Proprotein convertase furin inhibits matrix metalloproteinase 13 in a TGFβ-dependent manner and limits osteoarthritis in mice

Cartilage loss in osteoarthritis (OA) results from altered local production of growth factors and metalloproteases (MMPs). Furin, an enzyme involved in the protein maturation of MMPs, might regulate chondrocyte function. Here, we tested the effect of furin on chondrocyte catabolism and the development of OA. In primary chondrocytes, furin reduced the expression of MMP-13, which was reversed by treatment with the furin inhibitor α1-PDX. Furin also promoted the activation of Smad3 signaling, whereas activin receptor-like kinase 5 (ALK5) knockdown mitigated the effects of furin on MMP-13 expression. Mice underwent destabilization of the medial meniscus (DMM) to induce OA, then received furin (1 U/mice), α1-PDX (14 µg/mice) or vehicle. In mice with DMM, the OA score was lower with furin than vehicle treatment (6.42 ± 0.75 vs 9.16 ± 0.6, p < 0.01), and the number of MMP-13(+) chondrocytes was lower (4.96 ± 0.60% vs 20.96 ± 8.49%, p < 0.05). Moreover, furin prevented the increase in ALK1/ALK5 ratio in cartilage induced by OA. Conversely, α1-PDX had no effect on OA cartilage structure. These results support a protective role for furin in OA by maintaining ALK5 receptor levels and reducing MMP-13 expression. Therefore, furin might be a potential target mediating the development of OA.

Osteoarthritis (OA) is the most prevalent form of arthritis and a leading cause of pain and physical disability, especially in the elderly 1 . One major hallmark of the disease is the breakdown of articular cartilage matrix, especially type II collagen and proteoglycans, by several proteases 2 . Aggrecanases, such as a disintegrin and metalloproteinases with thrombospondin motifs (ADAMTS)−4 and −5, mediate cartilage remodelling 3 and their inhibition prevents OA-associated aggrecan degradation 4 . In addition, ADAMTS-5 deficient mice are protected against OA 5 . Similarly, matrix metalloproteinase (MMP)−13 is a major collagenase responsible for the degradation of type II collagen 6 . Increased expression of MMP-13 is observed in human OA cartilage 7,8 , and its knockout protects against cartilage destruction in murine OA 9 . Both Adamts5 and Mmp13 are target genes for transforming growth factor beta (TGFβ) in articular chondrocytes 10 . TGFβ, which is secreted by chondrocytes and stored within the matrix plays a crucial role in cartilage homeostasis 11,12 . Indeed, disruption of TGFβ signaling results in chondrocyte hypertrophy and cartilage breakdown 13 . The effects of TGFβ in chondrocyte metabolism are driven through its binding with TGFβ type II receptor (TGFβRII), leading to the recruitment of 2 activin receptor-like kinase (ALK) receptors. ALK1 activates the Smad1/5/8 pathway thereby leading to induction of chondrocyte hypertrophy and increased catabolism. By contrast, ALK5 receptor activates Smad2/3 signaling pathway involved in chondrocyte differentiation and anabolic response 14 . In addition, deletion of TGFβRII up regulates the expression of MMP-13 and ADAMTS-5 and results in OA-like phenotype in mice 10 .
ALK1/ALK5 balance is therefore crucial for chondrocyte metabolism and cartilage maintenance. A shift from ALK5 to ALK1 signaling occurs in OA and upon ageing 14 , leading to increased MMP-13 expression 15 .
Furin convertase, an enzyme involved in the maturation of chondrocytic proteases, also regulates TGFβ activity 16 . Furin was found to promote the in vitro maturation of MMPs in chondrocytes 17 and reduce collagen Furin activates ALK5/Smad3 signaling. Because MMP-13 is a target of TGFβ signalling, we investigated whether the downregulation of MMP-13 induced by furin was mediated by TGFβ pathway. In IL-1-stimulated chondrocytes, the ratio of ALK1/ALK5 increased compared to controls (1.8 ± 0.27 fold, p < 0.05). Furin restored an ALK1/ALK5 ratio similar to that of controls by reducing the expression of ALK1 and reversing the inhibition of ALK5 ( Fig. 2A,B). Additionally, furin promoted the phosphorylation of Smad3 (Fig. 2C) and the translocation of P-Smad3 into the nucleus (Fig. 2D). These effects were reversed by the TGFβ/ALK5 inhibitor SB431542. Interestingly, the protective effect of furin on the production of both proform and mature form of MMP-13 in IL-1-stimulated chondrocytes was lost in presence of SB431542 (Fig. 2E,F). We then knocked down ALK5 by siRNA which allowed a 90% reduction of Alk5 gene expression (not shown). In cells transfected with siALK5, furin had no effect on IL-1-induced MMP-13 expression (Fig. 2E). Overall, these findings demonstrate that ALK5/Smad3 pathway mediates the protective effects of furin on MMP-13 production induced by IL-1 in chondrocytes.
Furin reduces OA cartilage lesions and modulates ALK1/ALK5 ratio in mice. We then monitored the expression of furin in OA cartilage in mouse and human cartilage samples. Endogenous furin was expressed in healthy articular cartilage but not in damaged cartilage in mice (Fig. 3A). We further confirmed this finding in human samples (Fig. 3B), suggesting that the loss of furin activity might participate to cartilage degradation in OA. We thus assessed the effect of systemic administration of furin in a mice model of OA. Interestingly, OA score for cartilage lesions was lower with furin than vehicle (6.42 ± 0.75 vs 9.16 ± 0.6, p < 0.01) but not with the furin inhibitor α1-PDX (10.0 ± 0.51, p = NS). By contrast, furin had no impact on subchondral bone or synovium, suggesting that it elicited its effect specifically in cartilage during OA.

Discussion
Loss of cartilage that characterizes OA occurs mainly by a disrupted balance between the formation and degradation of extracellular matrix as a result of the maturation of growth factors and metalloproteases that are processed by proprotein convertases. Here, we evaluated the role of furin, the main proprotein convertase that metabolizes the activation of these mediators in OA 16,17,19 . We found that systemic treatment with furin alleviated OA cartilage lesions in mice, and that the inhibition of MMP-13 activation is a potential mechanism underlying this chondroprotective effect. Our results also suggest that the inhibition of MMP-13 expression in chondrocytes by furin occurs in an ALK5/Smad3 dependant manner. Indeed, we showed that furin maintains ALK5 and activates Smad3 signaling. This is consistent with previous study reporting that the TGFβ signaling regulates MMP-13 expression 10 and that Smad3 can repress MMP-13 expression in chondrocytes 20 .
The age-related increase in ALK1/ALK5 ratio observed in OA cartilage occurs mostly through a down-regulation of ALK5 14,21 . We show here that the inhibition of MMP-13 expression induced by furin was reversed by ALK5 knockdown in IL-1-stimulated chondrocytes. This suggests that furin inhibits MMP-13 expression by upregulating ALK5, in line with previous evidence that MMP-13 expression is increased after ALK5 knockdown in mice 15 . Consistently, furin maintained the expression of ALK5 in cartilage in our experimental model of OA, thus preventing the activation of MMP-13 and allowing the preservation of cartilage remodeling. These results are consistent with previous findings that deletion of Smad3 results in an OA phenotype in mice 13 . The shift from ALK5 to ALK1 signaling occurring with age is an important mechanism of cartilage damage in OA 22 . Here, furin maintained the levels of ALK5 and prevented the increased ALK1/ALK5 ratio in OA mice at the cell and tissue levels. Further evidence was provided by the furin inhibitor α1-PDX, which reversed the phenotype, and even promoted the expression of MMP-13 and aggrecanases. This supports a physiological role for endogenous furin to regulate cartilage catabolism and thus to limit post-traumatic OA in mouse knees by maintaining ALK5 and downregulating MMP-13. Consistently, we observed that furin was expressed in undamaged cartilage of humans and mouse while lost in OA cartilage, indicating that furin is involved in the protection of cartilage damage. However, the loss of furin expression in OA cartilage also suggest that this physiological mechanism is insufficient to prevent OA by itself, and supports the pharmacological relevance of the systemic administration of exogenous furin in OA. However, we cannot rule out an effect on other tissues because the administration was systemic but not local. The ability of furin to reduce MMP-13 expression in primary chondrocytes is mainly linked to furin-induced TGF processing and activity as a downstream effector.
Here, the administration of furin did not diminish the expression of aggrecanases in OA cartilage or in IL-1-treated chondrocytes. Despite the essential function of furin for cleavage of several proteases in chondrocytes and cartilage explants 19,[23][24][25][26] , furin might promote the processing of extracellular and intracellular proteins by several mechanisms 25,27 . Furin could have regulated the activation of ADAMTS in the matrix and facilitated the cartilage degradation by posttranscriptional processes, without affecting cellular expression 26 .
As discussed, our findings suggest that α1-PDX acted also on the endogenous Furin, without promoting Smad3 translocation or affecting ALK1/ALK5 ratio. The absence of regulation of TGFβ receptors by α1-PDX in primary chondrocytes suggests additional mechanisms induced by α1-PDX, a nonspecific inhibitor that might regulate several other proprotein convertases such as PACE4 24,26,28 . In addition to the direct effects in chondrocytes, systemic administration could have targeted other joint tissues, such as the synovium. Indeed, we have previously shown that low-grade synovitis was a common feature in the post-traumatic model of OA 29 . We have also previously demonstrated that α1-PDX was able to inhibit the regulatory T cells and promote synovitis and  joint damage in a mice model of collagen-induced arthritis 30 . Therefore, α1-PDX could have also promoted some level of inflammatory response that contributed to cartilage damage even in the OA mechanical model. Altogether, these data demonstrate that systemic administration of furin negatively regulates MMP-13 by maintaining ALK5 receptors and Smad3 signaling in OA, thus reducing cartilage degradation in mice. We here highlight the potential use of furin or other agonists for treatment of cartilage loss.
Animals. The experiments complied with the guidelines for animal experimentation issued by the local ethics committee on animal care and experimentation and approved (Ethical committee Lariboisière-Villemin, IRB n° 0000383, Paris). To induce joint instability of the knee, 10-week-old male C57/Bl6 mice (Janvier, France) underwent destabilization of the medial meniscus (DMM) of the right knee as previously described 33  The total amounts Furin and α1-PDX were provided in the same bulk by the manufacturer. Doses and schedule were determined in previous experiments and shown to be efficient for preventing arthritis in a collagen induced model 34 . In these conditions, mice did not show any abnormality in terms of growth, weight and behavior. At sacrifice, no macroscopic lesions were observed in any tissues examined.
Preparation of mouse and human joint samples. After sacrifice of the mice, whole knee joints were dissected free of soft tissues (n = 8 per group). All specimens were fixed with 4% paraformaldehyde (PFA; pH 7.4) for 24 h at 4 °C, and then decalcified with 1% PFA-0.2 M EDTA (pH 7.4, 4 °C) for 2 weeks, with the solution changed twice a week. The specimens were dehydrated with increasing concentrations of ethanol before being embedded in paraffin. Serial 5 µm thick sagittal sections were cut in the medial femoro-tibial compartment for histology and immunohistochemistry as described 35 .
Human cartilage samples were harvested from three patients who underwent total knee replacement surgery. Samples were obtained in accordance with the guidelines and regulations of the French National Authority Legislation for the collection of human tissues. Collections were approved by the ethical committee of the institution, informed consents were obtained from patients and stored in the medical record. Cartilage samples were collected from the femoral condyle at the posterior surface of the knees. We collected samples in a zone that appeared macroscopically undamaged defined by white and shiny cartilage without lesions and in a zone that appeared damaged defined by a discoloration with an irregular surface. All samples were fixed in 4% PFA, processed as described above and embedded in paraffin until processed for immunohistochemistry for Furin.
Histology and immunohistochemistry. Human cartilage samples were harvested from three patients who underwent total knee replacement surgery, in accordance with the French National Authority Legislation for the collection of human tissues. Cartilage samples were collected from the femoral condyle at the posterior surface of the knees. For each patient, we collected samples in 2 different zones: one sample in a zone that appeared macroscopically undamaged defined by white and shiny cartilage without lesions and one sample in a zone that appeared damaged defined by a discoloration with an irregular surface. All samples were fixed in 4% PFA and embedded in paraffin until processed for immunohistochemistry for Furin.
The blocks of the 4 groups of mice were processed (n = 8 per group). Serial 5 µm thick sagittal sections were cut, deparaffinized, rehydrated and washed twice by 5-min immersions in baths of distilled water. Sections Scientific REPORTS | (2018) 8:10488 | DOI:10.1038/s41598-018-28713-2 underwent Mayer-Hemalun staining for 5 min to stain the nuclei, and then counterstained with 0.125% Fast Green for 2 min for bone tissue. OA lesions were assessed in sections stained with Safranin-O. Sections were stained with 0.5% Safranin-O for quantification of OA scores 36 . Osteoarthritic lesions were scored by the OA scale ranging from stage 0 (normal) to 6 (vertical clefts/erosion to the calcified cartilage extending over >75% of the joint surface) in the medial compartment of tibia and femoral articular cartilage, for a global score of 0 to 12.
Immunohistochemistry was performed on serial sections as previously described 33 by use of the Vector kit (PK-6101, Abcys, France). Sections were incubated with the primary polyclonal antibodies for Furin (sc-20801, Santa Cruz Biotechnology, France) as well as the expression of ADAMTS using the following antibodies (murine ADAMTS-4, ab28285), ADAMTS-5, ab41037, Abcam, UK for both) and MMP-13 (ab3208, Abcam, France). Sections were counterstained with Toluidine blue. Slides with no primary antibody added were used as controls. Positive cells were counted on the articular cartilage of the tibia (80X magnification). We counted the number of cells expressing these metalloproteases and the number of total cells. Results are expressed as a percentage of positive cells to total cells.
The thickness of the synovium was measured in the serial sections stained for SO in the 4 groups of mice (n = 8). Synovial thickness was measured using a semi-automatic method with an image analyzer (Microvision, France). Data are expressed as mean ± SEM.
All the above experiments methods were carried out with the relevant guidelines and regulations and the protocols approved by the local scientific committee of the institution (Lariboisière-Villemin, IRB n° 0000383, Paris).

Statistical analysis.
Results are expressed as mean ± SEM of at least 3 separate experiments. The effect of DMM on mice was analysed by ANOVA followed by Wilcoxon rank or Mann-Whitney tests as appropriate. The latter test was also used to examine results of chondrocyte culture conditions. Level of significance was set at p < 0.05. Statistical analyses were performed using Statview software (SAS Inst., Cary, NC USA).