Primary human chondrocytes were isolated from cartilage of 11 OA patients (mean age of 62 years) undergoing knee arthroplasty after informed written consent from patients and approval by the local and national ethics committee (‘Cellule de bioéthique de la direction générale pour la recherche et innovation, Ministère de l’Enseignement Supérieur et de la Recherche’; registration number DC-2009-1052) were obtained, as described previously 22 . Cartilages from six healthy adult subjects (mean age of 53 years) were forensic waste from legal medicine with no need of informed consent after consultation with the national ethics committee and in strict agreement with French legislation. OA primary chondrocytes were cultured in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% fetal calf serum as described 23 . Primary OA chondrocytes (2.5 × 10⁵ cells) were pelleted by centrifugation in 15-mL conical tubes, placed in three-dimensional (3D) setting for 7 days in chondrogenic medium—DMEM supplemented with 0.1 μM dexamethasone (Sigma-Aldrich, St. Louis, MO, USA), 1 mM pyruvate sodium (Invitrogen, Paisley, UK), 0.17 mM ascorbic acid (Sigma-Aldrich), 0.35 mM Proline (Sigma-Aldrich), 1% Insulin Transferin Selenium (Lonza, Basel, Switzerland), 2 mM L-glutamine (Lonza), 100 U/mL penicillin, and 100 μg/mL streptomycin (Lonza)—supplemented with transforming growth factor-beta 3 (TGF-β3) at 10 ng/mL (R&D Systems, Minneapolis, MN, USA). Treatment with recombinant human IL-1β at 10 ng/mL (R&D Systems) was applied for the first 5 days. Wild-type or ink4a knockout mice (1 month old) were obtained as reported 24 . Mice were housed and cared for in accordance with the laboratory animal care guidelines. Approval was obtained from the regional ethics committee on animal experimentation before initiation of the study (approval CEEA-LR-10042). Experiments were performed in accordance with the regional ethics committee on animal research and care.

Total RNA was extracted from chondrocytes in micropellet treated (or not) with IL-1β by using a miRvana isolation kit (Ambion, Carlsbad, CA, USA). MiRNA expression profiling was performed by using Miltenyi (Bergisch Gladbach, Germany) microarray facilities. Labeling and hybridization were performed in accordance with the protocol of the manufacturer. Raw data were normalized and additional data analysis was performed as described previously 25 . Microarray data are available in the ArrayExpress database 26 under accession number E-MTAB-2229.

One microgram of Trizol-extracted total RNAs including microRNAs from the different samples were poly(A)-tailed with poly(A) polymerase (NEB M0276L). Then the polyadenylated RNA samples were reverse-transcribed as previously described using 50 units M-MLV Reverse Transcriptase (Invitrogen, Carlsbad, CA, USA) and either random primers or dTmiR adapter 27 . For microRNA and mRNA quantitative analysis, cDNA was mixed with Sybr Green Master Mix (Roche Diagnostics, Indianapolis, IN, USA) in 96-well plates containing specific primers for hsa-miR-24 (universal reverse + specific primer), interest genes or the ribosomal subunit protein-9, housekeeping gene (hRSP9). Quantitative polymerase chain reaction (qPCR) conditions as described 27 used the following primer-probe combinations: for hRSP9 sense 5′-GATTACATCCTGGGCCTGAA antisense 5′-ATGAAGGACGGGATGTTCAC; for Aggrecan (hACAN) sense 5′- TCGAGGACAGCGAGGCC anti-sense 5′-TCGAGGGTGTAGCGTGTAGAGA; for hCOL2A1, variant 2 (hCol2A1) sense 5′-CAGACGCTGGTGCTGCT anti-sense 5′-TCCTGGTTGCCGGA CAT; For hMMP13 sense 5′-TAAGGAGCATGGCGACTTCT anti-sense 5′-GTCTGGCGTTTTTGGATGTT; for hp16^INK4a sense 5′- GAAGGTCCCTCAGACATCCCC anti-sense 5′-CCCTGTAGGACCTTCGGTGAC; for hsa-miR-24 sense 5′-TGGCTCAGTTCAGCAGGAACAG Universal Reverse 5′- GCGAGCACAGAATTATACGACT.

Plasmids encoding for miR-24-2 promoter (−2041 base-pair) Luciferase and CMV β-galactosidase were provided by Charles Lecellier 28 . Empty vector or p16^INK4a encoding vector were purchased from Addgene 29 . For promoter activity assay, OA human primary chondrocytes were transfected at day 0, placed in pellet culture conditions and treated with IL-1β during 48 hours. Cells were then lysed according to the dual luciferase/βgal kit (Promega, Charbonnières-les-Bains, France). Firefly Luciferase and β-galactosidase activities were detected using specific substrates with MultiScan FC (Thermo Scientific, Loughborough, UK). Luciferase activity was normalized to β-galactosidase activity.

Human bone marrow-mesenchymal stromal cell (hBM-MSC) culture were established from bone marrow of patients undergoing Hip replacement surgery, after patient informed written consent and approval by the local and national ethics committee (“Cellule de bioéthique de la direction générale pour la recherche et innovation, Ministère de l’Enseignement supérieur et de la Recherche”; registration number DC-2009-1052). Human mesenchymal stromal cells (hMSCs) were isolated and amplified by using a complete alpha-minimum essential medium supplemented with 10% fetal bovine serum + 1 ng/mL of basic fibroblast growth factor. hBM-MSCs were positive for CD44, CD73, CD90, and CD105 but negative for CD14, CD34, and CD45. Chondrogenic differentiation of BM-MSCs was induced by 21-day culture in micropellet 30 . Chondrogenesis was monitored by measuring the expression of chondrocyte-specific markers by reverse transcription-qPCR (RT-qPCR) as described 30 .

Human chondrocytes (75 × 10⁴) were transfected with 15 μg of plasmid for 24 hours by using Transit-LT1 Reagent (Euromedex, Souffelweyersheim, France). Chondrocytes were transfected with 100 nM of AntagomiR control or AntagomiR-24 (purchased from Ambion) by using oligofectamine (Invitrogen, USA). After transfection, cells were trypsinized and pelleted in chondrogenic medium and cultured for 7 days.

For Western blotting, chondrocytes in pellet cultures were lysed in RIPA-Benzonase buffer 22 . After the addition of the lysis buffer, the samples were left on ice for 15 minutes with vortexing every 5 minutes for 10 seconds. Lysate protein samples were then sonicated for 5 minutes, followed by centrifugation at 7000 g for 15 minutes. The protein quantity loaded on Western blot gel corresponded to 25 × 10⁴ cells. Primary antibodies and dilutions were anti-CDKN2A/p16INK4a (ab54210; Abcam; 1:1,000) and anti-β actin (Sigma-Aldrich A228; 1:8,000). Secondary antibody used for Western blot analysis was goat anti-mouse IgG HRP conjugate (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA; 115-035-003; 1:80,000). Enzyme-linked immunosorbent assays (ELISAs) were performed by using kits (IL-6, IL-8, pro-MMP13, and MMP1) from R&D Systems on supernatants stored at −20°C until analysis. Data were normalized and expressed as picograms per milliliter.

Samples were fixed in 3.7% paraformaldehyde for 24 hours, washed in phosphate-buffered saline (PBS), and processed for routine histology. Paraffin-embedded sample sections (5 μm) were rehydrated through a gradient of xylene and ethanol. Samples were first incubated for 20 minutes at room temperature with pepsin solution (EmergoEurope, The Hague, The Netherlands) for antigen retrieval. Endogenous peroxidase blocking was done with 1% H₂O₂ for 20 minutes at room temperature. Samples were pre-incubated with blocking solution (PBS + 10% goat serum + 0.1% Triton) for 30 minutes at room temperature. Endogenous biotins were blocked by using a Streptavidin/Biotin blocking kit (Vector SP-2002) for 30 minutes. Primary antibody anti-CDKN2A/p16^INK4a monoclonal mouse antibody (1:200; Abcam ab54210) was incubated for 72 hours at 4°C. Incubation with Biotin-coupled secondary antibody, IgG (1:200; ABC kit Vector PK6100) was done for 1 hour at room temperature. Incubation with Avidin/Biotin complex (ABC kit Vector PK6100) was done for 30 minutes at room temperature. Immunolocalized antigens were detected by means of a DAB revelation kit (Sigma-Aldrich). Safranin-O staining was performed as described 31 . Quantification of proliferating cell nuclear antigen (PCNA)-positive or PCNA-negative hypertrophic chondrocytes was performed on four different sections of long bones of four mice of each genotype by using ImageJ software (D-0426) in accordance with the instructions of the manufacturer.

Experiments were performed with at least three independent individual chondrocyte or MSC samples. Comparisons of two conditions were done by using a paired Student t test. Unpaired Mann-Whitney test was applied for cartilage samples by using GraphPad Prism Software (GraphPad Software, Inc., La Jolla, CA, USA). Differences were considered significant when P values were less than 0.05.

We first wanted to establish an in vitro model based on mature chondrocytes, mimicking the senescence-like phenotypes found in OA cartilage. This model should associate a 3D setting reproducing that of chondrocytes within the tissue, expression of senescence markers such as p16^INK4a, and production of SASP factors. We therefore isolated primary chondrocytes from OA cartilage patients and that were placed in pellet culture conditions to maintain/reinduce the chondrocyte phenotype. We used IL-1β as one of the major cytokine inducers of cartilage degradation in OA 32 to induce senescence-associated phenotype. We observed that, in response to chronic IL-1β treatment for 5 days, mature chondrocytes lose (as expected) the expression of specific differentiation markers such as Aggrecan (Figure 1A) 33 , concomitantly with p16^INK4a accumulation (Figure 1B), phospho-active form of p38^MAPK (Figure 1C), and production of reactive oxygen species (data not shown), hallmarks of senescence-associated signaling pathways 34 . Furthermore, as expected, IL-1β-treated mature chondrocytes significantly produce senescence-associated secretory factors such as MMP1 and MMP13, two markers normally associated with chondrocyte terminal differentiation onset and IL-6 and IL-8, two pro-inflammatory cytokines (Figure 1D-G). Therefore, this 3D in vitro model recapitulates the p16^INK4a-associated secretory phenotype characterizing senescent-like chondrocytes found in OA cartilage.

Besides inflammatory cytokines, aberrant chondrogenic differentiation signals can also promote OA features. For instance, Notch activation and HIF-2α transcription factor, both inducers of OA, are also central in controlling chondrocyte terminal differentiation onset during in vitro chondrogenesis and in vivo endochondral ossification 9 10 . Thus, we next wanted to determine whether p16^INK4a could also be part of a normal in vitro and in vivo chondrogenic differentiation process as has been reported for Notch or HIF-2α. We therefore evaluated the p16^INK4a expression profile in human MSCs from three different donors undergoing TGF-β3-induced chondrogenic differentiation to recapitulate in vitro all stages of chondrogenesis. Surprisingly, p16^INK4a expression increases during chondrogenesis in parallel with a chondrocyte differentiation marker such as collagen IIB at days 7 to 14 and in a concomitant manner with MMP13, a chondrocyte terminal differentiation marker, at day 21 (Figure 2A-D). Of note, during chondrogenesis, p14^ARF mRNA, an alternative splicing form of p16^INK4a, was not detected at any time (data not shown). Thus, the senescence marker p16^INK4a, which is known to be required for astrocyte 4 or epidermal cell 5 differentiation, seems also to play a role during chondrogenesis. To dissect its in vivo function during chondrogenic differentiation, we compared the endochondral ossification process of transgenic mice deficient in ink4a locus with that of wild-type mice. PCNA and Safranin-O staining were used to quantify proliferative chondrocytes versus non-proliferative chondrocytes within the growth plate of 1-month-old mice. Without affecting the total number of chondrocytes, ink4a ablation reduces the height of hypertrophic terminally differentiated non-proliferative chondrocytes by 48% ± 2.2 (Figure 2E-G), demonstrating a role for ink4a locus in chondrocyte cell fate decision to engage in terminal differentiation. A main function of p16^INK4a is to specifically inhibit cell cycle progression by targeting G1 CDK4/6 activities 35 and therefore maintain retinoblastoma (pRb), p107, and p130 under their active unphosphorylated forms 36 . These three pocket proteins are known to control chondrocyte cell fate decision during bone growth 37 38 39 . Interestingly, two other CKIs—p27^KIP1 and p57^KIP2—are also part of the cell cycle regulation during terminal differentiation through CDK inhibition leading to pocket proteins activation 40 41 . But only p16^INK4a plays a major role in both cellular senescence and differentiation onset as revealed in ink4a knockout mice studies. Taken together, these findings point at the multiple roles of several CKIs and their indirect targets—pRb, p107, and p130—as orchestrators of the differentiation program and, in particular, in chondrocyte terminal differentiation during bone development.

To determine how Ink4a could participate in both OA initiation and chondrocyte terminal differentiation, we assessed whether p16^INK4a expression impacts the establishment of matrix remodeling secretome common in both events. We thus transiently transfected a p16^INK4a-encoding vector in human chondrocytes, before initiating the pellet culture for 7 days and in the absence of IL-1β. Overexpression of p16^INK4a was checked at the mRNA level (Figure 3A) and protein level (Figure 3B). Compared with the control, p16^INK4a-overexpressing mature chondrocytes produced significantly higher levels of MMP1 and MMP13 but did not modulate IL-6 and IL-8. Of note, our findings confirm recent published data showing that p16^INK4a is dispensable for the establishment of inflammatory secretome associated with senescent fibroblasts 42 . Taken together, our results revealed that the senescence-associated CKI, p16^INK4a, triggers the secretion of both MMP1 and MMP13 in mature chondrocytes. These two metalloproteases are matrix remodeling enzyme family members playing a central role in physiological and pathological processes occurring in cartilage 33 43 44 . Indeed, both are expressed during chondrocyte terminal differentiation within the growth plate to coordinate matrix remodeling that promotes bone growth 12 but also in OA articular cartilage during disease progression 45 . How could p16^INK4a control MMPs production? Firstly, ink4a locus could directly regulate MMP1 transcriptional activation through its described physical interaction with APA1/ZNF410 transcription factor which is bound to MMP1 promoter 46 . Secondly, by activating pRb-, p130-, and p107-dependent terminal differentiation programs through CDK inhibition, p16^INK4a accumulation could also contribute indirectly to MMPs production in OA and during endochondral ossification. Indeed, these three pocket proteins can modulate gene expression and cell fate decision through their interactions with chromatin-modifying enzymes 47 . One found, among these enzymes, histone deacetylase (HDAC) family members, which are controlling chondrocyte pre-hypertrophy/hypertrophy transition 16 17 . Remarkably, recent work from Culley and colleagues 48 reveals that pharmacological inhibition of HDACs prevents the expression of metalloproteases such as MMP13 by OA chondrocytes. Thus, p16^INK4a contributes to MMP secretion by activating the pocket protein-HDAC axis in pathological and physiological conditions.

We next wanted to identify putative regulators of p16^INK4a in mature chondrocytes. Since miRs have been shown to play an important role in cartilage physio-pathology 16 , we asked whether some miRs could regulate p16^INK4a expression in IL1β-treated chondrocytes. miR-array analysis was performed on small RNAs extracted from chondrocytes that were from three different donors and that were cultured in pellets in the presence or absence of IL-1β (Figure 4). Bioinformatic analysis revealed that 179 miRs (128 up and 51 down) are differentially expressed in response to IL-1β compared with untreated cells. These deregulated miRs have at least a 1.4-fold change and a P value of less than 0.05 (Figure 4A and data not shown). Interestingly, we found among them several previously OA-associated miRs such as miR-27b, miR-199, miR-29a, miR-26, and miR-365 16 17 .

Since inhibition of miR-processing enzymes such as Dicer induces senescence-associated phenotypes in primary cells 49 and promotes chondrocyte terminal differentiation in animal models 50 , we focused our attention on miRs that were repressed by IL-1β (Figure 4A). We found miR-24, a known negative regulator of p16^INK4a, through the presence of two binding sites for this miR within its encoding and 3′ untranslated region (UTR) 51 . We next confirmed, on three independent chondrocyte samples placed in 3D, that IL-1β significantly repressed miR-24 expression (Figure 4B) with a concomitant expected induction of p16^INK4a mRNA (Figure 4C).

miR-24 is encoded by two genes: miR-24-1 and miR-24-2 52 . These genes are organized in a cluster including three different miRs (miR-23a or b/27a or b/24). Each cluster is regulated by one promoter common for the three miRs of the cluster 52 . Our array analysis revealed that, upon IL-1β stimulation, chondrocytes show a reduced expression of several members of these two clusters (Figure 4A), suggesting a global repression of the transcription of the clusters. In keeping with this hypothesis, we confirmed the transcriptional repression of miR-24-2 promoter, upon IL-1β addition (Figure 4D), by using a reporter luciferase assay (Figure 4D) previously described 28 .

An increase in expression of p16^INK4a has been demonstrated in cartilage from patients with OA 13 . We then checked whether miR-24 expression could be reversely correlated with that of p16^INK4a in OA cartilage compared with healthy cartilage. By RT-qPCR on mRNA from OA (n = 5) versus healthy (n = 6) human cartilage samples, we revealed a significant miR-24 downregulation in OA cartilage (Figure 4E) while p16^INK4a is increased (Figure 4F). These results were confirmed at the protein level on serial sections of OA cartilage samples by using p16^INK4a immunohistochemistry and miR-24 in situ hybridization (Additional file 1).

Figure 2 shows that p16^INK4a mRNA accumulates throughout the time course of an in vitro chondrogenesis from days 7 to 21 (Figure 2C). We therefore evaluated, during chondrogenesis, whether the expression of miR-24 could also be reciprocal to that of p16^INK4a expression. By RT-qPCR, we revealed that, compared with days 0 to 7, miR-24 level is decreased at day 14 and significantly at day 21 (Figure 2D) in parallel with an increase in expression of the terminal differentiation marker, MMP13 (Figure 2B), while p16^INK4a remains elevated (Figure 2C). Taken together, these results demonstrate that the expressions of Ink4a and its epigenetic regulator are mutually exclusive in both in vitro and in vivo OA models and during the end of the chondrogenesis. Moreover, miR-24 downregulation seems to follow and sustain a high level of p16^INK4a rather than initiate p16^INK4a accumulation.

Finally, we aimed at determining whether modulation of miR-24 could be sufficient to promote p16^INK4a accumulation and p16^INK4a-dependent matrix remodeling secretome by using our 3D chondrocyte model. Because miR-24 overexpression has been reported to induce apoptosis by repressing DNA damage response pathways 53 , we relied on a loss-of-function experiment based on transfection of chondrocytes by either a specific inhibitor of miR-24 (antagomiR-24) or an irrelevant antagomiR as control. As expected, transfection of antagomiR-24 led to a dramatic downregulation of miR-24 level (Figure 5A) and the concomitant significant upregulation of p16^INK4a at both mRNA (Figure 5B) and protein levels as shown by immunostaining with p16^INK4a antibodies (Figure 5C). Furthermore, we showed that miR-24 downregulation is sufficient to promote a marked increase in MMP1 secretion (Figure 5D) but has no significant effect on MMP13, IL-6, or IL-8 secretion (Figure 5E-G), suggesting the existence of a direct axis miR-24-p16^INK4a-MMP1. The discrepancy between this result and p16^INK4a overexpression alone could be explained by mutual redundancy and interference between the signaling pathways. Nevertheless, miR-24 is part of a cluster containing two other miRs and regulated by the same promoter (Figure 4D). Therefore, miR-24 repression is in vivo always accompanied by that of miR27a/b and miR23a/b (Figure 4A and D). Remarkably, miR27b was recently shown to inhibit MMP13 expression in IL-1β-treated chondrocytes 21 and miR23a/b could negatively regulate Runx2, a transcription factor involved in chondrocyte terminal differentiation, OA, and osteoblastogenesis 54 . Thus, repression of these clusters during OA progression within articular cartilage would promote the appearance of several OA-induced features, including p16^INK4a, MMP1, MMP13, but also Runx2. These findings propose that miR-24-1/miR-24-2 clusters, together with the recently identified miR-140, which targets ADAMTS5 and HDAC4, two hypertrophic inducers 19 20 , are crucial in preventing chondrocyte terminal differentiation in OA.