A new long form of c-Maf cooperates with Sox9 to activate the type II collagen gene.

A new long form of the c-Maf transcription factor (Lc-Maf) was identified and shown to interact specifically with SOX9 in a yeast two-hybrid cDNA library screening. Lc-Maf encodes an extra 10 amino acids at the carboxyl terminus of c-Maf and contains a different 3'-untranslated region compared with c-Maf. The interaction between SOX9 and Lc-Maf was further confirmed by co-immunoprecipitation and glutathione S-transferase pull-down assays, which mapped the interacting domain of SOX9 to the high mobility group box DNA binding domain and that of Lc-Maf to the basic leucine zipper motif. In situ hybridizations showed that Lc-Maf RNA was coexpressed with Sox9 and Col2a1 RNA in areas of precartilaginous mesenchymal condensations during mouse embryo development. A DNA binding site of Lc-Maf was identified at the 5'-end of a 48-bp Col2a1 enhancer element near the high mobility group binding site of SOX9. Lc-Maf and SOX9 synergistically activated a luciferase reporter plasmid containing a Col2a1 enhancer and increased the transcription of the endogenous Col2a1 gene. In summary, Lc-Maf is the first transcription factor shown to interact with Sox9, to be coexpressed with Sox9 during an early step of chondrogenesis and to cooperate with Sox9 in activating a downstream target gene of Sox9.

In the multistep process of chondrogenesis, condensation of mesenchymal cells is the first identifiable morphological change and a pivotal step in chondrocyte differentiation (1). These mesenchymal condensations prefigure the shape of the cartilages that are the models for endochondral bone formation. Whereas the cells in the center of the condensations differentiate into mature chondrocytes, the more peripheral cells form a layer of cells that becomes the perichondrium. In the epiphyseal growth plate of endochondral bones, chondrocytes undergo a unidirectional proliferation that is mainly responsible for the longitudinal growth of bones and then change their genetic program and become hypertrophic. Several cytokines, growth factors, extracellular matrix components, and transcription factors have been shown to play important roles in discrete stages of chondrogenesis (2)(3)(4).
Sox9 is a typical transcription factor containing a high mobility group (HMG) 1 -box DNA binding domain and a potent transcription activation domain. Heterozygous mutations of SOX9 in humans cause campomelic dysplasia, a disease characterized by hypoplasia of most endochondral bones and often associated with sex reversal (5)(6)(7)(8)(9). More importantly, Sox9 null mutant cells in mouse embryo chimeras were excluded from mesenchymal condensations and failed to express Col2a1 and other chondrocyte-specific marker genes such as Col9a2, Col11a2, and aggrecan, indicating that Sox9 is required for mesenchymal condensations and subsequent cartilage formation (10). Heterozygous Sox9 mutant mice phenocopy the skeletal anomalies of patients with campomelic dysplasia. Histological analysis of these heterozygous mutant mice indicated that Sox9 haploinsufficiency results in defective cartilage primordia and premature skeletal mineralization. These results suggested the hypothesis that both mesenchymal condensations of cartilage primordia and the rate of transition of chondrocytes into hypertrophic chondrocytes were sensitive to Sox9 dosage (11). Recently, we also showed that phosphorylation of SOX9 by protein kinase A increased its DNA binding and transcriptional activities (12). The levels of phosphorylated Sox9 in the growth plate were shown to be highest in the prehypertropic chondrocytes, the same area where parathyroid hormone/parathyroid hormone-related protein receptor is expressed at high levels. Because parathyroid hormone/parathyroid hormone-related protein also increased the transcriptional activity of SOX9 in DNA transfection experiments, it was suggested that Sox9 may mediate some of the effects of parathyroid hormone-related protein in regulating the transition from proliferating chondrocytes into hypertrophic chondrocytes (13). Sox9 binds to specific sequences in enhancers or promoters of several chondrocyte-specific genes and activates these DNA segments in reporter constructions (14). Moreover, ectopically expressed SOX9 activates one of these genes, the endogenous Col2a1 gene, in some noncartilaginous sites in transgenic mice (15). Sox9 and Col2a1 are co-expressed in all chondrocyte precursors and chondrocytes but are not expressed in hypertrophic chondrocytes (16,17).
Two other members of the Sox family, a new long form of Sox5 (L-Sox5) and Sox6, were identified in chondrocytes and shown to be co-expressed with Sox9 during chondrogenesis (18). The three Sox proteins cooperated to activate expression of the Col2a1 and aggrecan genes in vitro (18). Whereas Sox5and Sox6-null mutants had only mild skeletal abnormalities, Sox5, Sox6 double null mutants showed a very severe chondro-dysplasia characterized by an almost complete absence of cartilage (19).
To identify proteins that interact with Sox9 during chondrogenesis, we used a yeast two-hybrid method consisting of a modified son of sevenless (SOS) recruitment system (20) to screen a chondrocyte cDNA library. One of the cDNAs coded for a new long form of c-Maf (Lc-Maf) that interacted specifically with SOX9. c-Maf is a proto-oncoprotein with a basic leucine zipper (bZip) motif (21) and belongs to a distinct subgroup of the bZip family of transcription factors (22). c-Maf and related proteins such as MafB and NF-AT play important roles in several developmental and cell differentiation processes (23). c-Maf forms both homodimers and heterodimers with other Maf family members and with AP-1 family proteins (24). It also interacts with other transcription factors that either enhance (25) or inhibit its transcriptional activity (26,27). Although a Maf consensus DNA binding sequence was identified by in vitro site selection (28), interaction of c-Maf with other proteins usually changes its binding specificity, hence producing a diversity of regulatory elements in different target genes. We show here that Lc-Maf, which has an extra 10 amino acids at the carboxyl terminus of c-Maf, interacted directly with SOX9. Lc-Maf and Sox9 were co-expressed during mesenchymal condensations, synergistically activated a Col2a1 chondrocyte-specific enhancer, and increased expression of the endogenous Col2a1 gene. Our results are consistent with the hypothesis that Lc-Maf and its interaction with SOX9 may play an important role during mesenchymal condensations, a critical step in chondrogenesis.

EXPERIMENTAL PROCEDURES
cDNA Library Construction and SOS Recruitment System Screening-The construction of a primary chondrocyte cDNA library and SOS recruitment system library screening were performed as described previously (12). One of the positive clones, which encoded the full-length Lc-Maf cDNA, was cotransformed with either the empty vector or pADNS-cJun-SOS or pADNS-SOS-SOX9 plasmid. The cotransformed cells were plated on glucose or galactose plates to test the specificity of interactions as described previously (12).
RNA Analysis-Total RNA extraction and Northern hybridization were performed as described previously (14). The Lc-Maf probe used was a 700-bp fragment cleaved by ApaI and EcoRI from the 3Ј-untranslated region (UTR) of a cDNA clone encoding the full-length Lc-Maf cDNA. Total RNA was extracted by using the RNeasy Mini Kit (Qiagen, Valencia, CA) according to the manufacturer's instructions. Fifteen micrograms of each RNA sample was eletrophoresed, and Northern hybridization with a Col2a1-specific probe was performed as described previously (14). The blot was striped and rehybridized with a glyceraldehyde-3-phosphate dehydrogenase probe (Ambion, Austin, TX) as loading control.
In Vitro Transcription-Translation-Full-length and deleted constructs of SOX9 and Lc-Maf were labeled with [ 35 S]methionine and generated by in vitro transcription-translation with the single-tube TNT protein system (Novagen, Madison, WI) according to the manufacturer's instructions.
Co-immunoprecipitation-COS-7 cells were cotransfected with pcDNA3.1 expression plasmids for SOX9 and Lc-Maf containing a FLAG tag sequence at its N terminus. Twenty-five microliters of cell lysates was incubated with 15 l of affinity-purified SOX9 antibody, SOX9 preimmune serum, or c-Maf antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), which also recognizes Lc-Maf, in phosphatebuffered saline containing 0.5% Nonidet P-40 for 3 h at 4°C. Then 2.5 l of protein A-Sepharose 4B (Sigma) resin was added in a final volume of 50 l. After incubation at 4°C for another 2 h, the resin was washed three times with 1 ml of incubation buffer. The resin was then boiled in SDS-PAGE loading buffer, electrophoresed in an SDS-10% polyacrylamide gel, and transferred to a nylon membrane. Western blotting was performed as described previously (12) with either mouse FLAG m2 antibody (Sigma) or rabbit SOX9 antibody diluted 1:1000.
GST Pull-down Assay-GST-SOX9 and GST-Lc-Maf were expressed and purified according to the method described previously (12). Two microliters of 35 S-labeled wild-type SOX9 or SOX9 deletions was incubated with 500 ng of either GST or GST-Lc-Maf in an interaction buffer containing 25 mM HEPES, pH 7.9, 100 mM NaCl, 1 mM EDTA, 0.5% Nonidet P-40, 0.5 mM phenylmethylsulfonyl fluoride, and 0.5 mM dithiothreitol for 30 min at room temperature; 20 l of glutathione-agarose resin (Sigma) that had been washed three times in interaction buffer was then added in a total volume of 100 l. After incubation at 4°C for another 30 min, the resin was washed three times with 500 l of the incubation buffer. The resin was then boiled in 10 l of SDS-PAGE loading buffer and run on an SDS-10 or 15% polyacrylamide gel, which was then dried and autoradiographed. A parallel SDS-polyacrylamide gel loaded directly with 2 l of each labeled protein sample was run as the loading control and analyzed in the same manner. 35 S-Labeled wild type Lc-Maf and Lc-Maf deletions with GST or GST-SOX9 were incubated in the same manner.
Electrophoretic Mobility Shift Assays-The R2 18-bp Col2a1 enhancer probe was described previously (14). Wild-type and mutant R1 20-bp probes containing 2-bp mutations were prepared by annealing the upper and lower chains of nucleic acid fragments synthesized by IDT (Coralville, IA) with overhanging 5Ј-end BamHI restriction sites. The probes were labeled with [ 32 P]dCTP as described previously (14). Two microliters of Lc-Maf, generated by in vitro transcription-translation, was incubated with these probes in DNA binding buffer as described previously (12). The samples were then fractionated by electrophoresis through a nondenaturing 5% polyacrylamide gel in 0.5ϫ Tris borate-EDTA at 150 V for 3 h, and the gel was autoradiographed.
Cell Culture and Transfection Experiments-COS-7 cells were cul-

FIG. 1. Interaction of Lc-Maf with SOX9 in a yeast two-hybrid assay.
A, specific interaction between SOX9 and Lc-Maf. A plasmid isolated from a positive colony by SOS recruitment system library screening, which encodes the full-length Lc-Maf cDNA, was retransformed into a cdc25-2 temperature-sensitive mutant yeast strain together with either empty pADNS, pADNS-cJun-SOS, or pADNS-SOS-SOX9. Four independent colonies generated from each plasmid combination were replica-plated onto galactose and glucose plates and grown at 37°C for 4 days. The expression of Lc-Maf was under the control of a GAL-1 promoter, and only the colonies cotransformed with SOS-SOX9 and Lc-Maf grew on the galactose plates but not on the glucose plate. B, alignment of c-Maf and Lc-Maf proteins, showing functional domains. LLLLLY represents the bZip motif. Lc-Maf had an extra 10 amino acids at the carboxyl terminus of c-Maf and a completely different 3Ј-UTR. C, alignment of the last 10 C-terminal amino acids of Lc-Maf (mouse), a long form of c-Maf (chicken), and V-Maf. Vertical lines show identical amino acids. tured as described previously (12). The cells were then transfected with luciferase reporter plasmids containing an 89-bp Col2a1 promoter without (p89Luc) or with four copies of a 48-bp (4ϫ48-p89Luc) or two copies of a 100-bp (2ϫ100-p89Luc) Col2a1 enhancer element and the plasmid pSV2-␤-gal as an internal control for transfection efficiency. The reporter plasmids and pSV2-␤-gal were transfected in a ratio of 3:1 as described previously (12). Expression plasmids for SOX9 (100 ng) and different c-Maf plasmids (including c-Maf, Lc-Maf, and the Lc-maf deletion mutant, Lc-Maf^TA, each 100 ng) were transfected by using FuGene6 (Roche Molecular Biochemicals) according to the manufacturer's instructions. Luciferase and ␤-galactosidase activities were assayed in cell lysates prepared as described previously (12). Reporter activities were reported as the average of triplicate cultures in one of several representative experiments as previously described (12).
In Situ Hybridization-The Lc-Maf probe was the same as that we used for Northern hybridization. The preparation of probes for Lc-Maf, Sox9, and Col2a1 was described previously (18). In situ hybridization of mouse embryo sections at different developmental stages was performed as described previously (18).
Plasmid Constructions-Full-length SOX9 and the deletion N9-304 were cloned into pcDNA 3.1 as described previously (12,14). Similarly, full-length SOX9 was cloned into GEX4T3 (Amersham Biosciences). All other SOX9 deletions were generated either by one-step PCR (N9-400, C9-304, N9-199, N9-HMG, N9-104) or by a two-step PCR (F9^PQA) as described previously (18). All deletions were inserted into pcDNA 3.1 between the BamHI and EcoRI restriction sites. Full-length cDNAs of c-Maf in pBluescript and c-Maf^TA in pT7␤plink were a gift from Dr. James I. Morgan. c-Maf was subcloned into pcDNA 3.1 vector between the HindIII and XbaI restriction sites, and c-Maf^TA released by EcoRI and SpeI digestion was subcloned into the pcDNA 3.1 between the EcoRI and XbaI sites. An XhoI fragment from a cDNA encoding the full-length Lc-Maf was inserted into the XhoI restriction site in pcDNA 3.1-c-Maf to generate pcDNA 3.1-Lc-Maf and in pcDNA 3.1-c-Maf^TA to generate pcDNA 3.1-Lc-Maf^TA. Lc-Maf-N and Lc-Maf-C were generated by PCR amplification and then inserted into pcDNA 3.1 plasmid in the same manner as described for the SOX9 deletions. pcDNA 3.1-FLAG-Lc-Maf and GEX4T3-Lc-Maf were constructed by inserting a PCR fragment that includes the full-length coding region of Lc-Maf between the BamHI and EcoRI sites in pcDNA 3.1-FLAG (14) and GEX4T3 in the same manner as described for SOX9 deletions. All constructions were verified by DNA sequencing.

RESULTS
Interaction of Lc-Maf with SOX9 in a Yeast Two-hybrid Assay-Using the full-length human SOX9 as bait, we screened a primary chondrocyte cDNA library by the using SOS recruitment system as described previously (12). We obtained two independent cDNA clones encoding c-Maf. One of these contained the entire coding sequence of a new isoform of c-Maf, designated here as Lc-Maf. We used this clone to test specific interaction with Sox9 (Fig. 1A). Rescue of the growth of cdc25-2 Saccharomyces cerevisiae cells on galactose at high temperature only occurred upon cotransformation of both SOS-SOX9 and Lc-Maf plasmids, indicating that there is a specific interaction between Lc-Maf and SOX9. The Lc-Maf cDNA had 5Јuntranslated and coding regions identical to the previously identified mouse c-Maf (21) but had an extra 10 amino acids at the carboxyl terminus and contained a completely different 3Ј-UTR (Fig. 1B). Protein alignment of the last 10 amino acids at the carboxyl terminus of Lc-Maf with those of a similar long form of chicken c-Maf (GenBank TM D28598) and v-Maf (29) (Fig. 1C) revealed that they were highly conserved. c-Maf and Lc-Maf probably result from differential splicing of a single gene.
RNA Expression of Lc-Maf-To investigate the RNA expression pattern of Lc-Maf, we used a 700-bp fragment from the 3Ј-UTR of Lc-Maf as a probe and performed a Northern hybridization of RNA from various cell lines, including chondrocytes, and various tissues from newborn mice (Fig. 2). The 4.5-kb Lc-Maf RNA was expressed strongly in primary chondrocytes, less in MC615 mouse immortalized chondrocytes, and even less in 10T1/2 fibroblasts, which express low levels of Sox9 and Col2a1. Lc-Maf transcripts were also present in EL4 lymphoma cells. Lc-Maf RNA was highly expressed in brain, skin, intestine, and kidney but at only very low levels or not at all in spleen, heart, testis, and tongue. Therefore, although Lc-Maf was widely expressed, it is not a ubiquitous factor.
Co-immunoprecipitation of Lc-Maf with SOX9 -To confirm the interaction between Lc-Maf and SOX9, we cotransfected COS-7 cells with Sox9 and an Lc-Maf expression plasmid containing an in-frame FLAG tag sequence at its N terminus and performed a co-immunoprecipitation. As shown in Fig. 3, A and B, Lc-Maf and SOX9 were efficiently expressed in COS-7 cells. Lysates of transfected cells were immunoprecipitated with a SOX9 antibody, and the precipitates were analyzed by Western blotting with an anti-FLAG m2 antibody. As shown in Fig. 3C, only the cell lysates cotransfected with SOX9 and Lc-Maf contained an immunoprecipitated polypeptide the size of FLAG-Lc-Maf. Next, we tested whether c-Maf antibody could also co-immunoprecipitate SOX9 in transfected COS-7 cell lysates. As shown in Fig. 3D, SOX9 was co-immunoprecipitated with c-Maf antibody but not with Sox5 antibody.
Interaction of the bZip Motif of Lc-Maf with the HMG DNA Binding Domain of SOX9 -To identify the interacting domains in SOX9 and Lc-Maf, we generated various deletions of SOX9 (Fig. 4A) and Lc-Maf (Fig. 4C). All deletions were labeled with [ 35 S]methionine by using an in vitro transcription-translation reaction. A GST pull-down assay was performed by incubating purified GST or GST-Lc-Maf with full-length SOX9 or with various deletions of SOX9. GST did not pull down either fulllength SOX9 (Fig. 4B, lane 0) or any SOX9 deletions (data not shown). All SOX9 constructions containing the HMG-box DNA binding domain were pulled down by GST-Lc-Maf (Fig. 4B,  lanes 1-4 and 6). A deletion construct containing the N terminus of SOX9 (lane 8) and a deletion containing the carboxylterminal transactivation domain of SOX9 (lane 5) could not be pulled down, indicating that the HMG-box DNA binding domain of SOX9 mediated the interaction with Lc-Maf. Likewise, GST-SOX9 pulled down both the full-length Lc-Maf and truncated Lc-Mafs without the N-terminal activation domain or a small segment of Lc-Maf containing the bZip motif (Fig. 4D,  lanes 1, 2, and 4). In contrast, a truncated Lc-Maf lacking only the carboxyl-terminal part of the bZip domain was not pulled down by GST-SOX9 (Fig. 4D, lane 3), indicating that the bZip domain of Lc-Maf is the interacting motif.
Binding of Lc-Maf to a GC-rich Sequence at the 5Ј-End of the 48-bp Col2a1 Enhancer Element-Previously SOX9 was shown to bind strongly to an HMG-related sequence at the 3Ј-end of the 48-bp Col2a1 enhancer element (18). However, no binding of Lc-Maf was observed by using an 18-bp segment at the 3Ј-end of this 48-bp segment (R2) to which SOX9 binds strongly (Fig. 5B). In contrast, Lc-Maf bound to a 20-bp fragment containing the 5Ј region of the 48-bp element (R1) (Fig. 5B). The sequence GGCTCTG, which is only 1 nucleotide different from the sequence GGCTCAG, which was previously found to be a binding site for c-Maf in the L7 Purkinje neuron-specific promoter (21), was present in this 20-bp Col2a1 segment. To narrow down the Lc-Maf binding sequence in the 48-bp element, we generated a series of 2-bp mutations shown in much more weakly, suggesting that the GGCTCTG sequence is important for Lc-Maf binding.
Synergistic Activation of a Col2a1 Chondrocyte-specific Enhancer by Lc-Maf and SOX9 -To test the functional relevance of the interaction between Lc-Maf and SOX9, we cotransfected the of Lc-Maf and SOX9 expression plasmids into COS-7 cells to test whether they cooperated in activating reporter plasmids containing four copies of a 48-bp or two copies of a 100-bp Col2a1 chondrocyte-specific enhancer (12). We have observed previously that SOX9 strongly activates the 4 ϫ 48-bp Col2a1 reporter (12). In comparison, the effect of Lc-Maf was more moderate (Fig. 6A). Coexpression of Lc-Maf and SOX9 produced more than an additive activation of the 4 ϫ 48-bp Col2a1 enhancer (Fig. 6A). This synergism was also observed with the 2 ϫ 100-bp Col2a1 enhancer, which Lc-Maf activated as strongly as SOX9 did, suggesting that Lc-Maf itself is a strong activator of this Col2a1-specific enhancer (Fig. 6B). Cotransfection of Lc-Maf and SOX9 produced 3 times more activation than did each plasmid alone. This joint activation was not observed with SOX9 and Lc-Maf^TA, a truncated Lc-Maf without the transcription activation domain (Fig. 6A), or with Lc-Maf and L-Sox5 (data not shown). The specificity of Lc-Maf activation of the Col2a1 enhancer was further demonstrated by the lack of response of M7, a mutant 48-bp Col2a1 enhancer containing a 2-bp point mutation in the Lc-Maf binding site (Fig. 6C). As shown in Fig. 5B, M7 decreased the binding of Lc-Maf. Moreover, with mutant enhancers containing either a 2-bp mutation located between the Lc-Maf and SOX9 binding sites or a 2-bp mutation in the SOX9 binding site, activation by Lc-Maf was unchanged. These experiments suggested that DNA binding was required for the activation by Lc-Maf of this Col2a1 enhancer. These results indicated that Lc-Maf was a potent activator of a Col2a1-specific enhancer and that Lc-Maf cooperated with SOX9 to activate this enhancer more strongly.
Increased Endogenous Col2a1 RNA Expression by Cotransfection of SOX9 and Lc-Maf Expression Plasmids-We next tested whether coexpression of Lc-Maf and SOX9 increased transcription of the endogenous Col2a1 gene. Transfection of either SOX9 (Fig. 7, lanes 2) or Lc-Maf (Fig. 7, lanes 3) expression plasmids alone into 10T1/2 cells, which express low levels of Col2a1, produced almost no increase of this transcript. However, cotransfection of Lc-Maf and SOX9 produced significantly more Col2a1 RNA (Fig. 7, lanes 4), suggesting that Lc-Maf and SOX9 cooperate, in accord with the previous transient transfection experiments. Similar results were obtained with MC615 cells (Fig. 7, lanes 5-8), in which expression of the endogenous Col2a1 is low after repeated passages in culture. These results strongly suggest that Lc-Maf and SOX9 cooperatively activate the endogenous Col2a1 gene.
Coexpression of Lc-Maf with Sox9 and Col2a1 Genes during Mesenchymal Condensation-To better understand the physiological significance of the interactions between SOX9 and Lc-Maf, we performed in situ hybridizations with sections from mouse embryos at various developmental stages by using probes specific for Lc-Maf, Sox9, and Col2a1. As shown in Fig.  8, Lc-Maf, Sox9, and Col2a1 were coexpressed in cartilaginous primordia during mesenchymal condensation between embryonic days 10.5 and 12.5 (E10.5 and E12.5), including the first and second branchial arches (Fig. 8A), forelimb and hind limb buds (Fig. 8A), vertebrae (Fig. 8B), and Meckel's cartilage and nose (Fig. 8C). In E13.5 embryos, expression of Lc-Maf was strong in the perichondrium, whereas Sox9 was strongly expressed in the center of the condensations, where cells undergo overt chondrocyte differentiation (Fig. 8D). This suggests that Lc-Maf and the interaction between Lc-Maf and Sox9 may play a role in chondrogenesis, mainly during mesenchymal condensations.

DISCUSSION
Using a modified yeast two-hybrid system to screen for polypeptides interacting with SOX9, we identified a new long form of c-Maf (Lc-Maf). Lc-Maf had 10 extra amino acids at the carboxyl terminus and a completely different 3Ј-untranslated region. A similar long form of c-Maf was identified in chickens as a differential splicing product (GenBank TM D28598), suggesting that the mouse Lc-Maf may be derived by the same mechanism. The functional difference between the short and long forms of c-Maf is still unknown. Very similar Northern and in situ hybridization patterns were observed by using 3Ј-UTR probes specific for either the short form of c-Maf or for Lc-Maf (data not shown), suggesting that these isoforms may have redundant functions. In transient transfection experiments, the short form of c-Maf also cooperated with SOX9 to synergistically activate the 48-bp Col2a1 enhancer element, although to a lesser degree than did Lc-Maf. Furthermore, that v-Maf also has a similar extra 10 amino acids suggests that v-Maf may be derived from Lc-Maf instead of c-Maf.
The interaction between Lc-Maf and SOX9 was further demonstrated by co-immunoprecipitation and GST pull-down assays. The GST pull-down assays not only further supported the were labeled with [ 32 P]dCTP. All labeled probes were adjusted with unlabeled probe to achieve identical specific activities. A protein mixture from an in vitro transcription-translation reaction using the empty vector pcDNA as template indicated by minus sign was used as a control.
direct interaction between SOX9 and Lc-Maf but also mapped the interacting segments to the HMG-box DNA binding domain of SOX9 and the bZip motif of Lc-Maf. These domains were also found to be used in the interactions of these proteins with other transcription factors. For example, the HMG DNA binding domain of SOX9 interacts with SF-1 (29), and the bZip motif of MafB interacts with the DNA binding domain of Ets-1 (30). The Lc-Maf binding sequence in the 48-bp Col2a1 enhancer element is not the typical Maf recognition element, which was determined by in vitro site selection (27). We have not yet determined whether interaction of Lc-Maf with Sox9 and perhaps other proteins increases its affinity for the 48-bp Col2a1 enhancer. Although Lc-Maf alone efficiently activated a Col2a1 chondrocyte-specific enhancer in transient transfection experiments, cotransfection of Lc-Maf with SOX9 resulted in activation that was more than additive, suggesting that there is a synergistic cooperation between these two factors. This was further confirmed in a Northern hybridization experiment showing an increased expression of endogenous Col2a1 when expression vectors for these two factors were cotransfected. However, forced expression of Lc-Maf and SOX9 in other fibroblasts in which the Col2a1 gene was silent failed to induce Col2a1 expression. It is possible that in these cells other factors or coactivators may be needed to switch on the Col2a1 gene.
To investigate the physiological relevance of the interaction between Lc-Maf and Sox9, we performed a detailed in situ hybridization using mouse embryo sections from different developmental stages. During chondrogenesis, Lc-Maf was coexpressed with Sox9 and Col2a1 in all areas of mesenchymal condensations, but after E13.5, when overt differentiation of chondrocytes occurs, expression of Lc-Maf was restricted to the perichondrium. We conclude that Lc-Maf and Sox9 may interact during chondrogenic mesenchymal condensation. After condensation, when cells overtly differentiate into chondrocytes, Lc-Maf expression was shut off, whereas the levels of Sox9, L-Sox5, and Sox6 remained high or increased. It should be noted that the studies that showed interactions between Lc-Maf and Sox9 were performed in COS-7 cells, which are not the  3 and 7), and SOX9 plus Lc-Maf (lanes 4 and 8). The cells were cultured for 36 h after transfection, and RNA was extracted for analysis by Northern hybridization with a Col2a1-specific probe. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH)-specific hybridization was used as the RNA loading control, and a Col2a1 RNA sample from RCS cells that had been hybridized in the same manner was used as a control (left panel). same cells in mesenchymal condensations, where these interactions are believed to occur. The conclusions of this study have thus to be qualified by this limitation.
Sox9 was previously shown to be required for chondrocyte differentiation and cartilage formation (10). Besides being expressed in chondrogenic tissues, Sox9 is also expressed in the notochord, central nervous system, heart, kidney, pancreas, genital ridges, early undifferentiated gonads, and Sertoli cells (31). Patients with campomelic dysplasia have other anomalies, including sex reversal and occasionally heart and kidney defects in addition to hypoplasia of endochondral bones. This suggests that Sox9 has additional roles in development other than chondrogenesis. An attractive hypothesis to account for its multiple functions is that Sox9 interacts with different transcription factors during organogenesis of different organs. For example, Sox9 and steroidogenic factor 1 were shown to interact in Sertoli cells and to regulate transcription of the human anti-Mullerian hormone gene (29).
L-Sox5 and Sox6, the genes for two other Sox proteins that belong to a different subfamily, are coexpressed with Sox9 during chondrogenesis. Although individual L-Sox5 and Sox6 mutants have relatively mild effects, L-Sox5-Sox6 double mutants have very severe chondrodysplasia and are characterized by an almost complete absence of cartilage. In these mice, chondrogenic mesenchymal condensations occur normally, but cells in the condensations are unable to differentiate into chondrocytes and to express the high level of cartilage extracellular matrix genes that are characteristic of chondrocytes (19). In addition to the Sox9 binding site, the 48-bp Col2a1 enhancer contains three HMG-box binding sites to which L-Sox5 and Sox6 presumably bind. Mutations in two of these sites outside the Sox9 binding site abolished enhancer activity in vivo. The Lc-Maf binding site overlaps with one of these additional HMG binding sites but not with that for Sox9 (18). The levels of L-Sox5 and Sox6 are relatively lower during mesenchymal condensations than at later stages of chondrogenesis. A binding competition assay showed that Lc-Maf competes with L-Sox5 to bind the 48-bp Col2a1 enhancer element (data not shown). Because the in situ hybridization results showed co-expression of Lc-Maf with Sox9 and Col2a1 RNAs during chondrogenic mesenchymal condensations, we propose that Lc-Maf may inhibit binding of L-Sox5 and Sox6 to the Col2a1 enhancer and similar enhancers in other chondrocyte genes during this early stage of chondrogenesis. After condensation, Lc-Maf was strongly expressed in the perichondrium but is no longer expressed in chondrocytes, in which L-Sox5, Sox6, and Sox9 are strongly expressed. Thus, Lc-Maf, L-Sox5, and Sox6 may function at different steps of chondrogenesis. We hypothesize that Lc-Maf cooperates with Sox9 to activate Col2a1 and perhaps other genes needed for chondrogenic mesenchymal condensations, whereas L-Sox5 and Sox6 would function together with Sox9, mainly at a subsequent step during overt chondrocyte differentiation when cartilage extracellular matrix genes are strongly expressed.