14-3-3 epsilon is an intracellular component of TNFR2 receptor complex and its activation protects against osteoarthritis

Abstract

Objectives:

Osteoarthritis (OA) is the most common joint disease; however, the indeterminate nature of mechanisms by which OA develops has restrained advancement of therapeutic targets. TNF signaling has been implicated in the pathogenesis of OA. TNFR1 primarily mediates inflammation, whereas emerging evidences demonstrate that TNFR2 plays an anti-inflammatory and protective role in several diseases and conditions. This study aims to decipher TNFR2 signaling in chondrocytes and OA.

Methods:

Biochemical co-purification and proteomics screen were performed to isolate the intracellular co-factors of TNFR2 complex. Bulk and single cell RNA-seq were employed to determine 14-3-3 epsilon (14-3-3ε) expression in human normal and OA cartilage. Transcription factor activity screen was used to isolate the transcription factors downstream of TNFR2/14-3-3ε. Various cell-based assays and genetically-modified mice with naturally-occurring and surgically-induced OA were performed to examine the importance of this pathway in chondrocytes and OA.

Results:

Signaling molecule 14-3-3ε was identified as an intracellular component of TNFR2 complexes in chondrocytes in response to progranulin (PGRN), a growth factor known to protect against OA primarily through activating TNFR2. 14-3-3ε was downregulated in OA and its deficiency deteriorated OA. 14-3-3ε was required for PGRN regulation of chondrocyte metabolism. In addition, both global and chondrocyte-specific deletion of 14-3-3ε largely abolished PGRN’s therapeutic effects against OA. Furthermore, PGRN/TNFR2/14-3-3ε signaled through activating ERK dependent Elk-1 while suppressing NF-κB in chondrocytes.

Conclusions:

This study identifies 14-3-3ε as an inducible component of TNFR2 receptor complex in response to PGRN in chondrocytes and presents a previously-unrecognized TNFR2 pathway in the pathogenesis of OA.

Introduction

Osteoarthritis (OA) is the most common cause of chronic disability and its prevalence is continuously increasing¹. Despite the high prevalence and morbidity of OA, effective disease modifying treatments capable of intervening this degradative cascade are not currently available, and the molecular mechanisms involved in OA’s initiation and progression remain poorly understood². Although it is unclear whether the primary cause of OA is cartilage damage, OA chondrocytes undergo a series of complex changes in the disease progression, impacting proliferation, catabolism, and ultimately death³. Chondrocytes themselves are major protagonists in this regulatory cascade - not just the target of external biomechanical and biochemical stimuli, but are themselves the source of proteases, cytokines, and inflammatory mediators that promote the deterioration of articular cartilage^4,5.

Accumulating evidences indicate that OA is a low-grade chronic inflammatory disease^6,7 and inflammation is thought to play a critical role in the pathogenesis of OA. TNFα signaling has received great attention due to its position at the apex of the pro-inflammatory cytokine cascade and its dominance in the pathogenesis of various diseases, including arthritis⁸. TNFα is one of the major pro-inflammatory cytokines detected in synovial fluid and is a widely studied regulator of catabolic processes in chondrocytes⁹. TNFα signals through two specific TNF receptors, TNFR1 and TNFR2. TNFR1 is ubiquitously expressed by nearly all cell types, and appears to be the dominant receptor responsible for mediating TNFα’s inflammatory activity and has been extensively studied^9,10. Conversely, TNFR2 exhibits a restricted expression, and knowledge concerning TNFR2 signaling remains largely unclear. Our global genetic screen led to the identification of TNFR2 as the high-affinity binding receptor of progranulin (PGRN)¹¹, a multifaceted growth factor known to regulate chondrocyte homeostasis and its deficiency causes susceptibility to OA^12–15. In contrast to TNFα, which demonstrates higher affinity for TNFR1 than TNFR2, PGRN exhibits 600 fold higher binding affinity to TNFR2 than TNFα¹¹.

Emerging evidences indicate that distinct from TNFR1, TNFR2 signaling plays anti-inflammatory and protective roles in several diseases and conditions, including neurodegenerative and cardiac diseases^16–18. TNFR2 was also reported to inhibit inflammation and prevent bone loss in inflammatory arthritis^19–21. Although we previously reported that PGRN’s protection against OA mainly depended on TNFR2¹², whether and how TNFR2 signaling is involved in chondrocyte metabolism and OA remain largely unknown. In this study, we took advantage of the knowledge gained through previous studies from several laboratories including ours, and performed several unbiased screens, including biochemical co-purification and proteomics screens, bulk RNA-seq analysis, single cell transcriptomic analysis, transcription factor activity screen, combined with various genetically-modified chondrocytes and mouse models, which led to the identification of the signaling molecule 14-3-3 epsilon (14-3-3ε) and transcription factor Elk-1 as essential components of TNFR2 signaling to mediate PGRN’s chondroprotective and therapeutic activities against OA.

Results

14-3-3ε is an intracellular component of TNFR2 receptor complex in response to PGRN in chondrocytes

Our previous findings that PGRN binds to TNFR2 with high affinity¹¹ and protects chondrocytes against OA¹² prompted us to identify additional components of the TNFR2 receptor complex in response to PGRN treatment. For this purpose, the intracellular domain (ICD) of TNFR2 was cloned into the PGEX-3X vector to express a fusion of GST to TNFR2ICD. As illustrated in Fig. 1a, GST (serving as a control) or GST-TNFR2ICD was affinity-purified on glutathione-agarose beads and used as a bait to trap proteins from PGRN-treated human chondrocytes. These samples were then analyzed by mass spectrometry and MS/MS spectra were searched against the Uniprot database. After subtracting the hits that were also trapped by the GST column, eight proteins were found to specifically bind to TNFR2 (Fig. 1a). Identification of TRAF1 and TRAF2, two known TNFR2-binding proteins, among the eight hits validated the technique. The protein ranking first was 14-3-3ε, a critical intracellular signaling mediator that belongs to 14-3-3 family^22–24.

Fig. 1. 14-3-3ε is an intracellular component of TNFR2 complex in chondrocyte and down-regulated in OA cartilage.

(a) Schematic of the experimental design to identify potential molecules recruited to TNFR2 intracellular domain (ICD) in response to PGRN stimulation. Summary of the hits that were specifically recruited to activated TNFR2 complexes in human C28I2 chondrocytes. (b) Chondrocytes isolated from WT or 14-3-3ε^Agc1 mice were treated with 10ng/ml TNFα or/and 200ng/ml PGRN for 30 min, then immunoprecipitated with 14-3-3ε or TNFR2 antibodies, and detection of TNFR2 and 14-3-3ε by immunoblotting. Results shown are representative of 3 biological replicates. (c) Volcano plots for gene expression of human OA (n = 4) versus normal (n=3) cartilage. Genes in red (up-regulated in OA) and blue (down-regulated in OA) have Benjamini–Hochberg adjusted P < 0.05. (d, e) Relative mRNA expressions of GRN (gene encoding PGRN) (d) and YWHAE (gene encoding 14-3-3ε) (e) in human OA versus normal cartilage by RNA-seq. (f) Unbiased clustering of scRNA-seq data from human nonarthritic (n = 3) and OA (n = 4) revealed 7 distinct cell clusters. (g-j) Expression of COMP (g), GRN (h), YWHAE (i) and TNFRSF1B (gene encoding TNFR2) (j) across the cell clusters. Each dot represents a single cell and colors correspond to the expression level of a gene in each cell. (k) qRT-PCR analysis of 14-3-3ε in human OA (n = 22) and normal (n = 21) cartilage. (l, m) Immunohistochemical staining of 14-3-3ε and quantification of 14-3-3ε positive cells in joint section collected from WT mice subjected to sham or DMM surgery (n = 8 mice per group). Scale bar, 50μm. (n) Relative 14-3-3ε mRNA level in cartilage isolated from sham or DMM operated mice (n = 8 mice per group). Data are mean ± SD.

To characterize the role of 14-3-3ε in chondrocytes, we generated inducible chondrocyte specific 14-3-3ε deficient mice (hereafter referred to as 14-3-3ε^Agc1) by crossing 14-3-3ε^f/f mice²⁵ with Agc1-Cre^ERT2 mice²⁶ in which Cre-mediated recombination is induced by tamoxifen (Supplementary Fig. 1). 14-3-3ε^Agc1 mice were born in a Mendelian ratio and tamoxifen treatment of 14-3-3ε^Agc1 mice resulted in no overt phenotype. Protein and gene analysis indicated the tamoxifen treatment could specifically delete 14-3-3ε in cartilage in 14-3-3ε^Agc1 mice (Fig. 1b, Supplementary 1b–d). Co-immunoprecipitation using both anti-14-3-3ε and anti-TNFR2 antibodies with the lysate of chondrocytes isolated from tamoxifen-treated 14-3-3ε^f/f and 14-3-3ε^Agc1 mice was performed to validate the interaction between TNFR2 and 14-3-3ε. The results revealed that TNFR2 was specifically detectable in the immunoprecipitated complex from 14-3-3ε^f/f but not 14-3-3ε^Agc1 chondrocytes in response to PGRN treatment, or PGRN plus TNFα treatment, but not to TNFα treatment alone. These results, together with our biochemical co-purification/MS results, indicate that 14-3-3ε was specifically recruited to TNFR2 following PGRN treatment in chondrocytes (Fig. 1b).

14-3-3ε is down-regulated in OA cartilage

In a separate effort to isolate OA-associated genes, we performed bulk RNA-seq analysis using cartilage isolated from normal nonarthritic and OA patients. Total RNA was isolated from 3 samples of nonarthritic cartilage and 4 samples of OA cartilage (Kellgren-Lawrence Grade 3 or 4). Genes (900 total, 600 up-regulated, 300 down-regulated) were differentially expressed in OA vs. normal (fold change > 2, FDR < 0.00001, adjusted p < 0.05) (Fig. 1c, Supplementary Fig 2a). Gene set enrichment analysis (GSEA) indicated altered gene expression pattern in OA cartilage compared with nonarthritic cartilage. Specifically, pathways known to be implicated in OA pathogenesis, including inflammatory response, interferon alpha and gamma response, apoptosis, and oxidative phosphorylation, were up-regulated in OA cartilage compared with nonarthritic cartilage (Supplementary Fig. 2b–f). Analysis of the genes differentially regulated between OA and nonarthritic cartilage, with a special interest in PGRN and 14-3-3ε, revealed that GRN (gene encoding PGRN) expression is significantly up-regulated in OA cartilage (Fig. 1d), which was in line with previous reports¹³, and intriguingly, YWHAE (gene encoding 14-3-3ε) exhibited a trend of reduced expression in OA cartilage (Fig. 1e).

To unravel the relative abundance and distribution of PGRN, TNFR2, and 14-3-3ε mRNA transcripts in different chondrocyte subpopulations, we performed single-cell RNA-seq (scRNA-seq) in chondrocytes isolated from human OA and normal cartilage. Similar to a recent study²⁷, unbiased clustering based on known cell specific markers identified 7 distinct cell clusters, including fibrocartilage chondrocytes (FCs), homeostatic chondrocytes (HomCs), prehypertrophic chondrocytes (preHTCs), hypertrophic chondrocytes (HTCs), proliferative chondrocytes (ProCs), effector chondrocytes (ECs), and regulatory chondrocytes (RegCs) (Fig. 1f). As expected, almost all chondrocyte clusters expressed cartilage oligomeric matrix protein (COMP), a known cartilage marker (Fig. 1g)²⁸. Both GRN and YWHAE appeared to be abundant across all cell clusters, whereas TNFRSF1B (gene encoding TNFR2) exhibited a much more restricted expression pattern, mainly in preHTCs (Fig. 1h–j, Supplementary Fig. 3). Thus, TNFR2 appeared to be the rate-limiting component in PGRN/TNFR2/14-3-3ε complex implicated in regulation of chondrocyte metabolism. Furthermore, independent validation by quantitative PCR (qPCR) revealed significantly decreased 14-3-3ε expression in human OA cartilage compared with nonarthritic cartilage (Fig. 1k). In line with the decrease at mRNA level, 14-3-3ε protein level was also reduced in human arthritic cartilage as compared to nonarthritic controls (Supplementary Fig. 4).

We next examined the expression of 14-3-3ε in the course of OA using the surgically induced destabilization of the medial meniscus (DMM) OA model in mice, and found that similar to observations with human OA cartilage, the levels of 14-3-3ε protein and mRNA were also reduced in the course of OA (Fig. 1l–n).

Aged 14-3-3ε deficient mice exhibit severer OA-like phenotype

Considering that 14-3-3ε was isolated as inducible component of TNFR receptor complex in response to PGRN treatment and its levels were downregulated in OA, we therefore explored the potential contribution of 14-3-3ε to the development of naturally occurring OA in aging mice. For this purpose, we generated inducible global 14-3-3ε knockout mice (hereafter referred to as 14-3-3ε−/−) by breeding 14-3-3ε^f/f mice with Rosa26-Cre^ERT2 mice in which Cre-mediated recombination is induced by tamoxifen (Supplementary Fig. 5a–c). PCR was implemented to confirm 14-3-3ε deletion efficiency in various tissues following tamoxifen administration in adult mice (Supplementary Fig. 5d). Thereafter, spontaneous changes in the histologic features of the articular cartilage were analyzed in 14-3-3ε−/− and 14-3-3ε^f/f littermates at ages 3 and 18 months.

As expected, there was no 14-3-3ε expression in the cartilage of 14-3-3ε−/− mice at either age (Fig. 2a, Supplementary Fig. 6a). Histological evaluations of H&E, Safranin O, and Movat pentachrome staining revealed that cartilage of 3-month-old 14-3-3ε−/− mice is indistinguishable from that of 14-3-3ε^f/f littermates (Supplementary Fig. 6a); 14-3-3ε^f/f and 14-3-3ε−/− cartilage show comparable cartilage features, including proteoglycan content, cartilage thickness and subchondral bone plate thickness at 3 months of age (Supplementary Fig. 6b–e). In general, histological staining of 18 months old 14-3-3ε^f/f mice displayed characteristic OA changes in joints, including proteoglycan loss, thinning of articular cartilage, thickening of the subchondral bone, and osteophyte formation (Fig. 2a–d)^29,30. Moreover, at 18 months of age, 14-3-3ε−/− cartilage exhibited a more severe OA phenotype, illustrated by a significantly greater degree of proteoglycan loss and reduction of articular cartilage thickness relative to 14-3-3ε^f/f littermates (Fig. 2a–d). Consistent with histological analysis, microCT analysis of undecalcified joint samples indicated that aging 14-3-3ε−/− mice have more osteophyte formation and severer subchondral bone sclerosis than 14-3-3ε^f/f littermates (Fig. 2e–g). Compared with 14-3-3ε^f/f cartilage, 14-3-3ε−/− cartilage appeared to have significantly increased levels of aggrecan neoepitope, COMP fragments, ColX and MMP13, indicators of cartilage degradation and degeneration (Fig. 2h). Consistent with the observations of immunohistochemistry staining, the transcript levels of catabolic markers, matrix metalloproteinase (Mmp13) and a disintegrin and metalloproteinases with thrombospondin type 5 motif (Adamts5), and inflammatory response markers, cyclooxygenase-2 (Cox-2) and inducible nitric oxide synthase (Nos2)³¹ were also significantly elevated in 14-3-3ε−/− cartilage relative to 14-3-3ε^f/f cartilage (Fig. 2i–l). Collectively, 14-3-3ε deficiency mice exhibited exaggerated age-associated, naturally occurring OA phenotype, thereby suggesting that genetic deletion of 14-3-3ε might contribute to age-related OA-like phenotype. It is also noted that the expressions of 14-3-3ε was markedly lower in cartilage from 18-months old 14-3-3ε^f/f mice than those in 3-months old 14-3-3ε^f/f mice (Fig. 2a, Supplementary Fig. 6a), suggesting that 14-3-3ε may be also associated with an aging phenomenon in addition to OA.

Fig. 2. 14-3-3ε deletion exaggerates naturally occurring phenotype with age.

(a) Immunohistochemical staining for 14-3-3ε, Safranin O, Movat pentachrome, and H&E staining in knee joint section collected from 14-3-3ε^f/f (n = 3) and 14-3-3ε−/− (n = 4) mice at age 18 months. Scale bar, 100μm. Representative image is shown. (b, c) Scoring of proteoglycan loss and cartilage thickness in 14-3-3ε^f/f and 14-3-3ε−/− mice at age 18 months, respectively. (d) Quantitation of the composition of the articular cartilage in 14-3-3ε^f/f and 14-3-3ε−/− mice at 18 months based on Movat pentachrome staining (yellow: bone; blue: cartilage). (e) Micro-CT scan and three-dimensional reconstruction of the knee joint from 18 months old 14-3-3ε^f/f and 14-3-3ε−/− mice, the region marked in red is osteophyte. (f, g) Three-dimensional microCT images and quantification of thickness for the medial compartment of the tibial subchondral bone of 18 months old 14-3-3ε^f/f and 14-3-3ε−/− mice. (h) Representative image of immunohistochemical staining for Aggrecan neoepitope, COMP fragment, ColX and MMP13 in WT and 14-3-3ε−/− knee section at age 18 months. Scale bar, 50μm. (i-l) Mmp13, Adamts5, Cox2 and Nos2 mRNA levels in cartilage from WT (n = 3) and 14-3-3ε−/− (n = 4) at age 18 months. Data are mean ± SD, * P < 0.05 or ** P < 0.01.

14-3-3ε is required for PGRN regulation of chondrocyte metabolism

Following isolation of 14-3-3ε as an effector recruited to the TNFR2 complex by PGRN, we sought to determine whether 14-3-3ε is involved in the regulation of chondrocyte metabolism and whether it is also important for PGRN/TNFR2 mediated regulation of chondrocyte metabolism. First, we generated 14-3-3ε knockout C28I2 human chondrocytes by employing CRISPR-Cas9 technique (Fig. 3a, b). Deletion of 14-3-3ε markedly inhibited the expressions of anabolic markers type II collagen (Col2a1), aggrecan (Acan) and COMP (Fig. 3c), and significantly enhanced TNFα-induced expressions of Adamts5 and Mmp13 (Fig. 3d). More importantly, PGRN/TNFR mediated stimulation of chondrocyte anabolism and inhibition of TNFα-induced catabolic/inflammatory response, including Cox2 and Nos2, were abolished in 14-3-3ε knockout human chondrocytes (Fig. 3c, d). Similar results were also observed in 14-3-3ε−/− mouse primary chondrocytes as compared to chondrocytes isolated from 14-3-3ε^f/f littermates (Supplementary Fig. 7a–c). Furthermore, PGRN’s regulatory effects upon chondrocyte metabolism were also blunted in 14-3-3ε−/− chondrocytes as compared to 14-3-3ε^f/f chondrocytes (Supplementary Fig. 7a–c).

Fig. 3. 14-3-3ε is required for PGRN regulation of chondrocyte metabolism.

(a) Schematic for generating 14-3-3ε−/− human C28I2 chondrocytes using CRISPR/Cas9 technology. (b) Western blotting to confirm the loss of 14-3-3ε in 14-3-3ε knockout C28I2 cells. Cell lysates were examined by immunoblotting with 14-3-3ε antibody. (c) mRNA levels of Col2, Acan and COMP in control and 14-3-3ε knockout C28I2 cells treated with or without 200ng/ml PGRN for 24hrs, assayed by qRT-PCR analysis. (d) mRNA levels of Mmp13, Adamts5, Cox2 and Nos2 in control and 14-3-3ε knockout C28I2 cells treated with 10ng/ml TNFα in the absence or presence of 200ng/ml PGRN for 24 hrs, assayed by qRT-PCR analysis. (e) Expression of Flag-14-3-3ε in control and 14-3-3ε knockout C28I2 cells, assayed by western blot. (f) mRNA levels of Col2 and Acan in PGRN (200ng/ml) treated control or 14-3-3ε knockout C28I2 cells with or without re-expression of 14-3-3ε, assayed by qRT-PCR analysis. (g) Control and 14-3-3ε knockout C28I2 cells with or without re-expression of 14-3-3ε were treated with 10ng/ml TNFα in the absence or presence of 200ng/ml PGRN for 24 hrs. mRNA levels of Mmp13, Adamts5, Cox2 and Nos2 were measured by qRT-PCR. Data are mean ± SD; n = 4 biological replicates; * P < 0.05 or ** P < 0.01.

We previously developed a PGRN-derived engineered protein called Atsttrin, composed of three TNFR2-binding fragments of PGRN, which exhibited therapeutic effects in both inflammatory arthritis and OA^11,32. Similar to PGRN, Atsttrin enhanced anabolism and inhibited TNFα induced inflammatory catabolism in control C28I2 cells while these effects were compromised in 14-3-3ε knockout C28I2 cells (Supplementary Fig. 8a, b).

To further characterize the necessity of 14-3-3ε in PGRN/TNFR2 regulation of chondrocyte metabolism, Flag-tagged 14-3-3ε was re-expressed in 14-3-3ε knockout C28I2 human chondrocytes to determine whether re-expression of 14-3-3ε could functionally rescue the 14-3-3ε deficiency phenotype (Fig. 3e). Re-expression of 14-3-3ε in 14-3-3ε knockout C28I2 cells reversed the phenotype induced by 14-3-3ε deficiency, more importantly, it could also restore PGRN mediated regulation of chondrocytes in terms of enhanced anabolism and suppressed inflammatory cytokine-induced catabolism and inflammation (Fig. 3f, g). Collectively, these results indicated that 14-3-3ε exerts chondroprotective effects as an essential mediator of PGRN/TNFR2 signaling in regulating chondrocyte metabolism.

14-3-3ε is required for PGRN’s therapeutic effects against OA in vivo

Following isolation of 14-3-3ε as an essential molecule mediating PGRN’s effects on chondrocyte metabolism, we assessed whether 14-3-3ε was also critical for PGRN’s protective and therapeutic effects in OA. To this end, we established the surgically induced DMM model in 14-3-3ε^f/f and 14-3-3ε−/− mice, followed by intra-articular injection of PGRN three times per week for a total of 8 weeks starting from 4 weeks after surgery and OA phenotypes were analyzed with a variety of techniques, including morphometric analysis, immunohistochemistry staining, ELISA and pain analysis (Fig. 4a).

Fig. 4. Global deletion of 14-3-3ε regulates OA pathogenesis and largely abrogates PGRN’s therapeutic effects against OA.

(a) Schematic of the experimental outline. 14-3-3ε^f/f and Rosa26-ERT2;14-3-3ε^f/f (i.e. 14-3-3ε−/−) mice are injected with tamoxifen at 10-weeks old, and DMM operation is performed on 3months old mice. n = 8 mice per group. (b) Representative images of Safranin O/Fast green stained sections of knee joints from 14-3-3ε^f/f and 14-3-3ε−/− mice treated with or without PGRN for 8 weeks. Scale bar, 50μm. (c) Quantitative analysis of OARSI score, osteophyte development, and subchondral bone plate (SBP) thickness in different group of mice. (d) Representative images of immunohistochemical staining for MMP13, Aggrecan neoepitope, COMP fragment, and ColX in knee joint sections of 14-3-3ε^f/f and 14-3-3ε−/− mice treated with or without PGRN for 8 weeks. Scale bar, 50μm. (e) Serum COMP fragment levels in 14-3-3ε^f/f and 14-3-3ε−/− mice treated with or without PGRN for 8 weeks. (f) 2min travel distance and von Frey pain assay in DMM-operated WT and 14-3-3ε−/− mice treated with or without PGRN at the indicated time after surgery. Data are mean ± SD; ** P < 0.01.

Compared with 14-3-3ε^f/f mice, 14-3-3ε−/− mice exhibited statistically severer cartilage erosion following DMM surgery (Fig. 4b). In addition, DMM-operated 14-3-3ε−/− mice exhibited slightly, yet statistically significant, higher OARSI scores and thickening of the subchondral bone plate, two critical characteristics of OA, relative to 14-3-3ε^f/f littermates (Fig. 4b, c). PGRN treatment ameliorated surgically induced OA pathogenesis in 14-3-3ε^f/f mice as evidenced by significant reduction of articular cartilage destruction, along with substantial inhibition of osteophyte formation and thickening of subchondral bone plate (Fig. 4b, c). In addition, PGRN-mediated protection against OA in 14-3-3ε^f/f mice, including reduction of articular cartilage destruction, osteophyte formation and thickening of subchondral bone plate, was almost abolished in 14-3-3ε −/− mice with DMM (Fig. 4b, c).

It is appreciated that concurrency of up-regulation of matrix-degrading enzymes and accelerated matrix degradation promotes OA³³, accordingly we assessed whether the expression of relevant effector molecules contributed to PGRN’s regulation of OA and its dependence on 14-3-3ε. Immunohistochemistry staining demonstrated that 14-3-3ε deficiency correlated with up-regulated MMP13, aggrecan neoepitope, and COMP fragment. In addition, 14-3-3ε deficiency enhanced the expression of ColX, a marker for hypertrophic chondrocytes (Fig. 4d, Supplementary Fig. 9a). Conversely, PGRN treatment following DMM markedly reduced the levels of MMP13, aggrecan neoepitope, COMP fragment, and ColX in 14-3-3ε^f/f mice, and these PGRN-mediated effects were markedly attenuated by 14-3-3ε deletion (Fig. 4d, Supplementary Fig. 9a). 14-3-3ε deletion also engendered significant elevation of COMP protein fragments in sera, largely unresponsive to PGRN treatment, while serum levels of COMP fragments were meaningfully reduced in PGRN treated relative to PBS treated 14-3-3ε^f/f mice (Fig. 4e). Additionally, DMM-induced OA pain was significantly reduced in PGRN treated 14-3-3ε^f/f mice, but not 14-3-3ε−/− mice, although 14-3-3ε deletion does not further enhance DMM induced pain as reflected by statistical equivalence of pain scores from PBS treated mice irrespective of genotype (Fig. 4f). Collectively, these results reinforce 14-3-3ε’s standing as a critical mediator of the PGRN/TNFR2 pathway in regulation of cartilage homeostasis and protection against OA pathogenesis.

We also examined whether PGRN and 14-3-3ε are involved in macrophage polarization. Deletion of PGRN and 14-3-3ε altered transcriptome of bone marrow derived macrophage which was stimulated with pro-inflammatory LPS/IFNγ (polarized to M1) or anti-inflammatory IL-4 (polarized to M2) (Supplementary Fig. 10a, b). GSEA analysis demonstrated that both 14-3-3ε deficiency and PGRN deficiency significantly up-regulated inflammatory response in macrophages compared with WT macrophages (Supplementary Fig. 10c–f). As OA is considered a chronic inflammatory disease, we then asked whether macrophage polarization regulated by PGRN and 14-3-3ε also contributed to OA pathogenesis. Immunohistochemistry staining of F4/80, a marker of general macrophage, showed undistinguishable macrophage infiltration among the mice with different genetic backgrounds treated with or without PGRN (Supplementary Fig. 11a, b). Further phenotypic characterization of macrophage in synovium revealed a significant decrease of M1 macrophage (iNos positive) and increase of M2 macrophage (CD206 positive) in the synovium of WT mice treated with PGRN compared with PBS. 14-3-3ε deletion skewed the macrophage toward M1 phenotype compared with WT, and remarkably, PGRN induced reduction of M1 and enhancement of M2 in WT mice was largely abolished in 14-3-3ε−/− mice (Supplementary Fig. 11a, c, d). No obvious difference was observed in populations of synovial CD4+ T cells and mast cells, the immune cells also reported to be involved in OA^34,35, between 14-3-3ε^f/f and 14-3-3ε−/− mice with DMM with or without PGRN treatment, as detected by immunohistochemistry staining (Supplementary Fig. 12a–c). Although no obvious difference for synovial CD4+ T cells was observed, whether 14-3-3ε is important for PGRN regulation of T cell subpopulations, including regulatory T cells³⁶, warrants further investigations. Collectively, these results suggested that macrophage phenotypic polarization modulated by PGRN and 14-3-3ε may also contribute to the regulations of OA by anti-inflammatory PGRN/TNFR2/14-3-3ε signaling complex.

Deletion of 14-3-3ε in chondrocytes exaggerates surgically induced OA and counteracts PGRN regulation of cartilage homeostasis

Having determined that global 14-3-3ε deficiency exaggerated OA and blunted PGRN-mediated protection against OA, we next investigated the role of chondrocyte-specific 14-3-3ε in the pathogenesis of surgically-induced OA. We thus established the DMM model in 14-3-3ε^Agc1 mice and their littermate controls and Safranin O staining revealed that cartilage degeneration was substantially progressed in both 14-3-3ε^Agc1 and 14-3-3ε^f/f mice following DMM surgery. Deficiency of 14-3-3ε in chondrocytes exaggerated cartilage destruction, with higher OARSI score in 14-3-3ε^Agc1 mice as compared with 14-3-3ε^f/f controls at 4w, 8w and 12w after DMM surgery (Supplementary Fig. 13a, b). Accordingly, deficiency of 14-3-3ε associated with elevated serum levels of COMP fragments, a biomarker correlated with severity of cartilage degradation³⁷, following DMM surgery as compared to 14-3-3ε^f/f controls (Supplementary Fig. 13c).

To evaluate whether chondrocyte-specific 14-3-3ε was also important for PGRN regulation of cartilage homeostasis, we also established DMM model in 14-3-3ε^Agc1 mice and their littermate controls and compared PGRN’s therapeutic effects between genotypes at 12w following surgery (Fig. 5a). Similar to 14-3-3ε−/− mice, genetic ablation of 14-3-3ε in chondrocytes elicited a slightly severer OA phenotype, including severe cartilage erosion, increased osteophyte development and thickening of subchondral bone plate as compared with 14-3-3ε^f/f controls at 12 weeks post-DMM (Fig. 5b, c). PGRN treatment substantially attenuated the OA phenotype by inhibiting articular cartilage destruction, osteophyte development and thickening of subchondral bone plate in 14-3-3ε^f/f mice while cartilage specific 14-3-3ε deficiency dampened PGRN’s protective effect against OA pathologies with reduced efficacy in preserving cartilage integrity and no ameliorative impact upon osteophyte maturity and thickness of subchondral bone plate relative to that observed in 14-3-3ε^f/f mice (Fig. 5b, c). Collectively, these results indicated that chondrocyte-expressed 14-3-3ε was required for maintaining cartilage homeostasis. Complimentary immunohistochemistry staining demonstrated marked reduction of MMP13, aggrecan neoepitope, COMP fragment and ColX observed in PGRN treated 14-3-3ε^f/f mice which was largely absent in 14-3-3ε^Agc1 mice (Fig. 5d, Supplementary Fig. 9b). Likewise, PGRN-mediated reduction of COMP fragments in serum of 14-3-3ε^f/f mice was abolished in 14-3-3ε^Agc1 mice (Fig. 5e). PGRN-triggered substantial reduction in DMM-induced OA pain in 14-3-3ε^f/f mice was also abolished in 14-3-3ε^Agc1 mice (Fig. 5f). In sum, the loss of PGRN’s therapeutic efficacy against OA observed following global 14-3-3ε knockout was closely recapitulated following chondrocyte specific deletion of 14-3-3ε, thereby confirming that chondrocyte-expressed 14-3-3ε not only primarily contributed to, but is also required to mediate PGRN/TNFR2’s protection against OA.

Fig. 5. Chondrocyte specific deletion of 14-3-3ε attenuates PGRN mediated protection against experimental OA.

(a) Schematic of the experimental outline. 14-3-3ε^f/f and Agc1-ERT2; 14-3-3ε^f/f (14-3-3ε^Agc1) mice are injected with tamoxifen at 10 weeks old, and DMM operation is performed on 3 months old mice. n = 8 mice per group. (b) Representative images of Safranin O/Fast green stained sections of knee joints from 14-3-3ε^f/f and 14-3-3ε^Agc1 mice treated with or without PGRN for 8 weeks. Scale bar, 50μm. (c) Quantitative analysis of OARSI score, osteophyte development, and subchondral bone plate (SBP) thickness in different group of mice. (d) Representative images of immunohistochemical staining for MMP13, Aggrecan neoepitope, COMP fragment, and ColX in knee joint sections of 14-3-3ε^f/f and 14-3-3ε^Agc1 mice treated with or without PGRN for 8 weeks. Scale bar, 50μm. (e) Serum COMP fragment levels in 14-3-3ε^f/f and 14-3-3ε^Agc1 mice treated with or without PGRN for 8 weeks. (f) 2min travel distance and von Frey pain assay in DMM-operated 14-3-3ε^f/f and 14-3-3ε^Agc1 mice treated with or without PGRN at the indicated time after surgery. Data are mean ± SD; ** P < 0.01.

We also examined PGRN regulation on macrophage plasticity in 14-3-3ε^Agc1 mice. Compared with PBS treated WT and 14-3-3ε^Agc1 mice, PGRN did not change the total macrophages presented in the synovium (Supplementary Fig. 14a, b). However, PGRN treatment regulated macrophage plasticity. Specifically, PGRN inhibited pro-inflammatory M1 macrophage, and skewed macrophages towards anti-inflammatory M2 macrophage (Supplementary Fig. 14a, c, d) in both WT and chondrocyte-specific 14-3-3ε^Agc1 mice, but not in global 14-3-3ε−/− mice (Supplementary Fig. 11a–d), highlighting the notion that 14-3-3ε is a critical mediator of PGRN/TNFR2 signaling in both chondrocytes and macrophages.

PGRN/TNFR2/14-3-3ε regulates chondrocyte metabolism by activating Elk-1 transcription factor

To further elucidate the molecular mechanisms by which the PGRN/TNFR2/14-3-3ε receptor complex regulates chondrocyte metabolism and OA, we performed transcription factor array to identify the transcription factor(s) activated by PGRN/TNFR2/14-3-3ε receptor complex. Among the 45 transcription factors examined, Elk-1, NF-κB and Stat3 showed more than 2-fold changes in transcriptional activity following PGRN treatment in WT primary chondrocytes, whereas these regulatory changes were abrogated in both TNFR2−/− and 14-3-3ε−/− chondrocytes (Fig. 6a), highlighting these 3 transcription factors as potential mediators of PGRN’s regulation of chondrocyte metabolism in TNFR2- and 14-3-3ε-dependent manners. Among the 3 isolated transcription factors, Elk-1 is the only one for which activity is enhanced, while NF-κB and Stat3 were inhibited, by PGRN through TNFR2 and 14-3-3ε. The transcription factor Elk-1 is known to act downstream of ERK, and ERK activation induces phosphorylation of Elk-1, leading to transcriptional activation of target genes³⁸. In addition, ERK signaling is also known to be required for PGRN/TNFR2 regulation of chondrocyte anabolism¹², we thus focused on examining the functional dependence of PGRN/TNFR2/14-3-3ε induced anabolism on the Elk-1 transcription factor. Indeed, Elk-1 luciferase reporter gene was activated by PGRN in WT articular chondrocytes, but this PGRN-mediated activation of Elk1 was completely lost in TNFR2−/− and 14-3-3ε−/− articular chondrocytes (Fig. 6b). Moreover, PGRN induced the phosphorylation of ERK and Elk-1in WT articular chondrocytes, and these activations were also abrogated in both TNFR2−/− and 14-3-3ε−/− articular chondrocytes (Fig. 6c, d).

Fig. 6. Transcription factor Elk-1 is indispensable for the regulation of chondrocyte anabolism by PGRN/TNFR2/14-3-3ε.

(a) Transcriptional activities of 45 transcription factors are scanned using the transcription factor array. Primary articular chondrocytes isolated from WT, TNFR2−/−, and 14-3-3ε mice are transfected with the construct included in the kit for 48hrs, followed by treatment with 200ng/ml PGRN for another 24 hrs. (b) Elk-1 transcription activity in primary articular chondrocytes isolated from WT, TNFR2−/−, and 14-3-3ε−/− mice treated with PGRN for 24hrs. n = 6 for each group. (c) Immunoblotting of indicated protein in WT, TNFR2−/− and 14-3-3ε−/− primary articular chondrocytes treated with PGRN for different time points, as indicated. n = 4 for each group. (d) Densitometry analysis of immunoblotting results shown in (c). (e) Immunoblotting of indicated protein in control and 14-3-3ε knockout human C28I2 cells transfected with Flag-14-3-3ε construct prior to treatment with PGRN for indicated time. n = 4 for each group. (f) Densitometry analysis of immunoblotting results shown in (e). (g) Elk-1 transcription activity analysis in human chondrocytes treated with 10μM U0126 for 1h prior to treatment with 200ng/ml PGRN for 24hrs. n =6 for each group. (h) Immunoblotting of indicated protein in human chondrocytes treated with 10μM U0126 for 1h prior to treatment with 200ng/ml PGRN for different time points, as indicated. n = 4 for each group. (i) Densitometry analysis of immunoblotting results shown in (h). (j) mRNA levels of indicated molecules in human chondrocytes treated with 10μM U0126 for 1h prior to treatment with 200ng/ml PGRN for 24hrs. n = 4 for each group. (k) A proposed model depicting the signaling pathway by which PGRN (its derivative Atsttrin as well) binds to TNFR2 and recruits 14-3-3ε to the receptor complex, leading to the activation of chondrocyte anabolism and protection against OA. Data are mean ± SD; ** P < 0.01.

We next re-expressed 14-3-3ε in 14-3-3ε knockout human C28I2 chondrocytes. Re-expression of 14-3-3ε efficiently restored PGRN-induced activations of ERK and Elk-1 (Fig. 6e, f), further indicating that 14-3-3ε represented an essential component in the PGRN/TNFR2 signaling cascade. Both pharmacological inhibition of ERK and siRNA knockdown of Elk-1 in human C28I2 chondrocytes markedly inhibited PGRN-activated Elk-1 transcriptional activity (Fig. 6g, Supplementary Fig. 15a). Accordingly, pharmacological inhibition of ERK significantly inhibited activation of ERK and Elk-1 by PGRN (Fig, 6h, i). In addition, siRNA knockdown of Elk-1 markedly reduced Elk-1 expression level and Elk-1 activation (Supplementary Fig. 15b, c). Notably, PGRN induced expressions of anabolic markers, including Col2, Acan and COMP, were abolished by U0126 and siRNA knockdown of Elk-1 (Fig. 6j, Supplementary Fig. 15d). These results confirmed that Elk-1 transcriptional activity was required for PGRN/TNFR2/14-3-3ε regulation of chondrocyte anabolism.

In addition to activating chondrocyte anabolism, PGRN also exerts anti-catabolic function in chondrocytes by down-regulating matrix-degrading enzymes and inflammatory response markers¹². Therefore, we utilized the same strategy to isolate the downstream transcription factor(s) implicated in mediating PGRN’s anti-catabolic activity. For this purpose, we treated chondrocytes with TNFα in the absence and presence of PGRN and performed transcription factor array. Among 45 transcription factors, PGRN inhibited TNFα-activated transcription factors NF-κB, STAT1 and STAT1/STAT2 in WT articular chondrocytes, and this inhibition was abolished in 14-3-3ε−/− chondrocytes (Supplementary Fig. 16a). TNFα displayed the most potent activation of the NF-κB transcription factor, and PGRN is known to inhibit TNFα-mediated activation of NF-κB in inflammatory arthritis¹¹. We thus selected NF-κB for functional validation in PGRN-mediated anti-catabolism and its dependence on 14-3-3ε. PGRN significantly inhibited TNFα-induced NF-κB phosphorylation and transcriptional activity, an effect that was lost in 14-3-3ε knockout human chondrocytes (Supplementary Fig. 16b–d). The selective Ikk-2 inhibitor SC-514³⁹ significantly inhibited TNFα-activated NF-κB phosphorylation, to a comparable extent as PGRN, in control human chondrocytes (Supplementary Fig. 16c). Despite SC-514’s effective inhibition of TNFα-activated NF-κB phosphorylation and expression of Mmp13 and Adamts5, PGRN lost these inhibitions in 14-3-3ε knockout human chondrocytes (Supplementary Fig. 16d–f). These results indicated that PGRN inhibited TNFα-activated NF-κB in a non-canonical 14-3-3ε-dependent anti-catabolic pathway in chondrocytes (Supplementary Fig. 16g).

Discussion

TNFR2 signaling plays a protective and anti-inflammatory role in joint destruction^19,20, and activation of TNFR2 by PGRN has been shown to protect against OA¹². In this study, combined use of biochemical co-purification and mass spectrometry led to the isolation of 14-3-3ε, an important intracellular signaling molecule, as a novel component recruited to TNFR2 complex in response to PGRN stimulation in human chondrocytes. By using multiple techniques including RNA-seq, single-cell transcriptomics, in vitro validations in mouse and human chondrocytes, alongside in vivo and ex vivo assessments of both spontaneous, age-related and surgically induced OA in genetically modified mice, we gain critical insights supporting the conclusion that 14-3-3ε is an essential mediator for the activation of the protective TNFR2 signaling by PGRN in OA.

We previously reported that both PGRN and its derivative Atsttrin promote chondrocyte anabolism through activating ERK^12,32. Herein, transcription factor array and in vitro validation isolated Elk-1 as the critical transcription factor in PGRN/TNFR2/14-3-3ε signaling and its ERK dependent activation is required for PGRN/TNFR2/14-3-3ε regulation of chondrocyte anabolism. Elk-1 is a transcription factor involved in various biological processes, such as cell growth, differentiation and survival, wound healing, and inflammation⁴⁰. Activation of Elk-1 has been shown to attenuate oxidative and apoptotic response in human chondrocytes⁴¹. Both pharmacological and siRNA knockdown of Elk-1 abrogated PGRN induced anabolism (Fig. 6). Future studies will lead to better understating of Elk-1 directed gene expression via PGRN/TNFR2/14-3-3ε/ERK signaling pathway in the context of OA. Nonetheless, Elk-1 appears to be indispensable for the regulation of chondrocyte anabolism by the PGRN/TNFR2/14-3-3ε receptor complex (Fig. 6k).

Besides activating chondrocyte anabolism, PGRN/TNFR/14-3-3ε signaling could also inhibit chondrocyte catabolism. PGRN’s promotion of anabolism and inhibition of catabolism rely on 14-3-3ε, and 14-3-3ε deficiency activates a catabolic cascade by up-regulating matrix-degrading enzymes, Mmp13 and Adamts5⁴². NF-κB and Stat3 are found to be inhibited by PGRN in a TNFR2- and 14-3-3ε-dependent manner in our transcription factor array. NF-κB and Stat3 activation by pro-inflammatory cytokines are shown to stimulate chondrocyte catabolism^9,43,44, thus inhibition of these two transcription factors by PGRN/TNFR2/14-3-3ε signaling contributes to PGRN’s anti-catabolic actions in chondrocytes. PGRN exhibits higher binding affinity to TNFR2 than does TNFα, and comparable binding affinity to TNFR1 and TNFR2¹¹. Interestingly, 14-3-3ε could also be recruited to TNFR1 upon stimulation by PGRN (data not shown) albeit TNFR1 and TNFR2 mediate distinct signaling pathways⁴⁵, suggesting that 14-3-3ε may act as a signaling switch of TNFRs in response to PGRN and TNFα stimulation. In addition, PGRN inhibited TNFα-activated NF-κB in a non-canonical 14-3-3ε-dependent manner in chondrocytes. Taken together, PGRN and its derivative Atsttrin exert their therapeutic and protective effects in OA through dual mechanisms: a) primarily activating PGRN/TNFR2/14-3-3ε/Elk-1 anabolic pathway independent of TNFα (Fig. 6k) and b) competing with TNFα to bind to TNFR1, thus simultaneously triggering PGRN/TNFR1/14-3-3ε/NF-κB anti-catabolic signaling (Supplementary Fig. 16g).

Emerging evidences demonstrate that accumulations of immune cells, particularly activated macrophages in the synovium of joints, also affect OA progression^46–50. Coincident with its role to mediate chondroprotective effects of PGRN/TNFR2 in chondrocytes, 14-3-3ε was also found to be required for PGRN regulation of macrophage polarization in the course of OA, which may also explain the more prominent blockade of PGRN effects observed in global 14-3-3ε deficient mice than seen in chondrocyte-specific 14-3-3ε deficient mice. In brief, roles of PGRN/TNFR/14-3-3-ε in regulating chondrocyte metabolism and macrophage polarization are all expected to contribute to the protective role of PGRN in the context of OA.

OA is a degenerative disease affecting the whole joints, including articular cartilage, subchondral bone, and synovium^51,52. In addition to deteriorating articular cartilage destruction, both global and chondrocyte-specific 14-3-3ε deletion caused more severe subchondral bone sclerosis, whereas activation of 14-3-3ε by PGRN through TNFR2 inhibited articular cartilage destruction, osteophyte formation and subchondral bone sclerosis. Although it is unclear how these events interact with each other and which event first occurs to initiate OA, the results provide genetic evidences that cartilage destruction, subchondral sclerosis and osteophyte development are highly correlated and targeted by PGRN/TNFR2/14-3-3ε signaling.

Pain is the common symptom of OA and a complex process involving structural changes in joint tissues, neuronal mechanisms and alterations of pain processing⁵³. Our results demonstrated that PGRN treatment could alleviate OA pain in a 14-3-3ε-dependant manner, albeit much remains to be learned about how PGRN/TNFR2/14-3-3ε signaling contributed to control OA pain. We previously reported that PGRN derived Atsttrin exhibited potent anti-inflammatory effects in several preclinical animal models of inflammatory arthritis, surpassing that of PGRN¹¹. Moreover, results from several laboratories, including ours, demonstrate that Atsttrin signals through TNFRs and protects against OA in both mouse and rat OA models^32,54. The current finding that Atsttrin’s regulation of chondrocyte metabolism also relies on 14-3-3ε, further supports a strong case for testing this reagent in a clinical trial.

Both global and chondrocyte specific 14-3-3ε mice demonstrate that 14-3-3ε mediates chondroprotective and anti-inflammatory effects of PGRN in OA. Consistent with these findings, deletion of 14-3-3ε favors OA development in vivo in both naturally occurring with age and surgically induced OA. Intriguingly, extracellular 14-3-3ε secreted by osteoblasts/osteocytes was reported to induce the release of catabolic factors by chondrocytes⁵⁵. This paradoxical controversy suggests that intracellular and extracellular 14-3-3ε might exert distinct effects on chondrocyte metabolism and may have different roles in the pathogenesis of OA.

In sum, this study reports discovery of intracellular 14-3-3ε as a crucial component of TNFR2 receptor complex in chondrocytes and OA, and establishes a novel TNFR2 signaling paradigm to orchestrate chondrocyte anabolism and combat the inflammatory/catabolism via PGRN/TNFR2/14-3-3ε/Elk-1 anabolic and PGRN/TNFR/14-3-3ε/NF-κB anti-catabolic cascade, respectively, thereby protecting against OA. The chondroprotective effects of PGRN on OA support the concept that targeted activation of TNFR2 signaling by PGRN, particularly its derivative Atsttrin, would be an effective therapeutic candidate for treating OA.

Key messages.

What is already known about this subject?

• TNFR2 was reported to inhibit inflammation and prevent bone loss in inflammatory arthritis. Whether and how TNFR2 signaling is involved in chondrocyte metabolism and OA remain largely unknown.

What does this study add?

• This study identifies the intracellular signaling molecule 14-3-3ε as a novel component of the TNFR2 receptor complex, and uncovers a new strategy for activating this key pathway of anti-inflammation in OA and other related diseases. This study also identifies Elk1 as a previously-unrecognized transcription factor which is required for TNFR2 anabolic signaling in chondrocytes.

• This study establishes a novel TNFR2 signaling paradigm to orchestrate chondrocyte anabolism and combat the inflammation/catabolism via PGRN/TNFR2/14-3-3ε/Elk-1 anabolic and PGRN/TNFR/14-3-3ε/NF-κB anti-catabolic cascade, respectively, thereby protecting against OA.

• This study advances our understanding of TNFR2 signaling pathway in chondrocytes and OA. In addition, the results of this study will also have broader application to the understanding of cartilage hemostasis and musculoskeletal degenerative diseases in general.

How might this impact on clinical practice or future developments?

• The chondroprotective effects of PGRN on OA support the concept that targeted activation of TNFR2 signaling by PGRN, particularly its derivative Atsttrin, would be an effective therapeutic candidate for treating OA.

Data availability statement:

Bulk and single-cell RNA-seq data that support the findings of this study have been deposited in Gene Expression Omnibus (GEO) with the accession codes GSE168505 and GSE169454. All the data relevant to the study are included in the article or uploaded as supplementary information.