CircRREB1 mediates lipid metabolism related senescent phenotypes in chondrocytes through FASN post-translational modifications

Osteoarthritis is a prevalent age-related disease characterized by dysregulation of extracellular matrix metabolism, lipid metabolism, and upregulation of senescence-associated secretory phenotypes. Herein, we clarify that CircRREB1 is highly expressed in secondary generation chondrocytes and its deficiency can alleviate FASN related senescent phenotypes and osteoarthritis progression. CircRREB1 impedes proteasome-mediated degradation of FASN by inhibiting acetylation-mediated ubiquitination. Meanwhile, CircRREB1 induces RanBP2-mediated SUMOylation of FASN and enhances its protein stability. CircRREB1-FASN axis inhibits FGF18 and FGFR3 mediated PI3K-AKT signal transduction, then increased p21 expression. Intra-articular injection of adenovirus–CircRreb1 reverses the protective effects in CircRreb1 deficiency mice. Further therapeutic interventions could have beneficial effects in identifying CircRREB1 as a potential prognostic and therapeutic target for age-related OA.


3.
Please add the group name in figure 2l.

Please check the magnifications and add the scale bars.
Reviewer #4 (Remarks to the Author): Gong et al. found that circRNA CircRREB1 expression is upregulated in aging chondrocytes, which in turn affects FASN expression. These events exacerbate senescence phenotypes and osteoarthritis progression. They claimed that CircRREB1 interacts with and stabilizes FASN by inhibiting the proteasomal degradation of FASN. They showed that overexpression of either CircRREB1 or FASN is sufficient to cause senescent phenotypes in chondrocytes while their inhibitions abolish all the senescence-associated changes in chondrocytes. Knockout of CircRREB1 ameliorates OA in various mouse models.
Major concerns: Although the authors claim that CircRREB1 influences FASN stabilization via two types of PTMs, the underlying mechanisms still remain as a "speculation" as they mentioned in the discussion (line 667). It appears that authors would like to suggest that CircRREB1 serves as a scaffold to recruit both FASN and PTM regulators (what specific deacetylase?). They need to provide more robust biochemical evidence to fully reveal the precise molecular function of CircRREB1.
They did not sufficiently show how mechanistically FASN-mediated lipid metabolism contributes to causing senescence of chondrocytes. This would be the key message that readers from senescence/OA fields would like to know from this paper with the most interest.
Meanwhile, there is a very high chance that investigators from the senescence field would not be fully convinced by their claims about CircRREB1-or FASN-mediated senescence. For instance, they showed that overexpression of either CircRREB1 or FASN is sufficient to cause senescent phenotypes in chondrocytes within "24 h" (as described in figures legends). However, generally speaking, mammalian cells rarely make a commitment to enter senescence within that short time window even after they are exposed to strong DNA-damaging agents. For human cells, it is not even unusual to take more than a week to develop senescence-associated secretory phenotypes (SASP). Therefore, what they observed in 24 h after CircRREB1 or FASN overexpression is unlikely to be a "senescence" response.
Another critical reason why this reviewer does not think what they observed is senescence-associated phenomena is that unlike the general notion that mTOR inhibition suppresses SASP factor expression in senescent cells (Judith Campisi and colleagues, Nature Cell Biology, 2015 volume 17), they report that the combined treatments of PI3K and AKT inhibitors increased SASP factor expression.
Specific comments 1. More description on metabolomics data (in Fig. 4a-d) such as alterations in each lipid type. Metabolite measurement is required to support the significance of FASN in aging and OA progression. Do the lipid levels alter in aged human or mice and CircRREB1 overexpression system? What are the implications of these lipids in OA? 2. How is circRREB1 upregulated in aging chondrocyte in the first place? The upstream regulatory mechanism for circRREB1 production in senescent chondrocytes remains unclear. In Fig. 1f and Extended Fig. 1, the expression of circRREB1 is upregulated in senescent or aged chondrocytes. How are the expressions of the host gene RREB1 in these conditions? 3. Introduction includes redundant description (for instance, line 108-111 and line 111115). In line 106-108, is there any previously identified association between FASN and aging-associated diseases other than cancer? 4. In line 161 and Supplementary Table 1, the result of RNA-seq summarized in the table described only 15 up-and down-regulated circRNAs, not 20. For the table, a statistical measurement such as FDR to test the significance of differential expression is missing. What does mean value indicate in the table? Read count? 5. Description of histological assessment method of human and mouse OA is missing, especially OARSI grading and synovitis scoring. How was the thickness of articular cartilage measured? 6. More detailed description of the origin of human cartilage samples is required since it lacks consistency throughout the manuscript. While online methods read "participants (50-60 years old) were included into the younger group and participants (70-89 years old) were included into the older group", other texts in Result and figures read "younger (50-65y) and older adults (70-85y)". 7. Related to Fig. 4e, human cartilage shows distinct zonal structure and chondrocytes in each zone exhibits different characteristics. The authors should have discussed the zonal origin of cartilage tissue and section method for the cartilage samples, and compared the protein level of fatty acid metabolism-related proteins within similar regions. 8. Related to Fig. 1c,d, the result of RNA-seq should be addressed in detail. How was total 3800 circRNAs were identified using CIRI2 and CIRC explorer? In the previous article published by the same author (Shen et al., Ann. Rheum. Dis. 2019), the author identified 12,738 circRNAs using three OA and control human cartilage samples and the same circRNA explorers. Where did the discrepancy come from? How was the expression of circRREB1 in the previous RNA-seq data? 9. Related to the screening of circRREB1 (hsa_circ_0001573) in Fig. 1d, the rationale for choosing hsa_circ_0001573 for further study among differentially expressed circRNAs is lacking the justification. For instance, in line 171-173, why was the top downregulated hsa_circ_0000211 excluded for further study? 10. Related to Fig. 4l and m, do CircRreb1 gKO mice show reduced synthesis of specific lipids in system level? 11. For Fig. 5d and Extended Fig. 4f,g, binding site prediction of circRREB1 with FASN is not clearly explained. How does the CatRAPID prediction result match with that from RIP assay? 12. Some of the statistical methods they used are inappropriate or inaccurately described. Especially, statistical analysis of nonparametric data such as histological grades should not have been conducted using 'two-side unpaired t-test' throughout the manuscript. The authors described that they used one-way ANOVA with LSD-t by assuming equal variance for multiple group comparisons. However, it says one-way ANOVA with Turkey's multiple comparison, which should be at least 'Tukey's HSD' although LSD and HSD use different significant difference. Overall, this reviewer cannot figure out the statistical method they used. 13. Related to Fig. 6a, the rational for selecting MDM2 as a candidate is insufficiently provided. 14. Related to mass spectrometry analysis of FASN in line 470-473, the method is completely missing and the explanation for MS data is not enough. Was the UPLC-MS/MS method performed focusing on detecting acetylation sites of FASN? Were K673 and K1065 the only acetylation sites on FASN? 15. In Fig. 6k, the author could not find the result of Flag-FASN MS. 16. Related to Extended Fig. 9-10, there are many attempts to reduce OA and SASP expression by inhibiting PI3K-Akt pathway (Wang et al., Am. J. Transl. Res. 2022) and the inhibition of this pathway is also known to promote autophagy (Xue et al., Biomed. Pharmacother. 2017). PI3K-Akt is generally known to activate mTOR and NF-kB pathway, which are considered as a pro-SASP transcription factors (Herranz et al., Nat. Cell. Bio. 2015;Salimenin et al., Cell Signal. 2012). 17. In Fig. 7h, the quality of IHC against FGFR3, FGF18, p-PI3K, and p-Akt are low (background levels are very different between samples), making it difficult to compare signals between cKO+DMM and cKO+DMM+CircRreb1 oe groups.

Reviewer #1 (Remarks to the Author):
This study examines the role of a circRNA in lipid metabolism related cellular senescence and osteoarthritis.
CircRREB1 was found by deep sequencing to be upregulated in aging chondocytes, and its role in mouse models of aging-related and surgically induced OA was studied.
Detailed mechanistic analyses revealed interactions with FASN and PI3k pathway.
The study is novel as it is the first to address CircRreb1 in the context of OA.
Most experimental approaches are suitable, but some have limitations and clarifications and changes in data interpretation are needed.
Response: Thank you for your efforts to review our manuscript. We appreciate your helpful comments to improve and correct this manuscript. We are very happy to edit our manuscript according to your constructive suggestions. We have done some experiments carefully to support these concerns and then we revised this manuscript with all changes highlighted in red. Thanks again for your help.

Suggestions for revision:
Authors refer to 'aging' chondrocytes, but the model is simply for culturing cells for two passages in monolayer. It is not appropriate to refer to this as a model of aging. It is only an effect of monolayer culture. Throughout the text this needs to be changed.
Response: Thanks for your constructive suggestions, we are very happy to further edit our manuscript according to your comments. We have changed aging chondrocytes into Response: Thanks for your comment for our manuscript. We cultured primary chondrocyte and subjected them to two passages (P2). At the station of P2, we found that most chondrocytes were highly expressed SA--Gal staining, suggesting P2 generation chondrocytes were at senescence station. A reference published in Science translational medicine by professor Jin-Hong Kim (Stress-activated miR-204 governs senescent phenotypes of chondrocytes to promote osteoarthritis development PMID: 30944169) confirmed that doxorubicin induced chondrocyte senescence. Only senescent cells expressed SA-J3-Gal and Doxo treatment significantly increased SA-J3-Gal positive staining. In this study, Kim and et al. used different model to investigate the senescence phenotypes. Hence, in our study, we would like to investigate the effects of CircRREB1 in different models. In Figure 2f, we used Doxo to induce senescence of human chondrocyte, followed by CircRREB1 knockdown to observe SA-J3-Gal staining, immunofluorescence staining of MMP13, p16, Sox9, and Alcian blue staining. In Figure   2a-d, for molecular detection, P2 generation OA chondrocytes were used. We have added this information in the figure legend. Would increased expression of these genes, for example by IL-1 also be suppressed by si CircRREB1?
Response: Thanks for your constructive suggestion. We used OA chondrocytes to investigate the role of CircRREB1 on MMP13 and ADAMTS5 expression. OA chondrocytes transfected with NC, CircRREB1#1, CircRREB1#2, CircRREB1#3 SiRNAs respectively. The CT value and power value of one sample are listed as follow: We have further performed an experiment to explore the role of CircRREB1 knockdown after IL-113 stimulation. Method: Human chondrocytes were seeded into 6-well plate and cultured with DMEM high glucose media containing IL-113 (10 ng/ml) for 48h, followed by CircRREB1 knockdown. Cells were collected and RNAs were extracted. Next, RT-qPCR was performed. β-actin (ACTB) rRNA was chosen as a housekeeping gene. Here, we listed the CT value and power value of MMP13 and ADAMTS5 of one sample as follows: The result of RT-qPCR was showed as follows. We found that IL-1β simulation significantly increased MMP13 and ADAMTS5 mRNA expression, however, these expressions were decreased after CircRREB1 knockdown.  The same concern applies to Extended Data Fig. 6. Intra-articular infection of FASN AAV.
Response: Thanks for your helpful suggestion. We performed CircRreb1 FISH staining in aging mice. Immunofluorescence pictures taken in the same batch of experiments under the same conditions were selected. We selected CircRreb1 immunofluorescence intensity to evaluate the fold increase of CircRreb1 in the joint tissues. The CircRreb1 immunofluorescence intensity was calculated by IntDen (correction) = IntDen (cell)-Area (cell) * Mean (background) through Image J software. We showed the relative fluorescence intensity as follows. The fold increase of CircRreb1 in 3M mice and 18M mice were showed in Extended Data Fig 5a- Response: Thanks for your constructive suggestion. In human normal cartilage, the superficial zone was defined as the first 10% of the tissue thickness, the translational zone as the next 10% and the radial zone as the remaining 80% of the tissue thickness (PMID:12127837). Hence, according to the distinct zonal structure of cartilage, we would like to explore the superficial zone (articular cartilage, AC) and the radial zone (calcified cartilage zone, CCZ) of tibia plateau according to this reference. According to your suggestion and another reviewer's comments, we showed the FASN, ELOVL5, ELOVL6, and SCD1 expression in articular cartilage (AC) and calcified cartilage zone (CCZ) respectively between younger individuals and older individuals (  The mRNA seq experiment is a strength as it adds an independent approach to investigate CircRreb1 mechanisms. The main finding that is further investigated is about PI3K-AKT signal Transduction and FGF3 and FGF18. Here it would be important to determine whether CircRreb1 changes mRNA levels of FGF3 and FGF18 directly (independent from FASN).
Response: Thanks for your constructive suggestion. We performed knockdown and gain of function experiment. We knockdown and overexpress CircRREB1 in HCs, then, RNA was extracted. RT-qPCR result confirmed that CircRREB1 knockdown upregulated FGFR3 and FGF18 mRNA expression, while CircRREB1 overexpression downregulated these expressions (Extended Data Fig. 14f). The result of RT-qPCR was showed as follows: In this manuscript circRREB1 is identified to be highly differently regulates in chondrocytes exposed to hypoxia compared to normoxia. circRREB1 were subsequently verified to be higher expressed in cells treated with doxo, an agent used to induce cellular senescence, and in mouse and human cartilage related to ageing or OA, respectively. The cellular and molecular role of circRREB1 is characterized by knockdown and transfection studies, and identified to be related to lipid metabolism.
Response: Thank you very much for your time and effort that you have put into reviewing our manuscript submitted in Nature Commutations. We appreciated that you give us this chance to better edit our manuscript according to your suggestions and comments.
Based on the comments, we carefully revised the manuscript with all changes highlighted in red. This group has previously identified differential regulated circRNA in samples from OA patients and normal controls (ref 29); however, circRREB1 is not listed in their findings in this article nor in a similar study based on patients' samples by Lin et al. 2017 to be among the highest expressed circRNAs. Expression levels do not have to reflect biological importance, but there is a lack of discussion related to this lack of identification in larger human patient material.
Response: Thanks for your comment to our manuscript. Your advice is beneficial to improve the quality and rationality of our study. We previously used 3 paired clinical OA and control tissues to perform CircRNA deep sequencing. There were no obvious difference of age, height, weight, and BMI between 3 paired clinical OA and control tissues. The difference between two groups is whether is OA. To further observe the relationship between CircRREB1 and age, we used sections of knee joints (Medial tibial plateau) from different age who received total knee joints replacement to perform CircRREB1 FISH staining. We collected 24 samples: 50-54y(n=4), 55-60y(n=3), 61-64y (n=4), 65-70y (n=3), 71-74y (n=3), 75-80y (n=4), 81-85y (n=3). CircRREB1 was labeled by Cy3. We showed the representative images of CircRREB1 FISH staining. We observed that CircRREB1 increased with aging. Next, we quantified the relative CircRREB1 intensity among these samples (intensity of CircRREB1 in 50y sample was used as control). The CircRREB1 immunofluorescence intensity was calculated by IntDen (correction) = IntDen (cell)-Area (cell) * Mean (background) through Image J software. The Reletive CircRREB1 intensity was showed as follow. Linear Hypoxia conditions are not described in method part.
Response: We are sorry for this missing part. We have added it into the method part in the revised manuscript. Chondrocytes were cultured in an incubator containing 5% CO2, and 3% O2 at 37°C (hypoxia condition). Please see line 921-924.

Scale bars lacking on in all figures with images of cells.
Response: We are thanks for your careful review. We are very pleasure to further edit our manuscript according to your advice. We have added scale bars in the revised manuscript.
In the extended data This study should incorporate one or more additional biomarkers of senescence, and careful consideration should be given to how the interventions being applied effects autophagy and the lysosomal activity of cells.
In addition, the pH values, which play roles in the results of SA-3-gal staining, should be described in the Method section.
Response: Thanks for your constructive suggestion. According to your suggestion, we performed some experiments to explore the roles of CircRREB1 on autophagy and the lysosomal activity (please see line 256-273). We collected and cultured human chondrocytes from OA individuals. Then, CircRREB1 was knockdown. Western blot analysis showed that CircRREB1 knockdown did not affect autophagy associated proteins such as p62 and LC3 (Extended Data Fig. 3c, d, and e). Although CircRREB1 shows no influence on autophagy pathway, we also explore the role of CircRREB1 on lysosomal activity by using chloroquine (Cq). After NC or CircRREB1 SiRNA transfection, chondrocytes were treated with or without Cq (5uM) for another 2 days. Western blot showed the LC3 and p62 expression with or without Cq stimulation. Autophagy flux index was used to evaluate the impact of CircRREB1 on lysosomal activity. Autophagy flux= (LC3B-II + CQ/3-actin)/(LC3B-II-CQ/3-actin). The result of Autophagy flux showed no difference between Si-NC and Si-CircRREB1 group (Extended Data Fig. 3f and g).
Furthermore, we used another senescence biomarkers such as p21 and -H2AX nuclear foci to support the anti-senescence role of CircRREB1 knockdown. CircRREB1 knockdown in chondrocytes decreased p21 and -H2AX nuclear foci expression (Extended Data Fig. 3a). Meanwhile, we have described the pH of SA-3-gal staining in the method part (please see 963).
We thank for this concerns to our manuscript, we also would like to explain that the kit used in this study is Senescence 13-Galactosidase Staining Kit. Senescent cells usually become larger and express high 13-Galactosidase enzyme activity at pH6.0, which suggests that senescent cells can be marked by this kit. Specific SA-13-gal staining kit has been widely used by many previous study (PMID: 28043053, 29941930, and 30944169).
We have also noticed Lysosomal 13-Galactosidase Staining Kit, which is used for acid 13-Galactosidase detection in lysosomal. We have also noticed that most normal cells express a high level of lysosome 13-Galactosidase activity level, which can be a negative control of Senescence 13-Galactosidase Staining. Hence, we performed experiment again to detect acid 13-Galactosidase after CircRREB1 knockdown in chondrocytes. The result of Lysosomal 13-Galactosidase Staining confirmed that no obvious difference was showed between Si-NC and Si-CircRREB1 group. All these results were showed as follows.
Extended Data Fig. 3  3. The human knee joints from younger (50-65y) and older adults (70-85y) but the mice joints from 8w, 3m, and 18m are evaluated. Why are these time points chose?
Response: The incidence rate of osteoarthritis gradually increases among people over 50 years old. We collected human knee joints samples from participants with different age.
The topic in this study is aging associated CircRNA, then, we would like to distinguish samples by age and classified them into two groups according to our previous study to improve our manuscript. There may be some certain errors between different individuals for this score evaluation, therefore, we further invited another two individuals who were blinded to our experiment to perform OARSI evaluation again in cKO mice and Fasn oe model (Fig. 7d, Extended Data Fig. 10 a).  Fig. 12b and d), suggesting that FASN mediates the function of CircRREB1 in age-related OA (please see line 500-513).
Secondly, chondrocytes were co-transfected with Flag-FASN and HA-UB plasmids and then treated with or without TSA and NAM. The acetylation level of FASN was increased by TSA (Fig. 6b), but not by NAM (Fig. 6c). Moreover, TSA treatment enhanced ubiquitination of FASN (Fig. 6b), thereby promoted FASN degradation. This result suggested that acetylation promoted FASN degradation through the ubiquitinproteasome pathway.
Thirdly, we would like to find out which specific deacetylase involved in FASN acetylation because the acetylation level of FASN was affected by an HDAC family deacetylase. In detail, we co-transfected Flag-FASN with different HDAC family deacetylase. Co-IP result showed that HDAC3 combined with FASN (Fig. 6d). Herein, we confirmed that HDAC3 is the specific deacetylase involved in FASN acetylation.
Next, we performed molecular docking to visualize the combination of CircRREB1, FASN, and HDAC3. We would like to explain that CircRREB1 serves as a scaffold to recruit HDAC3 to stabilize FASN protein level (Fig. 6h). In Fig. 6, part (b) showed the interaction sites between CircRREB1 and FASN, (c) and (d) showed the interaction sites between CircRREB1 and HDAC3. We found the K673 site on FASN, which is also the acetylation site on FASN detected by UPLC-MS/MS method (Fig. 6e). To confirm this conclusion, we also performed some experiments. We have confirmed that the interaction between CircRREB1 and FASN through RNA pulldown assay and RNA immunoprecipitation (RIP) assay. We also performed RIP assay to confirmed the interaction between CircRREB1 and HDAC3 (Fig. 6f). RIP result showed that CircRREB1 also interacted with HDAC3 (Fig. 6f). To further investigate the relationship between CircRREB1, FASN, and HDAC3, CircRREB1 knockdown and overexpression were performed respectively in chondrocyte and we found that CircRREB1 overexpression enhanced the interaction between FASN and HDAC3, and then decreased FASN acetylation, thereby decreased FASN ubiquitination (Fig. 6g). However, CircRREB1 knockdown showed the opposite result (Fig. 6g) Meanwhile, there is a very high chance that investigators from the senescence field would not be fully convinced by their claims about CircRREB1-or FASN-mediated senescence.
For instance, they showed that overexpression of either CircRREB1 or FASN is sufficient to cause senescent phenotypes in chondrocytes within "24 h" (as described in figures legends). However, generally speaking, mammalian cells rarely make a commitment to enter senescence within that short time window even after they are exposed to strong DNA-damaging agents. For human cells, it is not even unusual to take more than a week to develop senescence-associated secretory phenotypes (SASP). Therefore, what they observed in 24 h after CircRREB1 or FASN overexpression is unlikely to be a "senescence" response.
Response: We are very sorry for this statement. There must be a mistake because we First of all, we performed mRNA-seq (Si-CircRREB1 vs Si-NC) and found that CircRREB1 knockdown was highly associated with PI3K-AKT signaling transduction by regulating FGFR3 and FGF18 expression (Extended Data Fig. 14c) Fig. 14h), as well as downstream PI3K-AKT molecule (Extended Data Fig. 15c). However, we also found that FGFR3 and its related PI3K-AKT pathway are activated in many tumor models (PMID: 21078999, 24519156, 29299828,16767162 Fig. 15a). AKT inhibition also increased CDK inhibitors p16 and p21 in OA chondrocytes (Extended Data Fig. 15b). We speculated that CircRREB1 regulated AKT activation to regulate p21 pathway.
CircRREB1 knockdown, followed by AKT inhibition, Col2, MMP13, p16, and p21 expression were partially reversed, which suggested that CircRREB1 regulate p21 through PI3K-AKT signal pathway (Extended Data Fig. 15e). Thirdly, in our manuscript, we found that CircRREB1 and FASN knockdown enhanced p-PI3K and p-AKT expression, but not p-mTOR expression. Because mTOR pathway is highly associated with autophagy. Herein, we have done some more experiments to explore whether CircRREB1 regulate autophagy and lysosome activity in chondrocyte. We showed the Extended Data Fig.3 as follows. We collected and cultured human chondrocytes from OA individuals. Then, CircRREB1 was knockdown. Western blot analysis showed that CircRREB1 knockdown did not affect autophagy associated proteins such as p62 and LC3 (Extended Data Fig. 3c, d, and e). Although CircRREB1 shows no influence on autophagy pathway, we also explore the role of CircRREB1 on lysosomal activity by using chloroquine (Cq). After NC or CircRREB1 SiRNA transfection, chondrocytes were treated with or without Cq (5uM) for another 2 days. Western blot showed the LC3 and p62 expression with or without Cq stimulation (Extended Data Fig.   3f). Autophagy flux index was used to evaluate the impact of CircRREB1 on lysosomal activity. Autophagy flux= (LC3B-II + CQ/I3-actin)/(LC3B-II-CQ/I3-actin). The result of Autophagy flux showed no difference between Si-NC and Si-CircRREB1 group (Extended Data Fig. 3g). CircRREB1 knockdown also reduced p21 and gama-H2AX nuclear foci expression. Above all, we confirmed that CircRREB1 regulate chondrocyte senescence indicated by p21 and other senescence markers instead of regulating autophagy and lysosome activity. This result is consisted with the result that CircRREB1 did not affect mTOR pathway.

Extended Data Fig. 15 PI3K inhibitor and AKT inhibitor reverses the effects of
Extended Data Fig. 3  At last, we found that CircRREB1 knockdown decreased p21 expression and overexpression enhanced p21 expression ( Fig. 2a and h). We applied combination of PI3Ki and AKTi in OA chondrocytes and found that p21, p16 was upregulated. Then we performed rescue assay, combination treatment in chondrocyte with CircRREB1 knockdown and found that p21, Col3, MMP13, and p16 were rescued, suggesting CircRREB1 regulate p21 pathway through PI3K-AKT transduction. Although PI3K-AKT-mTOR pathway caused SASP in many senescent tumor cells, we would like to discuss it in the discussion part. CircRREB1 knockdown in chondrocyte showed little effect on autophagy, as well as little effect on mTOR activation. However, CircRREB1 knockdown actually activated AKT activation in OA chondrocytes and regulated senescence associated p21, hence, we discussed that the effect of CircRREB1 on regulating p21 pathway larger than mTOR pathway(please see line 843-867).

Specific comments
1. More description on metabolomics data (in Fig. 4a-d) 7a). Next, in order to investigate whether lipid metabolism also increased in aged mice, CircRREB1 overexpressed human chondrocytes or CircRreb1 gKO mice, we performed lipid metabolomics. In 18-month-old mice cartilage tissues, most types of lipid including FA, GL (DAG), GP (LPC, LPE, PC, PE), and SP (Cer, HexCer) were increased compared to 3month-old mice, suggesting that these lipids involved in aging mice (Extended Data Fig.   7b, c, and d). Human chondrocytes treated with Ad-CircRREB1 and Ad-vector and metabolic analysis showed that most types of lipid were also increased after CircRREB1 overexpression (Extended Data Fig. 7e, f, and g). Furthermore, cartilage samples from WT mice and CircRreb1 gKO mice were collected and lipid metabolomics showed that some lipids (FA, LPC, LPE, PC, and PE) were decreased after CircRreb1 knockout (Extended Data Fig. 7h, I, and j). Above all, we found that CircRREB1 mediated lipid metabolism in chondrocytes with aging (please see line 352-364).
As age is the most risk factor for OA. Next, we would like to explain the application of these lipid in OA. The current existing evidence suggests that OA is associated with obesity-related chronic inflammation as well as abnormal lipid metabolism in obesity, such as fatty acids ( 2. How is circRREB1 upregulated in aging chondrocyte in the first place? The upstream regulatory mechanism for circRREB1 production in senescent chondrocytes remains unclear. In Fig. 1f and Extended Fig. 1 staining. All sections were evaluated by Osteoarthritis Research Society International (OARSI) system and synovitis score to evaluate the OA degree by two independent individuals. OARSI grading and synovitis score were listed in Supplementary Table S9 and Supplementary Table S10.
For the thickness of articular cartilage measurement, maximal cartilage thickness measured from at least 5 sections (80 um apart) of each joint by two intendent individuals.
Then, measurements of each section of each joint were averaged to give an average joint thickness measure.
6. More detailed description of the origin of human cartilage samples is required since it lacks consistency throughout the manuscript. While online methods read "participants Response: Thanks for your suggestions for our manuscript. In human normal cartilage, the superficial zone was defined as the first 10% of the tissue thickness, the translational zone as the next 10% and the radial zone as the remaining 80% of the tissue thickness (PMID:12127837). Hence, according to the distinct zonal structure of cartilage, we would like to explore the superficial zone (articular cartilage, AC) and the radial zone (calcified cartilage zone, CCZ) of tibia plateau according to this reference (please see line 365-372).
Section method for the cartilage samples: Tibia plateau tissues were collected and fixed in 4% formaldehyde for 3 days and decalcified in 10% EDTA, pH=7.5 for one month. Tissues were dehydrated with a graded series of ethanol washed and embedded in paraffin and sectioned to 5-um slices.
According to advice, we performed IHC staining of FASN, ELOVL5, ELOVL6, and SCD1 again in the superficial zone (articular cartilage) and the radial zone (calcified cartilage zone) of tibia plateau respectively (Fig. 4e). We also compared the relative expression of FASN, ELOVL5, ELOVL6, and SCD1 of two distinct zones respectively (Extended Data Fig. 7k).
8. Related to Fig. 1c software predicts the position information before and after the formation of circular RNA, and further identifies the candidate circRNA. The specific analysis process is as follows: after the bwa mem comparison, CIRI2 processes the sam file twice. For the first time, the junction reads are detected through the PCC (paired chiastic clipping) signal.
The candidate circRNAs were obtained by preliminary filtration using PEM (paired-end mapping) and GT-AG sequence characteristics. The second time, detect additional junction reads again and further filter out false positive candidate circRNAs. The schematic diagram of CIRI2 software is as follows: CIRC explorer 2 analyzes the reverse splicing information from the Tophat-Fusion comparison results, and annotates the prediction results of circular RNA according to the comparison results. The specific analysis process is as follow: first of all, we filter out the reads that cannot be mapped by Tophat-Fusion, and then use Tophat-Fusion mapping to map these reads to the genome. Secondly, the nonlinear candidate reads in the Tophat-Fusion comparison results are re-compared to the genome, and the position of the junction site is more accurately determined with the help of the gene annotation file. The schematic diagram of CIRC explorer 2 software is as follows: Hence, the above are the method we identify the CircRNAs.
Next, we would like to explain that the CircRNAs sequencing we published in Ann. 12. Some of the statistical methods they used are inappropriate or inaccurately described.
Especially, statistical analysis of nonparametric data such as histological grades should not have been conducted using 'two-side unpaired t-test' throughout the manuscript. The authors described that they used one-way ANOVA with LSD-t by assuming equal variance for multiple group comparisons. However, it says one-way ANOVA with Turkey's multiple comparison, which should be at least 'Tukey's HSD' although LSD and HSD use different significant difference. Overall, this reviewer cannot figure out the statistical method they used.
Response: Thanks for your suggestions to our manuscript. We have corrected the statistical method according to your advice.
Statistical analysis was performed using GraphPad Prism software (version 8.0). Data from two groups were analyzed by using unpaired, two-tailed Student's t test for RT-qPCR and the percentages of positive cells. One-way analysis of variance (ANOVA) followed by Tukey's HSD test is used for multigroup comparisons. For nonparametric data, the Mann-Whitney U test was used for OARSI grade, Synovitis score, and osteophytes number. Parametric data were presented as mean ± SD and the calculated 95% confidence intervals (CIs) for nonparametric data. Statistical significance was set at P < 0.05 (please see line 1130-1138).
We also described the method in the figure legends. Thanks again for your helpful comment.
13. Related to Fig. 6a, the rational for selecting MDM2 as a candidate is insufficiently provided.
Response: We are very sorry for the missing part of MDM2 selection. Herein, we appreciated this chance and are very pleasure to explain why we select MDM2 as the candidate. First of all, we utilized the UbiBrowser (http://ubibrowser.ncpsb.org.cn) to predict E3 ligase of FASN (Fig. 6i). STUB1 and MDM2 are the top2 E3 ligase of FASN ( Fig. 6i). To further explore the effects of STUB1 and MDM2 on FASN protein level. We overexpressed STUB1 and MDM2 in chondrocytes respectively and western blot analysis indicated that MDM2 significantly reduced FASN protein level compared to STUB1 overexpression (Fig. 6j). Next, we performed Co-IP assay to confirm the interaction between FASN and MDM2 (Fig. 6k). FASN and MDM2 co-localized in cytoplasm (Fig. 6l).
Hence, above all, we consider that MDM2 as a potential E3 ligase of FASN.
14. Related to mass spectrometry analysis of FASN in line 470-473, the method is completely missing and the explanation for MS data is not enough. Was the UPLC-MS/MS method performed focusing on detecting acetylation sites of FASN? Were K673 and K1065 the only acetylation sites on FASN?
Response: We are appreciated your careful review and we are sorry for the missing part of MS method. Herein, we have described the MS method in the revised manuscript. We used UPLC-MS/MS method to detect acetylation sites of FASN. The method of UPLC-MS/MS is as followed: Chondrocytes overexpressed FASN were collected and cell protein solutions were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis, followed by coomassie blue staining. At first, for in-gel tripic digestion, gel pieces were destained in 50 mM NH4HCO3 in 50% acetonitrile (v/v) until clear. Gel pieces were dehydrated with 100 μl of 100% acetonitrile for 5 min, the liquid removed, and the gel pieces rehydrated in 10 mM dithiothreitol and incubated at 56 °C for 60 min. Gel pieces were again dehydrated in 100% acetonitrile, liquid was removed and gel pieces were rehydrated with 55 mM iodoacetamide. Samples were incubated at room temperature, in the dark for 45 min. Gel pieces were washed with 50 mM NH4HCO3 and dehydrated with 100% acetonitrile. Gel pieces were rehydrated with 10 ng/μl trypsin resuspended in 50 mM NH4HCO3 on ice for 1 h. Excess liquid was removed and gel pieces were digested with trypsin at 37 °C overnight. Peptides were extracted with 50% acetonitrile/5% formic acid, followed by 100% acetonitrile. Peptides were dried to completion and resuspended in 2% acetonitrile/0.1% formic acid.
Then, UPLC-MS/MS method was performed to detecte acetylation sites of FASN.
The gradient was comprised of an increase from 6% to 23% solvent B (0.1% formic acid in 98% acetonitrile) over 16 min, 23% to 35% in 8 min and climbing to 80% in 3 min then holding at 80% for the last 3 min, all at a constant flow rate of 400 nl/min on an EASY-nLC 1000 UPLC system. The peptides were subjected to NSI source followed by tandem mass spectrometry (MS/MS) in Q ExactiveTM Plus (Thermo) coupled online to the UPLC. The electrospray voltage applied was 2.0 kV. The m/z scan range was 350 to 1800 for full scan, and intact peptides were detected in the Orbitrap at a resolution of 70,000. Peptides were then selected for MS/MS using NCE setting as 28 and the fragments were detected in the Orbitrap at a resolution of 17,500. A data-dependent procedure that alternated between one MS scan followed by 20 MS/MS scans with 15.0s dynamic exclusion.
Automatic gain control (AGC) was set at 5E4.
The resulting MS/MS data were processed by using Proteome Discoverer 2.4. The modification setting is acetylation of lysine. Carbamidomethyl on Cys were specified as fixed modification and oxidation on Met was specified as variable modification. Peptide confidence was set at high, and peptide ion score was set > 20.
Next, we would like to explain the result of UPLC-MS/MS result. We matched two acetylation sites of FASN in this assay. One is K673 site and the other is K1065. We should explain that two acetylation sites on FASN were detected in our chondrocytes model.