Versican/PG-M Regulates Chondrogenesis as an Extracellular Matrix Molecule Crucial for Mesenchymal Condensation*

Mesenchymal cell condensation is an essential step for cartilage development. Versican/PG-M, a large chondroitin sulfate proteoglycan, is one of the major molecules expressed in the extracellular matrix during condensation. However, its role, especially as an environment for cells being condensed, has not been elucidated. Here we showed several lines of evidence for essential roles of versican/PG-M in chondrogenic condensation using a new chondrocytic cell line, N1511. Chondrogenic stimuli (treatment with parathyroid hormone, dexamethasone, 10% serum) induced a marked increase in the transcription and protein synthesis of versican/PG-M. Stable antisense clones for versican/PG-M, depending on suppression of the expression of versican/PG-M, showed different capacities for chondrogenesis, as indicated by the expression and deposition of aggrecan, a major chondrocytic cell product. The cells in the early stages of the culture only expressed V0 and V1 forms, having more chondroitin sulfate chains among the four variants of versican/PG-M, and treatment of those cells with chondroitinase ABC suppressed subsequent chondrogenesis. Furthermore, treatment with β-xyloside, an artificial chain initiator of chondroitin sulfate synthesis to consequently inhibit the synthesis on the core proteins, suppressed chondrogenesis. In addition, forced expression of the variant V3, which has no chondroitin sulfate chain, disrupted the deposition and organization of native versican/PG-M (V0/V1) and other extracellular matrix molecules known to be expressed during the mesenchymal condensation and resulted in the inhibition of subsequent chondrogenesis. These results suggest that versican/PG-M is involved in positively regulating the formation of the mesenchymal matrix and the onset of chondrocyte differentiation through the attached chondroitin sulfate chains.

Mesenchymal cell condensation is the first step of organogenesis in many tissues. Cartilage formation in limb buds, one of the most typical examples of organogenesis, starts with the condensation of chondrogenic mesenchymal cells (1). This step is thought to be essential for subsequent skeletal development in vertebrates (2) and requires the involvement of not only some cell growth and morphogenetic factors, such as transforming growth factor-␤ (3), growth differentiation factor-5 (4), and bone morphogenetic proteins (BMPs) 4 (5), but also some cellular adhesion and extracellular matrix molecules such as N-cadherin (6), neural cell adhesion molecule (7), fibronectin (8), tenascin (9), versican/PG-M (10), hyaluronan (11), syndecan (12), and perlecan (13).
Versican/PG-M was originally isolated as PG-M (medium-sized chondroitin sulfate proteoglycan) from the core (the mesenchymal cell condensation area) of chick embryonic limb bud at stage 23 (10). The cDNA of versican/PG-M was cloned as versican, a fibroblast proteoglycan (14), and also as one of the alternatively spliced forms of the PG-M core protein (15). Versican/PG-M has a molecular mass of more than 1,000 kDa and consists of two globular domains at the N and C termini (G1 and G3 domain, respectively) and the two chondroitin sulfate-attachment domains (CS-␣ and CS-␤) (16, also see Fig. 4A). The G1 and G3 domains are commonly found in proteoglycans belonging to the aggrecan family and are binding sites for hyaluronan and oligosaccharides, respectively (17,18). The CS-␣ and CS-␤ domains give unique properties to this proteoglycan in that multiple alternative splicing yields the following four variant forms with different numbers of the attached chondroitin sulfate chains: V0 having CS-␣ and CS-␤, V1 having CS-␤, V2 having CS-␣, and V3 having neither of the two (19 in chicken; 15,16, and 20 in mouse; 14 and 21 in human). An important role for versican/PG-M in the condensation process has been suggested by its unique properties as follows: 1) high transient expression in the mesenchymal cell condensation area during development of cartilage (10,(22)(23)(24), heart (25,26), hair follicles (25), and kidney (27); 2) specific binding to fibronectin (28), type I collagen (28), hyaluronan (11,29), and tenascin (9), also major matrix molecules in the mesenchymal cell condensation area; and 3) ability to inhibit cell adhesion to a variety of extracellular matrix molecules such as fibronectin or type I collagen through its chondroitin sulfate chains (30 -32). Versican/PG-M interacts with cell surface annexin VI and inhibits the subsequent cellspreading step that needs the rearrangement of cytoskeletal elements (33).
Exogenous treatment with versican/PG-M of mesenchymal cells isolated from chick embryonic limb buds at stage 22/23 inhibited their spreading onto the substrates and enhanced their aggregation and further differentiation into chondrocytes (34). Previous studies reported that chondrocyte shape influences the differentiation via changes in cytoskeletal elements (35), and it can be regulated by interactions between matrix substrate molecules and receptors on the cell surface under the optimum condition of high cell density. Versican/PG-M also modulates embryonic chondrocyte morphology (36). We hypothesize that versican/PG-M is greatly involved in the mesenchymal cell condensation and subsequent chondrogenesis by modifying cell-extracellular matrix molecule interactions so as to influence the cell shape. To address this hypothesis, null mutation of the versican/PG-M gene in murine was supposed to provide direct results. However, this mutation resulted in embryonic death at 10.5 days post-coitum because of a heart abnormality before the onset of the chondrogenic mensenchymal cell condensation (37), and so no direct evidence has been obtained for the above hypothesis.
The transfection of mesenchymal cells in culture with sense or antisense cDNA in expression vectors might also have provided some evidence. However, primary cell cultures are not ideal systems to pursue such experiments because of the limited number of mesenchymal cells at the same stage as well as the instability of their phenotype. We have recently established a chondrocytic cell line, N1511 without transformation (38). In N1511 cells, the entire process of chondrogenic differentiation, including mesenchymal cell condensation, can be reproduced by treatment with natural growth factors and cytokines, as characterized by changes in cell morphology and the onset of the expression of cartilage-specific genes such as type I, II, IX, and X collagen and aggrecan. These properties reflect well the in vivo chondrogenesis and further chondrocyte differentiation, suggesting that N1511 cells could provide an in vitro model system to study cartilage development from the step of a mensenchymal condensation as a process of groundwork for further differentiation.
In this study, we first demonstrate that, immediately after induction, N1511 cells expressed versican/PG-M (V0 and V1 forms) at a high level, resembling the in vivo phenotype of condensing mesenchymal cells in the early phase, and then differentiated into chondrocytes. We next examined the effect of abrogation of versican/PG-M expression and interruption of versican/PG-M deposition into the matrix during the multiple steps of the chondrogenesis of N1511 cells. The inhibition by antisense of the synthesis of versican/PG-M suppressed chondrogenesis assessed by the gene expression and deposition in the matrix of aggrecan. In addition, overexpressed V3, which is the smallest spliced variant form of versican/PG-M without both chondroitin sulfate attachment domains, affected the deposition of native (endogenously produced) versican/PG-M in the ECM and also resulted in the suppression of chondrogenic differentiation. Furthermore, reduction of chondroitin sulfate chains in the ECM by chondroitinase ABC digestion and inhibition of native synthesis of chondroitin sulfate chains on the core proteins by ␤-xyloside treatment suppressed chondrogenesis. These results indicate that V0 and V1 forms of versican/PG-M expressed in the early phase of chondrogenesis serve as ECM molecules involved in mesenchymal cell condensation and promote further chondrocyte differentiation. Evidence is also provided that chondroitin sulfate chains of those proteoglycans are functional sites for the modification of cell-matrix interactions during mesenchymal condensation and the subsequent steps of chondrogenesis.

EXPERIMENTAL PROCEDURES
Cell Line and Culture Conditions-N1511 cells, a cell line established from chondrocytic cells in the rib cage of a 4-week-old male p53-null mouse, reproduce each step of chondrogenesis (38). The cells were maintained in minimum essential medium ␣ (␣-MEM) supplemented with 10% fetal calf serum (FCS), penicillin, and streptomycin at 37°C in 5% CO 2 until 80% confluent. The cells were then plated at a density of 1.5 ϫ 10 6 cells/35-mm dish, and after 24 h (defined as day 0), chondrogenesis was induced by combined treatment with 1 ϫ 10 Ϫ6 M dexamethasone (Calbiochem) and 1 ϫ 10 Ϫ7 M rat PTH-(1-34) (Sigma) in the presence of 10% FCS. The medium was changed every other day.
Transfection-The cells at 40% confluency in 100-mm dishes were cotransfected with 3 g of vector containing antisense strand, vector containing sense strand, or vector alone (mock) and 0.1 g of PurR plasmid vector, using 12 l of TransIT-LT1 reagent (Mirus Corp., Madison, WI) and 500 l of Opti-MEM TM (Invitrogen) in 10 ml of ␣-MEM containing 10% FCS. The transfectants were screened with 5 g/ml puromycin for 3 days, and a pool of stable transfectants was again screened for 2 days using 5 g/ml puromycin. We also transfected the cells with vector containing the smallest spliced form (V3) of versican/ PG-M to examine the effect of a transient overexpression. To improve the transfection efficiency and to remove the effect of the untransfected cell debris on the cell adhesion, the transfection was performed at 40% confluency in 100-mm dishes. After 24 h of transfection, the cells were trypsinized and re-plated in the 35-mm dishes at a density of 1.5 ϫ 10 6 cells/dish.
Antibodies, Immunofluorescent Staining, and Alcian Blue Staining-For rabbit antibodies against the alternatively spliced domain, CS-␤, the cDNA fragment encoding this region was inserted into the protein A gene fusion vector, pRIT-2T (Amersham Biosciences). The fusion protein produced in transfected Escherichia coli N4830-1 was purified from cell lysate using IgG-Sepharose 6 Fast Flow and used for immunization. The antibodies obtained were designated anti-mouse CS-␤ domain antibodies. Rabbit polyclonal antibodies against rat aggrecan also proven to be specific for mouse aggrecan (38) were a gift from Dr. Yada in this laboratory. 2B1, a mouse monoclonal antibody against versican/ PG-M purified from human yolk sac tumor (39), and biotinylated HAbinding protein (40) were purchased from Seikagaku Corp. (Tokyo, Japan). A polyclonal antibody against mouse fibronectin was obtained from Santa Cruz Biotechnology (Santa Cruz, CA). For immunostaining, whole-mount cultures at appropriate time points were rinsed with PBS, fixed with 10% formalin/PBS for 10 min at room temperature, and then incubated for 1 h with the following primary antibodies at the indicated dilution, respectively: anti-mouse CS-␤ domain antibodies (dilution, 1:1000), anti-mouse aggrecan antibodies (1:100), monoclonal antibody for human versican/PG-M, 2B1 (1:100), and anti-fibronectin antibody (1:100). For the fluorescent staining, Alexa Fluor TM 488 and 596 of goat anti-rabbit IgG antibodies and goat anti-mouse IgG antibodies, respectively (Molecular Probes, Eugene, OR), were used for 1 h as secondary antibodies (1:1000). For the immunochemical staining, Histofine kits using biotinylated anti-rabbit or anti-mouse IgG antibodies as secondary antibodies and peroxidase-conjugated streptavidin and 3,3Ј-diaminobenzidine-HCl (DAB) as color-developing reagents (Nitirei, Tokyo, Japan) were used. Biotinylated HA-binding protein (1:500) (Seikagaku Corp., Tokyo, Japan) and FITC-conjugated streptavidin (1:1000; Molecular Probes, Eugene, OR) were used for HA staining as described previously (40). For Alcian blue staining, the cultures were stained with 0.5% Alcian blue 8GX (Sigma) in 0.1 M HCl, pH 2.0, overnight and destained with 0.1 M HCl for 10 min as described previously (38). Photographs of stained cultures were taken using a CCD camera under the same conditions. Immunostained and Alcian blue-positive areas (at least 50 areas that were obtained by three independent experiments performed in triplicate) were quantitatively measured with NIH IMAGE software (version 1.57).
Treatment with Chondroitinase ABC-Five units/ml protease-free chondroitinase ABC (Seikagaku Corp., Tokyo, Japan) was treated in the culture medium (2 ml of media/35-mm dish) up to day 4 every other day, and the cells were then rinsed with PBS several times and cultured in the regular induction medium for a further 2 weeks. For the digestion of the chondroitin sulfate chains, protein samples in the extracts with RIPA buffer containing protease inhibitors (50 mM Tris-HCl, pH 7.4, 150 mM NaCl, 1% Nonidet P-40, 0.25% sodium deoxycholate, 1 mM EDTA, 2 mM phenylmethylsulfonyl fluoride, and 3.6 mM pepstatin A) were treated with 5 units of chondroitinase ABC in the presence of protease inhibitors (41) at 37°C for 45 min. After being boiled for 2 min at 100°C, samples containing 50 mg of proteins were subjected to SDS-PAGE with a stacking gel of 3.75% (w/v) and a separation gel of 5% (w/v) acrylamide and transferred to polyvinylidene difluoride membranes as described previously (23). The core protein derivatives of versican/ PG-M were detected immunochemically by using the anti-mouse CS-␤ domain antibodies and peroxidase-conjugated goat anti-rabbit IgG, and blots were visualized using the ECL chemiluminescence detection system (Amersham Biosciences).
Treatment with ␤-Xyloside-␤-Xyloside, xyloside-S-(CH 2 ) 6 -CH 3 (42), was dissolved in ethanol and stored at Ϫ20°C (0.2 M). The cultures (35-mm dish) were treated with up to 2.0 mM of ␤-xyloside from day 0 to day 5, rinsed with PBS several times, and cultured in the regular induction medium for a further 2 weeks. The cell layer and medium were harvested at the indicated days. ␣-Xyloside was used for a negative control. The cultured medium was combined from day 1 to day 5 and precipitated with 3 volumes of cold ethanol containing 1.3% potassium acetate. These precipitates and cell layers were suspended in 2 or 1 ml of 0.2 M NaOH, respectively, incubated for 16 h at room temperature, neutralized with 4 M acetic acid added to a final concentration of 10 mM MgCl 2 , and then treated with DNase (100 g) and RNase (100 g) at 37°C for 2 h. Proteinase K (100 g) was added, and incubation was continued for 16 h at 37°C. The samples were centrifuged at 13,000 rpm for 10 min to remove insoluble materials. The supernatants were diluted with 1 volume of 20 mM Tris-HCl buffer, pH 7.2, and applied to a DEAE-Sephacel column equilibrated with 50 mM Tris-HCl buffer, pH 7.2, containing 0.2 M NaCl and subsequently washed with 10 column volumes of 0.2 M NaCl in 50 mM Tris-HCl buffer, pH 7.2. The fractions eluted with 2.0 M NaCl in the same buffer were collected. Three volumes of cold ethanol containing 1.3% potassium acetate were added to these fractions, and the glycosaminoglycans were recovered as precipitates by centrifugation at 13,000 rpm for 30 min at 4°C. The precipitates were dissolved in water, and portions of the solutions were digested with chondroitinase ABC as described previously (43). The digests were analyzed by a reversed phase ion-pair chromatography using Senshu Pak column. The amounts of CS were estimated from their absorbance on the chromatographic patterns using standard disaccharides as described previously (43).
Constructions of Vectors for Antisense Inhibition and for Expression of V3-The versican/PG-M antisense vector was constructed using the Ϫ99 to ϩ359 portion of the cDNA region as described previously (32). The fragment containing this region was amplified by PCR using primers (sense, 5Ј-GCTTCCTATGTGATCTTCCG-3Ј, and antisense, 5Ј-AGTTTGACCATGGTGAGGGA-3Ј), purified by agarose gel electrophoresis, and ligated into a SmaI site of the expression vector pKCR with the SV40 promoter. To confirm the direction of the inserted fragment, the plasmid was digested with PstI and sequenced.
A plasmid containing full-length mouse V3 cDNA was a gift from Dr. Zako in this laboratory (20). The cDNA fragment was excised with HindIII and EcoRI and subcloned into pSP72 cloning vector (Promega, Madison, WI). PCR was performed with 5Ј-GTGCGCCACCCTGT-GACTGTG-3Ј and 5Ј-ATCTTGCTCACAGAGTGCACCAAC-3Ј as primers and human V3 cDNA (44) as a template. The PCR fragment (340 bp) was digested with DraIII and ligated into pSP72, which was digested with DraIII and subsequently with HindIII and EcoRI, and the fragment was subcloned into an expression vector pKCR. This plasmid, designated mhV3/pKCR, expresses mouse V3 with a human-specific epitope of 2B1. The epitope of 2B1 was in the human G3 domain. 5 Northern Blotting, Quantitative Real Time PCR, and RT-PCR-Total RNA was prepared by using an RNeasy TM total RNA system (Qiagen Inc., Santa Clarita, CA). Twenty micrograms of total RNA was separated on a 1% agarose gel and transferred to a Duralon-UV membrane (Stratagene, La Jolla, CA). Filters were hybridized with mouse versican/ PG-M or rat aggrecan probe. Mouse versican/PG-M probe and rat aggrecan cDNA were gifts from Dr. Zako and Dr. Yada in this laboratory, respectively. Each probe was 32 P-radiolabeled using a Random Primed DNA labeling kit (Roche Applied Science). To normalize mRNA levels, 28 S ribosomal RNA was used.
At various time points, mRNA was isolated from N1511 cells using a Micro-FastTrack TM 2.0 kit (Invitrogen), and cDNA was synthesized with the SuperScript TM preamplification system (Invitrogen). The mouse versican/PG-M primers, mouse aggrecan primers, and their TaqMan probes were designed using Primer Express software (Applied Biosystems, Foster City, CA). The TaqMan probe contained a reporter dye at the 5Ј end and a quencher dye at the 3Ј end. The sequences of the primers and the probes for mouse versican/PG-M are as follows: forward primer, 5Ј-CCAGTGTGAACTTGATTTTGATGAA-3Ј, and reverse primer, 5Ј-AACATAACTTGGGAGACAGAGACATCT-3Ј; TaqMan probe, 5Ј-CACTCTAACCCTTGTCGGAATGGT-3Ј. Those for mouse aggrecan are as described previously (45). These sequences are based on data in GenBank TM , and the accession numbers for mouse versican/PG-M and mouse aggrecan are D28599 and U22901, respectively. As this primer-probe set was designed using the G3 domain of versican/PG-M, it can detect all the variants except mhV3. The RT-PCR was carried out with an ABI PRISM 7700. The reaction mixture (50 l) contained 40 ng of total RNA, 200 M of the primers, and 600 M of probe in 1ϫ TaqMan Universal PCR Master Mix (Rosh, Branchburg, NJ). TaqMan rodent GAPDH control reagents were used as the internal controls. The reverse transcription reaction was run, according to the manufacturer's recommended program. Data quantification and analysis were also performed according to the manufacturer's directions (PE Biosystems). Amplified fragments (108 and 80 bp, versican/PG-M and aggrecan) were sequenced for confirmation. Values represent levels relative to the control. Three independent experiments were performed in triplicate to obtain the values.

Expression of Versican/PG-M in the Early Phase of Chondrogenesis-
We previously optimized the conditions for a novel chondrocytic cell line, N1511, to undergo mesenchymal cell condensation and subsequently chondrocyte differentiation (38). Simultaneous treatment with PTH (P; 1 ϫ 10 Ϫ7 M) and dexamethasone (D; 1 ϫ 10 Ϫ6 M) in the presence of 10% serum (P/D) had a maximal effect on the chondrocyte differentiation, but separate treatment in the presence of 10% serum had no effect (PTH or dexamethasone, respectively) (see relative values (per cell) of [ 35 S]sulfate incorporation at day 8 ( Fig. 1A) and relative values (per cell) of Alcian blue-positive area at day 13 ( Fig. 1B)). Consistent with those results, P/D treatment brought about a marked expression of versican/PG-M even at day 2, whereas treatment with either PTH or dexamethasone yielded almost the same or higher but not maximal levels of versican/PG-M, compared with the expression without any inducing treatment (Ϫ) (see relative mRNA levels of versican/PG-M at day 2, Fig. 1C). In this study, all other experiments were performed using the combination treatment of P/D.
We then examined the effect of cell density on the expression. Cells at a density over 3 ϫ 10 4 cells per 35-mm dish exhibited high levels of versican/PG-M expression at day 2 ( Fig. 1D) and Alcian blue deposition at day 13 (Fig. 1E). On the other hand, cells at a density of 0.75 ϫ 10 4 cells per 35-mm dish showed the same levels of versican/PG-M at day 2 as those without the inducing treatment and only 15% Alcian blue-positive area of the cells cultured at a high density from the beginning at day 13, although the cells attained the confluency at day 8 (2 ϫ 10 5 cells per 35-mm dish). Because cell density is known to be a key factor for chondrogenesis, which may reflect the effect of mesenchymal condensation, the possible mechanism could be explained at least in part by the high level expression of versican/PG-M in the cells at high density. Taken together, the results suggest that the chondrogenic effects of the treatment and cell density on N1511 cells involve the strong expression and deposition of versican/PG-M in the early phase.
Changes of Versican/PG-M Expression and Cell Morphology during Chondrogenesis-When N1511 cells were plated at a density of 1.5 ϫ 10 6 cells/35-mm dish in ␣-MEM containing 10% FCS, they showed a fibroblastic or polygonal shape. One day later, the cells received the P/D treatment (defined as day 0 at induction). Typical cartilaginous nodules positive for Alcian blue staining and anti-aggrecan antibodies appeared at day 7 and gradually expanded in size with the accumulation of cartilage matrix as described previously (38). By using a well known cartilage development system, chick embryonic limb buds, versican/PG-M has been shown to be highly expressed in the mesenchymal cell condensation area at stage 23 but gradually decreased in an inverse correlation with the increase in the accumulation of aggrecan in the ECM of formed limb cartilage (10,22,23). To confirm that N1511 cells also display a sequential expression of versican/PG-M as observed in vivo, we performed immunostaining of the cultured cells. The deposition of versican/PG-M started in the ECM soon after plating and, with continuous P/D treatment, became more and more conspicuous at day 2 (Fig. 2, A  and B). When we doubly stained the cells with anti-versican/PG-M antibodies (DAB) and Alcian blue (aggrecan) at day 8, the Alcian bluestained area in the expanding nodules consistently merged with the versican/PG-M-positive area (Fig. 2C).
Northern blot analysis for the cells treated with P/D for several days also showed that the versican/PG-M expression preceded the gene expression of aggrecan (Fig. 2D), Col2␣1, and Col9␣1 (see Fig. 5A in Ref. 38). We also performed real time PCR to make a quantitative comparison of the transcription levels of the versican/PG-M and aggrecan genes. Versican/PG-M transcription was up-regulated 21-fold in response to the P/D treatment at day 2, and aggrecan transcription was 16-fold at day 13 (Fig. 2E). Without treatment, the cells did not undergo chondrogenesis (38), although they exhibited a transient low level of versican/PG-M expression with a peak at day 1 (data not shown). The results indicated that N1511 cells mimic the in vivo expression patterns of versican/PG-M and aggrecan when they receive chondrogenic stimuli, and the strong and temporal expression of versican/PG-M may be an essential step for the subsequent differentiation into chondrocytes. Together with all the results in the former section, N1511 cells could be considered an in vitro model system for studying the role of versican/ PG-M in chondrogenesis, although it is still unclear how well the system mimics mesenchymal condensation prior to chondrogenesis in vivo.
Inhibition by Antisense RNA of Versican/PG-M Synthesis Resulting in Suppression of Chondrogenic Differentiation in N1511 Cells-We next investigated how the expression of versican/PG-M is important for subsequent chondrocyte differentiation using a number of antisense-, sense-, and mock-transfected stable clones (93 antisense, 15 sense, and 94 mock clones). Each individual stable clone was stained with antiversican/PG-M antibodies for the CS-␤ domain of versican/PG-M at day 1, and the deposition level of it was measured by NIH image as described under "Experimental Procedures." We then classified stable clones into three groups according to their deposition levels shown in Table 1 (group I, no reduction; group II, moderate reduction; and group III, severe reduction). Most of the sense and mock transfectants were in group I, which showed similar levels of versican/PG-M deposition to the parental N1511 cells. In response to the change of versican/PG-M deposition among these three groups shown in TABLE ONE, chondrogenesis was studied by Alcian blue staining, and relative values (per cell) of the positive areas are measured by NIH image at day 18 (Fig. 3A). Antisense transfectants of the group II (AS-II) and group III (AS-III) showed dramatic reduction of the aggrecan accumulation up to 48 and 62%, respectively, compared with the parental N1511 cells (valued as 100%), whereas most of the sense and mock transfectants showed no significant reduction (Fig. 3A) 4A). The versican/PG-M expressed in N1511 cells after the chondrogenic induction must be V0 and/or V1 forms because antibodies used for the staining are specific to the CS-␤ domain, which only the V0 and V1 forms have (16,20,44). RT-PCR revealed that N1511 cells in the early phases after chondrogenic induction express mainly V0 at day 1 and V0/V1 at day 4 (Fig. 4B). Most interestingly, those cells at day 13, when most cells have already differentiated into chondrocytes, express the V2 form. The smallest form, V3 (20), was not detected at any stage (data not shown). Considering that the CS-␤ domain is larger in size and richer in potential chondroitin sulfate-attachment sites than the CS-␣ domain, it is plausible that the V0 and V1 forms have more chondroitin sulfate chains than the V2 form. Therefore, we hypothesized that the chondroitin sulfate chains themselves play an important role in the early phase of chondrocyte differentiation and, if the early phase reflects the case in vivo, in the mesenchymal condensation.
Possible Involvement of Chondroitin Sulfate Chains in Chondrocyte Differentiation-We showed previously that chondroitin sulfate chains immobilized to the substrate matrix inhibit the cell-spreading process of cell adhesion (28,32). It is therefore likely that the chondroitin sulfate chains themselves of versican/PG-M are implicated in the early phase of chondrocyte differentiation where they may have a mesenchymal condensation-like effect by inhibiting cell adhesion. Thus, we first examined the effect of chondroitinase ABC, which degrades the chondroitin sulfate chains of proteoglycans, including versican/PG-M, and leaves the core proteins in the ECM. When added to the culture in the early phase of chondrogenesis (day 0 -4), chondroitinase ABC (5 units/ml) fully degraded the chondroitin sulfate chains of versican/PG-M (Fig.  5A). It should be noted that the treatment with chondroitinase ABC was only for the initial 4 days of culture, during which the aggrecan expres-

TABLE 1 Numbers and versican/PG-M deposition levels of stable clones of antisense, sense, and mock transfectants
Clones of antisense, sense, and mock stable transfectants were isolated as described under "Experimental Procedures." Number of independent clones from the respective transfectants was shown below. Each exhibited the different level of versican/ PG-M deposition measured by NIH image after immunostaining with polyclonal antibodies for CS-␤ domain of versican/PG-M at day 1 as described under "Experimental Procedures." According to their levels, the clones were classified into three groups (group I, no reduction; group II, moderate reduction; group III, severe reduction). sion was not detected. However, it caused a marked decrease in the accumulation of aggrecan in the ECM by up to 50%, compared with the control culture (defined as 100%) after subsequent culture for a further 2 weeks (at day 18) (Fig. 5B), although the treated cultures did not give core bands stained with anti-aggrecan antibodies on the blotting membranes on any culture days (data not shown). When measured by real time quantitative RT-PCR 1 week after the subsequent culture (at day 13), the relative transcription level of aggrecan in the treated culture showed a 30% reduction compared with the normal culture (Fig. 5C). We demonstrated previously that ␤-xylosides with some of hydrophobic aglycons penetrated into the cells and acted as the artificial initiators for the chondroitin sulfate synthesis (42) to inhibit the native synthesis of chondroitin sulfate chain attached to the core proteins by depriving the synthetic machinery such as enzymes and precursor nucleotides (47). The inhibition is apparently dependent on the concentrations of ␤-xyloside, although many other factors appeared to be involved in the effect (47). In addition, it has also been shown that ␤-xyloside treatment of chick embryos in ovo induced a decrease in the growth rate of cartilage (42). Thus, we next examined the effect of treatment with various concentrations of ␤-xyloside in a culture where the treatment was expected for the cells to produce versican/PG-M with lesser chondroitin sulfate. When added to the culture in the early phase of chondrogenesis (day 0 -5), the amounts of CS recovered in cell layers were decreased at the concentration over 1.0 mM (Fig. 6A), whereas the amounts of CS in culture media were not altered at the concentrations from 0.05 to 2.0 mM (Fig. 6B). The CS in cell layer could reflect the native CS synthesized on the core proteins, and the one in media could include CS initiated by ␤-xyloside. In culture treated with ␤-xyloside at the   Table 1). Versican/PG-M and concentration over 1.0 mM, the cells morphologically changed to be fibroblast-like phenotypes at day 2 (data not shown), and Alcian bluepositive areas were reduced dose-dependently at the concentrations over 0.2 mM at day 18 (Fig. 6C). Also, transcription of aggrecan synthesis was suppressed at the concentration over 0.2 mM (Fig. 6D). ␣-Xyloside did not show significant effect (data not shown). ␤-Xyloside appeared to disrupt mesenchymal cell condensation phenotype through the alternation of CS synthesis and to inhibit chondrogenesis. These results suggested that the chondroitin sulfate portions of versican/PG-M contribute to the early phase of the chondrocyte differentiation.

Transfectants
Alteration of the Extracellular Matrix in the Early Phase of Chondrogenesis by Overexpression of the V3 Form-We also hypothesized that irregular expression of the V3 form of versican/PG-M, a spliced variant lacking the chondroitin sulfate-attachment domains, may suppress the deposition of V0 and V1 forms in the matrix in a competitive manner because of the sharing of the G1 and G3 domain binding functions with endogenous forms, which would have resulted in inhibition of the mesenchymal cell condensation-like process by the reduction of chondroitin sulfate chains in the matrix. We transiently transfected N1511 cells with the cDNA for mouse V3 with a human-specific epitope of 2B1 so that the expression of V3 coincided with that of native versican/ PG-M forms (V0 and V1), and also the localization of V3 and endogenous versican/PG-M was individually visualized by immunostaining of cultures with the 2B1 monoclonal antibody and anti-mouse CS-␤ antibodies, respectively. At day 2, about 17% of endogenous versican/PG-M in the ECM was colocalized with overexpressed V3 (Fig. 7, A and B), and some area only had V3. When V3 was deposited in the ECM, less endogenous versican/PG-M was deposited in the ECM than in a mock-transfected culture (Fig. 7C). The expression levels of the native forms (V0 and V1) in the early phase were not affected by the transcription of V3 because they were the same as the ones of the mock-transfected cells in the RT-PCR analysis (Fig. 8A). In addition, the quantitative RT-PCR demonstrated that relative mRNA levels of the native forms (V0 ϩ V1) did not differ between V3-and mock-transfected cells (Fig. 8B). The distribution of other ECM molecules such as hyalauronan and fibronectin, which have been shown to bind to the G1 domain of versican/PG-M The cells were first treated in cultures with 5 units/ml chondroitinase ABC, and the cell extracts were then exposed with or without chondroitinase ABC to confirm that the chondroitin sulfate chains of versican/PG-M were well digested in the culture media. Note that the core bands from the cells treated only "in cultures" are similar to those treated only "in extracts," in comparison with the sample with both treatments of in cultures and in extracts, and also with negative control (no treatment). B, effects of the treatment with chondroitinase ABC in the early phase of the cultures (day 0 -4) at different concentrations on the chondrogenic differentiation were shown by the relative Alcian blue-positive area (per well) at day 18. C, the effects of the treatment were also shown by the relative transcriptional levels of aggrecan at day 13, which were measured by quantitative real time PCR. Values in B and C were the means Ϯ S.D. (n ϭ 50 for B and n ϭ 9 for C), and the asterisks indicated a significant difference from control (0 IU/ml) (Student's t test; * , p Ͻ 0.01).  (31), was altered because endogenous versican/PG-M (stained with Texas Red) showed less overlap in its deposition with hyaluronan (stained with FITC) and fibronectin (stained with FITC) compared with mock transfectants on day 2 (Fig. 9A). Statistical analysis for the merged area of native HA or FN showed that both the areas were reduced in antisense-transfected cultures (Fig. 9B). These results suggest that abnormal expression of the V3 form affected the formation of ECM suitable for the early phase of chondrogenesis.
Suppression of Chondrogenesis by Overexpression of V3-We examined the effect of the overexpression of V3 on the chondrocyte differentiation. The V3-transfected cell culture at day 13 showed decreased Alcian blue staining of the ECM, compared with the mock-transfected cell culture (Fig. 10A). Statistical analysis also confirmed this reduction (Fig. 10B). In the V3-transfected cells, the deposition of aggrecan decreased where the V3 form was overexpressed at day 13 (Fig. 10C). Most of the aggrecan staining in V3-overexpressing cells was observed where V3 was not deposited. It is of note that the deposition of aggrecan in both the cultures was observed, where endogenous versican/PG-M had been heavily deposited as shown in Fig. 2C (data not shown). Furthermore, where the V3 form was overexpressed at day 13 in the V3-transfected cells, the deposition of native forms (V0/V1) decreased as shown in Fig. 7C (data not shown), and the aggrecan transcription level was decreased to one-fourth, compared with the mock-transfected cells at day 13 (Fig. 10D). The results suggest that the abnormal expression of the V3 form comprising only the G1 and G3 domains without any of the chondroitin sulfate-attachment domains interrupted the normal function of the wild type forms of versican/PG-M by decreasing the deposition in the ECM and suppressed the chondrocyte differentiation at the transcriptional level. This finding also suggested that the V0/V1 forms play a pivotal role in the chondrocytic differentiation of N1511 cells.
Considering the fact that the V3 form differs from the V0 and V1 forms in terms of the absence of the chondroitin sulfate-attachment domains as well as the negative effects of both ␤-xyloside and chondroitinase ABC treatments on the chondrogenesis, it is likely that versican/PG-M with chondroitin sulfate chains in mesenchymal ECM plays a crucial role in positively regulating mesenchymal cell condensation and subsequent chondrogenesis in vitro. The expression patterns of versican/PG-M and aggrecan in N1511 mimic quite well the in vivo mesenchymal condensation and subsequent chondrogenesis, that is the high level of expression and the deposition in the ECM of versican/PG-M and the subsequent gradual attenuation proceeded by the accumulation of aggrecan in the ECM (10,(22)(23)(24). The N1511 cells are induced to become chondrocytes not only by treatment with PTH and dexamethasone in the presence of serum but also by that of BMP-2 and insulin in the absence of serum. When exposed to BMP-2/insulin, N1511 cells show Col10␣1 expression and undergo further steps of chondrocyte differentiation toward hypertro-  phy and mineralization, i.e. terminal differentiation (38). Versican/ PG-M has also been shown to be strongly expressed in the ECM in the early phase of chondrocyte differentiation, when induced by BMP-2 and insulin (date not shown). N1511 cells have already expressed a low level of Sox9 without the induction of type II collagen and aggrecan gene expression (data not shown), and the cells appear to be in the differentiation phase committed to chondrogenesis, in comparison with other multipotential cell lines such as C3H10T1/2 and C1. In addition, N1511 cells expressed a type IIA procollagen mRNA at day 0, which is known to be preferentially expressed in prechondrocytes/chondroblasts (48). N1511 cells also expressed a PTH/PTHrP receptor, a critical regulator of chondrocyte proliferation and differentiation (49 -51), whereas undifferentiated and early stage ATDC5 cells lack a PTH/PTHrP receptor (52), and PTH treatment fails to up-regulate the proliferation of ATDC5 cells (53,54). Therefore, to our knowledge, the system using N1511 cells treated with PTH and dexamethasone may be the best in vitro system available for analyzing the role of versican/PG-M in the early phases of chondrocyte differentiation (mesenchymal cell condensation step).

DISCUSSION
N1511 cells in the early phases of the differentiation express N-cadherin, neural cell adhesion molecule, hyaluronan, type I collagen, tenasin, and fibronectin (data not shown and Fig. 9A), of which the former two molecules are known to be involved in mesenchymal cell-cell interactions (6,7,55), and the latter molecules are known to be typical ECM components during mesenchymal cell condensation and interact with versican/PG-M directly to form the mesenchymal ECM (28). Moreover, the expression of the versican/PG-M gene (Cspg2) is regulated by a promoter that includes a LEF-binding site, which is one of the targets of N-cadherin (25). N-cadherin, a cell-cell interaction molecule expressed in the mesenchymal cell condensation area has been shown to be involved in chondrogenesis (6,55,56). It has also been shown that N-cadherin-mediated cell adhesion is suppressed by activation of a socalled repulsive receptor, Roundabout (Robo), by the binding of its ligand, Slit (57). However, there is no evidence that versican/PG-M is involved in this Slit-Robo interaction.
The suppression of chondrogenesis of N1511 by the antisense expression for versican/PG-M and by forced expression of the variant V3 correlated well with the inhibition of the deposition and the abnormal organization of versican/PG-M in the ECM. Although the results could be explained simply by the possibility that the suppressed or abnormal deposition of endogenous versican/PG-M prevents aggrecan from binding with hyaluronan, real time quantitative RT-PCR has revealed that both the antisense expression for versican/PG-M and the forced expression of the V3 form caused a low level of aggrecan mRNA (Fig. 3D and Fig. 10D), indicating that versican/PG-M deposited in the ECM at the early phase of the differentiation signals the cells and regulates the gene transcription. In addition, our findings have revealed that the chondroitin sulfate chain portions of versican/PG-M may play a pivotal role. Four variants of versican/PG-M (V0, V1, V2, and V3) have different amounts and numbers of chondroitin sulfate chains (16). It is interesting that V0 (G1-CS-␣-CS-␤-G3) and V1 (G1-CS-␤-G3) are expressed in the heart and blood vessels where cell shapes are thought to be actively changed, whereas V2 (G1-CS-␣-G3) is mainly expressed in brain where the cell shape appears to be stable. In N1511 cells, early in the chondrocyte differentiation, V0 and V1 forms were expressed, and the V2 form was not expressed until late. It is also interesting that mesenchymal cell condensation largely involves changes in cell shape, whereas differentiated chondrocytes are thought to have a rather stable shape. It is also important to clarify the mechanism by which the variant form changes in a tissue-and differentiation stage-dependent manner, but this has not yet been studied.
The V3 form, composed of only G1 and G3 domains, has not been detected in vivo, but overexpressed V3 actually deposited in the ECM and partially colocalized with endogenous versican/PG-M. Our finding that overexpressed V3 inhibited chondrogenesis is in line with the observation that the exogenous addition of the G3 domain of versican/ PG-M to culture medium inhibited the chondrogenesis of mesenchymal cells (58), because the added G3 domain may have occupied sites on the mesenchymal ECM for the versican/PG-M binding, interrupted deposition of native versican/PG-M, and resulted in a significant alteration of matrix function. Moreover, there is a report that retrovirally mediated overexpression of the V3 form showed opposite effects to other variants of versican/PG-M in terms of cell adhesion, migration, and proliferation in arterial smooth muscle cells, i.e. the cells expressing V3 spread more on tissue culture dishes with larger areas of close and focal contacts, grew slower, and migrated a shorter distance than control cells (59).
The V3 form is more than 1000 amino acids shorter than V2 and V1 (Fig. 4A). Therefore, one possibility still remains that this significant amount of protein could have some biological effects on the ECM, resulting in inhibition of chondrogenesis. But results obtained from the ␤-xyloside treatment in cultures led us to think that this possibility is less likely than our hypothesis, because ␤-xyloside is known to affect directly and specifically the normal chondroitin sulfate chain synthesis on the core proteins. Besides, ␤-xyloside also yielded core protein-free chondroitin sulfate with short and undersulfated chains, and then nor- mal collagen synthesis is inhibited (42,60). It is interesting to see that temporal and significant stimulation of chondrogenesis was observed at the low concentration of ␤-xyloside (0.05 mM; Fig. 6, C and D) and well correlated with the increase of chondroitin sulfates in cell layers (0.05 mM; Fig. 6A). Because we showed previously that ␤-xyloside at such low concentrations yielded xyloside-derivatized chondroitin sulfate chains with the size and sulfation degree similar to the native ones (47), it is likely that the increased chondroitin sulfates in cell layers may be xyloside-derivatized chains.
Taken together, we could conclude that versican/PG-M with chondroitin sulfate chains is crucial for positively regulating mesenchymal condensation and subsequent chondrocyte differentiation. Moreover, we may hypothesize that chondroitin sulfate chains of versican/PG-M deposited in the mesenchymal ECM interact with cells via specific cell surface receptors or by providing cells with some physical environment suitable for migration and/or controlling cell shape, if they are active agents in induction of chondrogenesis. Although definite and convincing evidence for the receptor(s) for the chondroitin sulfate chains has not been available so far, there are a few candidates such as annexin VI and midkine that bind to the CS chains of versican/PG-M on the cell surface in vitro (33,61). Annexin VI has been shown to interact with the chondroitin sulfate chains in the ECM and to inhibit the subsequent cell spreading, although the mechanism is totally unknown (33). The chondroitin sulfate chains may be involved in mesenchymal condensation and further chondrogenesis by modifying cell-extracellular matrix molecule interactions so as to influence the cell shape.
Versican/PG-M-deficient mice died at embryonic day 10.5, because of a heart problem before the onset of chondrogenesis (37). This abnormality was caused by a defect in the transformation of the endothelium into the mesenchyme because of irregular cell migration and adhesion. This result is consistent with our finding that versican/PG-M can suppress interactions between cells and most ECM molecules such as fibronectin through the chondroitin sulfate chains (30 -32). Also, antiadhesive function of versican/PG-M may be affected by the treatment with ␤-xyloside in culture, because the morphological changes of cells are consistent with our previous data (30). To elucidate the function of versican/PG-M during chondrogenesis as well as organogenesis in which mesenchymal condensation is involved, studies with conditional knock-out mice and transgenic mice with tissue-specific or stage-specific expression may be necessary. Moreover, the relationships among cell-matrix interaction molecules, versican/PG-M, and cell-cell interaction molecules should be studied further, which may provide some idea as to the mechanism of action of versican/PG-M in chondrogenesis.