Alternative Splicing of Type II Procollagen Exon 2 Is Regulated by the Combination of a Weak 5′ Splice Site and an Adjacent Intronic Stem-loop Cis Element*

Alternative splicing of the type II procollagen gene (COL2A1) is developmentally regulated during chondrogenesis. Chondroprogenitor cells produce the type IIA procollagen isoform by splicing (including) exon 2 during pre-mRNA processing, whereas differentiated chondrocytes synthesize the type IIB procollagen isoform by exon 2 skipping (exclusion). Using a COL2A1 mini-gene and chondrocytes at various stages of differentiation, we identified a non-classical consensus splicing sequence in intron 2 adjacent to the 5′ splice site, which is essential in regulating exon 2 splicing. RNA mapping confirmed this region contains secondary structure in the form of a stem-loop. Mutational analysis identified three cis elements within the conserved double-stranded stem region that are functional only in the context of the natural weak 5′ splice site of exon 2; they are 1) a uridine-rich enhancer element in all cell types tested except differentiated chondrocytes; 2) an adenine-rich silencer element, and 3) an enhancer cis element functional in the context of secondary structure. This is the first report identifying key cis elements in the COL2A1 gene that modulate the cell type-specific alternative splicing switch of exon 2 during cartilage development.

Alternative splicing of the type II procollagen gene (COL2A1) is developmentally regulated during chondrogenesis. Chondroprogenitor cells produce the type IIA procollagen isoform by splicing (including) exon 2 during pre-mRNA processing, whereas differentiated chondrocytes synthesize the type IIB procollagen isoform by exon 2 skipping (exclusion). Using a COL2A1 mini-gene and chondrocytes at various stages of differentiation, we identified a nonclassical consensus splicing sequence in intron 2 adjacent to the 5 splice site, which is essential in regulating exon 2 splicing. RNA mapping confirmed this region contains secondary structure in the form of a stem-loop. Mutational analysis identified three cis elements within the conserved double-stranded stem region that are functional only in the context of the natural weak 5 splice site of exon 2; they are 1) a uridine-rich enhancer element in all cell types tested except differentiated chondrocytes; 2) an adenine-rich silencer element, and 3) an enhancer cis element functional in the context of secondary structure. This is the first report identifying key cis elements in the COL2A1 gene that modulate the cell typespecific alternative splicing switch of exon 2 during cartilage development.
Alternative precursor mRNA (pre-mRNA) 2 processing is an important mechanism to increase proteomic diversity in eukaryotes. Through this process two or more mRNA molecules are generated from a single gene, leading to the synthesis of proteins that differ in structure and/or biological function (1). Numerous reports have shown that some alternative splicing events are cell type-specific or developmentally regulated (2)(3)(4)(5)(6)(7). Constitutive removal of non-coding introns from pre-mRNA in the nucleus occurs via a complex set of reactions at exon-intron junctions called splice sites. These splice site sequences are recognized by specific small nuclear ribonucleoproteins and accessory protein factors that make up the spliceosome complex (8). Two bona fide intronic sequences are also required for constitutive splicing to occur in addition to the 5Ј and 3Ј splice sites; they are the branch point sequence and the polypyrimidine tract sequence, both present upstream of the 3Ј splice site (8 -10). The information content present in these canonical splicing signals is generally not sufficient to ensure correct assembly of the spli-ceosome, especially in the case of regulated exons. Therefore, additional signals exist in the form of auxiliary cis elements (11)(12)(13)(14), which can be present either within the exon or in the flanking introns. Subsequently, splicing can be affected in a positive or negative manner by trans-acting enhancer or silencer splicing factor proteins that bind to these cis elements (15)(16)(17)(18)(19)(20). In addition, other regulatory cis elements exist that are functional in the context of RNA secondary structure conformations (21)(22)(23).
Although it has been recently estimated that more than half of all human genes generate more than one mRNA due to alternative splicing, information on the molecular processes governing cell-type or developmentally regulated alternative splicing is limited. In this respect, the process of chondrogenesis is an attractive model to study alternative splicing since a number of important cartilage molecules are spliced during chondrocyte differentiation (24). In particular, the cartilage extracellular matrix proteins type II collagen (25), type XI collagen (26), fibronectin (27) and tenascin C (28) are all alternatively spliced during cartilage development where a specific exon(s) encoding potential binding domains are spliced (included) in mRNAs expressed by chondroprogenitor cells but are skipped (excluded) from mRNAs expressed by differentiated chondrocytes. Of these molecules, type II collagen represents the simplest model and the best described alternative splicing event that occurs during chondrogenesis. Type II collagen is the major collagen component of cartilage extracellular matrix and is synthesized as a procollagen molecule of three identical ␣ chains, ␣1(II), containing an amino and carboxyl propeptide (29). The amino and carboxyl propeptides are subject to cleavage resulting in mature, homotrimeric collagen fibers that form stable fibrils in the extracellular matrix. Only one of the 54 exons encoding COL2A1 is alternatively spliced, producing two mRNA isoforms, type IIA and type IIB procollagen (25). The type IIA procollagen mRNA isoform contains the regulated, cassette exon (exon 2) and is synthesized by chondroprogenitor cells, whereas type IIB procollagen mRNA, devoid of exon 2, is synthesized by differentiated chondrocytes. Transcription of type IIA procollagen occurs in other cell types during embryonic development (30 -33), but the developmentally regulated splicing switch from type IIA to type IIB procollagen apparently only occurs during chondrogenesis. In addition, the phenotype of a differentiated chondrocyte is defined by its expression of the type IIB procollagen isoform. Thus, the COL2A1 alternative splicing event essentially defines the process of chondrocyte differentiation and, as such, is an excellent model to study key mechanisms that control cartilage development.
Exon 2 encodes a cysteine-rich (CR) von Willebrand factor C-like domain within the amino propeptide of type II procollagen. Homologues of this CR domain are present in other fibrillar collagen amino propeptides as well as in extracellular matrix proteins including thrombospondins, connective tissue growth factor, and chordin (34,35). Pre-vious studies have shown that the CR exon 2-encoded domain of type II procollagen may have an important biological function during development by binding to growth factors such as bone morphogenetic proteins, similar to the function of chordin (35,36). The presence of the type IIA procollagen isoform in other non-cartilaginous embryonic tissues such as heart, lung, kidney, and eye (30,32,37,38) also suggests an important function for the CR domain during developmental processes. Furthermore, it has been reported that the immature IIA procollagen isoform is re-expressed during cartilage degradation, as seen in osteoarthritis (39), suggesting an additional function for the exon 2-encoded CR domain during tissue repair.
Two studies have been published on COL2A1 alternative splicing at the pre-mRNA level. One report (40) showed that a murine Col2a1 mini-gene was correctly spliced during insulin-dependent chondrocyte differentiation of murine ATDC-5 cells. The same group subsequently showed that deleting large portions of introns 1 and 2 still resulted in correct splicing of the Col2a1 mini-gene in ATDC-5 cells (41). However, to date, there are no reports of specific cis elements or trans-acting splicing factor proteins that are important in regulating human COL2A1 exon 2 alternative splicing.
Using a human COL2A1 mini-gene as a model system, the present study is the first to identify functional cis elements in intron 2 of COL2A1 pre-mRNA that modulate splicing of exon 2. RNA mapping analysis showed that a non-classical consensus splicing region adjacent to the 5Ј splice site of exon 2 contains RNA secondary structure in the form of a stem-loop. This is the first study to experimentally show the existence of a stem-loop directly adjacent to a weak 5Ј splice site of an exon that is regulated in a tissue-specific manner during development. We report that the double-stranded stem sequence, which is 100% conserved between species, contains both enhancer and silencer cis elements that are functional in regulating type II procollagen isoform expression during chondrocyte differentiation. From the data reported in the present study, we have proposed a model whereby the secondary structure of the stem-loop functions to mask the weak 5Ј splice site. Functionally, it is the interaction of enhancer and/or silencer transacting splicing factor proteins with cis elements in the stem-loop region that determines the pattern of exon 2 splicing at a specific phase of cartilage development.

MATERIALS AND METHODS
Construction of a Human COL2A1 Mini-gene-A human COL2A1 mini-gene was constructed spanning exons 1-3, including full-length intron 1 and intron 2 sequences (Fig. 1). Three separate fragments of the mini-gene were synthesized by PCR from human genomic DNA (Clontech) using the elongase amplification system (Invitrogen). Each fragment was amplified using the primer pairs listed in TABLE ONE containing specific restriction enzyme sites for sequential cloning into pcDNA3 vector (Invitrogen). The cloned mini-gene (ϳ5.9 kb) is under transcriptional control of the cytomegalovirus promoter. The DNA construct was sequenced to confirm correct orientation and the absence of mutations.
Transient Transfections of the COL2A1 Mini-gene-The following cell lines were transfected with the COL2A1 mini-gene: C3H 10T1/2 murine mesenchymal cells (ATCC), MC615 murine vertebral chondrocytes (a gift from Dr. Frederic Mallein-Gerin, Lyon, France), T/C28I2 chondrocytes from human costal cartilage (a gift from Dr. Mary Goldring, Harvard University), RCS (LTC) rat chondrosarcoma cells (42), and HEK-293 human embryonic kidney cells. COL2A1 mini-gene in pcDNA3 vector was transfected into each of the cell lines using FuGENE 6 reagent (Roche Applied Science) following the manufacturer's protocol. Briefly, 1-3 g of the mini-gene construct was transfected into each cell line at a ratio of 1:4 (g/l) DNA:FuGENE for 5 h in serum-free medium. Serum-containing medium was then added (final concentration, 10% fetal bovine serum), and the cells were cultured for a further 48 h until RNA isolation.
Analysis of Spliced mRNA Isoforms Derived from the COL2A1 Mini-gene-Total RNA was harvested from each cell line 48 h after transfection using the Qiagen RNeasy kit. Approximately 1 g of RNA was reverse-transcribed using random primers in a final volume of 20 l, and the resulting cDNA was diluted to 80 l with sterile water. 10 l of cDNA was used for quantitative PCR in the presence of [␣-32 P]dCTP (10 mCi/ml, 3000 Ci/mmol; Amersham Biosciences). The primers, pcDNA3-COL2A1-Exon1 (5Ј-CAAGCTTACATGATTCGC-3Ј) and sp6, were used to amplify cDNA derived only from the exogenously transfected COL2A1 mini-gene (Fig. 1). The linear range for these primers was established, and PCR was carried out for 20 cycles: 95°C for 30 s; 55°C for 30 s; 72°C for 30 s. 10 l of 6ϫ loading dye (30% glycerol, 0.025% (w/v) bromphenol blue, 0.025% (w/v) cyanol blue) was added to each reaction, and 7 l was electrophoresed at 200 V through 6% polyacrylamide gels. pBR322 DNA digested with MsbI was used as a size marker. Gels were dried and exposed to PhosphorImager screens (Amersham Biosciences) for 1 h and then scanned on the STORM TM 840 PhosphorImager (Amersham Biosciences). Bands corresponding to the type IIA (ϳ390 bp) and IIB (ϳ180 bp) mRNA isoforms were quantified using ImageQuant TM software. From these values, ratios of IIA:IIB mRNA spliced products were calculated for each cell type.
Detection of Aggrecan and Type I Collagen mRNA-Primers were designed (TABLE TWO) to amplify aggrecan or type I collagen from total RNA extracted from each of the five cell lines (HEK-293, C3H 10T1/2, MC615, T/C28I2, and RCS). RT-PCR was carried out in the linear range as determined for each primer pair. Briefly, 2 g of RNA was reverse-transcribed using random primers in a total reaction volume of 20 l. An equal volume of water was added to the RT reaction, and 3 l was used for PCR in the presence of [␣-32 P]dCTP (10 mCi/ml, 3000 Ci/mmol; Amersham Biosciences) in a total volume of 50 l. PCR products (6 l) were electrophoresed through 6% polyacrylamide gels. Primer pairs for amplification of the human COL2A1 mini-gene Three sets of forward (F) and reverse (R) primer pairs were used to amplify fragments 1, 2, and 3 of the COL2A1 mini-gene (Fig. 1). Restriction enzyme sites at the 5Ј and 3Ј ends of each amplified product are shown in bold and also underlined in the primer sequence. Numbers in parentheses denote the nucleotide positions of the region amplified from genomic DNA based on the numbering of the published COL2A1 sequence (accession number L10347 5-BamHI-EcoRV-3 R, GATAGGATATCTTGTATTGAATGCTGGGGAAG 3. In 2(5114)-Ex3(5908) F, AATACAAGATATCCTATCTCCCCTGCAGAG ϳ0.8

5-EcoRV-XhoI-3 R, CCGCTCGAGCTTTGGTCCTGGTTGCCCTGCAAGGA
Gels were dried and exposed to a PhosphorImager screen (Amersham Biosciences) and scanned on a STORM TM 840 PhosphorImager (Amersham Biosciences). Bands corresponding to aggrecan, type I collagen, or ␤-actin were quantified using ImageQuant TM software. Levels of aggrecan and type I collagen mRNA in each cell type were expressed relative to ␤-actin. Conservation Analysis of COL2A1 Genomic Sequence-The May 2004 genomic assembly of the human COL2A1 gene (chr12: 46, 679, 680, 700 -746,778) was accessed on the UCSC Genome Browser (genome.ucsc.edu). The species conservation tracks showing the pairwise alignments were obtained through the conservation link. Twelve pairwise-aligned sequence blocks derived from Blastz alignments (43) were scored by phastCons (44). The resulting annotation alignment using Multiz (45)  . Genomic sequence spanning the 5Ј and 3Ј splice site junctions of exon 2 was taken from blocks 3-9 and reverse-complemented to generate the final alignment ( Fig. 4).
Prediction of RNA Secondary Structure in Intron 2 of COL2A1-A stretch of nucleotides in intron 2 of COL2A1, directly adjacent to the 5Ј splice site of exon 2, was predicted to contain RNA secondary structure in the form of a stem-loop (29). The Zuker Mfold program (46,47) was used to locate the potential site of interest in intron 2. The mRNA sequence corresponding to nucleotides 4191-4455 of COL2A1 (accession number L10347) was entered into the Mfold program accessed via the Macfarlane Burnet Centre Mfold server (mfold.burnet.edu.au). This mRNA fragment corresponds to the entire 207 bp of human exon 2 and the first 58 nucleotides of intron 2. The predicted stem-loop site is between nucleotides ϩ4 and ϩ41 of intron 2.
RNA Mapping-RNase digestions were carried out to map the site of RNA secondary structure within intron 2 of COL2A1. A 265-bp fragment spanning exon 2 and the first 58 nucleotides of intron 2 were amplified by PCR using the wild-type COL2A1 mini-gene construct as the substrate. This fragment was subcloned into pcDNA3 using BamHI and EcoRI sites. RNA, corresponding to the antisense strand of the 265-bp exon 2-intron 2 fragment was synthesized by in vitro transcription (MAXIscript SP6 TM ; Ambion) in the presence of [␣-32 P]UTP (10 mCi/ml, 3000 Ci/mmol; Amersham Biosciences). The resulting, radiolabeled RNA probe was purified by phenol/chloroform extraction and ethanol precipitation in the presence of 3 M sodium acetate (pH 5.2). Before enzymic probing, RNA was heated for 1 min at 65°C and renatured by slowly cooling to 37°C. 1 l of either RNase T1 at 0.4 or 4 units/l (Invitrogen), S1 nuclease at 2 or 20 units/l (Invitrogen), or RNase V1 at 0.002 or 0.02 units/l (Ambion) was added to 1 g of radiolabeled RNA fragment in a total volume of 20 l and digested for 15 min at 30°C. A control aliquot of RNA without the addition of RNases was processed simultaneously with the digested samples. After digestion, RNA was purified by phenol/chloroform extraction, and 1 pmol was reverse-transcribed for 1 h at 37°C using Superscript TM II RNase H Ϫ reverse transcriptase (Invitrogen) with a sense primer that hybridized to a region in exon 2 (5Ј-GTGAAGACGTGAAAGACTGCCTCA-3Ј) (Fig. 6). Samples were then treated with RNase H (0.5 units) for 20 min at 37°C. After a final phenol/chloroform extraction, RNA was resuspended in gel loading buffer (95% formamide, 18 mM EDTA, 0.025% SDS, 0.025% (w/v) xylene cyanol, 0.025% (w/v) bromphenol blue), and 3 l was electrophoresed through 6% urea, polyacrylamide denaturing gel. The gels were dried and exposed to a PhosphorImager screen (Amersham Biosciences) overnight and scanned on a STORM TM 840 PhosphorImager (Amersham Biosciences). To localize the sites of RNA digestion, the dideoxy chain termination reaction was carried out on the original pcDNA3 construct containing the 265-bp cDNA sequence encoding exon 2 and the first 58 nucleotides of intron 2 using a commercially available kit (U. S. Biochemical Corp.). Reactions were carried out in the presence of 0.5 l [␣-32 P]dATP (10 mCi/ml, 3000 Ci/mmol; Amersham Biosciences), and samples were electrophoresed in parallel with reverse-transcribed RNA digests.
Synthesis of Mutant COL2A1 Mini-genes-A series of mutant minigenes was synthesized devoid of large regions (300 -500 bp) of intronic sequence. PCR was carried out to amplify two separate fragments of the mini-gene using specific primers containing ClaI restriction sites. The resulting fragments were gel-purified and ligated and a third PCR was done to amplify the ligated fragment devoid of the intronic sequence of interest. We named the mutant mini-gene with a 370-nucleotide deletion in intron 2 (adjacent to the 5Ј splice site of exon 2) deletion mutant 1 (Del 1). Primers used to synthesize Del 1 were PCR 1 (forward primer ϩ2542 BamHI, 5Ј-CTAGGGGATCCGGTTACGGC-3Ј, and reverse primer, Ϫ4406 ClaI, 5Ј-CCATCGATAATTACAACCAC-3Ј), PCR 2 (ϩ4778 ClaI, 5Ј-CCATCGATCGATACCTTGTCTTA-3Ј, and reverse primer, Ϫ5140 EcoRV, 5Ј-GATAGGATATCTTGTATTGAATGCT-GGGGAAG-3Ј), and PCR 3 (forward primer ϩ2542 BamHI and reverse primer Ϫ5140 EcoRV). The deletion mutant cDNA fragment was confirmed by DNA sequencing and ligated into the COL2A1 mini-gene construct using BamHI and EcoRV restriction sites to replace the wildtype 2.6-kb fragment.

Primers pairs for amplification of aggrecan and type I collagen mRNA
Each forward (F) and reverse (R) primer pair was used to amplify aggrecan or type I collagen from human (HEK-293, T/C28I2), mouse (C3H10T1/2,MC615), or rat (RCS) cells by RT-PCR. PCR cycle numbers were in the linear range as determined for each primer pair. cDNA was analyzed by PhosphorImager analysis and expressed relative to ␤-actin. The sequence of each primer is shown in the 5Ј-3Ј direction. Other mutant mini-gene constructs were synthesized by either substituting or deleting nucleotides near or within the apparent stem-loop sequence in intron 2. TABLE THREE lists all of the COL2A1 mutant mini-genes analyzed in the present study. Mutations were introduced using the QuikChange TM site-directed mutagenesis kit (Stratagene). Briefly, a complementary primer pair (purified by SDS-PAGE; Invitrogen) containing the desired nucleotide substitution or devoid of the nucleotide sequence of interest was added to ϳ20 ng of substrate DNA. Substrate DNA was prepared by sub-cloning the wild-type 2.6-kb fragment of the COL2A1 mini-gene into pSP73 vector (Promega) using BamHI and EcoRV restriction sites. This 2.6-kb fragment contains the specific region of intron 2 that was to be mutated. By doing this, the construct size was reduced from ϳ11.3 kb (size of COL2A1 mini-gene in pcDNA3) to ϳ 5 kb (the COL2A1 mini-gene 2.6-kb fragment in pSP73) to increase the efficiency of the in vitro mutagenesis procedure. PCR mutagenesis was carried out over 18 cycles (95°C for 30 s; 55°C for 1 min; 68°C for 5.30 min), and the resulting PCR products were digested with DpnI (1 l) for 1 h at 37°C to digest parental, methylated DNA. An aliquot (1 l) of digested DNA was transformed into XL-1 Blue Supercompetent Cells (Stratagene), and resulting colonies were screened for the presence of the correct mutation. Mutant colonies were selected, and the 2.6-kb fragment was re-ligated back into the COL2A1-pcDNA3 construct to create the mutant COL2A1 mini-gene. Transfections of these mutant mini-genes were done as described previously. Fig. 1 shows the human COL2A1 mini-gene containing exon 1, the regulated cassette exon (exon 2), exon 3, and full-length intervening intron 1 and intron 2. This genomic DNA fragment (ϳ5.9 kb) was cloned into pcDNA3 vector between T7 and sp6 RNA polymerase transcription initiation sites. Cloning of the mini-gene into pcDNA3 was done using the restriction enzyme sites shown. This human mini-gene contains the necessary bona fide sequences to ensure splicing (i.e. removal of introns 1 and 2) in vivo by any cell type. Transfection of this mini-gene construct and subsequent pre-mRNA splicing by cells used in the present study resulted in production of IIA and/or IIB mRNA isoforms that were distinguished by size difference based on the inclusion (IIA) or exclusion (IIB) of exon 2. RT-PCR using the specific primer pair (P1 and P2 in Fig. 1)-amplified cDNA fragments of ϳ390 and 180 bp that corresponded to the IIA and IIB mRNA spliced isoforms, respectively.

Alternative Splicing of the Human COL2A1 Mini-gene in Different Cell Types
Five different cell lines were selected to analyze splicing of the COL2A1 mini-gene. Human embryonic kidney (HEK-293) cells were chosen as a source of non-chondrocytes. C3H 10T1/2 cells were included as a source of chondroprogenitors as these cells can be induced to undergo differentiation in the correct culture environment (48,49). MC615 and T/C28I2 cells are transformed chondrocytes isolated from mouse vertebrae and human costal cartilage, respectively; these cells were expected to be in a de-differentiated state in the culture conditions used in the present study. Finally, rat chondrosarcoma (RCS) cells were chosen as a source of differentiated chondrocytes (42). To confirm the differentiation status of these cells, RT-PCR was carried out on RNA isolated from each of the cell lines to analyze the levels of aggrecan and type I collagen mRNA. Aggrecan is a chondrocyte marker, whereas type I collagen is a marker of de-differentiated cells in culture. Fig. 2 shows the levels of aggrecan and type I collagen mRNA relative to ␤-actin

COL2A1 mutant mini-genes
A range of nucleotide deletions and substitutions were introduced into the wild-type COL2A1 mini-gene to produce the series of mutant mini-genes analyzed in the present study. For each mutant mini-gene, the mutation type and nucleotide change are shown where applicable. Numbers in parentheses refer to the intron 2 nucleotide numbers deleted from or substituted in the mini-gene.

Mutant mini-gene Type and site of mutation
Del 1 Deletion in intron 2 (ϩ11 to ϩ380) ϩ5ЈSS Nucleotide substitutions, TGTA3AAGT (ϩ3 to ϩ6) SL-Del-1 Deletion of stem loop (ϩ7 to ϩ43) ϩ5ЈSS/SL-Del-1 Combination of nucleotide substitution and stem loop deletion SL-Del 2 Partial deletion of stem loop (ϩ7 to ϩ18) SL-Del 3 Partial deletion of stem loop (ϩ29 to ϩ41) CCC-1 Nucleotide substitutions, TTT3CCC (ϩ 8 to ϩ10) CCC-2 Nucleotide substitutions, TTT3CCC (ϩ8 to ϩ10) and TTT3CCC (ϩ12 to ϩ14) CCC-3 Nucleotide substitutions: TTT3CCC (ϩ8 to ϩ10), TTT3CCC (ϩ12 to ϩ14), and TTT3CCC (ϩ16 to ϩ18) GGG-1 Nucleotide substitutions, AAA3GGG (ϩ35 to ϩ37) GGG-2 Nucleotide substitutions, AAA3GGG (ϩ35 to ϩ37) and AAA3GGG (ϩ31 to ϩ33) CCC-1/GGG-1 Nucleotide substitutions, TTT3CCC (ϩ8 to ϩ10) and AAA3GGG (ϩ35 to ϩ37) mRNA for each cell line, and the ratio of aggrecan/type I collagen expression is also shown. As expected, RCS cells expressed the highest ratio of aggrecan/type I collagen, although undifferentiated C3H 10T1/2 and MC615 cells expressed the lowest ratios. The aggrecan/type I collagen ratio value shown for T/C28I2 cells suggests that these cells are in an intermediate stage of differentiation. Therefore, based on this knowledge, we would expect to see different patterns of COL2A1 minigene splicing where levels of the type IIA mRNA isoform derived from the mini-gene would exceed those of the type IIB mRNA isoform in chondroprogenitor cells and vice versa in chondrocytes. Fig. 3 shows that the mini-gene spliced products amplified by RT-PCR are consistent with the differentiation status of these cells. The C3H 10T1/2 and MC615 cells contained the highest ratio of IIA/IIB mRNA isoforms; T/C28I2 cells contained a lower IIA/IIB ratio in comparison, whereas the RCS cells spliced the mini-gene to produce more of the IIB mRNA isoform. To confirm the efficacy of using this COL2A1 mini-gene as a model system to study regulation of exon 2 alternative splicing, the ratio of endogenous type IIA and type IIB collagen isoforms in C3H 10T1/2, MC615, T/C28, and RCS cells was found to be similar to that derived from the COL2A1 mini-gene (results not shown). The non-chondrocyte HEK-293 cells spliced the mini-gene to produce ϳ2-fold more of the type IIA mRNA isoform than the type IIB mRNA isoform; levels of endogenous type II collagen were undetectable in these cells. Fig. 4 shows a sequence alignment comparison of a region of COL2A1 genomic sequence between human, chimp, mouse, rat, dog, zebrafish, and fugu (puffer fish). The alignment shows intronic sequence spanning the 3Ј and 5Ј splice sites of exon 2 as well as the first and last 20 nucleotides of exon 2. In all species, the 3Ј splice site is shown to conform to the classical consensus sequence, ( Ϫ4 N(T/C)AG2(G/A)N ϩ2 ), where the arrow denotes the exon-intron junction, and highly conserved nucleotides are shown in underlined bold font. Upstream of the 3Ј splice site is a long polypyrimidine tract sequence which may be an important feature in the regulation of exon 2 splicing. The 5Ј splice site sequence of the COL2A1 gene (e.g. Ϫ2 TG2GTTGTA ϩ6 in the human sequence) does not conform to the classical consensus sequence ( Ϫ2 AG2GU(G/ A)AGU ϩ6 ) and is, thus, referred to as a "weak" 5Ј splice site. This is the first report showing that a weak 5Ј splice site is present adjacent to the alternatively spliced exon in the type II procollagen gene from a number of different species. It was previously reported that a region directly downstream of the 5Ј splice site may contain RNA secondary structure in the form of a stem-loop (29). The nucleotides predicted to form the double-stranded RNA of the stem-loop are 100% conserved between all species analyzed (Fig. 4). Therefore, there is a high likelihood that this region of intron 2 contains regulatory cis elements involved in pre-mRNA splicing regulation due to 1) the location, adjacent to an alternatively spliced exon, 2) the conservation between species, and 3) the potential of secondary structure formation.

Altered Splicing of a COL2A1 Mini-gene Devoid of Intron 2 Sequence
To test for the presence of functional intronic splicing cis elements, a series of mutant COL2A1 mini-genes was synthesized devoid of large regions of introns 1 or 2. One of these deletion mutants, named Del 1, showed a marked difference in alternative splicing patterns compared with splicing of the wild-type mini-gene. Fig. 5 shows that the region of intron 2 deleted from the Del 1 mini-gene is a 370-bp fragment from intron 2 nucleotide numbers ϩ11 to ϩ380 directly adjacent to the 5Ј  splice site of exon 2. In the Del 1 mutant, the intronic 5Ј splice site nucleotides were not deleted, so that splicing to either include or exclude exon 2 could still potentially occur. The phosphorimage in Fig.  5 shows that all cell types processed the Del 1 mutant in a similar way, producing only the type IIB isoform. This suggests that the weak 5Ј splice site alone is not sufficient to promote exon 2 splicing (inclusion). In addition, the deleted region of intron 2 may also contain essential regulatory elements that modulate the distinct splicing patterns of exon 2 in cells at various stages of chondrocyte differentiation.

Analysis of RNA Secondary Structure in Intron 2 Adjacent to the 5 Splice Site
Using the Zuker Mfold program to predict secondary structure, it was found that the region in intron 2 directly adjacent to the 5Ј splice site of exon 2 (intron 2 nucleotide number ϩ4 to ϩ41) is likely to contain RNA secondary structure in the form of a stem-loop. Fig. 6 shows the location of this stem-loop and the predicted Mfold structure and ⌬G (Ϫ7.7 kcal). To experimentally determine the presence of an RNA stem-loop structure in intron 2, we performed RNase mapping analysis. RNase T1 digests single-stranded sites, preferentially adjacent to guanine residues, and S1 nuclease digests single-stranded RNA with no particular specificity, whereas RNase V1 digests sites of double-stranded RNA. The region of the COL2A1 mini-gene that was probed was a 145-nucleotide fragment containing the last 87 nucleotides of exon 2 and the first 58 nucleotides of intron 2. The phosphorimage of the polyacrylamide  sequencing gel in Fig. 6 shows results of the various RNase treatments. Localization of the digested sites was determined by referring to the sequencing reaction results shown on the left side of the phosphorimage. Specifically, RNase T1 or S1 nuclease digestion sites indicated that a single-stranded loop is present. Digestion sites corresponding to the regions of double-stranded RNA were seen in the lanes corresponding to RNase V1 digestions only. This is indicative of the stem region and confirms that RNA stem-loop secondary structure is present in intron 2 of human COL2A1, directly adjacent to the 5Ј splice site.

Effect of a Weak 5 Splice Site on Alternative Splicing of COL2A1 Exon 2
The COL2A1 genomic sequence alignment in Fig. 4 shows conservation of a potentially weak 5Ј splice site in all species. To determine the effect of the 5Ј splice site sequence on alternative splicing of exon 2, we synthesized a mutant mini-gene (named ϩ5Ј SS) with a four nucleotide substitution in intron 2 ( ϩ3 TGTA ϩ6 3 ϩ3 AAGT ϩ6 ) to create a strong splice site that conforms to the classical consensus sequence. The ϩ5Ј SS COL2A1 mini-gene was spliced similarly by all cell types to produce only the type IIA mRNA isoform (Fig. 7, second lane of each gel panel). This suggests that the presence of a weak 5Ј splice site is important to confer the differential cell type-specific splicing patterns of COL2A1 shown in the present study.

Effect of Intron 2 (stem-loop) Deletions on Alternative Splicing of COL2A1 Exon 2
To determine whether the stem-loop sequence (Fig. 6), located directly adjacent to the weak 5Ј splice site, contains regulatory elements that modulate exon 2 splicing, a series of deletion mutant COL2A1 mini-genes were synthesized (TABLE THREE). Stem-loop deletion 1 mutant mini-gene (SL-Del 1) was produced by deleting most of the stem-loop region (intron 2 nucleotides ϩ7 to ϩ43; Fig. 7) except the first three nucleotides (ϩ4 to ϩ6) that are part of the bona fide splice site sequence. Because the double-stranded stem sequence of the stem-loop showed 100% sequence similarity between species, we also produced deletion mutant mini-genes devoid of the uridine-rich region of the stem (SL-Del 2) or the opposite, adenine-rich region of the stem (SL-Del 3). These mutants are shown diagrammatically in the top panel of Fig. 7. In all of these deletion mutant mini-genes, the 5Ј splice site intronic sequence (ϩ1 to ϩ6) that binds to U1 snRNA (8) was intact. The Phos-phorImager gel pictures in Fig. 7 show that, compared with wild-type mini-gene splicing, deletion of the stem-loop sequence resulted in a marked inhibition of exon 2 splicing (inclusion), favoring type IIB mRNA production (SL-Del 1; third lane of each gel panel). This splicing pattern was displayed by all cells regardless of the cell type or differentiation status, again suggesting that the weak 5Ј splice site alone is not sufficient to induce exon 2 splicing and that a cis enhancer elementpromoting spliceosome assembly at the 5Ј splice site of exon 2 was removed. A mutant mini-gene combining both the strong 5Ј splice site sequence together with deletion of most of the stem-loop region (ϩ5ЈSS/SL-Del-1; Fig. 7, fourth lane of each gel panel) showed exclusively exon 2 inclusion to produce only the IIA mRNA isoform. This splicing pattern was shown by all cell types and confirms that the strong splice site sequence compensated for the deleted stem-loop region to promote exon 2 splicing. This suggests that 1) the stem-loop sequence does not contain bona fide nucleotides necessary for constitutive splicing (inclusion) of exon 2 and 2) the cell type-specific COL2A1 splicing patterns shown in the present study are dependent on the presence of both a weak 5Ј splice site and the adjacent stem-loop sequence.
Removal of the left side of the stem loop containing the uridine stretch sequence (AUUUAUUUAUUU; SL-Del 2), resulted in an apparent decrease in the ratio of IIA:IIB mRNA transcripts compared with splicing of the wild-type mini-gene in all cell types except RCS cells (Fig.  7, fifth lane of each gel panel). This suggests that an enhancer site is located within this uridine stretch that is not functional in RCS cells. However, by deleting the adenine-rich region of the stem (AUAAAUAAAU; SL-Del 3) the opposite effect was found whereby all cell types, including the differentiated chondrocyte RCS cells, spliced the mutant mini-gene to produce predominantly type IIA mRNA (Fig.  7, sixth lane of each gel panel). This suggests removal of a splicing Bands corresponding to the single-stranded loop region were seen after digestion with RNase T1 (0.4 or 4 units/l) and S1 nuclease (2 or 20 units/ l). Bands at the predicted double-stranded stem region were only seen after RNase V1 digestion (0.002 or 0.02 units/l). A sequencing reaction was done in parallel with the same primer (P1) to determine the RNase cleavage sites. The orientation of nucleotides within the stem loop region (ϩ4, ϩ16, ϩ29, and ϩ41) are also shown, each one represented by an asterisk (*) in the sequencing gel.
silencer element. Analysis of the intronic sequence after deletion of the 10 nucleotides comprising the adenine-rich side of the stem-loop showed that we had not created an alternative, stronger 5Ј splice site. This was confirmed by DNA sequencing of the amplified cDNA corresponding to the IIA mRNA isoform derived from SL-Del 3 mutant mini-gene (data not shown).

Effect of Stem-loop Nucleotide Substitutions on COL2A1 Exon 2 Alternative Splicing
Mutation of the Uridine-rich Stretch-To specifically localize regulatory cis elements within the RNA stem-loop, nucleotide substitutions were introduced on either side of the conserved double-stranded stem. Because of reports of splicing factor proteins that bind to intronic uridine-rich regions in pre-mRNAs to regulate alternative splicing, we first synthesized three mutant mini-genes named CCC-1, CCC-2, and CCC-3 that contained cytosines in place of the first, second, and third triplet set of uridine nucleotides within the double-stranded stem, respectively (TABLE THREE; Fig. 8). Analysis of the spliced mRNA products derived from the CCC-1 mutant mini-gene suggested that this mutation had either a minor effect or no effect in inhibiting exon 2 splicing (i.e. exon 2 skipping) in HEK-293, MC615, T/C28I2, and C3H 10T1/2 cells when compared with splicing of the wild-type mini-gene (Fig. 8). However, splicing of CCC-2 and CCC-3 mini-genes by these cell types showed a trend toward type IIB mRNA production, concomitant with an increasing number of uridine substitutions. This result is in agreement with that from splicing of the uridine-stretch deletion mutant (SL-Del2; Fig. 7) in these cell types, confirming the presence of a functional enhancer cis element, particularly within the second and third uridine triplets.
Interestingly, in RCS cells, splicing of the CCC-1 mutant mini-gene produced a significantly higher ratio of IIA:IIB mRNAs compared with splicing of the wild-type mini-gene. This splicing ratio was also higher than that achieved by splicing of the CCC-1 mini-gene in the other four cell types. In addition, the IIA:IIB mRNA splicing ratio decreased by mutating the second (CCC-2) and third (CCC-3) set of uridines. However, levels of IIA:IIB mRNA derived from these mini-genes did not fall below that derived from the wild-type mini-gene, suggesting that the uridine-rich region is not an enhancer site but, rather, a silencer element in differentiated chondrocytes. The observation that the CCC-3 splicing result was similar to that of the uridine-stretch deletion mutant (SL Del-2; Fig. 7) in RCS cells also supports the latter statement. In the case of CCC-1 mutant mini-gene splicing, it is possible that this mutation C(ϩ8 to ϩ10) may have created a new enhancer site that may only be functional in RCS cells. An alternative or stronger 5Ј splice site was not created by this mutation, as confirmed by DNA sequencing of the reverse-transcribed IIA mRNA derived from the CCC-1 mini-gene. It is difficult to predict if this potentially novel enhancer site is also functional in MC615 and C3H 10T1/2 cells since wild-type mini-gene splicing patterns in these cells are similar to CCC-1 splicing patterns in RCS cells. However, a lower IIA:IIB mRNA ratio by splicing of the CCC-1 mini-gene compared with wild type in HEK-293 and T/C28 cells is in keeping with disruption of a natural enhancer cis element.
Mutation of the Adenine-rich Stretch-To analyze potential regulatory elements on the opposite side of the stem, two mutant mini-genes were made, named GGG-1 and GGG-2, which contained guanines in place of one or both sets of adenine triplet nucleotides, respectively (TABLE THREE; Fig. 9). Splicing of these mini-genes by all cell types resulted in a higher IIA:IIB mRNA splicing ratio when compared with the corresponding wild-type mini-gene splicing pattern. The level of exon 2 inclusion was also slightly higher from the mini-gene-containing mutations in both sets of adenine triplets (GGG-2) compared with mutation of one adenine triplet (GGG-1) in all cell types. These splicing patterns are similar to those from the mutant mini-gene devoid of the adenine-rich region (SL-Del 3; Fig. 7), which suggests the presence of a functional silencer element within this region of the stem-loop. The IIA mRNA derived from the GGG-1 and GGG-2 mutant mini-genes was also reverse-transcribed and sequenced to confirm that an alternative 5Ј splice site was not created by these mutations.
Effect of Compensatory Mutations to Restore Secondary Structure-Another mutant mini-gene was synthesized (CCC-1/GGG-1; TABLE THREE and Fig. 9) that combined the CCC-1 and GGG-1 mutations to restore secondary structure to the stem-loop that would have been disrupted in the CCC-1 mutant mini-gene. The formation of a stem-loop by introducing the CCC and GGG mutations was verified by RNA mapping (results not shown), and Mfold analysis showed that the mutated CCC-1/GGG-1 stem loop was more stable (⌬G ϭ Ϫ13.6 kcal) than the wild-type stem-loop (⌬G ϭ Ϫ7.7 kcal). Individually, the CCC-1 and GGG-1 mutations resulted in a splicing pattern that favored exon 2 splicing (inclusion) in all cell types, and in most cases, splicing of these mutant mini-genes differed from the splicing pattern of the corresponding wild-type mini-gene. However, the presence of both mutations in the same mini-gene resulted in a significant change in COL2A1 isoform synthesis that favored exon 2 exclusion. This suggests that in addition to the presence of secondary structure, the specific nucleotides that form the stem-loop are also important in regulating exon 2 splicing. As shown in this study, the adjacent positioning of the U (ϩ8 to ϩ10) and A (ϩ35 to ϩ37) triplets within the double-stranded stem apparently functions as a splicing enhancer site. This enhancer function was also supported by the fact that the splicing pattern of CCC-1/GGG-1 was similar to that of the SL-Del 1 mini-gene (Fig. 7) in which the stem-loop, including the U (ϩ8 to ϩ10)/A(ϩ35 to ϩ37) triplet, was removed. This suggests that another level of complexity exists within the stem-loop to regulate COL2A1 exon 2 alternative splicing, where functional cis elements may exist as individual linear sites as well as within the context of secondary structure.

DISCUSSION
To identify key cis regulatory elements that modulate alternative splicing of exon 2 during chondrogenesis, we used a human COL2A1 mini-gene as a model system (Fig. 1). A recent report described a murine Col2a1 mini-gene that was successfully spliced by ATDC-5 cells during a 21-day chondrocyte differentiation assay system (40). However, it was not specified if the mini-gene was stably transfected to detect spliced products after 3 weeks in culture. Splicing of our human COL2A1 minigene was analyzed at one time point in cells at various stages of differentiation; C3H 10T1/2 cells were used as a source of chondroprogenitors, MC615 and T/C28I2 cells were a source of de-differentiated chondrocytes, and RCS cells were a source of differentiated chondrocytes. Variations in IIA-and IIB-spliced mRNA isoforms derived from the mini-gene correctly reflected the differentiation status of the cells. This indicated that the COL2A1 mini-gene contained the necessary cis elements required for correct splicing of exon 2 within different cellular contexts. The fact that the HEK-293 cells spliced the mini-gene to produce more of the type IIA isoform suggests that similar cis-and/or trans-acting factors are functional in non-chondrocyte cells as they are in chondroprogenitor and de-differentiated chondrocytes.
In the present study we show that splicing of a mutant mini-gene (Del 1; Fig. 5) devoid of 370 bp of intron 2 adjacent to the 5Ј splice site resulted in exon 2 exclusion in all cell types. The importance of intron 2 sequences in regulating exon 2 splicing was also reported by Nishiyama et al. (41), who found that processing of a murine Col2a1 deletion minigene devoid of ϳ92% of intron 2 by ATDC-5 cells resulted in IIA and IIB mRNA transcripts in addition to an abnormal increase in splicing intermediates. This mini-gene contained nucleotides downstream of the 5Ј splice site that were not present in the Del 1 mutant reported here. Therefore, the absence of exon 2 splicing (inclusion) in the Del 1 mutant mini-gene indicated that key, non-consensus splicing enhancer cis elements are present in intron 2 adjacent to the 5Ј splice site. A previous report predicted that a stretch of nucleotides in this region may contain secondary structure in the form of a stem-loop (29). By a series of RNase digestions, we showed that this region of intron 2 (nucleotide numbers ϩ4 to ϩ41) does indeed form a stem-loop. Because of location, sequence similarity between species, and secondary structure conformation, we hypothesized that the stem-loop is important in the regulation of exon 2 alternative splicing. Furthermore, nucleotides ϩ4 to ϩ6 of the stem-loop, which are part of the intronic sequence that interacts with U1 snRNA (8), does not conform to the classical 5Ј splice site consensus sequence. We identified this weak 5Ј splice site as another region of the type II procollagen gene that is conserved between species. Weak 5Ј and/or 3Ј splice sites are a common feature of many alternatively spliced exons. Similar to COL2A1, reports have been published of other genes containing regulated exons with weak splice sites that are differentially spliced during development, including fibronectin (27), cardiac troponin T (50), myosin phosphatase-targeting subunit-1 (MYPT-1) (51,52), and protein 4.1R (5). We showed that the presence of a weak 5Ј splice site is necessary for the cell type-specific splicing patterns of COL2A1 since conversion to a strong splice site resulted exclusively in exon 2 splicing (type IIA mRNA) regardless of the cell type and state of differentiation.
The functional significance of the stem-loop sequence in regulating exon 2 splicing was demonstrated by constructing a series of mutant mini-genes that contained either deletions or substitutions in this region. Deletion of the entire stem-loop, except nucleotides ϩ4 to ϩ6 that make up the intronic splice site sequence, resulted in a marked inhibition of exon 2 splicing in all cells (SL-Del 1; Fig. 7). Therefore, the weak 5Ј splice site alone is not sufficient to yield exon 2 splicing and, thus, requires additional cis elements present in the stem-loop. The stem-loop does not contain constitutive splicing elements since deletion of this region in combination with a strong splice site adjacent to exon 2 resulted exclusively in exon 2 splicing(ϩ5Јss/SL-Del 1; Fig. 7). Thus, the cis regulatory elements in the stem-loop are functional only in the context of a weak 5Ј splice site. The importance of non-consensus intronic splicing sequences downstream of the weak 5Ј splice site of a regulated exon has been reported in other genes. For example, pyrimidine-rich regions are required for inclusion of the alternatively spliced K-SAM exon in the FGFR-2 gene (53) or exon 6A in the ␤-tropomyosin gene (54). Similar to COL2A1 exon 2 splicing, these regulatory sequences are not functional when the 5Ј splice site sequence is optimized.
Studies of yeast splicing commitment complexes have identified a number of proteins that bind to non-consensus intronic regions downstream of 5Ј splice sites; the functional significance of these interactions in stabilizing the U1 small nuclear ribonucleoprotein-pre-mRNA complex to enhance splicing of an exon was also suggested by the authors (55,56). Therefore, we hypothesized that enhancer cis elements are present in the stem-loop to promote splicing of exon 2 and that these sites would be more functional in cells that naturally express more of the type IIA isoform. Mutations within the conserved double-stranded region of the stem-loop revealed that the uridine (U)-rich site of the stem-loop is a functional enhancer element in all cell types tested except the differentiated chondrocytes (Fig. 7, SL-Del2; Fig. 8). A U-rich enhancer element was identified downstream of the central regulated exon in the MYPT-1 gene which, like COL2A1 exon 2, contains a weak 5Ј splice site and is spliced in a developmentally regulated manner (57). Further studies showed that the splicing factor protein TIA-1 interacts with this MYPT-1 U-rich enhancer (58). By binding to U-rich elements, TIA-1 stabilizes the U1 small nuclear ribonucleoprotein complex association at the 5Ј splice site, thereby promoting splicing of the regulated exon (18,59). Furthermore, Shukla and co-workers (58) also reported that decreased in vivo expression levels of TIA-1 was concomitant with the alternative splicing switch (exclusion) of the regulated MYPT-1 exon. Therefore, TIA-1 is a potential candidate protein that may interact with the U-rich region of the stem-loop downstream of COL2A1 exon 2. A sequence corresponding to part of the COL2A1 U-rich element (AUUUAUUU) is also present within a larger splicing enhancer element identified downstream of the MYPT-1 alternative exon (57). This U-rich stretch is not in the context of secondary structure in the MYPT-1 gene, therefore suggesting the possibility that this element can function as a linear sequence in different cell types. Interestingly, this U-rich element apparently contains some silencing activity in the differentiated chondrocytes used in the present study. This points to the likelihood that different trans-acting splicing factors can bind to the same regulatory cis element and that the expression or regulation of specific splicing factor proteins changes during chondrocyte differentiation. This statement is also supported by splicing of the CCC-1 mutant mini-gene (Fig. 8) in differentiated chondrocytes compared with the other cell types. This mutation did not create an improved intron binding site for U6 snRNA, which displaces U1 snRNA during the constitutive splicing process (8). Here, we predict that this mutation created a novel enhancer site that was recognized by a trans-acting factor present or functional in the RCS cells only.
Mutations in the opposite, adenine (A)-rich region of the doublestranded stem suggested disruption of a functional silencer element in all cell types tested (Fig. 7, SL-Del 3; Fig. 9). The presence of both positive and negative splicing cis elements situated in close proximity within either a regulated exon or an intron has been identified in pre-mRNA encoding a number of proteins including fibronectin, MYPT-1, and tau, for example (57, 60 -62). Some of these regulatory sites can function as linear, independent elements, whereas others are functional only in the presence of the adjacent, antagonistic cis element (22,62). We cannot conclude from the present study that the U-rich and A-rich regions present in the stem-loop can function independently of each other in a linear context to regulate exon 2 splicing. For example, deletion or nucleotide substitution of the A-rich element that resulted in increased exon 2 splicing in all cell types may be due to 1) loss of the silencer activity functional in this region, independent of the U-rich site, 2) disruption of stem-loop secondary structure, thereby allowing the U-rich element to function better as a linear enhancer site, or 3) a combination of the above two scenarios.
The importance of stem-loop secondary structure in regulating COL2A1 exon 2 splicing is supported by the results obtained from the mutant mini-gene CCC-1/GGG-1 (Fig. 9). This mini-gene was constructed to restore secondary structure to the CCC-1 mutant minigene, thereby creating a more stable stem-loop structure (⌬G ϭ Ϫ13.6 kcal; wild-type stem-loop ⌬G ϭ Ϫ7.7 kcal). An altered splicing pattern derived from the CCC-1/GGG-1 mini-gene was noted in all cell types whereby exon 2 splicing was inhibited. This suggests the involvement of another enhancer protein or protein complex that binds specifically to double-stranded RNA in the stem-loop-containing adjacent U-A residues. Numerous reports have described the effects of RNA secondary structure on the regulation of splicing (63,64). In general, with respect to alternative exon splicing, the presence of stem-loop structures negatively regulates splicing by masking the splice site sequence, thereby preventing spliceosome formation at the exon-intron junction. For example, secondary structure was shown to sequester the alternative exon 6B in the chicken ␤-tropomyosin pre-mRNA, resulting in its exclusion from the final mRNA (65)(66)(67). A stem-loop situated downstream of the alternative exon in the human growth hormone gene influenced its splicing since mutations that stabilized the stem-loop resulted in use of an alternative splice site (68). To our knowledge, there are only two reports postulating the presence of stem-loop secondary structure overlapping a weak 5Ј splice site of an alternatively spliced exon (58,69). In both cases it was suggested that the hypothesized stem-loop functioned in masking the weak 5Ј splice site of the alternative exon. Importantly, the present study is the first to experimentally show the existence of a stem-loop at the weak 5Ј splice site of an alternative exon that is regulated in a tissue-specific manner during development. We are in agreement with the generally accepted function of the stem-loop in inhibiting splicing by masking the 5Ј splice site. As described above, the CCC-1/GGG-1 mutation in the stem-loop created a more stable secondary structure that inhibited exon 2 splicing. We hypothesize that this mutation altered an enhancer binding site within the double-stranded region of the stem-loop that potentially serves to unwind or melt the RNA secondary structure, thereby unmasking the 5Ј splice site. Subsequently, other splicing factor proteins may then interact with the appropriate regulatory elements (e.g. the U-rich or A-rich site), resulting in either exon 2 splicing or skipping. Potential candidates that may function to unwind the stem-loop include the family of RNA helicases (70), which have been implicated in altering RNA-RNA interactions and remodeling RNA-protein interaction. For example, two helicases, hPrp28p and p68, have been shown to unwind the RNA duplex formed between U1 snRNA and intronic nucleotides of the 5Ј splice site (71,72). Whether these or other proteins with similar function play a role in destabilizing the COL2A1 stem-loop remains to be determined.
The positive and negative cis elements identified in the present study that potentially serve as binding sites for specific splicing factor proteins/protein complexes are shown diagrammatically in Fig. 10. We hypothesize that cis elements present in the stem-loop are the major regulatory sites occupied by trans-acting splicing factors during COL2A1 splicing. In addition, the less conserved single-stranded loop region was not specifically analyzed for the presence of functional cis elements, and so we cannot rule out this possibility. It is now generally accepted that alternative splicing of a number of regulated exons involves a complex interplay between positive and negative-acting splicing factor proteins, some of which compete for the same cis elements (73)(74)(75)(76)(77). In addition, a further level of complexity exists from the increasing evidence that transcription and pre-mRNA splicing is coregulated (78). Interestingly, it was recently found that the SOX transcription factors important in chondrogenesis, SOX6 and SOX9, can regulate splicing (79) and may also influence what splicing factor proteins are recruited to the site of pre-mRNA splicing in the nucleus.
We predict that chondroprogenitor cells express a different subset of splicing factor proteins compared with differentiated chondrocytes. Subsequently, these splicing factor proteins may act cooperatively or antagonistically at enhancer or silencer sites within the stem-loop to ultimately determine whether the type IIA or type IIB isoform is expressed during cartilage development. We plan to explore this in future studies.