Primary cells (chondrocytes, synoviocytes, ASCs, and MSCs) were isolated from healthy or OA patients after informed consent. All subjects gave written consent in accordance with the Declaration of Helsinki. This study was carried out in accordance with the recommendations of Committee for Person Protection of Languedoc-Roussillon and approved by the French Ministry of Higher Education and Research (registration number: DC-2009-1052 and DC-2008-417). Protocols for isolation and characterization are described elsewhere (15). Typically, ASCs and MSCs were characterized by the expression of classical markers: CD13, CD73, CD90, and CD105 and the absence of hematopoietic and endothelial markers: HLA-DR, CD11b, CD14, CD31, CD34, CD45, and CD106. They were also shown to differentiate into the three lineages: adipocytes, osteoblasts, and chondrocytes [as previously shown in Ref. (19)].

Chondrocytes or synoviocytes were plated at high density (500,000 cells/well) on the bottom of 6-well plates and cultured with ASCs (70,000 cells/insert) (ratio 7:1) in cell culture inserts (PET membranes, 0.4 µm pore porosity, BD Biosciences, Le Pont de Claix). Cultures were maintained for 7 days in 3 mL of minimal medium [DMEM supplemented with penicillin (100 U/mL), streptomycin (100 µg/mL), proline (0.35 mmol/L), ascorbic acid (0.17 mmol/L), and sodium pyruvate (1 mmol/L)] as previously described (15). In some experiments, chondrocytes were treated with human recombinant THBS1 (rTHBS1) (R&D Systems, Lille) at different concentrations in minimal medium during 7 days. Chondrocytes and ASCs were collected for RT-qPCR analysis and supernatants were stored at −20°C.

Chondrocyte and ASC monocultures (10⁶ cells/60 mm culture dish) or cocultures (5 × 10⁵ each cell type/60 mm culture dish) were maintained in αMEM containing 2% platelet lysate overnight. Three different cell samples were used. They were then washed five times with PBS to eliminate platelet lysate (Macopharma, Tourcoing) and 2.5 mL of minimal medium were added. After 48 h, supernatants from cocultures and chondrocyte or ASC monocultures were centrifuged (300 g, 5 min), filtered (0.22 µm), and stored at −80°C until secretome analysis. Secretome analysis was performed on coculture supernatants and on mixed monoculture supernatants. The rationale for mixing monoculture supernatants was to accurately compare levels of proteins secreted by each cell type in non-induced conditions to those secreted in cocultures upon crosstalk. Proteins were concentrated by TCA-NLS precipitation and quantified by Lowry method (Bio-Rad DC) as described previously (20). Reduced/alkylated protein samples were separated on gel Nupage 4–12% and stained by Coomassie blue (Instant Blue, Invitrogen), cut into bands, and digested with trypsin. After extraction, peptides were analyzed by mass spectrometry on NanoLC/ESI LTQ-Orbitrap Velos MS/MS. Results were analyzed by consulting protein data bases with Mascot Daemon software and proteins were validated with Prosper software. Spectral counting approach (score) was used to compare cocultures and mixed monocultures conditions. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [1] partner repository with the dataset identifier PXD008146.

Chondrogenic differentiation was induced by centrifuging 2.5 × 10⁵ MSCs at 300 g for 5 min in 15 mL conical tubes. Chondrogenic medium (DMEM high glucose, dexamethasone 0.1 µM, sodium pyruvate 1 mM, ascorbic-2-phosphate acid 170 µM, proline 0.35 mM, ITS, TGF-β3 at 10 ng/mL) was changed every 3 days for 21 days. When tested, human rTHBS1 (10 or 100 ng/mL) was added in the medium at day 14 for the 7 last days of differentiation. At day 21, micropellets were washed in PBS and immediately processed or stored at −80°C.

Adipose stem cells were transfected at 60% of confluence with 100 nM of siRNA control (siCT) or THBS1 (siTHBS1) (Ambion, ThermoFisher Scientific, Illkirsch; see references in Table 1) using Oligofectamine reagent following the manufacturer’s instructions (Life Technologies, Courtaboeuf). Transfection was done on the day before coculture.

Splenocytes were collected from spleens of adult C57Bl6 mice and suspended in IMDM glutamax medium supplemented with 10% inactivated fetal calf serum, 2 mM glutamine, 100 U/mL penicillin, and 100 µg/mL streptomycin, 0.1 mM nonessential amino acids, 1 mM sodium pyruvate, 20 mM HEPES (N-2-hydroxyethylpiperazine-N′-2-ethanesulfonic acid), and 5 × 10⁵ M 2-mercaptoethanol. Human ASCs transfected with siCT or siTHBS1 were plated in 96-well flat-bottom plates at different density (5 × 10³, 1 × 10⁴, or 2 × 10⁴ cells/well). Splenocytes were added at 2 × 10⁵ cells/100 μL/well and mitogen-driven proliferation of T lymphocytes was induced by adding 5 µg/mL concanavalin A (ConA; Sigma-Aldrich, Saint-Quentin-Fallavier). When tested, rTHBS1 was added at different concentrations with stimulated splenocytes. Unstimulated splenocytes were used as negative control. After 3 days incubation, splenocyte proliferation was quantified using CellTiter-Glo Luminescent Cell Viability Assay (Promega, Charbonnières-les-Bains) following the manufacturer’s instructions. The inhibitory effect of ASCs on splenocyte proliferation was quantified subtracting the signal of unstimulated splenocytes and proliferation rate was calculated referring to 100% the value of ConA-stimulated splenocytes.

Total RNA was extracted from cells using the RNeasy kit (Qiagen, Courtaboeuf). RNA (0.5 µg) was reverse transcribed using the M-MLV reverse transcriptase (ThermoFisher scientific). PCR reaction was performed as previously described (15). All details for primer sequences (SYBR Green or TaqMan technologies) are described in Table 1. All values were normalized to RPS9 housekeeping gene since expression of this gene was the most constant among three others evaluated (GAPDH, β2 microglobulin, and β actin). Values were expressed as relative expression or fold change using the respective formulae 2^−ΔCT or 2^−ΔΔCt.

Supernatants from ASC/synoviocyte or ASC/chondrocyte cocultures were used to quantify the concentrations of interleukin (IL)-6, CXCL8/IL-8, CCL2/MCP-1, CCL3/MIP-1α, and CCL5/RANTES using multiplex bead-based sandwich immunoassay kits (Bio-Rad, Marnes-la-Coquette) following the manufacturer’s instructions. THBS1 was detected by specific Enzyme-linked ImmunoSorbent Assays (ELISA; R&D Systems).

The study was conducted in accordance with guidelines and regulations of the Ethical Committee for animal experimentation of the Languedoc-Roussillon (Approval 5349-2016050918198875). All experiments were performed after final approval given by the French Ministry for Education, Higher Education and Research. CIOA model was performed as previously described (17). Briefly, both knee joints of mice were injected with 1 U collagenase type VII from Clostridium histolyticum (Sigma-Aldrich) in 5 µL of saline at day 0 and day 2, causing disruption of the ligaments and local instability of the joint. At day 7, groups of 20 mice received saline, ASC-siCT, or ASC-siTHBS1 (2 × 10⁵ cells/5 μL of saline solution). Mice were sacrificed at day 42. Hind paws were collected and fixed in formaldehyde 4% for 7 days before histological processing.

After fixation, left hind paws were decalcified (formic acid 5%, 2 weeks), embedded in paraffin, cut (three slices of 7 µm each 140 µm), and stained with safranin O fast green staining. Cartilage degradation was quantified using the modified Pritzker OARSI score as described (17).

Right hind paws were dissected to carefully remove smooth tissues and expose articular cartilage of tibiae. Tibiae were scanned in a microCT scanner (SkyScan 1176, resolution 9 µm, 0.5 mm aluminum filter). Assessment of bone parameters was performed using NRecon and CTan softwares (SkyScan).

Quantification of cartilage damage was assessed after scanning the entire articular cartilage of tibiae through their depth in XYZ-mode, with a confocal laser scanning microscope (CLSM; TCS SP5-II, Leica Microsystems, Nanterre, France) with a voxel size of 6 µm, a 5× dry objective and a UV laser light source (l¼ 405 nm). Image stacks were then processed to evaluate articular cartilage degradations. Assessment of cartilage morphometric parameters was performed in both lateral and medial plateau of each tibia using Avizo software (FEI Visualization Sciences Group, Lyon).

Data were expressed as the mean ± SEM. Statistical analysis was performed using the GraphPad Prism software. The comparison between two different unpaired groups was analyzed with a Mann–Whitney test for a non-Gaussian distribution and with an unpaired t-test for a Gaussian distribution. The comparison between several groups was analyzed with a Kruskal–Wallis test or an ANOVA.

We here aimed at identifying factors that could mediate the chondroprotective and anti-inflammatory functions of ASCs that we have previously described in vitro (14, 15) using secretome analysis. We compared proteins produced in the supernatants from chondrocyte/ASC coculture versus supernatants from either monoculture or mixed supernatants from both monocultures. Comparison between cocultures and mixed supernatants aimed at better selecting proteins that were modulated by the crosstalk between cells. We identified 2,043 proteins, among which 62% were predicted to be secreted. Ninety-two proteins were statistically different between cocultures and mixed monocultures. We selected only 12 proteins on the basis of their identification by more than 1 peptide and of a potentially relevant function (Figure 1A; Table 2). Validation of gene expression for these 12 proteins has been performed by RT-qPCR (Figure S1A in Supplementary Material) but only 3 were significantly induced in ASCs when cocultured with chondrocytes (Figure S1B in Supplementary Material). Finally, we selected THBS1 because it was the highest induced protein in cocultures, and identified by the greater number of peptides (Figure 1A). THBS1 was also chosen because its over-expression was previously reported to be protective in a rat model of OA and its absence induced skeletal abnormalities in knock-out mice (21, 22). We therefore hypothesized that the chondroprotective effect of ASCs in OA could be mediated by THBS1.

We first validated the findings of secretome analysis by determining the protein level of THBS1 in chondrocyte/ASC cocultures by quantitative ELISA. We confirmed a significant increase of THBS1 in cocultures as compared to chondrocyte or ASC monocultures (Figure 1B). We then determined the expression level of THBS1 mRNA in chondrocytes and ASCs cultured independently or in cocultures. Expression of THBS1 mRNA was similar in ASC and chondrocyte monocultures as well as in chondrocytes when cocultured with ASCs (Figure 1C). In contrast, THBS1 mRNA was highly upregulated in ASCs when they were cocultured with chondrocytes. These data demonstrated that upregulation of THBS1 protein in ASC/chondrocyte cocultures was directly correlated with the upregulation of THBS1 mRNA in ASCs.

We first evaluated whether THBS1 may be involved in chondrogenesis. Here, we used BM-MSCs instead of ASCs because of their higher potential to differentiate into chondrocytes and confirmed similar expression of THBS1 in the two cell types (Figure 1D). We investigated THBS1 expression in BM-MSCs induced to differentiate into chondrocytes by the standard micropellet technique and measured the expression of classical chondrocyte markers at different time points. All four markers SOX9, COL2A1 variant B, COL10A1, and ACAN were upregulated after 21 days of pellet culture (Figure 2A). Although not significant, expression of THBS1 was also upregulated by a fourfold factor at late time points. The direct effect of THBS1 was then tested by adding the recombinant protein at the end of the differentiation process (between days 14 and 21) when expression of endogenous THBS1 is increased. Two doses of rTHBS1 were evaluated. Compared to the positive control (BM-MSCs cultured in presence of TGFβ3), addition of either doses of rTHBS1 significantly upregulated the expression of all chondrocyte markers (SOX9, COL2A1 variant B, ACAN, and COL10A1) by day 21 (Figure 2B). These results indicated that THBS1 enhanced BM-MSC differentiation into the chondrogenic pathway.

We then determined the potential effect of THBS1 on chondrocyte phenotype by evaluating the dose-dependent effect of rTHBS1 on OA primary chondrocytes and quantifying the expression of several chondrocyte markers by RT-qPCR. At low doses, rTHBS1 increased the expression of mature chondrocyte markers (SOX9, COL2A1 variant B, ACAN, and HAPLN1), which was not observed for high doses of rTHBS1 (Figure 2C). SOX9 expression was even significantly reduced at high doses of rTHBS1. Expression of the two hypertrophic chondrocyte markers MMP13 and ALPL differed. MMP13 was significantly decreased at the lowest dose of rTHBS1 but was not modulated at higher doses while ALPL increased in a dose-dependent manner. Finally, expression of two markers of fibrocartilage, COL1A1 and COL3A1, was increased whatever was the concentration of rTHBS1 in the culture medium. Indeed, rTHBS1 did not positively impact the altered phenotype of osteoarthritic chondrocytes.

Since we previously reported that ASCs modulate the phenotype of osteoarthritic chondrocytes by decreasing the expression of hypertrophic and fibrotic markers (15), we investigated the effect of THBS1 silencing in ASCs. Using a siRNA approach, we reproducibly downregulated THBS1 expression in ASCs by 30–37%, at the mRNA and protein level, respectively, as compared to untransfected ASCs or ASCs transfected with a control siRNA (siCT) (Figure 3A). Since THBS1 is part of thrombospondin family which includes 5 members (THBS-1, -2, -3, -4, and -5/cartilage oligomeric matrix protein), we checked whether downregulation of THBS1 could induce a modulated expression of other thrombospondin members. Downregulation of THBS1 expression in ASCs did not impact the expression of other thrombospondin family members (Figure 3B). As expected, addition of ASC-siCT decreased the expression of MMP13, APLP, COL1A1, and COL3A1 in cocultured chondrocytes (Figure 3C). COL2A1 variant B was also significantly reduced. By comparison, a similar effect of ASC-siCT and ASC-siTHBS1 was observed on chondrocytes indicating that THBS1 was not involved in the regulation of osteoarthritic chondrocyte metabolism by ASCs.

Our previous results showed that ASCs reduce the secretion of inflammatory mediators by osteoarthritic synoviocytes and chondrocytes (14). We therefore determined whether THBS1 was involved in this effect using siTHBS1-transfected ASCs. In coculture with synoviocytes, ASC-siCT tended to decrease the secretion of all tested factors but the decrease was only significant for RANTES (Figure 4A). Addition of ASC-siTHBS1 induced a similar reduction of inflammatory mediators. The level of IL-8 was even lower after addition of ASC-siTHBS1. In coculture with chondrocytes, both ASC-siCT and ASC-siTHBS1 decreased expression of IL-6 and tended to increase expression of TNF-α and IL-8 in chondrocytes (Figure 4B). The results therefore showed that THBS1 was not involved in the modulation of the inflammatory profile of osteoarthritic synoviocytes or chondrocytes.

We then investigated the possible anti-inflammatory effect of THBS1 on T cell response. We added different doses of rTHBS1 in a T lymphocyte proliferative assay and measured the proliferation of ConA-activated T cells in the splenocyte population. A bell curve was obtained with no suppressive effect of rTHBS1 at the lowest and highest tested doses (Figure 4C). However, the doses of 0.4 and 2 ng/mL of rTHBS1 significantly reduced T lymphocyte proliferation. We then compared the suppressive capacity of ASC-siCT and ASC-siTHBS1 on T cell proliferation. Using different ASC/splenocyte ratios, we observed that ASC-siTHBS1 were significantly less suppressive than ASC-siCT at the lowest ratio (Figure 4D). We evaluated that the concentration of THBS1 was approximately 2.5 ng/mL at 1/40 ratio. This amount was in the range of the suppressive activity of rTHBS1 (see Figure 4C) while highest ratios exerted no effect, suggesting a dose-dependent effect of THBS1. In addition, silencing of THBS1 in ASCs did not affect the expression of the immunosuppressive molecules TSG6, IL-6, and COX2 but significantly decreased IDO1 expression (Figure 4E). Our results might therefore suggest that secretion of THBS1 at least partly mediated the anti-inflammatory effect of ASCs through the expression of IDO1.

Finally, we determined whether THBS1 secretion plays a role in the anti-inflammatory and chondroprotective effect of ASCs in the CIOA pre-clinical model. To this end, we injected either ASC-siCT or ASC-siTHBS1 in the knee joints of mice induced to develop OA and evaluated cartilage and bone degradation at day 42 after OA induction. As previously shown (17), histological scoring on both femurs and tibias sections revealed that injection of human ASC-siCT protected mice from developing OA lesions as shown by representative pictures and OA score (Figures 5A,B). Histological analysis of knee joint sections revealed that ASC-siTHBS1 were significantly less efficient that ASC-siCT in protecting against cartilage degradation as shown by OA score (Figures 5A,B). Because histological analysis did not allow scoring the entire joint, we relied on morphometric analysis of cartilage using CLSM. Thanks to the autofluorescence property of cartilage, visualization and 3D reconstruction of entire articular cartilage on tibia plateaus are feasible. Using this technology, a highly significant preservation of cartilage volume was observed with ASC-siCT but not with ASC-siTHBS1 as shown in reconstructed 3D pictures (Figure 5C) and confirmed by volume quantification (Figure 5D). Other parameters, such as cartilage thickness and surface degradation (measured as surface/volume ratio), were preserved with both ASC-siCT and ASC-siTHBS1 even though ASC-siTHBS1 seemed to be less efficient (Figure 5D). At last, bone parameters were analyzed by μCT and 3D reconstructions of sub-chondral bone were performed using a color code for thickness representation. On 3D images of tibia epiphyses, sub-chondral bone thickness, as evaluated by larger red areas, was higher in mice treated with ASC-siCT compared to CIOA control or ASC-siTHBS1 treated mice (Figure 5E). These correlated with a significantly higher thickness measure of sub-chondral bone in ASC-siCT group compared to CIOA control (Figure 5F). Bone erosion (surface/volume) and bone mineral density parameters did not change in mice injected with ASC-siCT or ASC-siTHBS1 (Figure 5F). Altogether, the results suggested that ASCs expressing lower levels of THBS1 were less efficient to protect mice from developing OA symptoms.