Chondrogenic tumors that exhibit benign or malignant behaviors synthesize variable amounts of cartilage-like extracellular matrix. To define the regulators of these phenotypes, we performed a proteomic comparison of multiple human chondrogenic tumors, which revealed differential accumulation of the C-propeptides of procollagens Iα1 and II (PC1CP and PC2CP) in malignant versus benign tumors, respectively. Expression patterns of PC1CP correlated with levels of tumor vascularization, whereas expression patterns of PC2CP suggested its susceptibility to immobilization within the extracellular matrix. Prompted by these observations, we investigated the functions of recombinant PC1CP and PC2CP in the extracellular matrix in soluble or immobilized states. Each induced β1 integrin–mediated chondrocyte adhesion by distinct domains and efficacies, suggesting that they initiated distinct signaling pathways. Indeed, immobilized PC2CP, but not PC1CP, induced apoptosis of primary chondrocytes and EAhy926 endothelial cells. In contrast, soluble PC1CP, but not PC2CP, induced the migration of EAhy926 cells and increased vascular endothelial growth factor (VEGF) and CXCR4 expression in chondrocytes. Soluble PC2CP also increased VEGF expression, but along with a more pronounced effect on CXCR4 and matrix metalloproteinase 13 expression. Our findings suggest that PC1CP favors angiogenesis and tumor progression, but that PC2CP acts in a more complex manner, exerting antitumor and antiangiogenic properties through apoptosis induction when immobilized, but progression and metastasis when soluble. In summary, the relative levels of PC1CP and PC2CP and their interactions within the extracellular matrix contribute to tumor progression, angiogenesis, and metastasis in chondrogenic tumors. Cancer Res; 70(11); 4739–48. ©2010 AACR.
Cartilaginous tumors are characterized by the formation of cartilage-like extracellular matrix by neoplastic cells exhibiting a chondrocytic gene expression profile. Most chondrogenic tumors are benign (enchondromas) as they remain phenotypically very similar to differentiated chondrocytes, in respect of their gene expression profile and matrix composition (1, 2). Their malignant counterparts, chondrosarcomas, are divided into subgroups relating their matrix composition and behavior. The most distant from enchondromas in terms of behavior and phenotype, grouped under the term “unconventional chondrosarcoma,” exhibit phenotypes related to early, late, or dedifferentiation stages of chondrocytes (1). Conventional chondrosarcomas, which partly share the differentiated chondrocytic phenotype of enchondroma, tend to grow faster and produce metastases (2). They are classified into histologic grades, which tentatively relate their potential aggressiveness. The functional features distinguishing enchondromas from low-grade chondrosarcomas remain unidentified, however. The understanding of low-grade chondrosarcoma progression is poor (2) and the differential diagnosis of chondrosarcoma versus enchondroma remains difficult (3, 4).
We investigated the proteomes of multiple human enchondromas and chondrosarcomas of all types and grades in the search for molecular markers and regulators of chondrogenic tumors. We identified COOH-terminal propeptides of procollagen Iα1 (PC1CP) and IIα1 (PC2CP) as differentially accumulated polypeptides in malignant versus benign tumors. Therefore, we investigated their molecular functions in vitro as purified recombinant proteins.
Materials and Methods
Tissue procurement and processing, proteomic analysis, and protein spot identification
This study was performed in conformity with the Declaration of Helsinki principles. Written informed consent was obtained from the patients. Tissue preparation, protein extraction, two-dimensional electrophoresis, gel staining and analysis, protein spot identification, and Western blots were performed as described previously (5).
Chick anti-PC1CP and rabbit anti-PC2CP were generated at Eurogentech using the following peptides as immunogens: NH2-CWYISKNPKDKRHXWF-COOH (PC1CP), and NH2-SSKSKEKKHIWFC-COOH and NH2-ADQAAGGLRQHDAEC-COOH (PC2CP). Other antibodies were rabbit anti-G3PDH, goat anti–matrix metalloproteinase 13 (MMP13), and mouse anti-CD34 (QBEnd 10, DAKO).
Immunohistochemistry was performed on formalin-fixed, paraffin-embedded tissue sections using the streptavidin-biotin-peroxydase set up (DAKO). Antigen retrieval was processed in a Tris-citrate buffer (pH 6.0) for 30 minutes at 65°C. Proteoglycans were digested with 500 U/mL hyaluronidase and 1 U/mL chondroitinase ABC (Sigma). Anti-PC1CP and anti-PC2CP were diluted 1:200 and diaminobenzidine was used as a chromogen. Images were captured on a DMD108 microscope (Leica).
Reverse transcription and molecular cloning
RNA was extracted from cell cultures or tissues using the RNeasy kit (Qiagen). Reverse transcription was performed from 500 ng total RNA using M-MLV (Invitrogen). Open reading frames encoding PC1CP and PC2CP (as described in Fig. 4A) were cloned in pET15b (Novagen) or in pcDNA3.1 downstream of the MATN3 signal peptide (amino acids 1–28) and a 6xHis tag.
Protein production and purification
Bacterial PC1CP and PC2CP were produced and purified from the insoluble fraction in a urea-containing buffer on Ni-NTA columns (Qiagen). pcDNA3-PC1CP and pcDNA3-PC2CP plasmids were transfected into SW1353 cells using Exgen500 (Euromedex). For production, clones were grown as suggested by Ruggiero and Koch (6). Conditioned medium was clarified; equilibrated to 1 mol/L NaCl, 0.2% Triton X-100, and 5 mmol/L imidazole at 4°C; and incubated overnight with Ni-NTA resin. The resin was washed in dPBS containing 1 mol/L NaCl and 20 mmol/L imidazole, and elution was performed stepwise with dPBS containing 500 mmol/L NaCl and 500 mmol/L imidazole. Proteins were dialyzed into serum-free culture medium or PBS. We verified their identities by mass spectrometry and did not detect any known polypeptide other than PC1CP or PC2CP, respectively.
Quantitative PCR was performed as described in a previous publication (7). GAPDH was used as internal standard. Primers used were as follows: GAPDH_FOR, CGACCACTTTGTCAAGCTCA; GAPDH_REV, AGGGGAGATTCAGTGTGGTG; COL2_FOR, ATGACAATCTGGCTCCCAAC; COL2_REV, GAACCTGCTATTGCCCTCTG; COL1_FOR, GGCCCAGAAGAACTGGTACA; COL1_REV, CGCTGTTCTTGCAGTGGTAG; MMP13_FOR, AATTGTCCGGTTTGTCTTGGA MMP13_REV, GGGAAGTGCTGGGGGATTTT; CXCR4_FOR, GGTGGTCTATGTTGGCGTCT; CXCR4_REV, TGGAGTGTGACAGCTTGGAG; VEGF_FOR, GCAGAATCATCACGAAGTGG; and VEGF_REV, GCATGGTGATGTTGGACTCC.
Primary human chondrocyte and cell line culture
Primary human chondrocyte (PHC) handling and culture were performed as previously described (7). SW1353 and EAhy926 human cell lines were procured from the American Type Culture Collection and grown accordingly.
Cell adhesion assays
The colorimetric assay was described by Davies and colleagues (8). Blocking antibodies were used at 10 μg/mL. Anti-β1, anti-β3, and anti-α5β1 (MAB1959, MAB1957Z, and MAB1969) were from Millipore. Anti-α1 and anti-α2 (5E8D9 and P1E6) were from Santa Cruz. For morphologic analysis, plastic coverslips were coated with proteins overnight, rinsed in PBS, and nonspecific binding was blocked in 1% (w/v) bovine serum albumin (BSA). Cells (5 × 104) were added in 500 μL culture medium for 30 45 minutes. Attached cells were fixed with 4% (w/v) PFA, rinsed, and permeabilized in 0.2% (v/v) Triton X-100. Phalloidin-Alexa 488 (1 U/mL; Invitrogen) was added in 5% (w/v) BSA for 1 hour. Coverslips were rinsed four times and mounted on 4′,6-diamidino-2-phenylindole (DAPI)–containing Vectashield (Vector Laboratories). Images were captured on a DMI3000B microscope (Leica).
Cell death/proliferation assays
Cells (1 × 104) were grown in protein precoated 96-well plates for 48 hours in serum-containing medium, with 20 μg/mL bromodeoxyuridine (BrdUrd) for the last 24 hours, rinsed in PBS, and processed for Click-IT terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) Alexa488 assay (Invitrogen). Monoclonal anti-BrdUrd (Invitrogen, 1:200) was added in 5% (w/v) BSA in PBS for 2 hours. Secondary goat anti-mouse alexa594 (Invitrogen, 1:200) was incubated for 1 hour in blocking buffer. Cell enumeration and rates of TUNEL-positive and BrdUrd-positive cells were performed by fluorescence microscopy.
Cell migration assays
Boyden Chamber assays were directly derived from Palmieri and colleagues (9) except that 4 × 104 cells were used per insert.
All experiments were analyzed in triplicate and performed at least three times; experiments using PHCs were performed from three independent patients. P values were calculated by the paired or unpaired Student's t test in accordance with the situation.
Differential accumulation of C-propeptides of PC1CP and PC2CP in malignant versus benign chondrogenic tumors
Differential proteomic analysis of multiple human cartilage tumors and articular cartilage tissues was performed using a previously described method (5). Thirty-four samples (7), including articular cartilage, enchondromas, conventional chondrosarcomas of all grades, and nonconventional chondrosarcomas, were subjected to proteomic comparison by bidimensional electrophoresis.
One spot with a molecular weight (MW) of 34 kDa and an isolelectric point (pI) of ∼8.0 (Fig. 1A), identified from five tissues by mass spectrometry fingerprinting as the C-propeptide of PC2CP, was observed from all enchondromas and from 6 of 12 conventional chondrosarcomas of all grades (Fig. 1A, bottom right). Strikingly, in most enchondromas and one grade I chondrosarcoma, amounts of PC2CP reached levels much higher than in articular cartilage, as illustrated in Fig. 1A (top right). In contrast, 6 of 12 conventional chondrosarcomas (Fig. 1A, bottom center) and atypical chondrosarcomas (Fig. 1A, bottom right) lacked PC2CP.
Another spot with a MW of 34 kDa and a pI of ∼5.8 (Fig. 1A), identified from three tissues as the C-propeptide of PC1CP, was observed in 10 of 12 conventional chondrosarcomas of all grades (Fig. 1A, bottom), but not in any enchondroma. Based on these observations, we hypothesized that concomitant high PC1CP and low PC2CP levels might be signs of severity in chondrogenic tumors. We confirmed by Western blot (Fig. 1B) that PC1CP was restricted to malignant tumors, whereas PC2CP was absent from certain chondrosarcomas but not from any enchondroma.
Because PC1CP and PC2CP are products of COL1a1 and COL2a1 genes, their protein levels are considered to reflect mRNA levels of the corresponding gene (10, 11). This hypothesis was challenged by quantitative PCR from three tissue groups: osteoarthritic articular cartilage, enchondroma, and conventional chondrosarcoma grade II (Fig. 1C). COL2a1 expression followed PC2CP levels as enchondromas exhibited significantly higher mRNA levels than articular cartilage and chondrosarcomas (Fig. 1C). However, levels of COL1a1 mRNA in the three groups were not significantly different. Therefore, increased PC1CP levels in chondrosarcomas could not be attributed to elevated gene expression.
PC1CP accumulates in regions of highest vascularization
We investigated the immunohistochemical localizations of PC1CP and PC2CP in 72 chondrogenic tumors (Supplementary Data S1). Based on proteomic findings, it could have been suggested that tumors that exhibit either low PC2CP levels or high PC1CP levels would have a significant risk of malignancy. However, immunohistochemical observations did not confirm a straight relation between respective PC1CP or PC2CP expression and malignancy. PC2CP immunostaining was not detected in 4 of 29 enchondromas. In addition, PC1CP was found not only in the vast majority of chondrosarcoma samples, but also in many enchondromas (17 of 29, example shown in Supplementary Data S2). Therefore, the presence or absence of PC1CP and PC2CP did not allow distinction between benign and malignant tumors. In general, however, the staining for PC1CP increased with tumor grade (Supplementary Data S2). Interestingly, in most conventional chondrosarcomas (29 of 35), PC1CP staining was restricted to particular tumor compartments, which seemed less differentiated in terms of cell morphology and matrix content (Fig. 2A). Further, these compartments also exhibited higher vascularization, as estimated by the density of CD34 staining (Fig. 2B). Therefore, PC1CP was suspected to participate in tumor angiogenesis.
PC2CP either diffuses through the tumor or concentrates locally
PC2CP was observed in 14 conventional chondrosarcomas of all grades and in 25 of 29 enchondromas. PC2CP systematically localized to most differentiated cartilage-like areas, but exhibited highly variable repartition through the tissue. In 12 of 14 positive chondrosarcomas, its repartition was homogeneous throughout a tumor compartment (Fig. 2C). In contrast, in 11 of 25 positive enchondromas and in two low-grade chondrosarcomas, the staining was intensely concentrated to specific matrix areas (Fig. 2D). These very different staining patterns suggested that PC2CP either diffused freely through the tissue or was immobilized to particular extracellular components, depending on the tumor considered. Therefore, we investigated the functions of PC1CP and PC2CP as soluble and as immobilized proteins.
Characterization of the disulfide bonding of recombinant PC1CP and PC2CP produced in bacteria or human cells
To characterize the functions of PC1CP and PC2CP, we first purified them as recombinant proteins from bacteria. As shown by their apparent MW under nonreducing conditions (Fig. 3A), bacterial products exhibited two conformations: one with a MW identical to that observed under reducing conditions, suggesting that it was devoid of disulfide bonds, and the other with a lower MW, suggesting that it contained only intramolecular disulfide bonds. This, together with the fact that these proteins were purified under denaturing conditions, led us to suspect that PC1CP and PC2CP produced in bacteria would not possess all structural features of their natural counterparts. Therefore, PC1CP and PC2CP were produced in chondrosarcoma cell cultures, as secreted proteins. Under nonreducing conditions, the resulting proteins predominantly migrated with a MW close to 100 kDa (Fig. 3A), suggesting that they were trimeric, possessing intermolecular disulfide bonds, like their natural counterparts.
Immobilized recombinant PC1CP and PC2CP produced in bacteria induce cell attachment, whereas those produced in human cells induce adhesion through β1-integrin
Earlier studies showed that endogenous PC1CP and PC2CP were capable of inducing dose-dependent cellular attachment when immobilized to plastic surfaces (8). Therefore, the capacity of recombinant PC1CP and PC2CP produced in bacteria or human cells to promote PHC adhesion was investigated. (Fig. 3B). PC1CP and PC2CP from human cells induced saturating, dose-dependent cellular attachment with different capacities: first, the plateau value for PC1CP was 30% higher than that for PC2CP; in addition, the dose required to induce significant attachment was slightly lower for PC1CP. Recombinant PC1CP and PC2CP produced in bacteria also induced dose-dependent cell attachment. Here, in addition, the required concentration to achieve significant cell binding was slightly higher for PC2CP. Furthermore, the doses required to achieve significant attachment were 5-fold higher for proteins from bacteria than those from human cells. To define whether the receptor(s) for PC1CP and PC2CP were specifically expressed by PHCs, similar experiments were conducted using EAhy926 endothelial cells. The protein doses required to achieve significant EAhy926 cell attachment were approximately four times higher (data not shown) than for PHCs but remained convincing. Therefore, the receptor(s) for PC1CP and PC2CP was not specific to PHCs. We then compared the morphologies of attached cells induced by saturating doses of PC2CP from either bacteria or human cells (Fig. 3C). PHCs attached to bacterial PC2CP exhibited limited polymerized actin staining, which surrounded cell membranes. When attached to PC2CP from human cells, their polymerized actin staining was more intense and exhibited protrusions suggestive of adhesive structures. Similar results were obtained with PC1CP. In conclusion, PC1CP and PC2CP procured from bacteria were capable of interacting with cells, but not, like native PC1CP and PC2CP, of triggering proper adhesion. To gain insights into the identity of the molecular receptors of PC1CP and PC2CP, cell attachment experiments were performed after the incubation of cells with blocking antibodies directed against specific integrin subunits (Fig. 3D). Only an anti-β1 subunit antibody completely blocked cell attachment by either PC1CP or PC2CP. We concluded that adhesion promotion by PC1CP and PC2CP was dependent on β1 subunit–containing integrins. We could not ascertain whether they bound all β1-containing integrins independently on their α subunit, as antibodies directed against several specific α subunits were not tested.
Distinct peptides corresponding to PC1CP and PC2CP are capable of inducing cell attachment
To better understand the structural basis of the different capacities of PC1CP and PC2CP to induce cellular adhesion, we generated a library of 18-mer peptides encompassing whole primary structures of PC1CP and PC2CP, and challenged their respective abilities to promote cell attachment. The peptides were classified as corresponding to either conserved or divergent regions between PC1CP and PC2CP (Fig. 4A). None of the conserved peptides induced significant cell adhesion (data not shown). In contrast, several divergent peptides did. In particular, peptides D3 and D9, from both PC1CP and PC2CP, induced significant but differentially efficient cell attachment (Fig. 4B). Further, peptide D10 from PC2CP, but not that from PC1CP, induced significant cell attachment, suggesting that this region of the two proteins might participate in the discriminating structural features of PC1CP and PC2CP. Cells attached to peptides remained round and lacked apparent actin reorganization (data not shown). We concluded that the cell-adhesive properties of PC1CP and PC2CP originated from distinct structural determinants contained at least in regions D3, D9, and D10, the coordinated action of which was not fully recapitulated in bacterial proteins due to misfolding of the polypeptide or the lack of covalent multimers. The combined differential binding of these protein regions to their receptor(s), which seem to include β1 integrin subunit, may constitute the structural basis of the differential adhesive properties of PC1CP and PC2CP, and possibly of their downstream signaling functions.
Immobilized PC2CP but not PC1CP induces the apoptosis of PHCs and endothelial EAhy926 cells
The functional consequences of cellular adhesion promoted by PC1CP and PC2CP were then investigated. PC1CP, PC2CP, or fibronectin were coated in 96-well plates at concentrations allowing optimal adhesion and either PHCs or human endothelial EAhy926 cells were cultured on coated surfaces. Because there was a marked decrease of cell number remaining in PC2CP-coated wells, we performed cell enumeration and TUNEL/BrdUrd incorporation experiments. After a 48 hours of culture, the number of cells of both types remained at similar levels in control, fibronectin-, and PC1CP-coated wells, but decreased by ∼50% in PC2CP-coated wells (Fig. 5A). The proportion of remaining cells, which were positive for BrdUrd incorporation, was similar in all groups (data not shown). In the TUNEL assay (Fig. 5B), remaining cells grown on fibronectin or PC1CP exhibited a slightly, but nonsignificant decrease in apoptosis compared with controls, whereas those grown on PC2CP exhibited an ∼3-fold increase in apoptosis rate. We concluded that immobilized PC2CP, but not PC1CP, induced apoptosis of both PHCs and endothelial EAhy926 cells, without obviously affecting cell growth per se. When PC2CP was administered as a soluble protein, or when bacterial PC2CP was used instead, no changes were observed. Therefore, apoptosis induction by PC2CP required its native structure and immobilization.
PC1CP, but not PC2CP, promotes migration of EAhy926 cells
The heterotrimeric COOH-terminal propeptide of procollagen I, which is very similar to PC1CP, induces chemotactism of EAhy926 cells (9). Therefore, we investigated the capacity of PC1CP and PC2CP to promote migration in a classic Boyden chamber assay (Fig. 5C). PC1CP induced obvious promotion of EAhy926 migration in a dose-dependent manner. In contrast, PC2CP induced little, if any, migration, even at the highest doses tested. We concluded that PC1CP, but not PC2CP was capable of inducing migration of these cells. We asked whether PC2CP was able to inhibit the PC1CP-induced migration of EAhy926 cells but could not show such inhibition (data not shown). These data suggested that PC1CP, but not PC2CP, might promote angiogenesis through induction of endothelial cell motility.
Soluble PC1CP and PC2CP differentially modulate gene expression in PHCs
The downstream signaling of PC1CP and PC2CP was then investigated in terms of gene expression regulation. Again based on previous data about the heterotrimeric COOH-terminal propeptide of procollagen I, we focused our investigations on VEGF, MMPs, and CXCR4, which were previously found upregulated by this molecule and involved in tumor metastasis and vascularization (12–15). When chondrocytes were stimulated with immobilized PC1CP and PC2CP in serum-free medium (SFM), no obvious changes in their gene expression could be evidenced (data not shown). When stimulated with 20 μg/mL soluble PC1CP or PC2CP, chondrocytes exhibited gene regulation, however (Fig 6). Both PC1CP and PC2CP significantly induced VEGF expression with comparable efficacies (Fig. 6). PC2CP, but not PC1CP, induced MMP13 expression (Fig. 6). Finally, CXCR4 mRNA was increased 10-fold by PC1CP, but 200-fold by PC2CP (Fig. 6). No significant differences in the expression of cartilaginous extracellular matrix components were observed. These data show that soluble PC1CP and PC2CP differentially modulated in PHCs the expression of genes potentially implicated in vascularization and metastasis of chondrosarcoma. Soluble PC2CP favors the expression of genes that are considered as acting toward extracellular matrix destruction, vascularization, and metastasis to an even greater extent than PC1CP.
Over the years, extracellular matrix degradation fragments and by-products have been considered as important signaling molecules. In the context of cancer, fragments of several collagen types modulate tumor behavior. Endostatin (16, 17), the COOH-terminal fragment of Collagen XVIII, inhibits endothelial cell migration and is a potent inhibitor of angiogenesis in vitro and in vivo. Fragments of collagen IVα decrease the proliferation and induce the apoptosis of endothelial cells in vitro and inhibit tumor growth in vivo (17, 18). Conversely, the heterotrimeric propeptide of procollagen I, also termed C3, induces migration (9, 12) in vitro and favors breast carcinoma tumor vascularization in xenograft mice (15). Therefore, either protumoral or antitumoral functions have been attributed to polypeptides of collagen origin. Although the detailed mechanisms of function of these polypeptides remain controversial, they all apparently imply integrin signaling (9, 17, 18).
Physiologically, the C-propeptide of procollagen I consists primarily in a heterotrimer of two α1 and one α2 subunits, also termed C3 (9). An imbalance in the levels of COL1a1 and COL1a2 expressions can result in the synthesis of procollagen I devoid of the Iα2 subunit (19), however. Our proteomic search for proteins upregulated in chondrosarcoma revealed increased C-propeptide of PC1CP (Fig. 1), but not Iα2. RT-PCR measurements of the expression of Iα2 in chondrogenic tumors suggested that it was 30 or more times lower than that of Iα1 (data not shown). Therefore, in the context of chondrogenic tumors, investigating the function of PC1CP was more pertinent than that of C3. We postulated, however, based on the strong homology between these two proteins, that PC1CP might favor angiogenesis and tumor progression.
Cartilage is an avascular tissue and therefore has been proposed as a source for antiangiogenic factors (20). PC2CP, also termed chondrocalcin, is one of the most highly synthesized proteins of articular cartilage as part of procollagen II (21) and was found in our proteomic study to accumulate mainly in benign tumors (Fig. 1). Therefore, we thought that it would be a good candidate as an inhibitor of angiogenesis from cartilaginous origin. There is a well-accepted relationship between the differentiation level of chondrogenic tumors and their potential malignancy (1, 3). However, a direct functional link between the two is lacking. PC1CP and PC2CP are translated as part of the precursors of COL1α1 and COL2, respectively, and they act as structural scaffolds important for the trimeric assembly of procollagens (22). Upon secretion, PC1CP and PC2CP are cleaved from their respective procollagen molecules to allow fibrillation (23). COL2 is considered the best marker of articular chondrocyte differentiation, whereas COL1 is a marker of chondrocytic dedifferentiation (3, 24). This is in accordance with our observations, as PC1CP seems to increase whereas PC2CP decreases with dedifferentiation (Figs. 1 and 2). Further, in most chondrosarcomas, PC1CP localized primarily to most dedifferentiated tumor areas, which also exhibited higher vascularization (Fig. 2). More surprisingly, depending on the tumor considered, PC2CP either localized diffusely throughout the tumor or concentrated to very particular sites (Fig. 2C and D). Based on these observations, we investigated the functions of either soluble or immobilized (on plastic), purified, recombinant PC1CP and PC2CP. We preferred to use PHCs as a cellular model rather than available immortalized chondrosarcoma cell lines, which exhibit profoundly dedifferentiated phenotypes (25) and most likely defective cell death regulation.
When immobilized, recombinant, native PC1CP and PC2CP induced cellular adhesion in a β1 subunit in an integrin-dependent manner (Fig. 3) with different efficacies. Their adhesive properties depended on the structural arrangement of at least three distinct, divergent subdomains (Fig. 4). We did not determine whether both proteins bind identical or respectively distinct subsets of β1 integrins. Nevertheless, it was likely that the PC1CP- or PC2CP-mediated adhesion would result in differential downstream signaling. Indeed, immobilized PC2CP induced apoptosis of PHCs and EAhy926 cells, whereas immobilized PC1CP did not (Fig. 5A and B). This finding favored the hypothesis that PC2CP, but not PC1CP, acted against tumor progression and vascularization when immobilized. We could not show a function of immobilized PC1CP. Indeed, in tumors, PC1CP was never observed in such a concentrated pattern as PC2CP. Soluble PC1CP, however, induced the dose-dependent migration of EAhy926 endothelial cells in the Boyden chamber assay (Fig. 5C), as earlier shown for C3 (9), whereas PC2CP did not. Soluble PC1CP induced the expression of VEGF and CXCR4 (Fig. 6). VEGF is a major actor of angiogenesis and its expression correlates to chondrosarcoma grade and proliferating capillary index (26). CXCR4 and its cognate chemokine SDF-1/CXCL12 contribute to local and distant metastasis and to angiogenesis in a variety of cancers (27) and promote chondrosarcoma cell invasion in vitro (13). Therefore, all observed functions of PC1CP suggested that it could favor chondrogenic tumor vascularization and progression, whereas immobilized PC2CP could inhibit it. However, soluble PC2CP induced VEGF expression by chondrocytes similarly to PC1CP (Fig. 6). Not only so, but it induced CXCR4 expression to a much higher extent (Fig. 6) and that of MMP13, whereas PC1CP did not. MMP13 is a major catabolic enzyme of cartilage extracellular matrix, which was proposed as contributing to chondrosarcoma progression (28). Therefore, although immobilized PC2CP induced the apoptosis of PHCs and endothelial cells, soluble PC2CP regulated gene expression in chondrocytes in a manner that may favor tumor progression and angiogenesis to a greater extent than PC1CP. Of note, the intratumor localization of PC2CP suggested that it was predominantly immobilized in enchondroma, whereas its diffusing appearance was better represented in chondrosarcoma. Our data do not prove that the differential localizations of PC2CP translate variable solubility. Nevertheless, the localization patterns observed for PC2CP in many tumors were impressively stringent (Fig. 2D). Furthermore, our functional data suggest differential functions of soluble and immobilized PC2CP. This was also observed for Endostatin, which exerted antagonistic functions toward cell survival and migration through integrin receptors, depending on whether it was administered as soluble or immobilized (29). Altogether, our data point to previously underestimated consequences of collagen metabolism over chondrogenic tumor behaviors. Not only the respective levels of PC1CP and PC2CP in chondrogenic tumors, but also their interactions with the surrounding extracellular matrix may modulate tumor progression, angiogenesis, and metastasis.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant Support: Fondation pour la Recherche Médicale, Association pour la Recherche contre le Cancer, Ligue Régionale contre le Cancer, Région Lorraine and Fédération de Recherche 3209.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 108 U.S.C. Section 1734 solely to indicate this fact.