrated the Col2a1-␤geo construct was obtained as described under "Materials and Methods" for each of the RCS (squares), 10T1/2 (circles), and NIH/3T3 (triangles) cell lines. Expression of the ␤geo gene in these cell pools was determined by measuring the growth of the cells during 9 days of culture in the presence of G418. 3 ϫ 10 5 cells of each pool were plated in 50-cm 2 dishes, 500 g/ml of G418 was added to the medium, and the number of cells excluding trypan blue was estimated at 3-day intervals in a hemocytometer after release of the cells from plastic with trypsin/EDTA. Three similar independent experiments were carried out with essentially the same results, i.e. only the stably transfected RCS cells survived in the presence of G418. C, expression of ␤-galactosidase. Pools of either RCS, 10T1/2, or NIH/3T3 cell clones were obtained that had stably integrated either a pPGK-HYG construct only or both the pPGK-HYG and the Col2a1-␤geo constructs (see "Materials and Methods"). ␤-Galactosidase was assayed after selection with hygromycin (ϩHYG) and, also, for RCS cells transfected with both pPGK-HYG and Col2a1-␤geo constructs, after selection with hygromycin, followed by 1 week of selection with 500 g/ml G418 (ϩHYG/ϩG418). ␤-Galactosidase activities are presented in relative chemiluminescent units per g of protein. The activity of ␤-galactosidase in cells transfected with pPGK-HYG alone corresponds to the endogenous level of ␤-galactosidase activity in each cell type; these endogenous ␤-galactosidase levels were higher in RCS cells than in 10T1/2 and NIH/3T3 fibroblasts. Only the RCS cells stably transfected with the Col2a1-␤geo construct contained levels of ␤-galactosidase above the endogenous levels. *, ␤-galactosidase assays could not be performed since these cells did not survive in the presence of G418. Similar results were obtained in two independent experiments. was expressed, the pooled colonies were first tested for resistance to neomycin. As shown in Fig. 4B, RCS cells continued to grow in the presence of G418, whereas 10T1/2 and NIH/3T3 fibroblasts rapidly died. In control experiments, RCS cells that did not contain the Col2a1-␤geo transgene died within a few days in G418 selection medium (data not shown). Results similar to those in Fig. 4B were also obtained when a lower G418 concentration (350 instead of 500 g/ml) was used (data not shown).
We then tested expression of ␤-galactosidase in RCS, 10T1/2, and NIH/3T3 cells stably transfected with pPGK-HYG alone or with both pPGK-HYG and Col2a1-␤geo (Fig. 4C). After selection with hygromycin, no difference in ␤-galactosidase activity was observed whether 10T1/2 and NIH/3T3 fibroblasts were transfected with one or with both plasmids. RCS cells contained endogenous ␤-galactosidase activity that was considerably higher than that of 10T1/2 and NIH/3T3 cells. However, the RCS cell ␤-galactosidase activity was much higher in cells containing the Col2a1-␤geo construct than in control cells. After G418 selection, these levels increased by an order of magnitude in RCS cells transfected with the Col2a1-␤geo construct. This increase was probably due to the G418 selection which eliminated cells that were transfected by the hygromycin vector only and probably also the cells that expressed levels of ␤geo below the G418 threshold. These functional experiments indicated that RCS cells were able to support the activity of a stably transfected chondrocyte-specific chimeric transgene whereas in 10T1/2 and NIH/3T3 cells the transgene was inactive. These results are consistent with the notion that, unlike 10T1/2 and NIH/3T3 fibroblasts, RCS cells contain the transcription factors needed for the activity of a chondrocyte-specific transgene.
Initial Identification of Sequences in Intron 1 of the Col2a1 Gene That Show Enhancer Activity Specifically in Chondrosarcoma Cells-RCS cells were then used for transient expression experiments with the DNA constructions shown schematically in Fig. 5. An 886-bp fragment of the mouse Col2a1 gene con-taining 767 bp of 5Ј-flanking promoter sequences and 119 bp of exon 1 was placed upstream of the luciferase reporter gene in the pA 3 LUC vector. Various deletion fragments of intron 1 were cloned approximately 800 bp 5Ј to the promoter. This 800-bp vector sequence contains three poly(A) sites to stop transcription from upstream vector sequences and from intron 1 sequences. These constructions were transfected into RCS cells, 10T1/2 fibroblasts, and C 2 C 12 myoblasts.
The results of the luciferase assays are presented as multiples of the luciferase values obtained with the promoter alone (Fig. 5). Activity of the promoter fragment alone was about 5 to 15 times higher in 10T1/2 cells and about 30 times higher in C 2 C 12 myoblasts than in RCS cells. A 3-kb intron 1 fragment (ϩ306 to ϩ3325) produced a modest 3-fold increase in RCS cells, but a 5-to 7-fold decrease in 10T1/2 and C 2 C 12 cells. This intron fragment is similar to the one that was used in conjunction with a 3-kb mouse Col2a1 promoter segment in stable transfections of Fig. 4 and also in transgenic mice. 2 In contrast to the modest increase in activity seen in transiently transfected RCS cells, mice harboring the latter construction showed strong expression of the transgene that was completely chondrocyte-specific. This difference could reflect differences between transient expression in tissue culture cells and stable integration in transgenic mice or differences in the placement of the intron sequences, which were located in their natural position 3Ј of the transcription start site in the construction used for transgenic mice. In other experiments we have shown that when a 300-bp promoter instead of a 3-kb promoter was used, no change occurred in chondrocyte-specific expression in transgenic mice. 2 The 5Ј-half of the 3-kb intron fragment (ϩ306 to ϩ1880) showed no enhancing activity, whereas the 3Ј-half (ϩ1637 to ϩ3251) showed 10 times more activity than the 3-kb fragment in RCS cells (Fig. 5). It is possible that the 5Ј-half of the 3-kb fragment contained inhibitory sequences that decreased the activity of the 3-kb intronic enhancer. Further fragmentation of this active segment showed that a 953-bp fragment (ϩ1394 to ϩ2346) was an active enhancer in RCS In the pI9ЈP1L and pI10P1L constructs, two copies of the intron fragments were cloned head to tail, as indicated between brackets (2ϫ). All these DNA constructs were transiently transfected into 10T1/2, C 2 C 12 , and RCS cells. Luciferase activities were measured in cell extracts and normalized for transfection efficiency (see "Materials and Methods"). Transcriptional activation of the promoter induced by intron fragments is given relative to the activity obtained with the pP1L construct (considered as 1) and is presented as the average Ϯ S.E. of all transfection assays performed, with the number of assays indicated in parentheses. cells but that a 908-bp fragment (ϩ2344 to ϩ3251) located 3Ј of this segment was much less active. Further work was focused on smaller fragments. A 546-bp fragment (ϩ1880 to ϩ2425) showed a 14-fold increase specifically in chondrosarcoma cells but smaller fragments were clearly less active. However, we found that when two copies of a 465-bp fragment (ϩ1880 to ϩ2344) were cloned tandemly a 44-fold increase in activity occurred in chondrosarcoma cells whereas practically no increase occurred in the other two cell types. Similarly, two tandem copies of a 231-bp fragment (ϩ2113 to ϩ2343) produced a 100-fold increase in activity in RCS cells, but much less in 10T1/2 and C 2 C 12 cells. It is possible that several modular elements in the first intron contribute to the strength of the chondrosarcoma cell-specific enhancer and that, when smaller fragments are generated, duplication of one or more of these modular elements is necessary to compensate for the loss of other active elements.
DNase I Footprint Studies Using the 231-bp Intron Fragment-The active 231-bp enhancer fragment was used in DNase I footprint experiments to compare the patterns of protection by nuclear proteins of RCS and 10T1/2 cells. Fig. 6A shows two adjacent areas of protection on the upper DNA strand, indicated as FP1 and FP2, which were located toward the 3Ј end of the 231-bp fragment. No other clear footprints were detected elsewhere in this fragment. Footprints covering the same area of DNA were observed on the lower strand, and again no other footprints were seen elsewhere in this fragment (data not shown). The 3Ј part of FP1 includes part of a 10-bp sequence (CACAATGCAT, see Fig. 6B) which in the rat Col2a1 gene was previously implicated in chondrocyte-specific en-hancer activity and was reported to be a binding site for a chondrocyte-enriched protein (4). However, as shown in Fig.  6A, we found no difference in FP1 between 10T1/2 and RCS cell nuclear extracts even when different concentrations of nuclear extracts were used. Similarly, FP2 showed no differences with these two extracts. Since the 5Ј part of FP1 and the protected DNA of FP2 were very GC-rich, it is possible that the footprints were due to GC-binding proteins in extracts of both cell types. Interestingly, most of the region corresponding to FP2 is not present or not well conserved in the human and rat genes (Fig. 6B).
Additional Deletions of the 231-bp Enhancer Fragment-We then asked whether a 97-bp subfragment (ϩ2248 to ϩ2344) of the 231-bp fragment, containing the two protected regions identified by DNase I footprinting, had enhancer activity. When cloned as 2 tandem copies in the 886-bp promoter vector, this 97-bp fragment failed to activate the promoter in transient expression experiments using RCS cells (Fig. 7A). Similarly either four tandem copies of a fragment containing FP1 or four tandem copies of a fragment containing FP2 were inactive (Fig.  7A). Thus, each of the two segments containing the individual footprints as well as a 97-bp fragment containing the two footprints were unable by themselves to act as chondrocytespecific enhancers.
In subsequent experiments, a shorter 433-bp (Ϫ314 to ϩ119) promoter fragment was used to replace the 886-bp (Ϫ767 to ϩ119) promoter. The activity of this promoter was reduced about 5 to 10 times compared to the 886-bp promoter in both FIG. 6. Protein binding and nucleotide sequence analysis of the 231-bp Col2a1 enhancer. In A, a DNase I protection assay was performed on the coding strand of the Col2a1 231-bp enhancer fragment in the absence of nuclear extracts (no N.E.) or after preincubation with nuclear extracts from 10T1/2 or RCS cells, as indicated at the top of the lanes. The first lane shows a G ϩ A sequencing of the same DNA strand. Two regions protected by nuclear proteins are bracketed and labeled FP1 and FP2. In B, the coding strand of the 231-bp enhancer is aligned with analogous regions in the human and rat type II collagen genes. Only 96 bp of the corresponding rat sequence are available from the GenBank TM data base. Only bases in the human and rat sequences that differ from the mouse sequence are shown; identical bases are indicated by a dot. Gaps, marked by a dash, have been added to maximize alignment. A decamer sequence in the rat gene, that was postulated to have a role in chondrocyte-specific expression, is shown in italics. The regions footprinted in DNase I protection assays are underlined in the mouse sequence and labeled FP1 and FP2. The numbers ϩ2113 to ϩ2343 above the mouse sequence indicate the distance in nucleotides from the mouse Col2a1 transcription start site. Repeated DNA sequencings have indicated that nucleotide 2280 in the mouse sequence is a C residue rather than a T residue as reported earlier (23) .   FIG. 7. Further deletions of the RCS-specific 231-bp Col2a1 enhancer. In A, the pP1L and pI9ЈP1L plasmids are as described in Fig. 3. Plasmids pI11P1L, pI12P1L, and pI13P1L were obtained by cloning 2 (2ϫ) or 4 (4ϫ) tandem copies of small Col2a1 intron sequences in the BglII site of pP1L. In B, plasmid pP2L was obtained by cloning a 443-bp fragment of the mouse Col2a1 promoter (box with grid) in the SmaI site located just upstream of the luciferase gene (box labeled LUC) in the pA 3 LUC vector. Plasmids pI9ЈP2L and pI14P2L to pI17P2L contain two copies (2ϫ) of different fragments of the Col2a1 first intron (boxes with stripes) cloned head to tail in the BglII site of the pP2L vector. In A and B, the sizes of the promoter and intron fragments are indicated inside their respective boxes, and the position of their first and last nucleotides relative to the Col2a1 transcription start site is indicated at the bottom left and right sides of the boxes. All these DNA constructions were transiently transfected in RCS cells (A and B) and 10T1/2 cells (B). Luciferase activities were measured in cell extracts and normalized for transfection efficiency (see "Materials and Methods"). Transcriptional activation of the promoter induced by intron fragments is given relative to the activity obtained with the corresponding promoter constructs (considered as 1) and is presented as the average Ϯ S.E. of all transfection assays performed, with the number of assays indicated in parentheses.
10T1/2 and RCS cells (data not shown), despite the fact that this shorter promoter removed two potential silencer elements that were identified in the rat gene (18). Sequence comparison of the rat, human, and mouse genes failed to show conservation of these silencer elements (33). When two tandem copies of the 231-bp fragment (ϩ2113 to ϩ2343) were cloned upstream of this promoter, an 88-fold activation was seen in transient expression experiments using RCS cells and no activation using 10T1/2 cells (Fig. 7B). Additional deletions of the 231-bp enhancer fragment were tested in constructions containing two tandem copies of these subfragments. A 5Ј deletion producing a 156-bp fragment resulted in even greater activity than the 231-bp fragment in RCS cells (Fig. 7B). When 37 bp at the 3Ј end of this 156-bp fragment were deleted, which removed most of FP2, the remaining 119-bp fragment was still active in RCS cells albeit at a severalfold lower level. A further 30-bp deletion of 3Ј sequences in this 119-bp fragment, which removed FP2 and the 3Ј part of FP1, including the previously described decamer sequence, leaving an 89-bp fragment (ϩ2188 to ϩ2276), resulted in complete loss of activity in RCS cells. Similar results were obtained with a construction that deleted both 5Ј and 3Ј sequences (ϩ2232 to ϩ2285).
Hence, the RCS-specific enhancer was located in a 156-bp fragment in which both 5Ј and 3Ј sequences appeared to be required simultaneously for activity since neither the 3Ј 97-bp segment, containing FP1 and FP2, nor the 5Ј 89-bp segment was able to show enhancer activity by themselves. Deletion of the 3Ј-DNA corresponding to most of FP2 in the 156-bp fragment significantly decreased the activity of the enhancer, but RCS specificity was maintained indicating that the elements necessary for chondrocyte enhancer activity were contained in a 119-bp segment.
Deletions in the Mouse Col2a1 Promoter-To determine whether sequences in the promoter were needed for RCS cellspecific expression and also to identify sequences in the promoter that were important for its activity, a series of promoter deletions was generated (Fig. 8). Like the Ϫ767 to ϩ119 promoter (Fig. 5), the Ϫ314 to ϩ119 and the Ϫ314 to ϩ7 promoters by themselves were considerably more active in 10T1/2 cells than in chondrosarcoma cells (Fig. 8). Like a deletion from Ϫ767 to Ϫ314, which reduced promoter activity severalfold (data not shown), deletions to Ϫ159 and to Ϫ89 further reduced promoter activities in both cell types, suggesting that the deleted segments contain binding sites for transcriptional activators. Cloning of two tandem copies of the 231-bp intron fragment produced a large increase in activity with each of the different promoter deletions in chondrosarcoma cells (Fig. 8). The absolute activities of these different promoter-enhancer constructions were similar, resulting in a several thousandfold stimulation with the Ϫ159 and Ϫ89 promoters. Hence, the promoter sequences upstream of Ϫ89 are not needed to achieve optimal activation by the 231-bp enhancer segment in chondrosarcoma cells. In contrast, the 231-bp enhancer had no effect on promoter activity in 10T1/2 cells with the Ϫ314 promoter and only a small effect with the two smaller promoters (Fig. 8).
To determine whether any sequences in the Ϫ89 Col2a1 promoter were needed for chondrosarcoma cell specificity, the Col2a1 promoter was replaced by a short segment of the adenovirus major late promoter (Ϫ46 to ϩ10) (Fig. 9). In the absence of enhancer, this promoter was essentially inactive in either RCS or 10T1/2 cells. Addition of the 231-bp enhancer increased promoter activity in RCS cells to levels similar to those observed with constructions in which Col2a1 promoter fragments were present. This enhancement was much less in 10T1/2 cells (Fig. 9). This experiment showed that no specific element in the Col2a1 promoter was required to produce selec-tive activation by the 231-bp enhancer in RCS cells.
Transient Transfections in Mouse Primary Chondrocytes-We wanted to verify whether the 231-bp intron fragment which showed high RCS cell-specific activity was also active in primary chondrocytes. The plasmid containing a 95-bp Col2a1 promoter (Ϫ89 to ϩ6) and the plasmid containing the same promoter with two tandem copies of the 231-bp intron fragment were transfected into mouse rib primary chondrocytes less than 24 h after release from cartilage at a time when the cells still express an essentially differentiated phenotype (Ref. 17 and Fig. 3). The construction containing the 231-bp intron fragment caused a thousandfold increase of promoter activity (Table I). Hence, this 231-bp fragment has a strong enhancing activity in both RCS cells and primary chondrocytes, but is inactive in fibroblasts. iments has unique properties not previously described in other chondrocytic cell lines. The advantage of this cell line which has been established in long-term culture from cells obtained from the rat Swarm chondrosarcoma tumor is that its chondrocytic phenotype is completely stable in standard tissue culture conditions in contrast to that of primary chondrocytes. Importantly, as for chondrocytes in intact animals, these RCS cells did not contain any type I collagen but expressed the chondrocyte-specific type II, IX, and XI collagens. In addition, they also synthesized large amounts of alcian blue-stainable cartilagespecific proteoglycans. The absence of ␣1(X) collagen RNA suggested that these cells were frozen in the pathway of chondrocyte differentiation at a stage preceding transition to hypertrophic cells. These cells were then stably transfected with a chimeric gene in which sequences of the mouse Col2a1 gene drove expression of the ␤geo reporter gene, a construction that displayed essentially perfect chondrocyte-specific expression in transgenic mice. The stably transfected RCS cells exhibited clear G418 resistance in contrast to similarly transfected 10T1/2 and NIH/3T3 fibroblastic cells which were unable to survive after G418 addition. This experiment strongly suggested that, like chondrocytes in intact embryos, the chondrosarcoma cells contained transcription factors needed to selectively activate the Col2a1-␤geo chimeric gene, whereas 10T1/2 and NIH/3T3 fibroblasts did not.

Rat Chondrosarcoma Cells Display a Stable Chondrocytic
Delineation of Chondrosarcoma-specific Enhancer Sequences in the Mouse Col2a1 Gene-The RCS cells were then used in transient expression experiments to systematically dissect which of the segments of the mouse Col2a1 gene first intron were able to stimulate the expression of a luciferase reporter gene in chondrosarcoma cells but not in 10T1/2 fibroblasts and C 2 C 12 myoblasts. We also sought to analyze the effects of successive deletions in the promoter of this gene and to determine whether any Col2a1 promoter sequences were needed for the effects of the intron sequence to appear.
Our experiments demonstrated that the presence of two tandem copies of a 156-bp segment of intron 1 in a construction containing a 314-bp Col2a1 promoter segment provided an almost 200-fold promoter activation in RCS cells but had no effect on promoter activity in 10T1/2 cells. By comparison, previous transient expression experiments from another laboratory using primary chondrocytes indicated that a 260-bp minimal active enhancer sequence of the intron 1 of the rat Col2a1 gene provided a much more modest 6-fold cell-specific enhancement in activity (4). This rat DNA segment contained sequences that are in part homologous to the mouse 156-bp sequence. In transient expression experiments in mouse primary chondrocytes, a construction containing a 231-bp intron fragment, which included the 156-bp fragment, showed a high level of expression of the reporter gene which paralleled its activity in RCS cells. In separate experiments separate from our laboratory, a fragment of 182 bp, which included the 156-bp fragment, conferred essentially perfect chondrocyte-specific expression of a lacZ reporter gene in transgenic mice. 2 In the 3Ј part of the 156-bp segment, we identified two adjacent DNase I footprints, FP1 and FP2, but these same footprints were found with nuclear extracts from chondrosarcoma cells and 10T1/2 fibroblastic cells. The 5Ј footprint (FP1) included part of a decamer sequence that had previously been postulated to be involved in chondrocyte-specific expression of the rat Col2a1 gene (4). In nuclear extracts of primary chick chondrocytes, Wang et al. (4) identified a protein binding to this decamer sequence that was enriched in chondrocytes, although it did not appear to be completely chondrocyte-specific. Removal of most of the FP2 sequence decreased the activity of the enhancer severalfold, but the remaining 119-bp DNA was still RCS cell-specific. Since the FP2 sequence in the mouse gene is not present at an analogous location in the rat and human sequences, we speculate that in these species a similar sequence present at some other location can substitute for this GC-rich sequence. Removal of the sequence of FP1, leaving an 89-bp fragment at the 5Ј end of the 156-bp segment, abolished the activity of the enhancer. Importantly, a 97-bp fragment containing the 3Ј part of the 156-bp segment including both FP1 and FP2 was also completely inactive. Hence, our deletion analysis indicates that sequences both in the 5Ј and in the 3Ј portions of this 119-bp fragment are needed together for RCS cell-specific enhancer activity. The 5Ј-DNA segment of the 119-bp fragment contains an inverted repeat sequence which is completely conserved in the mouse and human genes. The sequence of this inverted repeat contains 11 bp in one arm of the repeat and 10 bp out of 11 that are conserved in the other arm. We propose a model whereby proteins that bind to the 3Ј part of the 119-bp segment and proteins that bind to the 5Ј part of this segment cooperate with each other either in DNA binding or in transcriptional activation to provide RCS cell-specific enhancing activity. Analysis of the Col2a1 Promoter-Our deletion and substitution experiments using the promoter produced three novel results. First, successive deletions between Ϫ314 and Ϫ89, which removed two conserved GC-rich segments, indicated that the promoter contained elements between these end points that could activate transcription when the promoter itself was tested in transient expression experiments. We also noted that constructions containing these promoters without intron 1 sequences were much more active in 10T1/2 fibroblasts than in RCS cells, suggesting that the transcription factors that interact with these segments are more active or more abundant in fibroblastic cells than in RCS cells. In transgenic mice, constructions containing either a 767-bp or a 314-bp promoter without intron sequences were unable to direct activity of a reporter gene in chondrocytes. 2 Second, when an active intron enhancer segment was cloned upstream of the promoter, the levels of activity of the promoter were greatly increased in RCS cells, but they were approximately the same with different lengths of promoter. This suggested the hypothesis that the protein interactions needed for optimal activation of the promoter in transient expression experiments occur between enhancer-bound proteins and proteins bound to the promoter segment between Ϫ89 and ϩ6. The segment upstream of Ϫ89 might be needed to control activation of the endogenous Col2a1 gene in chromatin, to mediate responses to extracellular stimuli, to effect interactions with other parts of the gene, or to regulate expression in nonchondrocytic cells. Lastly, our transient expression experiments demonstrated that no specific elements in the Col2a1 promoter were required for activity of the RCS-specific enhancer. Indeed, the response of a minimal adenovirus major late promoter to the 231-bp enhancer was at least as high as that of Col2a1 promoters, and this construction showed very little activation in 10T1/2 fibroblasts. This conclusion was supported by results obtained in transgenic animals, since mice harboring a construction in which a 182-bp enhancer fragment was cloned upstream of a minimal ␤-globin promoter displayed high levels of chondrocyte expression of a lacZ reporter gene. 2 In summary, our results establish the unique usefulness of a new RCS cell line as a tissue culture system to study chondrocyte-specific regulatory elements of the Col2a1 gene and presumably other chondrocyte-specific genes. Our experiments indicate that both 5Ј and 3Ј sequences in a 119-bp intron 1 segment are needed for RCS cell-specific enhancer activity. We also demonstrate that the Col2a1 promoter sequences are dispensable for this cell-specific enhancer activity. Further work is needed to better characterize the binding sites within the 119-bp segment for proteins involved in the chondrocyte-specific expression of the Col2a1 gene and to isolate cDNA clones for these proteins.