CircHYBID Regulates Hyaluronan Metabolism in Chondrocytes by Promoting the Expression of the hsa-miR-29b-3p Target Gene TGF-β1

Background: Hyaluronan (HA) metabolism by chondrocytes is essential for cartilage development and homeostasis. However, the function of circular RNAs (circRNAs)in HA metabolism is limited. In this study, we proled the role of the novel HA-related circRNA circHYBID in the progression of osteoarthritis. Methods: The function of circHYBID in HA metabolism in chondrocytes was investigated using gain-of-function experiments. The mechanism of HA-binding protein involved in hyaluronan depolymerization (HYBID) was conrmed through bioinformatics analysis and luciferase assays. The expression correlation of the circHYBID–hsa-miR-29b-3p–transforming growth factor (TGF)-β1 axis was examined by qRT-PCR and western blotting. CircHYBID, TGF-β1, and HA levels in cartilage samples were evaluated using immunohistochemistry. ELISA assay was used to assess HA accumulation in chondrocyte supernatant. Results: CircHYBID expression was signicantly downregulated in damaged cartilage samples comparing with that in the corresponding intact cartilage samples. CircHYBID expression was inversely correlated with the Mankin score and positively correlated with HA expression. Interleukin-1β stimulation in chondrocytes downregulated circHYBID expression and decreased HA accumulation. Gain-of-function experiments revealed that circHYBID overexpression in chondrocytes increased HA accumulation by regulating HA synthase 2 and HYBID expression. Further mechanism analysis illustrated that circHYBID upregulated TGF-β1 expression by sponging hsa-miR-29b-3p. Conclusions: Our results describe a novel HA-related circRNA that could promote the synthesis and accumulation of HA. The circHYBID–hsa-miR-29b-3p–TGF-β1 axis may play a powerful regulatory role in HA metabolism and Osteoarthritis (OA) progression. Thus, these ndings will provide new perspectives for studies on OA pathogenesis, and circHYBID may serve as a potential target for OA therapy.


Background
Osteoarthritis (OA) is the most common degenerative joint disease, and its socioeconomic impact is growing because of population aging and the rising prevalence of obesity (1). OA is also one of the major diseases leading to functional disability in the elderly (2).The chronic in ammatory response associated with OA is mediated by multiple factors, such as age, body weight, metalloproteinases (MMPs), cytokines, signaling pathways, metabolic disorders, and non-coding RNA (3,4). However, the pathogenesis of OA is unclear. At present, early interventions for clinical OA mainly consist of anti-in ammatory analgesia and cartilage-protecting treatment, but there are no effective drugs to reverse the progression of OA (5).
Degradation of the cartilage extracellular matrix (ECM) and secondary osteogenesis are the most prominent characterizations of OA (6). The cartilage ECM is a highly specialized and dynamic threedimensional scaffold in which chondrocytes reside in tissues. The ECM is composed of a variety of brillar components, such as collagens, bronectin, and elastin, and non-brillar molecules, such as proteoglycans, hyaluronan (HA), and glycoproteins including matricellular proteins (7). Homeostasis of ECM components including HA is tightly controlled by synthesis and degradation in cartilage tissue, and these processes are altered under pathological conditions. The initial pathological change of articular cartilage in patients with OA is the depletion of the HA-aggrecan(a major cartilage proteoglycan) network structure, which is followed by the degradation of collagen brils (8). HA is a non-sulfated glycosaminoglycan composed of disaccharide units containing N-acetylglucosamine and glucuronic acid (9). HA metabolism plays an important role in maintaining tissue homeostasis and regulating cell behavior (10,11). However, the regulatory mechanism of HA metabolism in chondrocytes remains to be further explored. Elucidation of the mechanism regulating HA metabolism in chondrocytes is expected to provide a new theoretical basis for OA pathogenesis.
Circular RNAs (circRNAs) are endogenous non-coding RNAs that play central regulatory roles by functioning as miRNA sponges or competing endogenous RNAs (ceRNAs) that naturally sequester and competitively inhibit miRNA activity (12). Previous studies illustrated that multiple circRNAs that are differentially expressed in cartilage tissue play central regulatory roles in ECM degradation, in ammatory response, apoptosis, and intracellular signaling pathways (13)(14)(15). Another study has revealed that downregulation of circ_001653/miR-486-3p/CEMIP axis promoted proliferation of NPCs and ECM synthesis of nucleus pulposus cells in intervertebral disc degeneration (16). Cell migration-inducing hyaluronan binding protein (CEMIP), also known as HYBID, is closely associated with cellular HA metabolism. However, there is no clear evidence concerning whether circRNA can regulate HA metabolism in cartilage tissues.
Therefore, this study investigated the function of circRNAs in HA metabolism. The ndings revealed a small number of up-and down-regulated circRNAs from different regions of cartilage tissue. Of interest, we speci cally proposed a novel circRNA, named circHYBID (has_circ_00003893), which is related to HA metabolism in chondrocytes, and systematically explored its role in the pathogenesis of OA. We present the following article in accordance with the MDAR reporting checklist.

Patients and specimens
All cartilage with primary damage and the corresponding normal cartilage were isolated from patients who underwent total knee replacement. The articular cartilage of femoral heads was separated using a knife blade within 24 h of cartilage isolation. The chondrocytes were separated using a primitive enzyme digestion method for primary culture and subculture, and the remaining cartilage tissue was frozen in liquid nitrogen for subsequent histological examination. Informed consent was provided by all organizational donors. This study was approved by the Human Ethics Committee of Meizhou People's Hospital.

Histological examination
The isolated cartilage was xed in 4% paraformaldehyde at 4°C for 2 days and then decalci ed in 15% EDTA for 2 weeks. Cartilage specimens were dehydrated with ethanol and xylene, embedded in para n, and cut into 5-mm sections. The prepared sections were stained with toluidine blue, saffron O, and hematoxylin-eosin according to the routine procedures. Cartilage specimens were graded according to the Mankin score. Scores <4 and >6 indicated that the cartilage samples were intact and damaged, respectively (17).

RNA sequencing and differential expression analysis
After RNA isolation, the quantity of total RNA in both sample types was measured using a Qubit RNA Assay Kit (Life Technologies, Waltham, MA, USA.), and the quality of RNA was assessed using the Agilent 2100 Bioanalyzer and RNA6000 Nano Kit (Agilent Technologies, Santa Clara, CA, USA). The sequencing library was prepared asper the manufacturer's instructions for the NEBNext Ultra™ RNA Library prep kit for the Illumina platform. Next, we analyzed the functions of differentially expressed genes. OA and normal cartilage were selected using log2 ratio ≥ 1 and the false discovery rate of multiple hypothesis testing (p < 0.05).

Bioinformatics analysis
According to previously reported sequencing data, speci c circRNAs were selected to construct the network. In particular, the potential target circRNAs were selected according to the circRNA microarray data, and a co-expression network was constructed according to the standardized signal intensity of circRNAs and mRNAs in the original microarray data. 6 circRNAs were selected to construct a circRNA-miRNA-mRNA network. In further research, the target circRNA, circHYBID, were selected to construct another circRNA-miRNA-mRNA network. The interactions of circRNAs and mRNAs with miRNAs were predicted with the Arraystar miRNA target prediction software based on TargetScan (http://www.targetscan.org/vert_72/) (17).

Primary culture of chondrocytes
Donor chondrocytes were isolated as previously described. Chondrocytes (5 ×10 5 /well) were obtained at passage 2 and grown in ProNectin F-coated Bio ex six-well culture plates (Flexcell International) to 80% con uence (17). The chondrocytes were treated with the addition of 10 ng/ml of interleukin (IL)-1β in the culture medium for 12 h. Untreated chondrocytes were used as controls.

Plasmid construction and transfection
CircHYBID and TGF-β1 overexpression plasmid were purchased from RiboBio (Guangzhou, China) and transfected into cells to overexpress circHYBID and TGF-β1. The wild-type (WT) or mutated type (MUT) plasmid of circHYBID and TGF-β1, miR-29b-3p mimics were also purchased from RiboBio (Guangzhou, China) for subsequent experiments. All plasmids above were veri ed by direct sequencing. Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) was utilized for all transfections in accordance with the manufacturer's instructions. The cells with the highest transfection e ciency were obtained after 48 h for subsequent experiments.
Quantitative real-time polymer chain reaction (qRT-PCR) Total RNA was isolated from cartilage tissues or primary chondrocytes using Trizol reagent. The gDNA was extracted from tissues or cultured cells according to the PureLink™ Genomic DNA Mini Kit protocol (Thermo Fisher Scienti c). RNA was reverse-transcribed using HiScript II Q RT SuperMixfor Q-PCR (+genomic DNA [gDNA] wiper) (Vazyme, Nanjing, China). AmpliTaq DNA Polymerase (Life Technologies) was used for PCR. The complementary DNA (cDNA) and gDNA PCR products were observed using 2% agarose gel electrophoresis. AceQqPCR SYBR Green Master Mix (Vazyme) was used for QPCR. GAPDH was used as internal controls. All primer sequences are listed in Table S1. The relative expression levels were determined using the 2 −ΔCt or 2 −ΔΔCt method.

Dual-luciferase assay
The sequence of circHYBID including the potential binding sites of miR-29b-3p was cloned into the dualluciferase reporter vector pmirGLO (Promega, China).The WT/MUT plasmids of circHYBID or TGF-β1 were co-transfected into cells with miR-29b-3p mimics respectively, together with the circHYBID overexpression vector or pcDNA3.1 empty vector, and changes in luciferase activity were analyzed using the Dual Glo Luciferase Assay System (Promega) following the manufacturer's instructions. Each experiment was performed at least three times independently.

Enzyme-linked immunosorbent assay (ELISA)
The HA concentration in the supernatant of chondrocytes cells in the presence or absence of interleukin-1 (IL)-1β) were determined using ELISA kits (R&D Systems, MN, USA) according to the manufacturer's instructions.

Statistical analysis
Statistically signi cant differences between groups were calculated using ANOVA or the non-parametric Wilcoxon signed ranks test. The results from two groups were compared using a paired two-tailed Student's t-test. The strength of the association between continuous variables was tested with the Spearman correlation. The results are reported as the mean ± SEM. p < 0.05 was considered statistically signi cant. All experiments were performed and analyzed in triplicate. Data analysis was performed using SPSS 25.0 software (Abbott Laboratories, Chicago, IL, USA).

HA accumulation was signi cantly decreased in damaged cartilage
Forty-eight pairs of cartilage samples from intact and damages areas were assessed histologically ( Figure 1A) and graded using a modi ed Mankin score (Table S2). The mean score for intact cartilage was 4, compared with 9 for damaged cartilage (p <0.0001, Figure 1B and Table S3). Regarding IHC, positive staining was observed primarily in the nuclei of chondrocytes, although positive staining was occasionally identi ed in the cytoplasm ( Figure 1A). The staining index was used to access the level of HA accumulation. The median staining index of intact cartilage was 6.5, compared with 3.4 for damaged cartilage (p < 0.0001, Figure 1C and Table S3). Pearson's analyses performed using pathological and IHC data revealed a negative correlation between the Mankin score and HA expression in OA cartilage tissues ( Figure 1D).

CircRNA expression pro les in intact and damaged cartilage
Based on the histologic examination, three pairs of cartilage samples were collected for further analysis (Figure 2A). Among these selected samples, the Mankin score of intact cartilage was ≤4, versus ≥9 for damaged cartilage (Table S4). According to the results of qRT-PCR, IL-1β and IL-6 expression was upregulated in damaged cartilage ( Figure 2B).
Sequencing was performed to characterize the expression pro les of circRNAs in cartilage. We identi ed 450 differentially expressed circRNAs between paired intact and damaged cartilage (Table S5). Differentially expressed circRNAs between damaged and intact cartilages were subjected to hierarchical clustering analysis (>1-fold difference in expression, p < 0.05). Among these differentially expressed circRNAs, 200 were upregulated in damaged tissue samples, whereas 250 were downregulated ( Figure  2C). The column chart revealed the frequency distribution of circRNA length in the cartilage samples ( Figure 2D). The volcano map was constructed according to fold changes and p-value, illustrating the varied circRNA expression between paired intact and damaged cartilage ( Figure 2E).
CircHYBID was predicted to be related to HA metabolism To further screen candidate circRNAs, we analyzed the host genes of the differentially expressed circRNAs. The result revealed that multiple circRNA-associated host genes were involved in the pathogenesis of OA, such as those involved in ECM degradation, collagen secretion, HA metabolism, and immune in ammatory response (Table S4).
Among these circRNAs, we noticed a special cartilage-associated circRNA (circHYBID, circRNA ID hsa_circ_0003893 in circBase; http://circbase.org), and its 398-bp gene was located at chr15:81229014-81230320. The symbol of the associated gene is KIAA1199, which encodes HA-binding protein involved in HA depolymerization (HYBID) through linear expression. Because HYBID was previously identi ed as an important enzyme involved in HA degradation, we assumed that circHYBID might also play an important role in HA metabolism and OA pathogenesis. Therefore, circHYBID was chosen for further investigation.
To further characterize circHYBID, divergent and convergent primers were designed to amplify the circular and linear transcripts, respectively, in both cDNA and gDNA. The PCR results revealed that the circular form was ampli ed using the divergent primers only from cDNA, whereas convergent primers ampli ed both cDNA and gDNA ( Figure 3C). Reverse splicing of circHYBID was successfully con rmed using Sanger sequencing ( Figure 3D).

CircHYBID expression and HA accumulation were downregulated by IL-1β
To further validate the role of CircHYBID, primary chondrocytes were successfully isolated (NC group) and stimulated by IL-1β to construct the in vitro OA chondrocyte model (IL-1β group). QRT-PCR illustrated that TNF-α expression was downregulated after IL-1β stimulation, con rming that the OA chondrocyte model was successfully constructed ( Figure 4A). CircHYBID expression of the chondrocytes was downregulated after IL-1β stimulation, which was consistent with the sequencing results of cartilage tissue ( Figure 4B).
The mRNA expression of HA synthase 2 (HAS2) was downregulated and HYBID was upregulated in chondrocytes treated with IL-1β. However, there is no signi cant difference of the HAS1 expression under IL-1β stimulation ( Figure 4C). WB data revealed that the protein expression of HAS2 was downregulated and HYBID was upregulated in OA chondrocytes ( Figure 4D). HA level in cell culture supernatant was increased after IL-1β stimulation ( Figure 4E).

CircHYBID increases HA accumulation by regulating the expression of HA-metabolizing enzymes
To investigate the relationship between circHYBID and HA metabolism, a circHYBID overexpression vector was constructed and transfected into chondrocytes. CircHYBID upregulation of circHYBID overexpression chondrocytes was veri ed by qRT-PCR ( Figure 5A). CircHYBID-overexpressing chondrocytes (OV-circHYBID+ IL-1β group) were further incubated with IL-1β. TNF-α expression was upregulated in standard chondrocytes with IL-1β stimulation. However, there is no signi cant TNF-α upregulation under IL-1β stimulation in circHYBID-overexpressing chondrocytes ( Figure 5B). Overexpression of circHYBID shows chondroprotective effect.
As mentioned previously, HAS2 expression was downregulated and HYBID expression was upregulated in IL-1β group. Comparing to IL-1β group, HAS2 downregulation was restored, and the upregulation of HYBID was offset in OV-circHYBID+ IL-1β group ( Figure 5C). The protein expression of HAS2 and HYBID was consisted with the mRNA results ( Figure 5D). HA content in the cell culture supernatant decreased after IL-1β stimulation, and HA production was recovered by circHYBID overexpression ( Figure 5E). Base on the above results, circHYBID increases HA accumulation of the chondrocytes.
CircHYBID upregulates the expression of the hsa-miR-29b-3p target gene TGF-β1 Given that many circRNAs commonly function as miRNA sponges that competitively sequester and suppress miRNAs, we assumed that circHYBID may also bind to miRNAs as a sponge and regulate targets via the ceRNA mechanism. Thus, the circRNA-miRNA-mRNA network of circHYBID was constructed to explore the molecular mechanisms of circHYBID ( Figure 6). Among the potential downstream targets of circHYBID, hsa-miR-29b-3p has caught our attention because it has been demonstrated to participate in the progression of OA (19). The expression of hsa-miR-29b-3p was detected using qRT-PCR, and the result show that hsa-miR-29b-3p expression was upregulated in chondrocytes treated with IL-1β. However, the expression of hsa-miR-29b-3p was inhibited by circHYBID overexpression ( Figure 7A). Hsa-miR-29b-3p may be a potential binding target miRNA of circHYBID. For further con rmation, the sequence of circHYBID was separately cloned into dual-luciferase reporter vectors (circHYBID-WT), and then the potential binding sites of hsa-miR-29b-3p were mutated (circHYBID-MUT). The aforementioned plasmids were co-transfected into cells, and changes in luciferase activity were analyzed. Luciferase activity was signi cantly downregulated in cells co-transfected with circHYBID-WT and hsa-miR-29b-3p mimics but obviously restored in cells co-transfected with circHYBID-MUT and hsa-miR-29b-3p mimics, indicating that the sequence is the potential binding site of circHYBID and hsa-miR-29b-3p ( Figure 7B).
From the circRNA-miRNA-mRNA network, we found that TGF-β1 was a potential downstream target of hsa-miR-29b-3p ( Figure 6). TGF-β1 has been reported to be closely related to HA metabolism and OA pathogenesis (20). Furthermore, a previous study has described the interaction between hsa-miR-29b-3p and TGF-β1 (21). Thus, hsa-miR-29b-3p was selected as the target mRNA for further research. TGF-β1 mRNA and protein expression was detected using qRT-PCR ( Figure 7C) and WB ( Figure 7D), respectively. The result illustrated that IL-1β inhibits the expression of TGF-β1 in chondrocytes, whereas circHYBID overexpression restored TGF-β1 expression. TGF-β1 expression was further veri ed in 48 pairs of cartilage samples using IHC. The median staining index of intact cartilage was 7.3, versus 4.1 in damaged cartilage. TGF-β1 expression was downregulated in damaged cartilage samples ( Figure 7E). Meanwhile, pearson's analyses indicated a positive correlation between relative circHYBID and TGF-β1 expression in OA cartilage tissues ( Figure 7F). Dual luciferase reporter assays as previous described was used to con rm the binding sites of hsa-miR-29b-3p and TGF-β1. Similarly, the luciferase activity of cells cotransfected with TGF-β1-WT and hsa-miR-29b-3p mimics was signi cantly downregulated but obviously restored in cells co-transfected with TGF-β1-MUT and hsa-miR-29b-3p mimics, indicating that the sequence is the potential binding site of TGF-β1 ( Figure 7G). Altogether, these results reveal that circHYBID upregulates the expression of TGF-β1, which is the potential target gene of hsa-miR-29b-3p.

Discussion
OA is a common disabling joint disease. No existing clinical options exist for preventing early OA from progressing to severe disease. To improve this situation, the underlying mechanisms of OA must be further explored to identify strategies to promote the biological regeneration of cartilage tissue in the early stage of OA (22).
HA is a ubiquitous high-molecular-weight polymer of repeated disaccharides of glucuronic acid and Nacetylglucosamine, and it is majorly found in the cartilage ECM between chondrocytes (23). HA constitutes the ECM and stabilizes its integrity. HA participates in varied biological processes, such as cellular and tissue development, migration, and repair after injury or in ammation, by binding to secretory proteins and cell surface receptors (24).
The synthesis and degradation metabolism of HA are maintained by the dynamic regulation of key enzymes such as CD44, HAS, hyaluronidase (HYAL), and HYBID (25,26).Dysregulated HA metabolism is an important factor in cartilage ECM degeneration and the development of OA. However, little is known about the functions of HA in the development and progression of OA. In the present study, IHC results using cartilage specimens from patients with OA revealed that HA levels were lower in damaged cartilage specimens. The staining index of HA was inversely correlated with the Mankin score, suggested that HA levels in cartilage gradually decrease with OA progression. Thus, assessing the mechanism of HA metabolism in OA may provide a new perspective for clarifying the pathogenesis of OA.
The role of circRNAs in the pathogenesis of OA has drawn increasing attention (17,27,28). Previous studies revealed that circRNAs play regulatory roles in cartilage ECM degradation, in ammatory response in chondrocytes, and apoptosis (12,29). This study investigated the pathogenesis of OA from the perspective of the regulation of HA metabolism by circRNAs. Among the identi ed differentially regulated circRNAs, we identi ed circHYBID as a key downregulated circRNA related to HA metabolism in OA cartilage for the rst time.
CircHYBID is generated by back splicing of the KIAA1199 gene, which comprises the protein coding sequence of HYBID mRNA. KIAA1199 has a key role in the binding and depolymerization of HA, and the encoded protein is also named HYBID or cell migration-inducing protein (26). HYBID can catabolize HA via the clathrin-coated pit pathway and play an important role in ECM homeostasis. HYBID is essential for the degradation of HA in skin, arthritic synovial broblasts, nucleus pulposus cells and cartilage independently of HYAL1 and HYAL2/CD44. HYBID is highly expressed by chondrocytes in the HAdepleted area of OA cartilage (4). Our data indicated that circHYBID plays a protective role against OA progression.
Accumulating evidence supports the key regulatory role of the ceRNA mechanism in many diseases, including OA. The mechanistic ceRNA network of circHYBID was explored in this study. Our data indicated that TGF-β1 was a potential target for circHYBID. TGF-β1 was recently demonstrated to play an important role in HA metabolism. Speci cally, TGF-β1 has a stimulatory effect on equine chondrocytes, enhancing HA synthesis and promoting ECM generation (20).Shen et al reported that CircCDK14 promote Smad2 expression in TGF-β signaling pathway, which protect against OA by maintaining the ECM of chondrocytes and regulating chondrocyte apoptosis and proliferation (30).
Based on our study ndings, we propose a mechanism by which circHYBID acts as an hsa-miR-29b-3p sponge to upregulate TGF-β1 expression. This axis promotes HA anabolism and suppresses HA catabolism, thereby increasing HA levels in cartilage.

Conclusion
we identi ed the circHYBID-hsa-miR-29b-3p-TGF-β1 axis as a novel mechanism for HA metabolism in OA. CircHYBID overexpression elevated HA accumulation, suggesting its potential use as a target for preventing and treating OA. Deciphering the precise molecular mechanisms of circHYBID in OA is critical for understanding the pathogenesis of OA and exploring new potential therapeutic targets.

Consent for publication
Not applicable.

Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.