Expression of p16^INK4a is a biomarker of chondrocyte aging but does not cause osteoarthritis

We determined the expression of p16^INK4a with aging in murine chondrocytes by performing quantitative RT–PCR on cartilage tissue isolated from the proximal femur. In accord with observations in other murine tissues (Krishnamurthy et al., 2004), there was a significant increase in p16^INK4a expression (48.4‐ and 43.5‐fold, respectively) in the 18‐month and 22‐ to 27‐month groups as compared to skeletally mature mice of 4 months of age (p < .01, Figure 1a). A significant but less robust upregulation was seen in the related gene product p19^ARF, with an average fold increase of 24.4‐ and 12.9‐fold at 18 months and 22–27 months of age, respectively (p < .01, Figure 1a). These data support the concept that the physiological stresses of aging induce expression of the Ink4a/Arf locus, even in a hyporeplicative cell type such as chondrocytes. To determine whether aged murine chondrocytes display other characteristic features of senescent cells, we assessed senescence‐associated β‐galactosidase (SA β‐gal) activity (Dimri et al., 1995). Consistent with reports in human chondrocytes (Martin & Buckwalter, 2001), chondrocytes from 22‐ to 24‐month‐old mice displayed a higher percentage of SA β‐gal‐positive cells as compared to 5‐ to 9‐month‐old mice (p = .03, Figure S1).

The expression of p16^INK4a with aging was also evaluated in primary human chondrocytes isolated from the cartilage of 57 cadaveric donors without evidence of OA. The donors were between 17 and 72 years old and age was responsible for 27% of the variation in p16^INK4a levels (Figure 1b, r² = .2682, p < .0001). The expression of ARF (denoted as p14^ARF for human cells) was also significantly correlated with age (Figure 1b, r² = .186, p < .001) but to a lesser extent than p16^INK4a. Expression of p16^INK4a and p14^ARF showed a strong correlation with each other (r² = .6962, p < .001, data not shown). The expression of additional candidate genes with potential involvement in senescence or cell cycle inhibition was examined for a relationship to age, but only p16^INK4a and p14^ARF showed a significant correlation to age (Table 1). Thus, the analysis of human chondrocytes suggests that INK4a/ARF expression is particularly affected by the process of chondrocyte aging.

The proliferation of human chondrocytes is restrained in vivo by the dense extracellular matrix of cartilage, but these cells retain the potential for proliferation upon culture in low density monolayer conditions. To determine whether increased p16^INK4a expression with aging has the potential to alter chondrocyte proliferation, we analyzed cell cycle entry in a set of young and older chondrocyte donors (24.25 ± 2.6 vs. 64 ± 2.1 years old) using a pulse of 5‐ethynyl‐2′‐deoxyuridine (EdU). In this cohort, older donors had a 3.5‐fold average increase in p16^INK4a gene expression (Figure 2a, left, p < .05) and displayed a significantly reduced number of cells in S phase as compared to the younger donors (Figure 2a, right, 7.9% vs. 16.1%, p < .05). To explore a potential mechanistic role for p16^INK4a in regulating proliferation in this setting, we used pharmacological inhibition of CDK4/6 to mimic high expression of p16^INK4a. Palbociclib inhibits the binding of CDK4/6 to D type cyclins and has been developed as a promising therapeutic for cancers with altered function of the Cyclin D‐CDK4/6‐p16^INK4a pathway (Fry et al., 2004; Otto & Sicinski, 2017). Palbociclib completely inhibited the proliferation of human chondrocytes (p < .001, Figure 2b, c). Inhibition of CDK1/2/5/9 with Dinaciclib was similar to Palbociclib in preventing cell cycle entry, indicating that human chondrocytes are sensitive to inhibition of multiple CDKs. These results suggest that aging reduces the potential for chondrocytes to enter S phase when stimulated in monolayer culture, and that these effects on proliferation could be caused by altered regulation of several pathways including the Cyclin D‐CDK4/6‐p16^INK4a axis.

In addition to increased expression with chronological age, p16^INK4a responds to physiological demands that accelerate the rate of aging. This has allowed p16^INK4a expression to serve as a biomarker of molecular aging, which can be used to measure the senescence burden and predict cellular function in some settings (Koppelstaetter et al., 2008; Liu et al., 2009; Wood et al., 2016). To determine this relationship in primary human chondrocytes, we analyzed whether the expression of p16^INK4a was associated with markers of chondrocyte dysfunction independent of chronological aging (Figure 3). Expression of several SASP markers showed a positive correlation to p16^INK4a levels despite no correlation to age: insulin‐like growth factor binding protein 3 (IGFBP3, r² = .0971, p = .018), matrix metalloproteinase 1 (MMP‐1, r² = .1247, p < .01), and a strong trend for MMP‐13 (r² = .0667, p = .054). The expression of Aggrecan (ACAN), which is a positive marker of extracellular matrix synthesis, showed a significant reduction in donors with high levels of p16^INK4a expression (r² = .1155, p < .01). This gene expression pattern suggests that p16^INK4a levels may represent the degree of dysfunction in human chondrocytes.

We analyzed the expression of SASP genes that showed a positive correlation to p16^INK4a levels in the context of INK/ARF knockdown. Within one day of transfection, pooled siRNA targeting both genes reduced expression of p16^INK4a and ARF to undetectable levels and less than 10% expression of scrambled siRNA controls, respectively. The expression of IGFBP‐3, MMP‐1, and MMP‐13 was unaffected by knockdown of p16^INK4a and ARF (Figure 3c). When Palbociclib was used to mimic p16^INK4a‐mediated cell cycle arrest, these same SASP factors were also unaffected (Figure 3d). These data support the concept that while SASP markers correlate with p16^INK4a expression, p16^INK4a or cell cycle arrest is not required for production of the SASP.

Given the correlation of p16^INK4a expression with chronological aging and the SASP, as well as the potential for CDK4/6 inhibition to restrain chondrocyte proliferation, we sought to determine the effects of somatic p16^INK4a loss in the chondrocyte compartment. Toward that end, we utilized Cre recombinase to eliminate exon 1α of p16^INK4a in chondrocytes in vivo at skeletal maturity using a previously reported floxed p16^INK4a allele (p16^L; Monahan et al., 2010) and an inducible, chondrocyte‐specific Aggrecan Cre driver (Acan^{tm1(cre/ERT2)Crm}; Henry et al., 2009). A loxP‐stop‐loxP ZsGreen fluorescent reporter allele was also included in a subset of mice to facilitate fluorescent activated cell sorting (FACS) of chondrocytes that had undergone Cre‐mediated recombination. To test the efficacy of this system, chondrocytes were isolated by FACS 1 month after tamoxifen injection, showing undetectable levels of p16^INK4a by qPCR in cells sorted from Acan^{tm1(cre/ERT2)Crm}p16^L/L (p16^INK4a loss) mice, with retained expression in Acan^{tm1(cre/ERT2)Crm}p16^INK4a+/+ (p16^INK4a intact) controls (Figure S2A). The consistency of recombination in the articular cartilage after tamoxifen injection was demonstrated with anti‐tdTomato immunohistochemistry in mice containing Acan^{tm1(cre/ERT2)Crm} and loxP‐stop‐loxP tdTomato fluorescent reporter alleles (Figure S2B).

We examined the effects of p16^INK4a loss on the expression of chondrocyte mRNA markers, expecting little or no effect of p16^INK4a deletion in young mice due to the low expression at this age. To determine whether p16^INK4a loss would affect any age‐related increase in SASP factor production, we also analyzed gene expression of murine chondrocytes from 9‐ to 18‐month‐old animals. Mmp‐13 showed significantly increased expression in older mice (p < .001, Figure S2C), but there was no effect of p16^INK4a loss in either age group. Expression of Igfbp3 showed a similar trend but was not statistically significant. In addition to gene expression, Mmp‐13 protein was detected by immunohistochemistry in joints with and without p16^INK4a loss at 18 months (Figure S2D). These data indicate that p16^INK4a can be efficiently deleted from CreER^T2‐expressing chondrocytes via tamoxifen induction in adult mice, but that this deletion does not inhibit the increased production SASP factors during physiologic aging in mice.

The somatic loss of p16^INK4a can initiate increased cell division and cancer in a cell‐type‐specific fashion (Liu et al., 2011), leading us to investigate chondrocyte proliferation and neoplasia in this cohort. We did not observe neoplasia in the articular cartilage, but did note a high rate of medullary neoplasia accompanied by abundant intramedullary bone formation in Acan^{tm1(cre/ERT2)Crm}p16^L/L mice (Figure S3). This neoplastic process did not appear to alter the histological features of the articular cartilage or underlying subchondral bone, but may have had indirect effects on joint function. To determine whether Acan^{tm1(cre/ERT2)Crm}‐driven loss of p16^INK4a initiated widespread proliferation of articular chondrocytes at the joint surface, we provided BrdU through the drinking water for long‐term pulsing experiments. Articular chondrocytes with p16^INK4a loss remained refractory to proliferation even in the context of cartilage damage induced by destabilization of the medial meniscus (DMM) that might stimulate attempted repair, as 3 weeks of continuous BrdU treatment only marked areas of early osteophyte formation (Figure S4A). This low rate of proliferation is one reason that Acan^{tm1(cre/ERT2)Crm}‐driven recombination persists in the articular cartilage over time, as demonstrated by immunohistochemistry targeting the product of a loxP‐stop‐loxP reporter allele 6 months after initiating recombination with tamoxifen (Figure S4B). These data suggest that articular chondrocytes exhibit little if any proliferation in adult murine cartilage, even in the setting of p16^INK4a loss.

We also used in vitro culture of murine chondrocytes to explore the effect of p16^INK4a loss on expansion rate and the development of replicative senescence. In these studies, we isolated chondrocytes from 3‐week‐old mice and expanded the cells through four passages. Cells with p16^INK4a loss showed similar expansion rates and lost replicative potential at the same passage as chondrocytes from littermate controls that retained p16^INK4a expression (Figure S5A). Additional features of senescence such as extensive staining for SA β‐gal (Figure S5B) and increased expression of the SASP marker Igfbp3 were also not affected by somatic loss of p16^INK4a (Figure S5C).

The functional effects of p16^INK4a loss in chondrocytes in vivo were further explored by analyzing the extent of age‐related OA in animals with or without p16^INK4a deletion in chondrocytes. Littermate cohorts of Acan^{tm1(cre/ERT2)Crm}p16^INK4a+/+ and Acan^{tm1(cre/ERT2)Crm}p16^L/L male mice were treated with tamoxifen at 4 and 12 months of age to induce p16^INK4a loss. The development of OA was evaluated in mice sacrificed at 18 months of age using established histological scoring systems based on the loss of Safranin‐O staining and osteophyte formation in the load‐bearing cartilage of the femoral condyle and tibial plateau (McNulty et al., 2011). Spontaneous age‐related OA occurs with variable progression in male C57BL/6 mice, and our histological results underscored this variability by demonstrating mild, moderate, and severe OA at 18 months of age (Figure 4a). The Safranin‐O staining scores showed a similar distribution in both cohorts, indicating that inducing p16^INK4a loss in chondrocytes is insufficient to prevent cartilage degradation with age (p > .05, Figure 4b). The effect on osteophyte formation, which is another marker of OA development in both human and murine joints, also showed no change with somatic inactivation of p16^INK4a (p > .05, Figure 4c).

Given that no effect of p16^INK4a deletion was observed in aging‐associated OA, we sought to provoke a more severe and phenotypically homogeneous form of OA through DMM. Applying our genetic approach to a joint injury model was also motivated by the demonstration that p16^INK4a expression increases with transection of the anterior cruciate ligament and that elimination of senescent cells can mitigate post‐traumatic OA (Jeon et al., 2017). DMM surgery was performed on littermate cohorts of Acan^{tm1(cre/ERT2)Crm}p16^INK4a+/+ and Acan^{tm1(cre/ERT2)Crm}p16^L/L male mice at 12 months of age and we assessed cartilage degradation 8 weeks after surgery. Histological evidence of cartilage degradation was present in the medial compartment of DMM hindlimbs (Figure 5a). As expected, DMM surgery increased both the Safranin‐O and osteophyte scores as compared to the contra‐lateral control hindlimbs in both the p16^INK4a intact and p16^INK4a loss cohorts (p < .05, Figure 5b,c). As was the case for age‐induced OA, however, there was no difference in the degree of OA when comparing the DMM hindlimbs of p16^INK4a loss mice to the DMM hindlimbs of p16^INK4a intact mice (p > .05, Figure 5b,c). These results indicate that p16^INK4a loss in chondrocytes is insufficient to protect from the development of age‐induced or post‐traumatic OA.

All animal experiments were performed under protocols approved by the Institutional Animal Care and Use Committee of the University of North Carolina at Chapel Hill. The Acan^{tm1(cre/ERT2)Crm} allele (Henry et al., 2009) (received from Dr. Benoit de Crombugghe, now available as stock #019148; Jackson Labs, Bar Harbor, ME, USA) was crossed to a p16^L conditional allele (Monahan et al., 2010). For OA studies with p16^INK4a intact (Acan^{tm1(cre/ERT2)Crm}p16^INK4a+/+) and p16^INK4a loss (Acan^{tm1(cre/ERT2)Crm}p16^L/L) cohorts, male mice of C57BL/6 background were used due to the known development of hindlimb OA. For flow cytometry cell sorting, the loxP‐stop‐loxP ZsGreen reporter mouse (Gt(ROSA)26Sor^{tm6(CAG‐ZsGreen1)Hze}/J stock # 007906; Jackson Labs) was crossed into mice with or without loss of p16^INK4a. For lineage tracing and some flow cytometry cell sorting, mice containing Acan^{tm1(cre/ERT2)Crm} were crossed to the loxP‐stop‐loxP tdTomato reporter mouse (Gt(ROSA)26Sor^{tm14(CAG‐tdTomato)Hze}/J stock # 007914; Jackson Labs). For in vitro monolayer expansion studies, a constitutive type II collagen Cre driver (Col2a1‐Cre) was used to recombine the p16^L conditional allele in chondrocytes without the necessity of tamoxifen induction (Sakai et al., 2001).

Cartilage tissue was dissected from the proximal end of the femur from C57BL/6 mice with the following ages: 4, 10–12, 18, and 22–27 months. Tissue was placed in ceramic bead tubes (MP Bio, Santa Ana, CA, USA) containing Trizol (Thermo Fisher Scientific, Waltham, MA, USA) and isolated using a Precellys^® 24 homogenizer (Bertin Corp, Rockville, MD, USA). RNA was isolated using phenol chloroform extraction and NucleoSpin^® column clean‐up (Macherey‐Nagel, Düren, Germany). Reverse transcription was performed using the ImProm‐II™ system (Promega Corporation, Madison, WI, USA) according to the manufacturer's instructions and quantitative RT–PCR was performed with TaqMan™ Universal Master Mix on a QuantStudio™ 6 Flex machine (Applied Biosystems, Foster City, CA, USA). Custom TaqMan™ primers specific to murine p16^INK4a (Assay ID: AIMSG0H; F: CGGTCGTACCCCGATTCAG; R: GCACCGTAGTTGAGCAGAAGAG; probe AACGTTGCCCATCATCA) and p19^ARF (Assay ID: AIMSH0Y; F: TGAGGCTAGAGAGGATCTTGAGAAG; R: GTGAACGTTGCCCATCATCATC; probe: ACCTGGTCCAGGATTC) were used, with data normalized to murine Tbp as a housekeeping control (Mm00446973_m1; Applied Biosystems). For studies involving gene expression analysis of murine chondrocytes with and without p16 loss, chondrocytes were sorted based on Cre‐driven fluorescent reporters directly into Trizol, RNA was cleaned with NucleoSpin^® columns, and RNA reverse transcribed with qScript XLT reagent (Quantabio, Beverly, MA, USA). Taqman primer probes for Mmp13 (Mm00439491_m1) and Igfbp3 (Mm01187817_m1) were used for SASP analysis.

For studies on cell expansion, chondrocytes from 3‐week‐old Col2a1‐Cre; loxP‐stop‐loxP ZsGreen mice were digested overnight using 0.4 mg/ml Collagenase P. Cells sorted as zsGreen‐positive chondrocytes were plated at 10,000 cells/cm² and passaged weekly. In the final passage, some wells were stained for SA β‐gal (Cell Signaling Technologies, Danvers, MA) according to the manufacturer's recommendations. After DAPI counterstain, five matched bright field and fluorescent images were taken for each independent cell preparation to quantify the percentage of positively stained cells.

At four months of age (and again at 12 months for the aging study), mice received three doses of 25 mg/kg tamoxifen (Sigma‐Aldrich, St. Louis, MO, USA) in corn oil (Sigma‐Aldrich) by intraperitoneal injection to activate the Cre recombinase. The DMM surgery (Glasson, Blanchet & Morris,) was performed on mice at 12 months of age as described previously (Loeser et al.,). Briefly, mice were anesthetized with isoflurane, access to the joint space was obtained through a para‐patellar incision, and the medial meniscotibial ligament was transected using a scalpel. Mice recovered with normal movement and were sacrificed 8 weeks after the surgery. 2007 2012

Murine hindlimbs were dissected at sacrifice and fixed for 3–4 days in 4% paraformaldehyde (Electron Microscopy Sciences, Hatfield, PA, USA) at 4°C before decalcification in Immunocal for 3–4 days at room temperature (age‐related OA study) or 19% EDTA for 2 weeks at room temperature (DMM study). Joints were embedded in paraffin and coronal sections of 4–5 μm were taken from the mid‐coronal plane for histological assessment. Slides stained with Safranin‐O (proteoglycans)/Fast Green (collagen)/Hematoxylin (nuclei) were used to analyze a Safranin‐O score as previously described (McNulty et al., 2011). Briefly, the four quadrants of the load‐bearing region (medial and lateral sides of the femoral condyle and tibial plateau) were summed after grading on a 0–12 scale, with 12 representing complete loss of staining or tissue over more than 2/3 of the surface. An articular cartilage structure score was also generated as previously described (McNulty et al., 2011), which showed similar results to the Safranin‐O staining score but with less sensitivity to subtle changes in OA progression (data not shown). Slides stained with hematoxylin and eosin were used to assess osteophyte score, with each quadrant graded on a 0–3 scale for osteophyte size as previously described (Rowe et al., 2017). For immunohistochemistry with a primary antibody targeting tdTomato (cat # 600‐401‐379; Rockland Immunochemicals, Limerick, PA) or targeting Mmp‐13 (ab84594; Abcam, Cambridge, UK), antigen retrieval was performed with heated sodium citrate and sections were stained using the VECTASTAIN^® Elite ABC‐HRP kit (Vector Laboratories, Burlingame, CA).

Human ankle tissue was provided by Dr. Susan Chubinskaya at Rush University Medical Center (Chicago, IL) through the Gift of Hope Organ and Tissue Donor Network (Elmhurst, IL). Donors with OA (Collins grade ≥ 3) were excluded. As previously described (Loeser, Pacione & Chubinskaya, 2003), cartilage tissue was dissected and sequentially digested with Pronase and Collagenase to obtain chondrocytes. Cell pellets were snapped frozen and stored at −80°C until RNA isolation with RNeasy columns (Qiagen, Hilden, Germany). RNA was reverse transcribed using the ImProm‐II™ system (Promega) and analyzed for gene expression as indicated in Table 1.

Isolated primary chondrocytes were plated in six‐well plates (Corning, Corning, NY, USA) at 20,000 cells/cm² in DMEM/F12 media (Gibco, Life Technologies, Carlsbad, CA, USA) with 10% fetal bovine serum (Sigma‐Aldrich), penicillin/streptomycin (Gibco), and amphotericin B (Sigma‐Aldrich). For studies comparing young (24.25 ± 2.6 years old) and older donors (64 ± 2.1 years old), a 4‐hr pulse of EdU was added 22–26 hr after a media change to label cells in S phase. Cells were harvested by trypsinization, fixed in 1% paraformaldehyde, and permeabilized with 0.1% Saponin (Sigma‐Aldrich). Incubation was performed with Click‐It™ EdU Alexa Fluor™ 555 according to the manufacturer's instructions and DAPI was used to label DNA for assessment of 2n and 4n content. Cells were analyzed on either a CyAn flow cytometer (Beckman Coulter, Brea, CA) or Attune NxT (Thermo Fisher) flow cytometer.

For assessment of cell cycle inhibition in five donors (67 ± 5.7 years old), 1 μm Palbociclib (PD‐0332991 HCl, ChemShuttle, Wuxi, China) or 50 nm Dinaciclib (Selleck Chem, Houston, TX) was delivered to chondrocytes for 26 hr, with the inclusion of EdU for the final 4 hr. For knockdown of the INK/ARF locus, chondrocytes from four donors underwent Amaxa Nucleofection (Lonza, Basel, Switzerland) for treatment with 1 μm siRNA SMARTPool (Dharmacon, Lafayette, CO) that targets both p16^INK4a and ARF or a scrambled control. Gene expression for SASP markers was analyzed 3 days after transfection.

Statistical analysis and plotting were performed using Prism 7 (GraphPad, La Jolla, CA, USA) and flow cytometry data were processed with FCS Express (De Novo Software, Glendale, CA, USA). Data are plotted as individual points with bars indicating mean ± standard error of the mean (SEM). Correlations for human chondrocyte gene expression to age and p16^INK4a were performed using linear regression of data normalized to the housekeeping gene YWHAZ, with r² indicating goodness of fit. For Safranin‐O and osteophyte scores, data were analyzed using nonparametric tests (Kruskal–Wallis for one‐way ANOVA, Mann–Whitney test for comparing two unmatched groups, and Wilcoxon signed rank test for paired analysis of control vs. DMM limbs). All other analysis was performed on normally distributed data using t test or ANOVA with Tukey's post hoc analysis.