Up-regulated Endonuclease Regnase-1 Suppresses Osteoarthritis by Forming Negative Feedback Loop of Catabolic Signaling in Chondrocytes


 Background: Ribonucleases (RNases) play central roles in the post-transcriptional regulation of mRNA stability. Our preliminary results revealed that the endonuclease Regnase-1 is specifically up-regulated in osteoarthritic chondrocytes. We herein explored the possible functions and regulatory mechanisms of Regnase-1 in a mouse model of osteoarthritis (OA).Methods: The expression and target genes of Regnase-1 were identified by microarray analysis in primary-culture mouse articular chondrocytes. Experimental OA in mice was induced by destabilization of the medial meniscus (DMM). The function of Regnase-1 in DMM-induced post-traumatic OA mice was examined by adenovirus-mediated overexpression or knockdown in knee joint tissues, and also by using Regnase-1 heterozygous knockout mice (Zc3h12a+/-). Results: Among the RNases, Regnase-1 was exclusively up-regulated in chondrocytes stimulated with OA-associated catabolic factors. Adenovirus-mediated overexpression or knockdown of Regnase-1 alone in joint tissues did not cause OA-like changes. However, overexpression of Regnase-1 in joint tissues significantly ameliorated DMM-induced post-traumatic OA cartilage destruction, whereas knockdown or genetic ablation of Regnase-1 exacerbated DMM-induced cartilage destruction. Mechanistic studies showed that Regnase-1 appears to suppress cartilage destruction by modulating the expression of matrix-degrading enzymes in chondrocytes. Conclusion: Our results collectively suggest that up-regulated Regnase-1 in OA chondrocytes may function as a chondro-protective effector molecule during OA pathogenesis by forming negative feedback loop of catabolic signaling such as expression of matrix-degrading enzymes in OA chondrocytes.


Background
Osteoarthritis (OA) is a whole-joint disease characterized by cartilage destruction, synovial in ammation, osteophyte formation, subchondral bone remodeling, etc [1]. Among these manifestations of OA, progressive articular cartilage degradation is a hallmark of OA. This degradation is primarily regulated by chondrocytes through the production of matrix-degrading enzymes and/or the down-regulation of cartilage extracellular matrix (ECM) molecules [2]. Among the matrix-degrading enzymes, matrix metalloproteinase 3 (MMP3), MMP13, and ADAMTS5 are known to play important roles in OA cartilage destruction [3−5]. In chondrocytes, the expression levels of these molecules are regulated by extracellular catabolic regulators, such as the pro-in ammatory cytokine, interleukin (IL)-1β [6]. We previously identi ed cellular catabolic mediators in chondrocytes, including the transcription factor, hypoxiainducible factor (HIF)-2α [7], the zinc importer, ZIP8 [8], and the cholesterol hydroxylase, CH25H [9]. These cellular mediators exert catabolic functions by up-regulating matrix-degrading enzymes and/or downregulating ECM molecules in chondrocytes.
The expression of OA-related catabolic and anabolic factors in chondrocytes can be regulated at multiple checkpoints, including transcriptional and post-transcriptional levels. The ribonucleases (RNases) comprise a broad class of RNA-degrading enzymes that play central roles in the post-transcriptional regulation of mRNA stability [10]. However, the roles of speci c ribonucleases and their targets in OA pathogenesis have not yet been clearly elucidated. Here, we initially used a microarray analysis to screen for RNases whose expression were modulated in chondrocytes upon stimulation with OA-associated catabolic factors, such as IL-1β treatment, ZIP8 overexpression, or HIF-2α overexpression [6−8]. Among the examined RNases, our initial screening analysis revealed that Regnase-1 (a zinc nger CCCH-type containing 12A encoded by Zc3h12a in mouse) was exclusively up-regulated in chondrocytes stimulated with OA-associated catabolic factors. Regnase-1, which is also known as MCPIP1 (monocyte chemotactic protein-induced protein 1), is an endonuclease that destabilizes various target mRNAs by recognizing a stem-loop structure within their 3´-UTRs [11,12]. Regnase-1 is known to play diverse functions in different cell types, such as mitigating in ammation by down-regulating the IL-6 and IL-12B mRNAs in monocytes [11]; restricting T cell activation by targeting the c-Rel, Ox40, and IL-2 mRNAs in T cells [13,14]; and promoting cancer cell apoptosis by regulating mRNAs of Bcl2L1, Bcl2A1, RelB, and Bcl3 [15].
One of the known Regnase-1 target in chondrocytes is IL-6 [16,17], which plays an important role in OA pathogenesis [18,19]. However, the role of Regnase-1 in OA pathogenesis in vivo has not been investigated to date. Thus, we herein explored the possible functions and regulatory mechanisms of Regnase-1 in mouse models of post-traumatic OA. Our gain-of-function (adenovirus-mediated overexpression in joint tissues) and loss-of-function (Zc3h12a +/− ) approaches clearly indicated that up-

Methods
Mice and experimental RA C57BL/6J mice were used for the experimental OA studies. C57BL/6J-background Regnase-1-KO mice (68-bp deletion in exon 2) were generated by ToolGen, Inc. (Seoul, Korea). Because homozygous KO mice (Zc3h12a −/− ) die shortly after weaning [20], we used heterozygous Zc3h12a +/− mice for our experimental OA studies. All experiments were approved by the Animal Care and Use Committee of Gwangju Institute of Science and Technology. Post-traumatic OA was induced in 12-week-old male mice by destabilization of the medial meniscus (DMM), and the mice were sacri ced at the indicated weeks after the surgery [21].
Experimental OA was also induced by IA injection (once weekly for 3 weeks) of adenovirus [1 × 10 9 plaque-forming units (PFUs) in a total volume of 10 µl] expressing Regnase-1 (Ad-Regnase-1), SAA3 (Ad-SAA3), ZIP8 (Ad-ZIP8), or HIF-2α (Ad-HIF-2α). Mice were sacri ced 3 weeks after the rst injection of adenovirus [7−9]. When performing IA injection after DMM surgery, the rst injection was started at 10 days after surgery, and IA injections were performed weekly for a total of three injections. The mice were sacri ced at the indicated weeks after the surgery.

Histological analysis
Mouse knee joint samples were xed with 4% paraformaldehyde, decalci ed in 0.5 M EDTA, embedded in para n, and sectioned frontally at 5-µm thickness. Sections were depara nized in xylene, hydrated with graded ethanol, and stained with Safranin-O [8,9]. Cartilage destruction was scored by four observers under blinded conditions using the OARSI scoring system (grade 0-6) [22]. Synovitis was determined by Safranin-O and hematoxylin staining, and synovial in ammation (grade 0-3) was scored [23]. Osteophytes were identi ed by Safranin-O staining, and osteophyte size was measured with an Aperio ImageScope Version 12.4.0 (Leica Biosystems) [4]. We measured the thickness of the subchondral bone plate (SBP) using an Aperio ImageScope to assess subchondral bone sclerosis [24]. Skeletons of E18.5 whole-mouse embryos were stained with Alcian blue and Alizarin red [8,9]. The expression of GFP in cartilage tissues of mice IA-injected with GFP-tagged Ad-Regnase-1 was detected by standard uorescence microscopy Primary culture of mouse articular chondrocytes Mouse articular chondrocytes were isolated from the femoral condyles and tibial plateaus of 5-day-old WT or Zc3h12a −/− mice by 0.2% collagenase digestion [25]. The cells were maintained as a monolayer in Dulbecco's modi ed Eagle's medium (DMEM; Gibco) supplemented with 10% fetal bovine serum and antibiotics (penicillin G and streptomycin). On culture day 2, cells were infected with the indicated MOI (multiplicity of infection) of empty adenovirus (Ad-C), Ad-Regnase-1, Ad-shRegnase-1, Ad-SAA3, Ad-HIF-2α, or Ad-ZIP8 or treated with IL-1β as indicated in each experiment.

Microarray analysis
Our microarray data from chondrocytes stimulated with IL-1β treatment or those overexpressing HIF-2α (Ad-HIF-2α infection) or ZIP8 (Ad-ZIP8 infection) were previously deposited to the Gene Expression Omnibus under accession codes GSE104794 (HIF-2α), GSE104795 (ZIP8), and GSE104793 (IL-1β). We also performed microarray analysis in chondrocytes infected with 800 MOI of Ad-Regnase-1 or Ad-shRegnase-1 for 36 hours. Brie y, total RNA was extracted from mouse articular chondrocytes using a Purelink RNA mini kit (Ambion) and analyzed using Affymetrix Gene Chip arrays (Affymetrix GeneChip Mouse Gene 2.0 ST Array) in accordance with the Affymetrix protocol (Macrogen Inc.). The probe signals in the raw data were normalized with respect to the RMA (Robust Multi-array Average) for each separate data set (infection of Ad-C, Ad-Regnase-1, Ad-shC, or Ad-shRegnase-1). These microarray data were deposited to the Gene Expression Omnibus under accession codes GSE153179 (for Ad-Regnase-1 and Ad-shRegnase-1).
Reverse transcription-polymerase chain reaction (RT-PCR) and quantitative RT-PCR (qRT-PCR) analysis Total RNA was extracted from primary-culture chondrocytes using the TRI reagent (Molecular Research Center, Inc.). Total RNA was reverse transcribed, and the resulting cDNA was PCR ampli ed using the PCR primers and experimental conditions summarized in Supplementary Table 1. qRT-PCR was performed using an iCycler thermal cycler (Bio-Rad) and SYBR premixExTaq. Transcript levels were normalized with respect to those of β-actin and expressed as fold changes relative to the control.

Luciferase assay
The SAA3-3´-UTR reporter plasmid, which contained the 5´-AATAAATACTTGTGAAATGCA-3´ sequence of 3 -UTR of SAA3, was purchased from GeneCopoeia. Primary-culture mouse chondrocytes were pretreated with hyaluronidase type I-S (Sigma) for 3 hours in serum-free DMEM, and transfected by incubation for 6 hours with SAA3-3´-UTR reporter vector (0.05 µg) and Lipofectamine 2000, as described by the manufacturer. The cells were co-transfected with 0.05 µg of empty vector (Origene), WT-Regnase-1 expression vector (Origene), or D141N Regnase-1 (which lacks RNase activity) [13]. The cells were harvested at 24 hours after treatment, and re y luciferase and Renilla luciferase activities were measured using a Dual-Luciferase Assay System (Promega).

Statistical analysis
For statistical comparison of experimental groups, the data were analyzed by the Shapiro-Wilk test for normality and Levene's test for homogeneity of variance. Two groups of non-parametric data based on the ordinal grading systems (OARSI grade, synovitis, osteophyte maturity) were compared using the Mann-Whitney U test, whereas the Kruskal-Wallis test was used to compare multi-groups of nonparametric data. We also used the Mann-Whitney U test for the direct comparison of pairs of groups among the multi-groups of non-parametric data. Parametric data collected from two independent experimental groups were compared by two-tailed t-test. For comparisons of three or more groups of parametric data, one-way analysis of variance (ANOVA) with Bonferroni's post-hoc test was used. Each n number indicates the number of biologically independent samples or mice per group. Signi cance was accepted at the 0.05 level of probability (p < 0.05). Each bar represents the s.e.m. for parametric data and the calculated 95% con dence interval (CI) for nonparametric data.
Overexpression of Regnase-1 in joint tissues ameliorates post-traumatic OA cartilage destruction in mice Because the above results suggest that up-regulated Regnase-1 in OA-like chondrocytes may play a role in OA pathogenesis, we examined the functions of Regnase-1 in OA development by overexpressing it in mouse knee joint tissues via intra-articular (IA) injection of GFP-tagged Ad-Regnase-1. We previously reported that an adenoviral system can be used to effectively deliver transgenes into joint tissues [7][8][9]. Consistent with this, detection of GFP uorescence con rmed the overexpression of Regnase-1 in cartilage tissue ( Fig. 2A). However, this overexpression of Regnase-1 did not cause any OA-like change in the cartilage or synovium of the joint tissues (Fig. 2B). Consistently, Regnase-1 overexpression in primaryculture chondrocytes did not affect mRNA levels of matrix-degrading enzymes (MMP3, MMP13, and ADAMTS5), cartilage ECM molecules (type II collagen and aggrecan), or cellular catabolic regulators (HIF-2α and ZIP8) (Fig. 2C).
We further examined the effects of Regnase-1 overexpression in knee joint tissues under DMM-induced post-traumatic OA. Unlikely to the effects of Regnase-1 overexpression alone, adenoviral overexpression of Regnase-1 in joint tissues in DMM-operated mice signi cantly suppressed post-traumatic OA cartilage destruction ( Fig. 2D and E). Other examined manifestations of OA, such as osteophyte formation and thickness of the subchondral bone plate (SBP), also tended to be reduced by Regnase-1 overexpression, but the differences did not reach the level of statistical signi cance ( Fig. 2D and E). These ndings suggest that overexpression of Regnase-1 alone does not cause OA pathogenesis; instead, up-regulated Regnase-1 in OA chondrocytes appears to play a protective role in DMM-induced post-traumatic cartilage destruction in mice.
To further elucidate the functions of Regnase-1 in OA pathogenesis, we generated Zc3h12a KO mice or by knocking down Regnase-1 in whole-joint tissues via IA injection of an adenovirus expressing shRNA against Regnase-1 (Ad-shRegnase-1). IA injection of Ad-shRegnase-1 alone did not cause cartilage damage (Supplementary Fig. 2A). Additionally, Regnase-1 knockdown in primary-culture chondrocytes did not affect the expression levels of matrix-degrading enzymes or cartilage ECM molecules ( Supplementary  Fig. 2B). However, knock-down of Regnase-1 via IA injection of Ad-shRegnase-1in DMM-operated mice signi cantly enhanced DMM-induced OA manifestations such as cartilage destruction, osteophyte formation, and thickening of the SBP plate ( Fig. 3A and B). This is consistent with the idea that up-regulated Regnase-1 in OA chondrocytes has chondro-protective effects in DMM-induced OA pathogenesis.
We further validated the functions of Regnase-1 in OA pathogenesis by using a Regnase-1 KO mouse generated by deleting 68 bp from exon 2 of the Zc3h12a gene ( Supplementary Fig. 3A and B). Because homozygous KO mice (Zc3h12a −/− ) exhibit pre-mature death within 12 weeks of birth [20], we used heterozygous Zc3h12a +/− mice for our experimental OA studies. Zc3h12a +/− mice exhibited marked decreases in the mRNA levels of Regnase-1 in the chondrocytes of their cartilage tissues, but had normal skeletal development in E18.5 embryos (Supplementary Fig. 3C). Compared with WT littermates, Zc3h12a +/− mice exhibited signi cant enhancement of DMM-induced cartilage erosion but no signi cant change in osteophyte formation or SBP thickness ( Fig. 3C and D). These results additionally indicate that up-regulated Regnase-1 exerts protective functions in DMM-induced post-traumatic OA cartilage destruction in mice.

Regnase-1 modulates the expression levels of matrix-degrading enzymes in chondrocytes
To elucidate possible mechanisms underlying the protective effects of Regnase-1, we examined whether Regnase-1 modulates the expression levels of matrix-degrading enzymes and cartilage ECM molecules in chondrocytes. For this purpose, Regnase-1 was overexpressed in primary-culture chondrocytes via Ad-Regnase-1 infection, and the chondrocytes were stimulated with IL-1β. Among the examined molecules, the mRNA level of MMP3 (but not those of MMP13, ADAMTS5, or cartilage ECM molecules) was slightly but signi cantly decreased by the overexpression of Regnase-1 in IL-1β-treated chondrocytes ( Fig. 4A and  B). Similarly, the HIF-2α-induced increase in the mRNA level of MMP3 was also slightly but signi cantly decreased by Regnase-1 overexpression, whereas the mRNA levels of the other examined catabolic and anabolic factors were not modulated ( Fig. 4C and D). Although overexpression or knockdown of Regnase-1 alone did not modulate expression of matrixdegrading enzymes in chondrocytes (Fig. 2C, Supplementary Fig. 2B), up-regulation of matrix-degrading enzymes by OA-associated catabolic factor such as IL-1β and HIF-2α was signi cantly inhibited by Regnase-1 overexpression (Fig. 4) and potentiated by Regnase-1 knock-out in chondrocytes (Fig. 5). Therefore, it is likely that up-regulated Regnase-1 in OA chondrocytes may suppresse cartilage destruction by forming negative feedback loop of catabolic signaling such as expression of matrix-degrading enzymes in chondrocytes.
Of the putative targets of Regnase-1, serum amyloid A3 (SAA3) was the most markedly up-regulated by Regnase-1 knockdown and down-regulated by Regnase-1 overexpression (Fig. 6B). SAA3 is a member of the SAA protein family of apolipoproteins. Different isoforms of SAA are expressed constitutively (constitutive SAAs) at different levels or in response to in ammatory stimuli (acute-phase SAAs) [26]. We rst used the SAA3-3´-UTR reporter plasmid to test whether SAA3 was a bona de direct target of Regnase-1. We found that the expression of WT Regnase-1 reduced luciferase activity, whereas that of D141N Regnase-1, which lacks RNase activity [13], did not (Fig. 6C). In addition to SAA3, we also found that SAA1, SAA2, and SAA4 also tended to be up-regulated by Ad-shRegnase-1 and down-regulated by Ad-Regnase-1 (Fig. 6B, Supplementary Fig. 4). Among the SAA family members, SAA3 was the most abundant in chondrocytes (Fig. 6D) and the most signi cantly modulated by Regnase-1 expression (Supplementary Fig. 4). We therefore selected SAA3 as a Regnase-1 target for functional analysis in OA pathogenesis.
SAA3, a Regnase-1 target, does not modulate OA pathogenesis The function of SAA3 in OA pathogenesis was examined by its overexpression in joint tissues via IA injection of Ad-SAA3. Compared with Ad-C injection, IA injection of Ad-SAA3 in mice did not affect cartilage homeostasis or cause synovial in ammation (Fig. 6E). Consistently, overexpression of SAA3 in primary-culture chondrocytes did not modulate the mRNA levels of matrix-degrading enzymes (MMP3, MMP13, and ADAMTS5) or cartilage ECM molecules (type II collagen and aggrecan) (Fig. 6F). Furthermore, IA injection of Ad-SAA3 in sham-or DMM-operated mice did not affect post-traumatic OA manifestations, such as cartilage destruction, osteophyte formation, and synovitis ( Fig. 6G and H). Our results collectively indicate that SAA3 overexpression in joint tissues did not cause or modulate OA pathogenesis in mice, suggesting that the chondro-protective functions of Regnase-1 in post-traumatic OA cartilage destruction is mediated by unidenti ed other target(s) of Regnase-1, rather than SAA3, in chondrocytes.

Discussion
The post-transcriptional regulation of mRNA stability is an important checkpoint for gene expression, and ribonucleases play a central role in this process. Consequently, abnormalities in ribonucleases have been associated with multiple disease [12,27,28]. Indeed, we previously demonstrated that ZFP36L1, an RNAbinding protein that binds to AU-rich elements (AREs) in the 3'-UTRs of its target mRNAs, regulates OA pathogenesis by modulating the mRNA decay of HSP70 family members [29]. Here, we demonstrate that up-regulated Regnase-1 in OA chondrocytes protects against post-traumatic OA cartilage destruction by modulating the expression levels of matrix-degrading enzymes. Therefore, our results indicate that upregulated Regnase-1 in OA chondrocytes forms a negative feedback loop to function as a chondroprotective effector molecule in OA cartilage destruction.
One of the key ndings of our study is that Regnase-1 inhibits the expression levels of matrix-degrading enzymes, such as MMP3, MMP13, and ADAMTS5, in chondrocytes, and thereby suppresses DMMinduced cartilage destruction. Regnase-1 overexpression suppresses the IL-1β-or HIF-2α-induced upregulations of MMP3, whereas KO of Regnase-1 potentiates the up-regulations of MMP3, MMP13, and ADAMTS5 with corresponding modulation of cartilage destruction. Regnase-1 causes decay of target mRNAs by speci cally binding to UAU and UGU loops of the 3'-UTR located more than 20 nucleotides away from the stop codon [30]. According to RegRNA2.0 analysis [31], the mRNAs of mouse MMP3, MMP13, and ADAMTS5 have loops that meet these criteria. However, unlike the regulation of these enzymes in the presence of catabolic stimuli (i.e., IL-1β and HIF-2α), overexpression or knockdown of Regnase-1 in unstimulated chondrocytes did not markedly modulate the expression levels of these enzymes (Supplementary Tables 3-6). Therefore, our results indicate that Regnase-1 regulates the expression levels of matrix-degrading enzymes via indirect regulatory mechanisms, rather than direct targets of Regnase-1.
Regnase-1 regulates a variety of different target mRNAs in different cell types [11][12][13][14][15]. The mRNA of IL-6 has been identi ed as a Regnase-1 target in multiple cell types, including macrophages [20,32], T cells [13,14], and chondrocytes [16]. In unstimulated human OA chondrocytes, overexpression of Regnase-1 decreases the IL-6 mRNA level whereas expression of D141N mutant Reganse-1 (lacking RNase activity) does not [16]; this suggested that the IL-6 mRNA is a target of Reganse-1, although this was not con rmed in the previous study. Because IL-6 plays a critical role in OA pathogenesis [19], we initially speculated that Regnase-1 may exert its effects by modulating the IL-6 mRNA. However, similar to our observations regarding matrix-degrading enzymes, we found that overexpression or knockdown of Regnase-1 in unstimulated chondrocytes did not modulate the mRNA expression of IL-6 ( Supplementary  Tables 3-6). This suggests that IL-6 mRNA is not a direct target of Regnase-1 in mouse articular chondrocytes and does not play a role in the ability of Regnase-1 to regulate post-traumatic OA cartilage destruction. This is consistent with the fact that Regnase-1 targets vary dramatically among different cells and conditions [11,33]. It has been suggested that the process of Regnase-1 recognition is likely to be determined by multiple factors and distinct layers of regulation that may differ among cell types [11,33].
To identify mediator(s) through which Regase-1 regulates OA cartilage destruction, we tried to identi ed Regnase-1 target mRNAs via microarray analysis in chondrocytes with overexpression or knockdown of Regnase-1. We found that SAA family members met the criteria for Regnase-1 targets: they were downregulated by Regnase-1 overexpression and up-regulated by Regenase-1 knockdown, and contain UAU and UGU loops in their 3'-UTRs. SAA1 and SAA2 are classic acute phase response serum proteins in humans and mice; SAA3 is a pseudogene in humans that is transcribed but not translated; and SAA4 is expressed constitutively in humans [26]. The SAAs have been suggested to play roles in in ammatory diseases, such as atherosclerosis, rheumatoid arthritis, and chronic in ammatory bowel disease, and may function in cholesterol transport [34]. SAAs have also been reported to induce the expression of proin ammatory genes via NF-κB pathways [35], and recombinant SAA3 was reported to induce MMP13 expression in rabbit articular chondrocytes [36]. Here, we characterized the possible functions of SAA3 in the Regnase-1 regulation of OA pathogenesis, as it was the most abundantly expressed and signi cantly modulated SAA isoform in our system. However, our results revealed that overexpression of SAA3 in mouse articular chondrocytes does not modulate the expression of matrix-degrading enzymes or cartilage ECM molecules, and OA cartilage destruction is not modulated by the overexpression of SAA3 in mouse joint tissues. Therefore, it is likely that unidenti ed Regnase-1 targets other than SAA3 may mediate the chondro-protective functions of Regnase-1 during OA pathogenesis.

Conclusions
The present results suggest that up-regulated Regnase-1 in OA chondrocytes may functions as a chondro-protective effector molecule during OA pathogenesis by forming negative feedback loop of catabolic signaling such as expression of matrix-degrading enzymes in OA chondrocytes.

Availability of data and materials
The data supporting the conclusions of this article are included within the article.
Ethics approval and consent to participate