NOS inhibition reverses TLR2-induced chondrocyte dysfunction and attenuates age-related osteoarthritis

TLR stimulation has been shown to induce catabolic reactions of human chondrocytes (15–18), whereas knockout mice of TLR family genes yielded ambiguous results in experimental OA models (19, 20). To date very few studies have investigated the expression of only select TLR family members in human chondrocytes by using a histological approach (21–23). However, the interpretation of the data has remained difficult, and there is a lack of completeness. To gain a global comprehension of the expression of TLR molecules in human chondrocytes, we used a systematic approach to assess the mRNA expression of TLR family members and their signaling adaptors MyD88 and TRIF in human OA cartilage ex vivo. We isolated RNA directly from fresh cartilage tissue obtained from OA patients undergoing knee arthroplasty and performed RNA-sequencing (RNA-seq) analysis. We detected the expression of TLR1, 2, 3, 4, 5, 6 and to a lower degree of TLR7, 8, 9, and 10 as well as expression of the TLR signaling adaptor molecules MYD88 and TRIF (Fig. 1A). This observation is supported by a publicly available RNA-seq data set of knee chondrocytes from healthy and OA-diseased donors (GSE114007) (24), in which we found comparable expression of various TLR genes already in the healthy state as much as in the OA samples, indicating that chondrocytes express TLR genes prior to OA onset (SI Appendix, Fig. S1A). We further confirmed the expression of the TLR-encoding genes by TaqMan quantitative PCR (qPCR) (Fig. 1B). Moreover, by flow-cytometric analysis of freshly isolated human chondrocytes, we detected the expression of TLR1, 2, 4, 5, 6, and 9 at protein level (SI Appendix, Fig. S1 B and C). In addition, we confirmed TLR2 mRNA expression (Fig. 1C) and TLR2 protein expression (Fig. 1D) in situ in chondrocytes of freshly obtained cartilage tissue using RNAScope hybridization and immunofluorescence analysis, respectively. Taken together, primary human chondrocytes express the genes and produce the proteins of many TLR family members ex vivo and thus are likely to respond to TLR stimulation.

To evaluate the effects of TLR stimulation on primary human OA chondrocytes, we first set up a three-dimensional spheroid culture system to maintain optimal chondrogenesis (25, 26). In addition, we maintained the oxygen level at 4% to achieve a physioxia growth environment (27). To these cultures, we added agonists for TLR1/2, 3, 4, 5, 2/6, 7, 8, or 9 (for TLR10 a defined agonist is not yet available; SI Appendix, Fig. S2A). After 18 h, the expression of NFKB, a known target of most TLR signaling, was upregulated as assessed by qPCR (SI Appendix, Fig. S2B), confirming that human chondrocytes are responsive to TLR agonists. By the end of a 28-d culture, spheroid weights were determined. TLR1/2 and 2/6 stimulation suppressed the growth of chondrocyte spheroids most strongly, with a weight reduction of about 80%. The activation of TLR4 and 5 resulted in a less pronounced suppression, while addition of TLR3, 7, 8, and 9 agonists had no impact on spheroid growth (Fig. 2A). Alcian blue staining revealed the lowest accumulation of extracellular matrix (ECM) in TLR1/2- and 2/6-stimulated spheroids. The difference was less pronounced in TLR4- and 5-stimulated samples (Fig. 2B and SI Appendix, Fig. S2C). The reduction of ECM production elicited by TLR1/2 and TLR2/6 stimulation was further confirmed by reduced glycosaminoglycans (GAGs)/DNA ratios of each spheroid after a 28-d culture (SI Appendix, Fig. S2D). This pattern in the reduction of spheroid growth was reflected by the diminished expression of transcripts encoding the matrix proteins collagen 2A1 (COL2A1) and aggrecan (ACAN), as well as enhanced expression of the cartilage-degrading enzymes MMP3 and a disintegrin and metalloproteinase with thrombospondin motifs 5 (ADAMTS5) (Fig. 2 C–F). Beyond the previously reported IL-6 induction by TLR2 (16, 17), we found a broader effect of TLR stimulation on the inflammatory phenotype of chondrocytes: TLR1/2 and TLR2/6 stimulation induced secretion of inflammatory cytokines including macrophage inflammatory protein-1α (MIP-1α/CCL3), MIP-1ß (CCL4), TNFα, interferon-γ (IFNγ), and triggered high-level production of IL-6, IL-8, and granulocyte colony-stimulating factor (G-CSF) (Fig. 2 G−I and SI Appendix, Fig. S2E). Of note, in many aspects, TLR1/2 stimulation and TLR2/6 stimulation elicited very similar effects: 1) A similar inhibition on chondrocyte spheroid growth (Fig. 2 A and B and SI Appendix, Fig. S2 C and D); 2) a similar suppression of expression of the cartilage-anabolic factors COL2A1 and ACAN and a similar enhancement of the catabolic factors MMP3 and ADAMTS5 (Fig. 2 C-F); 3) a similar induction of the inflammatory cytokines IL-6, IL-8, G-CSF and other cytokine secretion (Fig. 2 G−I and SI Appendix, Fig. S2E). Thus, TLR2 signaling, which is elicited downstream of both TLR complexes with TLR2 participation, i.e., TLR1/2 and TLR2/6, strongly impairs ECM generation while simultaneously inducing a proinflammatory phenotype in human chondrocytes.

Considering these observations, we further investigated the function of TLR1/2 signaling in suppressing chondrocyte spheroid growth, starting with a kinetic analysis of a 4-wk culture, in which spheroids were sampled every 3.5 d. Stimulation with the TLR1/2 agonist Pam3CSK4 (P3C4) suppressed the growth of spheroids from early on, which was associated with decreased COL2A1 and increased MMP3 expression (SI Appendix, Fig. S3 A–C). P3C4 also tended to increase the secretion of the inflammatory cytokines IL-6, IL-8, and G-CSF throughout the stimulation period (SI Appendix, Fig. S3 D–F).

Since OA chondrocytes have been reported to feature reduced ATP production (12), we assessed whether TLR1/2 stimulation was sufficient to induce energy deficiency in human chondrocytes. Four days after stimulation with P3C4, spheroid cellular ATP production was significantly reduced compared to control spheroids (Fig. 3A). Comparable MitoSpy staining intensities between control and P3C4-stimulated spheroids indicated a similar total mitochondrial mass per cell (Fig. 3B). However, a significant reduction of the mitochondrial membrane potential was detected by TMRM staining (Fig. 3C). Mitochondrial membrane potential is a key indicator of mitochondrial activity because it reflects the process of electron transport and oxidative phosphorylation (OXPHOS), the driving forces behind mitochondrial ATP production (28). We therefore conducted a Seahorse Spheroid Mito Stress Test on day 4 of culture, when the spheroid weight and chondrocyte cell number, viability, and expansion rates were not yet affected by P3C4 stimulation (SI Appendix, Fig. S3 A and G–I). Similar matrix densities (SI Appendix, Fig. S3J) likely allowed comparable spheroid infiltration of oligomycin, FCCP, and rotenone/actinomycin A, which are used to quantify OXPHOS activity and glycolytic efficiency by assessing the oxygen consumption rate (OCR) and extracellular acidification rate (ECAR). P3C4 stimulation strongly diminished OXPHOS activity-as shown by the reduction of both basal and maximum OCR – and ATP-linked oxygen consumption (Fig. 3D). Basal and maximum glycolytic rates as well as glycolytic reserves were comparable, indicating that glycolysis was not affected (Fig. 3E). To confirm that the observed stimulatory effects of P3C4 were propagated by TLR2, we treated chondrocytes with a blocking antibody against human TLR2, prior to P3C4 stimulation. Indeed, TLR2 antibody pretreatment largely prevented the P3C4-mediated impairment of mitochondrial respiration (Fig. 3F), as well as the loss of spheroid matrix components (Fig. 3G). To investigate whether TLR1/2 stimulation also has an impact on mature chondrocyte spheroids, we first cultured chondrocyte spheroids for 4 wk to establish maturity, then added P3C4, and performed metabolic analyses 4 d later. TLR1/2 stimulation impaired the OXPHOS activity without affecting glycolysis (SI Appendix, Fig. S4A) and suppressed total cellular ATP production also in mature chondrocyte spheroids (SI Appendix, Fig. S4B). Furthermore, COL2A1 and ACAN expression decreased, while MMP3 and ADAMTS5 expression increased (SI Appendix, Fig. S4C) together with IL-6, IL-8, and G-CSF (SI Appendix, Fig. S4 D–G). Thus, P3C4 stimulation of mature chondrocyte spheroids for 4 d is sufficient to suppress ECM anabolic gene expression, enhance transcription of catabolic genes, and induce inflammatory cytokine secretion. In addition, P3C4 stimulation selectively impairs mitochondrial respiration but not glycolysis in both mature and just 4-d–old chondrocyte spheroids.

We have described above the detrimental effect of P3C4-mediated TLR1/2 stimulation on human chondrocytes. To investigate the OA-associated pathogenic impact of TLR1/2 signaling triggered by an endogenously occurring agonist, we stimulated chondrocyte spheroids with 32-mer, which is a peptide released from the matrix protein aggrecan after several enzymatic digestion steps and which can activate TLR2 (10). We observed that 32-mer tended to reduce the expression of COL2A1 and ACAN and increase the expression of MMP3 and ADAMTS5 (Fig. 4A) and of the inflammatory cytokines IL8 and GCSF (Fig. 4B). In addition, 32-mer significantly reduced the mitochondrial respiration capacity of chondrocytes (Fig. 4C). Overall, endogenous 32-mer stimulation yielded similar results as P3C4 stimulation, although the effect of 32-mer was less pronounced when compared to P3C4.

Increased mitochondrial reactive oxygen species (mROS) results in oxidation of mitochondrial proteins and thus mitochondrial dysfunction (29). Mitochondria-targeted antioxidants attenuated cholesterol-induced OA (30). In addition, it has been reported that mROS promote TLR-induced inflammatory cytokine production (31–34). To understand whether the impaired mitochondrial function and the enhanced inflammatory phenotype induced by TLR1/2 stimulation resulted from over-accumulation of mROS, we measured mROS using MitoSOX staining. In contrast to our hypothesis, this analysis showed mildly but significantly reduced mROS levels in P3C4-stimulated chondrocytes (Fig. 5A). Next, we determined total cellular ROS production via DCFDA staining and found a reduced total cellular ROS production upon TLR1/2 stimulation (Fig. 5B). Thus, neither mROS nor total cellular ROS appear to be the cause of TLR1/2-induced mitochondrial dysfunction and inflammatory cytokine secretion.

Energy deficiency has been reported to metabolically hinder type II collagen synthesis during murine growth plate development (35). To determine whether the reduced mitochondrial ATP production caused the perturbation of ECM homeostasis and the inflammatory cytokine production induced by P3C4 stimulation, we partially suppressed mitochondrial respiration by adding decreasing amounts of the complex I inhibitor rotenone. At concentrations of 0.8 to 20 nM, rotenone treatment reduced the chondrocytes’ ATP production to a similar extent as TLR1/2 stimulation without affecting cell survival (Fig. 5C and SI Appendix, Fig. S5A). Yet, rotenone influenced neither the expression of ECM-anabolic and -catabolic factors (Fig. 5 D–F), nor the expression and secretion of inflammatory cytokines (Fig. 5 G–I and SI Appendix, Fig. S5B). Thus, the reduction of ATP production is not sufficient to recapitulate the TLR1/2-induced cartilage-degrading and inflammatory phenotype in chondrocytes.

To identify molecular mediator(s) responsible for the TLR1/2-induced dysfunction of chondrocytes, we sequenced RNA isolated from 3.5-d–old P3C4-stimulated and unstimulated spheroids. We found that 1,462 genes (including MMP3 and ADAMTS5) were significantly upregulated and 1,419 genes (including COL2A1 and ACAN) were significantly downregulated (FDR < 0.05 and fold change > 1.3) in P3C4-stimulated spheroids (Fig. 6A). As expected, the upregulated genes were enriched in inflammatory pathways of the immune system (SI Appendix, Fig. S6A). Importantly, NOS2, together with the afore-detected inflammatory cytokines IL6, IL8, and GCSF were among the top 15 differentially upregulated genes that are also part of the Reactome gene set “cytokine signaling in the immune system” (R-HSA-1280215) (Fig. 6 A and B). Since nitric oxide (NO) has been shown to modulate mitochondrial respiration in macrophages and chondrocytes (36–39), we quantified the upregulation of NOS2 expression in the spheroids and the increase of NO production in the supernatants of P3C4-stimulated chondrocytes (Fig. 6C). This NO induction was partially prevented by preblocking TLR2 signaling via anti-TLR2 indicating that P3C4 increased NO production indeed through activating TLR1/2 (Fig. 6D). Moreover, the endogenous TLR2 activator 32-mer also increased the NO production of human chondrocytes (Fig. 6E). Taken together, TLR1/2 stimulation promotes the NO production capacity of human chondrocytes. Gene set enrichment analysis confirmed a significant positive enrichment of genes that represent a response to NO (Fig. 6F). Concurrent with impaired mitochondrial respiration and reduced matrix density following TLR1/2 stimulation, the downregulated genes were enriched for terms relevant to metabolic regulation, matrix organization, and translational regulation (SI Appendix, Fig. S6B). In addition, gene set enrichment analysis revealed a significant negative enrichment in gene sets related to a variety of metabolic processes (Fig. 6G), which provides molecular support for the observed mitochondrial dysfunction and energy deficiency induced by TLR1/2 stimulation. Interestingly, here we detected a downregulation of the NADPH oxidase 4 (NOX4), an important source of ROS (40), and an upregulation of the antioxidant thioredoxin reductase 1 (TXNRD1) (41) and the mitochondrial ROS-specific scavenger superoxide dismutase 2 (SOD2) (42) (SI Appendix, Fig. S6C). This may explain, at least partially, why TLR1/2 stimulation reduced ROS accumulation in chondrocytes (cf. Fig. 5 A and B).

TLR2 stimulation as well as TLR4 and 5 stimulation, increased NOS2 expression, enhanced NO production (SI Appendix, Fig. S7 A and B), and impaired mitochondrial respiration in chondrocytes (SI Appendix, Fig. S7 C and D). This was in line with the reduced expression of cartilage-anabolic mediators and increased expression of cartilage-catabolic and inflammatory factors along with the suppressed growth of chondrocyte spheroids (cf. Fig. 2). In contrast, stimulation with TLR3, 7, 8, or 9 neither affected NOS2 expression, NO production, nor mitochondrial function (SI Appendix, Fig. S7). To determine whether the increased production of NO was the cause of the impaired mitochondrial function and the enhanced inflammatory phenotype induced by TLR1/2 stimulation, we added the NOS inhibitor L-NAME to P3C4-stimulated chondrocyte spheroids. We observed a dose-dependent suppression of NO production, and NO levels comparable to controls were achieved at 10 µM L-NAME (Fig. 7A). Likewise, blockade of NO production profoundly rescued mitochondrial membrane potential (Fig. 7B), and restored the ATP production (Fig. 7C). Moreover, the addition of L-NAME restored the expression of 250 genes that were altered by P3C4 stimulation (SI Appendix, Fig. S8A). Among the restored genes, 42 genes functionally belong to the mitochondrial gene set (SI Appendix, Fig. S8B), suggesting that NO plays an important role in TLR1/2-mediated mitochondrial dysfunction. Functional enrichment analysis (GSEA) of the 26 mitochondroid genes that were downregulated by P3C4 and then restored by L-NAME addition revealed that these genes were involved in hypoxia response, the generation of precursor metabolites and energy, and ADP metabolic processes (Fig. 7 D and E and SI Appendix, Fig. S8B). Blockade of NO production dose-dependently restored OXPHOS activity, as both basal and maximal OCR values gradually recovered (Fig. 7F). Addition of the NOS inhibitor did not affect glycolysis or mROS accumulation (SI Appendix, Fig. S8 C and D).

Furthermore, NOS inhibition significantly reduced the expression of TLR1/2-induced inflammatory factors (NFKB, IL6, IL8, and GCSF), and ECM-catabolic factors (MMP3 and ADAMTS5) (Fig. 8A). However, it failed to rescue the expression of COL2A1 and ACAN (Fig. 8A). Nevertheless, L-NAME restored the growth of chondrocyte spheroids to a significant extent (Fig. 8B and SI Appendix, Fig. S9). Importantly, NOS depletion also showed cartilage-protective effects in vivo. When applying histopathological scorings on 2-y–old wildtype (WT) and Nos2^−/− mice, we observed that Nos2^−/− mice exhibited less signs of OA compared to WT animals, with reduced cartilage erosion (Fig. 8C) and synovitis severity (Fig. 8D). In line with previous reports on varying incidences of naturally occurring OA in mice (43, 44), about half of the WT mice developed a fully osteoarthritic phenotype, which did not occur in Nos2^−/− mice. Thus, 2-y–old Nos2^−/− mice were significantly less affected by age-related OA compared to WT mice despite similar body weight (Fig. 8E), indicating a role for NOS2 in age-related OA development.

Age-matched male C57BL/6 and Nos2^−/− mice on C57BL/6 background were maintained under specific-pathogen-free (SPF) conditions until they reached the age of 2 y. Mice were then humanely sacrificed, and hind limb knee joints were removed and processed for OA severity evaluation. For details see SI Appendix, Supplementary Material. All animal experiments were performed in accordance with the German law for animal protection with permission from the local veterinary offices.

In total, 134 patients (74 women and 60 men; mean age, 70.78 ± 12.01y), who had a clinical and biopsy-proven diagnosis of OA and gave their informed consent for the use of clinical data and samples for research purposes, participated in this study. The study was approved by the Ethics Committee of Charité—Universitätsmedizin Berlin (EA1/032/16) (Table 1).

Femoral condyles of OA patients were collected immediately after removal during knee arthroplasty in the Center for Musculoskeletal Surgery of Charité—Universitätsmedizin Berlin. Cartilage was separated from bone and digested by collagenase II to release chondrocytes. After a round of monolayer culture, chondrocytes were subjected to spheroid generation, hypoxia (4% O₂) cultivation, and TLR stimulation. For details see SI Appendix, Supplementary Material.

Chondrocyte spheroids were individually placed in the center of Agilent Seahorse XFe96 Spheroid Microplate wells. OCR and ECAR measurements were performed every 5 min prior to and after sequential addition of oligomycin, FCCP, or rotenone/actinomycin A. Data were analyzed using Wave (Agilent).

For more method details see SI Appendix, Supplementary Material.

Statistical analysis was performed using GraphPad Prism (v5.02 and v7). Data were first examined for normality. If normal distribution was found, significance was determined using paired or unpaired two-tailed t test for two-group comparisons and one-way ANOVA was used for multiple group comparisons. In case of non-normally distributed groups, comparisons were performed using nonparametric tests with corresponding corrections (paired t test: Wilcoxon correction; unpaired t test: Mann–Whitney U test; one-way ANOVA: Friedman test). For comparison of two groups in kinetic analyses, two-way ANOVA was used.

P.S. and M.L. designed research; P.S., P.W., X.L., Y.D., N.D.-H., D.T.M.N., M.D., K.N., Y.L., and V.H. performed research; P.S., M.F., T.M., M.-J.R., C.G., T.J., T.W., M.P., C.P., and M.L. contributed new reagents/analytic tools; P.S., S.S., T.M.B., G.A.H., F.H., P.D., G.K., E.Z., and M.-F.M. analyzed data; and P.S. and M.L. wrote the paper.

The authors declare no competing interest.