Up-regulation of FOXD1 by YAP alleviates senescence and osteoarthritis

Cellular senescence is a driver of various aging-associated disorders, including osteoarthritis. Here, we identified a critical role for Yes-associated protein (YAP), a major effector of Hippo signaling, in maintaining a younger state of human mesenchymal stem cells (hMSCs) and ameliorating osteoarthritis in mice. Clustered regularly interspaced short palindromic repeat (CRISPR)/CRISPR associated protein 9 nuclease (Cas9)-mediated knockout (KO) of YAP in hMSCs resulted in premature cellular senescence. Mechanistically, YAP cooperated with TEA domain transcriptional factor (TEAD) to activate the expression of forkhead box D1 (FOXD1), a geroprotective protein. YAP deficiency led to the down-regulation of FOXD1. In turn, overexpression of YAP or FOXD1 rejuvenated aged hMSCs. Moreover, intra-articular administration of lentiviral vector encoding YAP or FOXD1 attenuated the development of osteoarthritis in mice. Collectively, our findings reveal YAP–FOXD1, a novel aging-associated regulatory axis, as a potential target for gene therapy to alleviate osteoarthritis.


Introduction
Mesenchymal stem cells (MSCs) are widely distributed in adult tissues and have the capacities of self-renewal and differentiation into multiple cell lineages, such as chondrocytes, osteoblasts, and adipocytes [1]. MSCs are involved in tissue repair and homeostatic maintenance [2,3]. Over time, MSCs exhibit an age-associated decline in their number and function [4][5][6], namely, MSC senescence, which may be implicated in the loss of tissue homeostatic maintenance and leads to organ failure and degenerative diseases [7][8][9][10]. Therefore, an understanding of the mechanisms underlying MSC senescence will likely reveal novel therapeutic targets for ameliorating degenerative diseases.
Osteoarthritis is a prevalent aging-associated disorder that is characterized by the progressive deterioration of articular cartilage [11,12]. In osteoarthritis joints, degenerative changes start with cellular disorganization, gradual stiffening, and irregular surface of superficial zone followed by loss of matrix, clefts, and osteophyte formation in the deep articular cartilage [13,14]. Accordingly, disruption of the superficial zone of cartilage is an onset of osteoarthritis. Previous reports have demonstrated that cells isolated from the superficial zone of mouse and human articular cartilage express MSC markers, including cluster of differentiation (CD) 105, CD166, CD29, and exhibit MSC characteristics [15][16][17][18][19][20]. Cell death induced by oxidative stress or wound occurs primarily at the surface zone of cartilage [21,22]. When such cell death is inhibited by chemicals, cartilage disorganization and matrix loss are greatly reduced [23]. Therefore, MSCs or chondrocyte progenitor cells residing in the superficial zone of cartilage may be a critical target for the prevention of osteoarthritis. Although the transplantation of ex vivo cultures of MSCs into the osteoarthritic joint has been shown to improve the symptoms [24][25][26], the rejuvenation of endogenous senescent MSCs may also be a therapeutic option for osteoarthritis. The localized nature of osteoarthritis, which has no major extra-articular or systemic manifestations, makes it an ideal candidate for local, intra-articular gene therapy [27,28]. However, gene therapy strategies aiming at alleviating senescence, particularly MSC senescence, for treating osteoarthritis have not yet been reported.
Yes-associated protein (YAP) and transcriptional coactivator with PDZ-binding motif (TAZ) are primary targets of the Hippo signaling pathway, which plays important roles in the regulation of development, homeostasis, regeneration, and so forth [29][30][31]. The Hippo kinase cascade phosphorylates YAP and TAZ, resulting in their cytoplasmic retention and proteolytic degradation. When the Hippo pathway is inactive, YAP and TAZ translocate into the nucleus and interact with transcription factors to regulate the expression of target genes [32]. YAP and TAZ, as paralogs, have been demonstrated as key regulators in organ size control [33] and essential transducers of mechanical signals [34]. Here, we identified a critical role for YAP, but not TAZ, in regulating human MSC (hMSC) senescence. YAP exerted a geroprotective effect on hMSCs through the transcriptional activation of forkhead box D1 (FOXD1) in a TEA domain transcription factor (TEAD)-dependent manner. Gene therapy with lentiviral vectors encoding YAP or FOXD1 prevented cellular aging and attenuated osteoarthritis in mice. Our data suggest that YAP and its downstream target FOXD1 are novel suppressors of hMSC senescence and that the YAP-FOXD1 regulatory axis represents a potential therapeutic target for osteoarthritis.

YAP suppresses hMSC senescence in a TEAD-dependent manner
Given that YAP and TAZ displayed distinct functions in regulating hMSC senescence, we next examined whether there were differences in the subcellular localizations of YAP and TAZ. In hMSCs, YAP was predominantly located in the nucleus, whereas TAZ was in the cytoplasm ( Fig 3A). It has been shown that nuclear YAP binds transcription factors, including TEAD family of transcription factors (TEAD1, 2, 3, and 4), as a transcriptional coactivator to induce target gene expression and thus regulate a series of cellular processes [44]. To test whether the nuclear YAP acted in conjunction with TEAD to regulate hMSC senescence, we blocked the activities of all the members of TEAD family in hMSCs as confirmed by immunoblotting analysis (referred to as TEADs knockdown [KD] and KO hMSCs, TEADs KD/KO hMSCs; Fig  3B). Similar to YAP-deficient hMSCs, TEADs KD/KO hMSCs also showed major phenotypes of premature senescence, such as an increased number of SA-β-gal-positive cells (Fig 3C),
Western blotting verified the down-regulation of FOXD1 expression in YAP −/− hMSCs ( Fig  3F) as well as its up-regulation upon the reintroduction of YAP (Fig 3G), suggesting that FOXD1 was transcriptionally controlled by YAP. We examined the FOXD1 promoter region, including 1,500 bp upstream of the transcriptional start site (TSS) and identified 4 putative TEAD binding sites between −1,500 and −1,000 bp and 1 between −1,000 bp and the TSS ( Fig  3H). Accordingly, we detected these 2 regions followed by chromatin immunoprecipitation (ChIP) using YAP and TEAD4 antibodies, revealing that YAP and TEAD4 bound predominantly within 1,000 bp upstream of the FOXD1 TSS, where there was a putative TEAD binding site (Fig 3I and 3J). Next, we cloned this promoter region (−1,000 bp to the TSS) as a transcriptional element upstream of a basic Luc reporter. Reporter activity was lower in YAP −/− hMSCs than in WT cells ( Fig 3K) and was increased upon YAP or TEAD4 overexpression. Luc activity was even higher upon the expression of a constitutively activated YAP mutant (YAP-S127A) and was further enhanced by coexpression of YAP and TEAD4 (Fig 3L). The high levels of Luc activity were significantly abolished when we mutated the predicted TEAD binding site ( Fig  3M). By contrast, ChIP assay demonstrated that TAZ did not bind to the FOXD1 promoter (S4F Fig), and the Luc activity was insensitive to cellular TAZ levels (Figs 3K and S4G). Therefore, the YAP-TEAD pathway, but not TAZ, transcriptionally activates FOXD1 expression.
FOXD1 was initially implicated in renal development [47], but there was a lack of evidence for a link between FOXD1 and cellular senescence. To investigate whether FOXD1 participated in YAP deficiency-induced accelerated senescence of hMSCs, we knocked out FOXD1 in hMSCs using a lentiviral vector-dependent CRISPR/Cas9 system [42,43] (Figs 4A and S5A).  (Fig 4C). In addition, we also examined the gene expression profile of FOXD1 KO hMSCs using RNA-seq (S5 Data). FOXD1 KO decreased the expression of genes that were mainly associated with cell division and DNA replication, which ultimately contributed to the senescence phenotypes ( Fig 4D and S6 Data). Combined analyses with WT and YAP −/− hMSCs showed that FOXD1-deficient and YAP −/− hMSCs were similar to each other at the transcriptomic level (S5D and S5E Fig). Many differentially expressed genes were overlapped between YAP −/− (compared to WT) and FOXD1 KO (compared to NTC-transduced) hMSCs, including 116 up-regulated genes accounting for 20% of the total up-regulated genes in YAP −/ − hMSCs and 276 down-regulated genes accounting for 32% of the total down-regulated genes in YAP −/− hMSCs (Fig 4E and 4F), implying an important role for FOXD1 in mediating YAP deficiency-induced premature cellular aging. Of note, many of those commonly down-regulated genes were elevated upon ectopic expression of FOXD1 in YAP −/− hMSCs ( Fig 4G). Conversely, ectopic expression of YAP in FOXD1 KO or TEADs KD/KO hMSCs did not exert obvious rescue effect on the senescence phenotypes (S5F and S5G Fig). Taken together, these data indicate that down-regulation of FOXD1, an effector of YAP-TEAD signaling, contributes to the premature senescence induced by YAP deficiency. The protein levels normalized with β-actin were shown as fold change relative to Ctrl hMSCs (right). Data are presented as the mean ± SD, n = 3, � P < 0.05. (C) SA-β-gal staining of Ctrl and TEADs KD/KO hMSCs. Scale bar, 100 μm. Data are presented as the mean ± SD, n = 3, ��� P < 0.001. (D) A heat map showing relative mRNA expression levels of the differentially expressed genes in YAP −/− hMSCs. Genes were sorted by the fold change and P value (fold change > 2 or < 0.5, P < 0.01). Corresponding gene expression profiles obtained from TAZ −/− hMSCs were also shown. (E) The bioinformatics analysis predicted that 476 (55%) of 862 genes down-regulated in YAP −/− hMSCs were potential YAP-TEAD targets. Among these genes, FOXD1 was the most down-regulated gene. (F) Western blot analysis of FOXD1 in WT, YAP −/− , and TAZ −/− hMSCs. GAPDH was used as a loading Ctrl (left). The protein levels normalized with GAPDH were shown as fold change relative to WT hMSCs (right). Data are presented as the mean ± SD. n = 3, ��� P < 0.001. (G) RT-qPCR showing elevated expression of FOXD1 in YAP −/− hMSCs transduced with a lentivirus encoding YAP. Data are presented as the mean ± SD, n = 3, ��� P < 0.001. (H) TEAD binding sites were examined in the FOXD1 Pro, and the putative binding sites are depicted as triangles. (I) ChIP-qPCR for YAP enrichment within different FOXD1 Pro regions (Pro1 and Pro 2) containing putative TEAD binding motifs. Data are presented as the mean ± SD, n = 3, ��� P < 0.001. (J) ChIP-qPCR for TEAD4 enrichment within different FOXD1 Pro regions (Pro1 and Pro 2) containing putative TEAD binding motifs. Data are presented as the mean ± SD, n = 3, � P < 0.05, �� P < 0.01. (K) The FOXD1 Pro region (Pro 2) was cloned upstream of a Luc reporter, and Luc activity was detected in WT, YAP −/− , and TAZ −/− hMSCs. Data are presented as the mean ± SD, n = 3, � P < 0.05. (L) The FOXD1 Pro-mediated Luc activity was detected in YAP −/− hMSCs in the indicated experiments. Data are presented as the mean ± SD, n = 3, � P < 0.05, �� P < 0.01, ��� P < 0.001. (M) The mutant FOXD1 Pro2 driven Luc activity was detected in the indicated experiments. Data are presented as the mean ± SD, n = 3. The numerical data underlying this figure are included in S8 Data. ChIP, chromatin immunoprecipitation; Ctrl, control; FOXD1, forkhead box D1; FPKM, fragments per kilobase per million mapped fragments; GAPDH, glyceraldehyde-3-phosphate dehydrogenase; GFP, green fluorescent protein; hMSC, human mesenchymal stem cell; IgG, Immunoglobulin G; KD, knockdown; KO, knockout; Luc, luciferase; MSC, mesenchymal stem cell; mut, mutant; ns, not significant; Pro, promoter; RT-qPCR, reverse transcription quantitative polymerase chain reaction; S127A, serine127alanine; SA-β-gal, senescence-associated-β-galactosidase; TAZ, transcriptional coactivator with PDZ-binding motif; TEAD, TEA domain transcriptional factor; TSS, transcriptional start site; WT, wild type; YAP, Yes-associated protein.

Overexpression of YAP or FOXD1 attenuates post-traumatic osteoarthritis in mice
Mesodermal cellular aging has emerged as a fundamental hallmark of aging-related disorders, including osteoarthritis, one of the most common degenerative diseases, the incidence of which increases significantly with age. Dysfunction of MSCs residing in the superficial zone of cartilage precedes osteoarthritis [15][16][17][18][19]21,22] that is characterized by articular cartilage degradation [49][50][51]. To validate a role of hMSC senescence in driving osteoarthritis, we injected young hMSCs, RS hMSCs, and RS hMSCs overexpressing YAP or FOXD1, respectively, into the joints of immunodeficient mice and performed histological assessment of the joints 1 month later (S7A Fig). In line with a previous report [52], Safranin O staining revealed the delamination of the articular surface and erosion of articular cartilage in the RS hMSC-administrated joints (S7B and S7C Fig). However, no osteoarthritis-related features manifested in the joints transplanted with young hMSCs or RS hMSCs overexpressing YAP or FOXD1. RT-qPCR further demonstrated that RS hMSCs, rather than young hMSCs, and RS hMSCs overexpressing YAP or FOXD1 induced aging markers in mouse joints (S7D Fig). These results suggest that accumulation of senescent MSCs in joints contributes to the development of osteoarthritis, which can be eliminated by YAP or FOXD1 overexpression.
The elimination of local senescent cells using pharmacological or genetic approaches have been proven effective in attenuating age-associated bone loss and development of post-traumatic osteoarthritis in rodents [51,53]. Given the ability of YAP or FOXD1 to rejuvenate senescent MSCs, we hypothesized that intra-articular injection of lentiviral vectors expressing YAP or FOXD1 might exert a therapeutic effect on osteoarthritis. To test this, we performed an anterior cruciate ligament transection (ACLT) surgery widely used to trigger osteoarthritis in mice and then administrated the lentiviruses expressing flag-tagged Luc, YAP, or FOXD1 intra-articularly ( Fig 6A). The lentiviral vectors steadily expressed exogenous proteins in and around the joints receiving virus injection for at least 7 weeks (S8A Fig). High expression levels of YAP and FOXD1 were detectable by RT-qPCR (Fig 6B and 6C); immunohistochemical analysis of the flag-tagged Luc, YAP, and FOXD1 further verified the persistent infection of the lentiviruses and expression of indicated proteins primarily in the superficial zone of articular cartilage (S8B Fig). As expected, ACLT induced the accumulation of P16-positive senescent cells in the articular cartilage, particularly in the superficial zone of cartilage, of the osteoarthritis mice (Fig 6D), which was accompanied by decreased levels of YAP and FOXD1 (S8C and S8D Fig). YAP or FOXD1 gene therapy reduced the number of senescent cells and alleviated ACLT-induced articular cartilage erosion and clefts (Fig 6D and 6E). Consistently, a substantial proportion of gene expression changes in the joints induced by ACLT were reversed by YAP or FOXD1 gene therapy (S8C- S8F Fig and S7 Data). For instance, increased expression  (Dapk1, Casp4, etc.) were observed in ACLT-induced osteoarthritis joints, and the expression levels of most of these genes were diminished upon YAP or FOXD1 treatment. Moreover, the YAP or FOXD1 treatment enhanced the expression of proliferation markers (Ki67, Aspm, etc.) and chondrocyte differentiation-related genes (Col2a1, Acan, etc.) (Fig 6F). Taken together, these data suggest that the YAP-or FOXD1-mediated alleviation of cellular senescence in local bone joints helps create a prochondrogenic environment and alleviates disease symptoms.

Discussion
Cellular senescence and stem cell exhaustion are hallmarks of aging [54]. Accelerated attrition of the MSC pool has been observed in human stem cell and mouse models of premature aging disorders, including WS and Hutchinson Gilford progeria syndrome (HGPS) [37,55]. Transplantation of mesoderm-derived stem cells from young animals increases the lifespan of progeroid mice [56]. Quercetin has been shown to alleviate MSC senescence [57], improve physical function, and increase lifespan in aged mice [58]. From this perspective, senescent MSCs could be good therapeutic targets for aging-associated degenerative disorders. In this study, we presented several lines of evidence supporting a geroprotective role of YAP and FOXD1 in rejuvenating hMSCs: (1) YAP is required for preventing premature senescence of hMSCs; (2) YAP transcriptionally activates FOXD1 expression whereas YAP deficiency results in down-regulation of FOXD1, which contributes to the early-onset of cellular aging; and (3) lentiviral gene transfer of YAP or FOXD1 alleviates cellular senescence and osteoarthritis. Our findings define a critical role of the YAP-FOXD1 axis in regulating hMSC aging, which highlights new avenues for translation into geriatric and regenerative medicine.
The Hippo-YAP/TAZ signaling pathway is an evolutionarily conserved pathway that regulates cell proliferation and apoptosis. Here, we focused on the function of YAP and/or TAZ in regulating hMSC senescence. We generated isogenic YAP-or TAZ-deficient hMSCs. Compared with WT cells, TAZ −/− hMSCs showed minimal effect in cell growth, whereas YAP −/− hMSCs exhibited accelerated senescence. The cytoplasmic localization of TAZ underlays its inactivation in hMSCs, whereas nuclear YAP was essential for counteracting hMSC senescence. Consistent with our observations, emerging studies have revealed the differences between YAP and TAZ. For example, TAZ promotes the myogenic differentiation of myoblasts at late stages of myogenesis, whereas YAP inhibits this process in mice [59]. However, in-depth insights into the molecular mechanisms underlying these functional differences require further investigations.
https://doi.org/10.1371/journal.pbio.3000201.g005 YAP-FOXD1 axis alleviates senescence and osteoarthritis  is a new downstream target of YAP, loss of which mediates the senescent phenotype of YAP-deficient hMSCs. As a member of the forkhead box family of transcription factors, FOXD1 is known to regulate kidney development during organogenesis [60,61]. Recently, FOXD1 has been shown to promote cell proliferation by targeting the sonic hedgehog pathway and cyclin-dependent kinase inhibitors [62,63]. FOXD1 also facilitates the reprogramming of mouse embryonic fibroblasts (MEFs) into induced pluripotent stem cells (iPSCs) [64]. Here, we identified a geroprotective role for FOXD1 as a transcriptional target of YAP in rejuvenating hMSCs. Overexpression of YAP or FOXD1 delayed replicative and pathological senescence, implying a therapeutic potential of targeting the YAP-FOXD1 axis to relieve agingassociated degenerative diseases.
In a therapeutic context, we provided a proof-of-concept evidence that intra-articular lentiviral transduction of a single protein exerted therapeutic effects on ACLT-induced osteoarthritis, an age-related disorder. Because ACLT-induced osteoarthritis is accompanied by the accumulation of senescent cells [65], efforts have been made on chemical-induced elimination of senescent cells to alleviate osteoarthritis in mouse models [50,51]. With gene therapy offering novel therapeutic options for osteoarthritis [66], intra-articular injection presents a minimally invasive procedure that avoids conventional barriers to joint entry, increases bioavailability, and lowers systemic toxicity [67]. For the first time, our study shows that the intra-articular injection of lentiviruses expressing YAP or FOXD1 reduces the number of senescent cells, inhibits articular inflammation and cartilage erosion, and ameliorates the pathological symptoms. Therefore, gene therapy via the introduction of geroprotective factors aiming at rejuvenating senescent cells may represent a new avenue to treating osteoarthritis in the future.

Generation of YAP −/− and TAZ −/− hESCs
CRISPR/Cas9-mediated gene targeting was performed using previously described methods, with some modifications [69]. The YAP or TAZ gRNA was cloned into the gRNA vector (Addgene #41824). The donor plasmid for homologous recombination containing homology arms and a neo cassette was described previously [70]. Briefly, 5 × 10 6 H9 ESCs were mixed with the plasmid cocktail and electroporated. After electroporation, cells were plated on a G418-resistant MEF feeder layer. Two days later, cells were treated with 100 μg/ml G418 (Gibco, 10131027) for screening. After 2 weeks of selection, G418-resistant clones were manually picked, transferred to 96-well plates, and expanded for genotyping. Gene-targeted clones were identified using genomic PCR. gRNA sequences and primers are listed in S1 Data.

Generation of TEADs KD/KO hMSCs
We generated 2 lentiviral constructs to silence the expression of TEAD1, 2, 3, and 4: one containing an shRNA targeting TEAD1, TEAD3, and TEAD4 and the other containing an sgRNA targeting TEAD2 [73]. The shRNA was cloned into the PLVTHM vector (Addgene #12247), and the sgRNA was cloned into lentiCRISPRv2. We then cotransduced these lentiviral constructs into hMSCs. Seventy-two hours later, transduced cells were enriched by treatment with 1 μg/ml puromycin (Gibco, A1113803). The targeting sequences are listed in S1 Data.

Immunofluorescence staining
Cells were fixed with 4% paraformaldehyde for 30 minutes, washed with PBS, permeabilized with 0.4% Trion X-100 in PBS, and then blocked with 10% donkey serum (Jackson ImmunoResearch Labs, West Grove, PA). Afterwards, cells were incubated with primary antibodies in blocking solution at 4˚C overnight, followed by an incubation with the corresponding secondary antibodies and Hoechst 33342 for 1 hour at room temperature.

SA-β-gal staining
The SA-β-gal staining of hMSCs was conducted using a previously described method [74].

Protein, RNA, and DNA analyses
For western blotting, cells were lysed in RIPA buffer containing a protease inhibitor cocktail (Roche) and quantified with a BCA kit. Generally, 20 μg of cell lysate was subjected to SDS-PAGE and electrotransferred to a PVDF membrane (Millipore, Billerica, MA). Then, the membrane was incubated with primary and HRP-conjugated secondary antibodies. Western blot data were quantified with Image Lab software for the ChemiDoc XRS system (Bio-Rad, Hercules, CA). For RT-qPCR, cellular total RNA was extracted using TRIzol (Thermo Fisher Scientific), and genomic DNA was removed with a DNA-free Kit (Ambion, Austin, TX), followed by cDNA synthesis with the GoScript Reverse Transcription System (Promega, Madison, WI). RT-qPCR was performed with qPCR Mix (TOYOBO, Tokyo, Japan) in a CFX384 Real-Time system (Bio-Rad). For genomic PCR, genomic DNA was extracted with a DNA extraction kit (TIANGEN, Beijing, China), and PCR was conducted using PrimeSTAR (TAKARA, Tokyo, Japan).

Clonal expansion assay
Two thousand cells were seeded in each well of a 12-well plate and then cultured until clear cell colonies formed to determine the clonal expansion abilities of hMSCs. The relative colony area was then determined by performing crystal violet staining and measured using ImageJ software.

Dual Luc assay
The indicated fragments of the FOXD1 promoter were amplified by PCR and cloned into the pGL3-Basic vector (Promega). The mutant of pGL3-FOXD1 promoter 2-Luc was constructed with the Fast Mutagenesis System kit (FM111; Transgen Biotech, Beijing, China). PGL3-FOXD1 promoter 2-Luc or PGL3-FOXD1 promoter 2(mut)-Luc was transfected into hMSCs together with vectors expressing the proteins of interest and Renilla-Luc, which was used to normalize the transfection efficiency. For detection of the 8 × GTIIC-Luc activity, the 8 × GTIIC reporter (Addgene #34615) and Renilla-Luc plasmids were cotransfected into hMSCs. Cells were harvested 72 hours later using the Dual-Luciferase Reporter Assay System (Vigorous Biotechnology, Beijing, China) and assayed according to the manufacturer's instructions.

ChIP
ChIP was performed using a previously reported protocol with minor modifications [75]. Briefly, cells were cross-linked with 1% (v/v) formaldehyde for 15 minutes at room temperature, and the reaction was terminated by the addition of 125 mM glycine and an incubation for 5 minutes at room temperature. Then, cells were scraped and lysed in lysis buffer. After sonication, protein-DNA complexes were incubated with antibody-coupled Protein A beads at 4˚C overnight. After elution and reverse cross-linking at 68˚C, DNA was purified by phenol/ chloroform extraction and ethanol precipitation and then subjected to qPCR analysis. Antibodies for ChIP included anti-YAP (14074; Cell Signaling Technology), anti-TAZ (4883; Cell Signaling Technology), anti-TEAD4 (101184; Santa Cruz Biotechnology), and normal rabbit IgG (2027; Santa Cruz Biotechnology) as a negative control.
For RS-hMSC-induced osteoarthritis, we transplanted PBS, young hMSCs, RS hMSCs, and RS hMSCs overexpressing YAP or FOXD1 intra-articularly into the joints of NOD-SCID mice (6 to 8 weeks, male). Firstly, the mice were anaesthetized using isoflurane, and skin around the joints were shaved. For each injection, the needle was inserted beneath the middle patellar ligament, and a volume of 10 μl containing either PBS or 3 × 10 6 cells was injected intra-articularly. One month later, the mice were euthanized, and the joints were collected for mRNA quantification and histological assessments.
For surgically induced osteoarthritis, we performed ACLT surgery on 8-week-old male C57BL/6 mice. Animals were anaesthetized, and their hindlimbs were shaved. After the opening of the joint capsule, the anterior cruciate ligament was transected with microscissors under a surgical microscope. After irrigation with saline to remove tissue debris, the skin incision was closed. Then, 7 days later, a total volume of 10 μl of the indicated lentivirus was injected intra-articularly. At week 8, the mice were euthanized, and the joints were collected for mRNA quantification and histological assessments.

Histology
Mouse joints were fixed with 4% paraformaldehyde overnight, decalcified with 5% methanoic acid for 7 days, and embedded in paraffin. Sections (5 μm) were cut from the paraffin blocks and stained with Fast Green FCF (0.02%) and Safranin O (0.1%). Joint pathology was quantified using the OARSI scoring system [13].

Immunochemistry
For immunohistochemical staining, paraffin-embedded tissue sections were subjected to a heat-mediated antigen retrieval procedure, and then endogenous peroxidases were blocked with hydrogen peroxide. Next, tissue sections were incubated with a primary antibody overnight. Finally, the appropriate secondary antibody (ZSGB-BIO, Beijing, China) was added to the sections, which were then incubated for 30 minutes. Antigen-positive cells were visualized using the DAB Substrate kit (ZSGB-BIO). Anti-P16 antibody (54210; Abcam) and anti-flag (166355; Santa Cruz Biotechnology) were used as the primary antibodies.

CNV identification
First, genomic DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen, Duesseldorf, Germany) according to the manufacturer's instructions. DNA was sheared into fragments of approximately 300 bp using Covaris, and then the library of the fragmented DNA was constructed using the NEBNext ultra DNA Library Prep Kit for Illumina (NEB, Beverly, MA), according to the manufacturer's protocol. The libraries were sequenced on an Illumina HiSeq 4000 platform. For CNV identification, we used the published R package HMMcopy [76]. Briefly, the genome was binned into consecutive 1 Mb windows with read Counter, and then we calculated the absolute number of reads detected in each window. We estimated the copy number with GC and mappability corrections with HMMcopy.

RNA-seq library preparation and sequencing
Total RNA was extracted from cultured human cells or mouse joints using the RNeasy Mini Kit (Qiagen) according to the manufacturer's protocol. For cells, 1 × 10 6 cells were analyzed in biological triplicate. For mouse joints, we mixed the RNA extracted from the sample group, and then divided the sample into 3 technical replicates. One to two micrograms of total RNA was used to construct sequencing libraries using the NEBNext Ultra RNA Library Prep Kit for Illumina (NEB). The libraries were sequenced on an Illumina HiSeq 4000 platform. RNA-seq reads were aligned to the hg19 or mm10 reference genome using TopHat2 software [77]. The analysis of differentially expressed genes was performed using DESeq2 [78] based on read counts.