Abstract Patients suffering from rheumatoid arthritis (RA) and related autoimmune joint diseases exhibit cyclic episodes of resolution and exacerbation of joint inflammation, referred to as flares. Fibroblast-like synoviocytes (FLSs) that are epigenetically transformed by chronic inflammation are implicated as the orchestrators of these flares. In this study, we compared the cellular and molecular features of the FLSs during the inflammatory and resolution phases of RA progression. We performed histopathological evaluations of the joints from an inducible tumor necrosis factor-alpha (TNF-α) transgenic mouse model to reveal that phenotypic RA hallmarks including synovial hyperplasia, increased angiogenesis, and macrophage infiltration, were all reversed upon the initiation of resolution. However, the FLSs from the resolution phase joints exhibited a transcriptomic signature reminiscent of a highly inflammatory state. They exhibited a G0/G1 cell cycle arrest accompanied by reduced viability. In addition, factors secreted from the resolution FLSs, induced cell death, and decreased the angiogenic potential in human microvascular and umbilical cord endothelial cells. These findings indicate that the secretome of the resolution phase FLSs impairs endothelial cell function and suggest that understanding the interaction between the FLSs and endothelial cells during the resolution phase of RA is essential for achieving complete remission. Subject terms: Transcriptomics, Inflammation __________________________________________________________________ Fibroblast-like synoviocytes from the inflamed joints of the Tumor necrosis factor alpha-induced arthritis undergo transcriptomic changes that persist during the resolution phase, leading to perturbation of endothelial cell function. Introduction Fibroblast-like synoviocytes (FLSs) are a unique mesenchymal cell population that resides in the joint synovium. Under physiological conditions, the FLSs are essential for the maintenance of joint homeostasis^[36]1–[37]3. However, the FLSs become the main drivers of persistent inflammation in arthritic joints and play a critical role in the initiation, propagation, and recurrence of chronic autoimmune arthritic diseases such as rheumatoid arthritis (RA), psoriatic arthritis, and gout, among others^[38]4–[39]11. Epigenetic alterations have been shown to govern the disease-promoting inflammatory phenotype of FLSs during the development of these inflammatory joint diseases. They have been strongly implicated in providing a strong molecular bias for the increased expression of inflammatory genes, development of resistance to apoptosis, induction of cartilage degeneration, osteoclastogenesis, and aberrant synovial angiogenesis^[40]12–[41]19. During the onset phase of inflammatory arthritis (first hit), the naïve non-primed FLSs get activated under the influence of pro-inflammatory mediators. However, the recurrent inflammatory insults and danger sensing (second hit) lead to tissue priming, resulting in more pronounced and persistent arthritis. Apart from somatic mutations in mitochondrial and nuclear DNA, the FLSs in the inflammatory state of arthritis behave like imprinted aggressors by undergoing epigenomic reprogramming, encompassing chromatin remodeling, active histone marks, distinct methylation signature, and altered expression of miRNAs^[42]7,[43]16,[44]20–[45]23. However, there is currently insufficient data on the status of FLSs during the resolution state, which is a critical phase that the patients encounter between the first and second or the subsequent hits of the inflammatory phase. This resolution phase determines both the disease course and future treatment plan for autoimmune arthritis patients as it possesses all the hallmark epigenetic changes that result in the aggressive response towards recurrent inflammatory insults. The results from this study indicate that the aggressive phenotype of FLSs persists even after the inflammation has subsided and may abundantly co-stimulate and crosstalk with various other cells in proximity, particularly the endothelial cells. The role of chronic inflammation in increased angiogenesis and neovascularization has been established in multiple diseases, including various forms of cancers and autoimmune diseases such as RA^[46]24–[47]26. During RA progression, angiogenesis is considered an early event that initiates pannus tissue formation by the recruitment of leukocytes and induction of synovial hyperplasia^[48]27. Subsequently, the number and density of blood vessels in the RA synovium continue to increase and the crosstalk between the synovial tissue and the vasculature further drives the exacerbation of synovial hyperplasia^[49]28. Thus, angiogenesis influences the disease’s course on several levels^[50]29–[51]31. Moreover, when endothelial cells are activated by chronic inflammation, they produce inflammatory mediators and express surface molecules that potentially alter their proliferation and survival. These processes collectively contribute to the complex pathomechanisms of the RA joints^[52]32–[53]37. Therefore, angiogenesis inhibition therapy in combination with standard treatment of RA is expected to exhibit enhanced therapeutic potential. Recent studies report that the intercellular communication between FLSs and the endothelial cells critically regulates the pathophysiology of both cell types and altogether plays a role in exacerbating disease pathogenesis. RA FLSs that are adjacent to the vasculature have been shown to influence endothelial cell function and the angiogenic potential by both autocrine and paracrine mechanisms^[54]38–[55]41. Direct intercellular communication via NOTCH3 signaling was shown to establish vascular endothelium and the RA synovial fibroblast interaction^[56]38. Similarly, the secretion of extracellular vesicles loaded with micro RNAs from RA FLSs to influence the angiogenic properties of endothelial cells was also reported^[57]39. It was also suggested the endothelial cells enable the RA FLSs to migrate to distant locations^[58]40,[59]41. However, whether or not after the inflammation has subsided, the imprinted FLSs from the resolution phase retain their aggressive character to stimulate the vascular endothelial cells remains unknown. We, therefore, hypothesized that as a consequence of long-term inflammatory memory, the RA FLSs will likely continue to negatively influence the endothelial cells even during the resolution phase. Thus, our study focused on understanding and elucidating the mechanisms underlying the endothelial cell dysfunction induced by activated FLSs in the quiescent phase of inflammatory arthritis. Results Phenotypic changes in inflammation and angiogenesis were reversible in the resolution phase We utilized the previously described doxycycline-induced inflammatory arthritis mouse model (iTNF-tg) to mimic the homeostasis, inflammatory, and quiescent phases of disease progression^[60]42. Oral administration of doxycycline was shown to result in the ubiquitous production of human TNF and consequently, the development of inflammatory arthritis phenotype in those mice. Withdrawal of doxycycline treatment in this model has resulted in the resolution of the arthritic pathology^[61]22,[62]42,[63]43^, thus making them an ideal model to study the resolution phase of RA. We adapted this model to our study requirement. We treated 6-week-old iTNF-tg mice with doxycycline in 5% sucrose water for 3 weeks to induce the inflammatory phase. The resolution phase was established approach by a doxycycline withdrawal for 3 additional weeks following the inflammatory phase. The control mice that exhibited homeostatic (healthy) joints were obtained by treating them with 5% sucrose in their drinking water (Fig. [64]1A), Histologic analysis of the distal interphalangeal joints from the forepaws was performed. Toluidine blue staining of paw sections in the inflammatory revealed hyperplasia of synovial tissue as determined by increased cellularity in the synovial tissue regions adjacent to the joints. (Fig. [65]1B). Additionally, the considerable volume of pannus tissue also penetrated the adjacent bone tissue. Histomorphometric quantification of the number of cells in the synovial tissue adjacent to the joints indicated a 2.5-fold increase in the cell numbers upon inflammation, whereas the area of destaining cartilage showed a robust 20-fold increase compared to the homeostatic joints (Fig. [66]1B, C). According to our expectations, withdrawing doxycycline reversed the progression of these defects and led to improvement in the histological and morphometric parameters of the synovium and articular cartilage, to the extent that was not significantly different from the homeostatic mouse joints (Fig. [67]1B, C). We next confirmed that the resolution of inflammation indeed resulted in the reduced infiltration of the macrophage and myeloid lineage cells by performing immunohistochemistry for F4/80 protein. Our data show an increased presence of F4/80 positive cells adjacent to inflamed joints but not the homeostatic and resolved phase joints (Fig. [68]1D). Similar results were also obtained regarding the changes in endothelial cell density adjacent to the joints. The immunohistochemical staining pattern of an EC marker, CD31 (also referred to as platelet endothelial cell adhesion molecule 1, PECAM-1), among the three groups of mice revealed indicated increased angiogenesis adjacent to inflamed joints, but not in the resolved joints (Fig. [69]1E). Together these data demonstrate that the exacerbation of inflammation and angiogenesis were effectively resolved during the resolution phase of joint degeneration in the iTNF-tg mice. Fig. 1. Inflammation in distal interphalangeal joints of iTNF-tg was reversible at the histopathological level. Fig. 1 [70]Open in a new tab A Schematic drawing describing the experimental design of doxycycline-inducible human TNFα–transgenic (iTNF-tg) arthritis mice model. Inflammation phase mice were treated with 1 mg/ml doxycycline and 5% sucrose for 3 weeks; Resolution phase mice were treated with 1 mg/ml doxycycline and 5% sucrose for an initial 3 weeks, after which the antibiotic was removed for the next 3 weeks; The homeostasis group included untreated control mice which were kept on only sugar (5% sucrose) water for 3 weeks. B Images of toluidine blue staining. Synovitis (asterisks) and cartilage destaining (arrow) were observed in the inflammation group. Scale bar = 100 μm. C Quantification of inflammation by estimation of No. of synoviocytes/10^6 sq.µm (left) and cartilage destaining (right) by histomorphometry (n = 5). ∗∗ p < 0.01; ∗∗∗∗p < 0.0001. Values in C are shown in mean ± SD. D Representative images of F8/40 and (E) CD31 immuno-stained sections. Scale bar: 100 μm. Samples were counterstained with Fast Green FCF. Transcriptomic changes acquired by the FLSs under inflammatory conditions persist despite histopathological resolution of disease progression FLSs are known to be epigenetically primed due to chronic inflammation leading to the development of a long-term memory of the inflammatory insult within their chromatin^[71]16,[72]21,[73]22. Therefore, our next goal was to determine whether the FLSs in the resolution phase joints, which exhibited no signs of inflammatory pathology at the histological level in the iTNF-tg mouse model were also resolved at their transcriptomic level. We isolated synoviocytes from the paw joints of the control and inflammatory and resolution phases of iTNF mice (Fig. [74]1A). To identify the various cell populations, we performed flow cytometry analysis of the lice cells isolated synoviocytes cultures utilizing PDGFRA (pan-fibroblast marker), CD45 (pan-hematopoietic cell marker), and CD31 (pan-endothelial cell marker). Our data revealed that these cultures contained less than 3% of CD45+ and less than 6% of CD31+ cells and were predominantly enriched for PDGFRA+ cells indicating their fibroblastic phenotype (Fig. S[75]1). We performed bulk RNA-sequencing of these FLS-enriched cultures to compare the transcriptomic signatures of the homeostatic, inflammatory, and resolution phases. Principal component analysis (PCA) showed that the 3 groups were distinct in the first three principal component dimensions whereas samples of the same group clustered together, suggesting unlike histological observations the gene expression profiles of the resolution phase FLSs were distinct from that of the homeostatic FLSs (Fig. [76]2A). Pearson correlation analysis of all expressed genes revealed that the transcriptomic signature of the resolution phase FLS was closer to the inflammatory phase than that of the homeostasis phase (Fig. [77]2B). We then performed differential gene expression analysis of the 36,930 expressed genes with signal intensity greater than or equal to 5 and absolute fold-change greater than 2.0. Analysis of volcano plots of both inflammatory vs homeostasis groups and resolution vs homeostasis groups revealed differential expression of several thousands of genes between these conditions (Fig. [78]2B, C). Our analysis revealed that 1077 of the differentially expressed genes (DEGs) were common to both inflammatory and resolution phase conditions (Supplementary Data [79]1). Interestingly, hierarchical cluster analysis of normalized gene expression of the commonly deregulated genes in the inflammatory and resolution phases compared to homeostasis revealed that a majority of the transcriptomic changes acquired during the inflammatory phase were not reversible, as approximately 52.92% of upregulated genes and 40.85% of downregulated genes in the inflammation group showed similar trends of differential expression even after turning off the doxycycline treatment. Meanwhile, only 5.94% of the upregulated and 0.27% of downregulated genes upon inflammation were restored to their homeostatic expression level (Fig. [80]2D). We next evaluated the clinical relevance of our findings by comparing 1077 DEGs from the iTNF-tg FLSs with the DEGs identified in our previous study comparing human RA FLS and OA FLSs ([81]GSE217012)^[82]44. Despite the likely differences in culture conditions and disease states of the human FLS, we were able to identify 113 overlapping genes (Fig. [83]2E). Our next goal was to predict the upstream transcription factors that likely regulate the expression of the 113 overlapping genes that are relevant to human RA. We performed an upstream transcription factor binding prediction analysis by ChEA3 using the ENCODE ChIP-seq database. We found that STAT1/2 family transcription factors with a well-established role in FLSs transformation as a top hit. In addition, we also found the Ikaros family of zinc-finger protein, IKZF2, whose role in FLSs pathophysiology remains to be determined as a potential upstream regulator (Fig. [84]2F). Reactome Pathway analysis to predict the biological roles of the 113 overlapping genes identified interferon signaling, vitamin and co-factor metabolism, and activation of the complement system as the most significantly dysregulated pathways (Fig. [85]2G). Thus, the differential gene expression that persists in the resolution phase of iTNF-tg mouse FLSs is clinically relevant to the molecular pathology of human RA FLSs. Fig. 2. The transcriptomic signature of the FLS from the iTNF-tg was largely irreversible. [86]Fig. 2 [87]Open in a new tab A Principal component analysis (PCA) of normalized gene expression in the primary FLS showing segregation of inflammation, resolution, and homeostasis groups. B Pearson correlation average linkage hierarchical cluster analysis of normalized gene expression. C Genes with differential expression levels (DEGs) greater than twofold (false discovery rate [FDR] p-value < 0.05) were visualized as volcano plots. D Clustergrammer was used to visualize the clustering of normalized gene expression (logCPM) of the DEGs from inflammation, resolution, and homeostasis. E Venn diagram showing the overlap between upregulated DEGs in human RA FLS vs OA FLS RNA-seq data set and the DEGs commonly upregulated to inflammation vs homeostasis and resolution vs homeostasis groups. F Top 5 transcription factors predicted from the ENCODE ChIP-seq library database regulating the expression of the overlapping gene set in panel E. Data was visualized using ChEA3. G Top 5 Reactome pathways enriched in the overlapping gene set in panel E. DEGs common to inflammation and resolution phase were primarily associated with cell cycle regulation Next, we predicted the potential cellular mechanisms regulated by 1077 commonly deregulated genes that may result in persistently pathological gene expression in the resolution phase. We performed Reactome pathway analysis by the Enrichr, an integrative web-based gene-list enrichment analysis tool of the upregulated and downregulated genes underlying the different phases of arthritis (inflammatory and resolution) in iTNF-tg FLSs. The most critical upregulated pathways were related to the immune system, cytokine signaling, glutathione conjugation, interferon signaling, and interleukin-18 signaling (Fig. [88]3A). The most significant downregulated pathways were related to cell cycle and cell cycle checkpoints (Fig. [89]3B). A subset of the genes exhibiting the most significant differential expression that were identified in the Recatome pathway analysis (Fig. [90]3A, B) was further validated by real-time qRT-PCR in additional preparations of FLS primary culture. Our data show a high degree of correlation between the fold changes observed in the RNA-seq and qRT-PCR experiments in both the inflammatory state and resolution state (Fig. [91]3C, D). Taken together, these data confirm that the histological resolution of inflammation does not ensure the molecular resolution of inflammation. Signaling pathways that support a pro-inflammatory phenotype remain activated in the resolution phase FLSs despite the apparent phenotypically healthy joint structures. Fig. 3. Molecular Pathway enrichment prediction analysis of the DEGs in FLSs from different phases of arthritis. [92]Fig. 3 [93]Open in a new tab A Top 10 upregulated canonical pathways among different groups. B Top 10 downregulated canonical pathways among different groups. C Quantitative real-time PCR (RT-qPCR) validation of selected differentially expressed genes (DEGs) detected between inflammation and homeostasis groups. Linear regression analysis between gene expression ratios derived from RNA-Seq and RT-qPCR data. Significant difference at p ≤ 0.05. D Quantitative real-time PCR validation of selected DEGs detected between resolution and homeostasis groups. Linear regression analysis between gene expression ratios derived from RNA-Seq and RT-qPCR data. Significant difference at p ≤ 0.05. Resolution phase FLSs showed exacerbated deregulation of genes associated with cytokine and innate immunity responses The Reactome pathway analysis predicted cytokine responsiveness as one of the significantly dysregulated pathways in the inflammatory and resolution phase FLSs (Fig. [94]3A). We therefore delineated the phase-specific changes in cytokine responses. Among the 1077 common DEGs from the inflammatory and resolution phase FLSs, we delineated a subset of 64 DEGs based on their co-occurrence with the term ‘cytokine’ in literature-supported statements describing functions of genes from the Gene Reference Into Function (GeneRIF) biological term annotations on Harmonizome 3.0^[95]45,[96]46 (Fig. [97]4A). Our analysis revealed that the upregulated cytokine genes included secreted pro-inflammatory factors such as interleukins Il7 and Il18, growth factors Igf2, Hgf, and Fgf2, complement system protein-coding genes, Cfh and C3, and damage-associated molecular pattern protein S100a8. The downregulation of pro-anabolic factor genes belonging to the BMP family Bmp4, Bmp7, and Gdf5 may additionally contribute to the pathology of the resolution phase FLSs. Interestingly the pro-resolution mediator GAS6 remained upregulated in the resolution phase suggesting a potential coping mechanism against acute inflammation (Fig. [98]4B). Fig. 4. Expression of cytokine and adaptive-immune inflammatory genes. [99]Fig. 4 [100]Open in a new tab A Heat map of normalized expression of genes co-occurring in literature reports with term “cytokine” in GeneRIF database. B Normalized expression changes in RNA-seq of upregulated and downregulated ligand coding genes. Two-way ANOVA and Tukey’s multiple comparisons test were used to determine statistical significance. *p  <  0.05, **p  <  0.01, ***p  <  0.001, ****p  <  0.0001. C Gene expression analysis quantified by real-time PCR (qPCR) in FLSs treated with or without TNF. Gene expression levels were normalized to the housekeeping gene (Tbp). Data are expressed using the 2^−∆∆Ct method. Statistical analyses were performed using unpaired Student´s t-test to identify differences between groups. *p  <  0.05, **p  <  0.01, ***p  <  0.001, ****p  <  0.0001. It was previously reported that epigenetic priming of the FLSs, by a primary exposure to inflammation (first hit), predisposes them to an exacerbated response toward a secondary exposure to inflammatory conditions (second hit). We, therefore, isolated inflammatory, and resolution phase FLS from the iTNF-tg mice that were already exposed to a first hit of inflammatory insult, followed by a secondary treatment with recombinant human TNF (rhTNF) to mimic the second hit of inflammatory insult to recapitulate the recurrent flares in RA patients. We analyzed the resulting changes in the expression of a panel of genes involved in the innate immunity response by qRT-PCR. The expression of the innate immunity regulating cytokines including Tnf, Cxcl10, Mip1a, and Il17a was significantly upregulated in response to the rhTNF treatment in both inflammatory and resolution phase FLSs, whereas the healthy FLSs from the homeostasis condition showed a muted response to rhTNF treatment (Fig. [101]4C). We also evaluated anti-inflammatory innate immunity gene expression in response to the secondary TNF stimulation. Bmp4, which was previously reported to be downregulated in the RA FLS, remained repressed in the resolution phase FLSs. However, immunomodulatory factor genes, such as Il10 and Tgfb were significantly upregulated (Fig. [102]4C). The upregulation of the immunomodulatory factors suggests that the resolution phase FLSs, despite the exacerbated expression of pro-inflammatory mediators, attempted to counteract inflammation by increasing the production of immunomodulatory factors. Resolution phase FLSs exhibit reduced viability, cell death induction, and G0/G1-phase arrest Another significantly altered biological pathway associated with the differentially expressed genes was related to cell cycle and cell cycle checkpoints (Fig. [103]3A). To provide functional evidence for the alteration in the predicted deregulation of cell cycle pathways, we monitored the FLSs from all three groups for viability, proliferation, and apoptosis at 24, 48, 72, and 96 hours of culture. Although cell numbers increased with time in all three groups, we could see a significant decline in cell viability, particularly in the resolution FLS group (Fig. [104]5A, B). We further found that the reduced viability of resolution phase FLSs might be due to an increase in the apoptotic cells as indicated by an increase in the sub-G0/G1 phase DNA content (Fig. [105]5C and Fig. S[106]2). Our cell cycle analysis also revealed a significant time-dependent decrease in the percentage of the G0/G1 phase (Fig. [107]5D and Fig. S[108]2) and a significantly decreased percentage of the S and G2/M phases after 72 hours in DOX-OFF FLSs (Fig. [109]5E and Fig. S[110]2). Collectively, these data corroborate the results from the pathway prediction analysis regarding the persistence of cell cycle perturbation that continues from the inflammatory to the resolution phase of the FLSs. Fig. 5. Cell viability and cell cycle analysis of FLSs at different phases of arthritis. [111]Fig. 5 [112]Open in a new tab A DAPI-based nuclear staining assay quantification by imaging cytometry to assess changes in FLSs’ number at 24, 48, 72, and 96 h among different groups. B Changes in FLSs’ viability at 24, 48, 72, and 96 h among different groups. C Increase in the percentage of sub-G0/G1 phase FLSs from resolution phase at 24, 48, 72, and 96 h by imaging cytometry. D Decrease in the percentage of G0/G1 phase FLSs from resolution phase FLSs at 24, 48, 72, and 96 h by imaging cytometry. E Decrease in the percentage of S and G2/M phase FLSs from resolution phase at 72 and 96 h by imaging cytometry. FLSs in the resolution phase impair endothelial cell tube formation and proliferation The in vitro formation of capillary-like tubes by endothelial cells on a basement membrane matrix is a powerful method to screen for various factors that promote or inhibit angiogenesis^[113]47. Since the resolution phase FLSs expressed various secreted factor genes that could potentially influence angiogenesis, we utilized the human umbilical vein endothelial cells (HUVECs) as a model system to perform the in vitro angiogenesis assay. Treatment of HUVECs with the conditioning medium from FLSs under the inflammation phase did not show any significant effect on the angiogenic potential as compared to the homeostasis group (Fig. [114]6A, B). However, the treatment of HUVECs with a resolution phase conditioning medium decreased the number of capillary-like tube structures and resulted in broken and shortened tubes compared with the control group. Also, there was a significant reduction in the number of nodes, segments, junctions, and meshes in HUVECs treated with the factors secreted from the resolution phase FLSs (Fig. [115]6A, B). Thus, our results implied that resolution phase FLSs may possess the ability to alter the cellular properties of the endothelial cells. Additionally, in an independent set of experiments, we incubated HUVECs with 20% conditioning medium from the FLSs of each group and quantified cell viability and the cell cycle progression based on their DNA content. A time-dependent and increasing trend was observed in cell numbers among the HUVECs treated with the conditioning medium from the FLSs of each group with no significant differences between groups (Fig. [116]6C). However, we observed a significant reduction in the viability at the initial 24-hour time point, which was rescued at late time points of 3- and 4-days post-treatment in the resolution phase FLS conditioning treatment condition, as indicated by increased cell viability (Fig. [117]6D). Next, we performed an adherent cell imaging cytometry-based analysis of the cell cycle phases to understand these dynamic changes observed in terms of viability. We observed that resolution phases of the FLSs conditioning medium induced a significant accumulation of HUVECs with a low molecular weight DNA content at 24 h, lower than in the G0/G1 phase, indicative of apoptotic cell death (sub-G0/G1 phase HUVECs) (Fig. [118]6E and Fig. S[119]3). This finding was associated with a reduction in the G0/G1 phase cells, further supporting the decreased cell viability observed at 24 h (Fig. [120]6D). Finally, rescue in cell viability is reflected by the increase in the S and G2/M phase HUVECs at 4 days post-treatment with resolution phases FLS conditioning medium (Fig. [121]6E and Fig. S[122]3). Thus, FLSs belonging to the resolution phase of arthritis may likely decrease cell viability and arrest cell cycle progression by inducing apoptosis and G0/G1 phase arrest in the endothelial cells. Fig. 6. Effect of FLS secreted factors on HUVEC cellular properties. [123]Fig. 6 [124]Open in a new tab A HUVECs were seeded on extracellular matrigel and exposed to 20% FLSs conditioned media for in vitro tube formation assay. Representative bright field micrographs of extracellular matrigel angiogenesis after 4 h of stimulation with 20% FLS conditioned media. These images were used to perform analysis with the Angiogenesis Analyzer for ImageJ. Scale bar: 200 μm. B Angiogenic parameters quantified with Angiogenesis Analyzer were reported as a measurement of tube forming ability: total tube length, number of nodes, number of segments, segment length, and total isolated branch length (n = 3, in triplicate). One-way ANOVA and Dunnett’s multiple comparisons test were used to determine statistical significance. *p  <  0.05, **p  <  0.01, ***p  <  0.001, ****p  <  0.0001. C Changes in HUVECs number at 24, 48, 72, and 96 h after incubation with 20% FLSs conditioned media among different groups. D Changes in HUVECs viability after incubation with 20% FLSs conditioned media among different groups at 24, 48, 72, and 96 h. E An early increase in the percentage of sub-G0/G1 phase HUVECs after incubation with conditioning medium of resolution phase FLSs at 24 h. A decrease in the percentage of G0/G1 phase HUVECs after incubation with secretome of resolution phase FLSs at 24 h; (n = 3, in triplicate). Two-way ANOVA and Tukey’s multiple comparisons test were used to determine statistical significance. *p  <  0.05, **p  <  0.01, ***p  <  0.001, ****p  <  0.0001. Microvascular endothelial cells are a major constituent of the synovial vasculature that play essential roles in homeostasis, as well as inflammatory pathology of the joints^[125]48. We utilized the human dermal microvascular endothelial cells (HDMECs) as a model to validate the negative effects of the resolution phase FLSs on microvascular endothelial cells tube formation. We co-cultured HDMECs with the FLSs by plating the HDMECs on a Matrigel matrix at the bottom and FLSs on transwell inserts to facilitate the continuous exchange of the secretome between the two cell types. Our data show that compared to homeostasis phase the inflammatory phase FLSs promoted a trend towards increased tube formation as indicated by an increase in the number of branches (p = 0.06) and total branch length (p = 0.04) (Fig. [126]7A, B). Similar to the HUVEC cells, the co-culture of HDMECs with the resolution phase FLSs decreased the number of capillary-like tube structures and resulted in broken and shortened tubes compared with the homeostatic group. Also, there was a significant reduction in the number of junctions, extremities, branches, nodes, and total branch length in HDMECs co-cultured with the resolution phase FLSs (Fig. [127]7A, B). Thus, our results implied that resolution phase FLSs may possess the ability to inhibit the angiogenic potential of the macro and microvascular endothelial cells. Fig. 7. Effect of FLS secreted factors on HDMEC tube formation. [128]Fig. 7 [129]Open in a new tab A HDMECs were seeded on extracellular matrigel and co-cultured with FLSs for 4 h for in vitro tube formation assay. Representative fluorescence micrographs of extracellular matrigel angiogenesis treated with calcein AM prior to assay. These images were used to perform analysis with the Angiogenesis Analyzer for ImageJ. Scale bar: 200 μm. B Angiogenic parameters quantified with Angiogenesis Analyzer were reported as a measurement of tube forming ability: number of junctions, number of extremities, number of branches, number of nodes, and total branch length (n = 3, in triplicate). One-way ANOVA and Dunnett’s multiple comparisons test were used to determine statistical significance. *p  <  0.05, **p  <  0.01, ***p  <  0.001, ****p  <  0.0001. FLSs in the resolution phase inhibit angiogenesis and promote pro-inflammatory gene expression by endothelial cells We assessed the potential interaction between FLSs and the molecular features of the endothelial cells by studying the effects of factors released by FLSs on HUVECs by analyzing a representative set of pro-angiogenic inflammatory genes which were previously reported to be altered by inflammatory cytokines in endothelial cells^[130]49,[131]50. Interestingly, we observed that there was an upregulation in the expression levels of pro-inflammatory genes (ACE, MEST, CSF1, CXCL1) and a downregulation of pro-angiogenic genes (BMP4, MMP10, SELE, CCL2) in HUVECs treated by supernatant of resolution phase FLSs as compared to the control group (Fig. [132]8). Although not statistically significant, the mean axis fold-change of pro-inflammatory gene RELB showed a higher trend in HUVECs treated by supernatant of resolution FLSs as compared to the homeostasis group. Also, the mean axis fold-change of pro-angiogenic genes PIK3CD, ICAM1, and VCAM1 were not statistically significant but showed an upregulated gene expression in HUVECs treated by conditioning medium from inflammatory phase FLSs as compared to the control group. Collectively, our data from the experiments on the effects of FLS conditioning medium on HUVEC cellular and molecular properties revealed that the FLSs from the resolution phase secrete factors that possess the potential to impair endothelial cell function during the resolution phase of RA. Fig. 8. Pro-angiogenic and pro-inflammatory gene expression levels in endothelial cells treated with FLS conditioned media. [133]Fig. 8 [134]Open in a new tab Gene expression analysis quantified by real-time PCR (qPCR) in HUVECs stimulated by 20% FLSs conditioned media. Gene expression levels were normalized to the housekeeping gene (TBP). Data are expressed using the 2^−∆∆Ct method, and statistical significance was calculated by Ordinary one-way ANOVA and Dunnett’s multiple comparisons test. *p  <  0.05, **p  <  0.01, ***p  <  0.001, ****p  <  0.0001. Endothelial cells and FLSs communicate via a distinct cell communication pattern in the resolution phase of the K/BxN serum transfer RA mouse model Our in vitro data suggest that the FLSs from the resolution phase secrete factors that impair endothelial cell functions including angiogenesis and proliferation (Figs. [135]6 and [136]7). Our next goal was to compare the in vivo molecular signatures of the intercellular communication between FLSs and endothelial cells during the various states of RA progression and resolution. For this, we utilized a published single-cell RNA-seq data set ([137]GSE230145^[138]51) consisting of CD45 negative synovial cell populations from homeostasis (control), peak (inflammation), and resolved (resolution) phases of the K/BxN serum transfer RA mouse model. We first subjected the CD45 negative synovial cell populations to UMAP cell clustering analysis based on previously described gene signatures^[139]38,[140]51. Fibroblast-1 (Cd34, Cxcl5 and Il1rl1; sublining fibroblasts), Fibroblast-2 (C3, Cxcl14, Col1a1high sublining fibroblasts), Fibroblast-3 (Prg4high and Htra4) and pan-endothelial cells (Pecam1) and pan-fibroblasts (PDGFRA). (Fig. [141]9A). We next used the CellChat program to identify the differentially expressed ligand-receptor pairs by each cell cluster, followed by the prediction of cell communication patterns^[142]52. Dot plot representation of the incoming and outgoing cell communication pattern shows that the fibroblast subtypes and the endothelial cells exhibit distinct communication patterns at each stage of the disease. In particular, the endothelial cells received incoming signals from ligands of the SEMA3, TGFb, VISFATIN, VEGF, and KIT pathways. Additional signals from the IL2 were received during the inflammatory phase. However, during the resolved phase endothelial cells only received ligand inputs from SEMA3 and KIT pathways. These analyses revealed that the endothelial cells communicated with other cells by far fewer ligand-receptor interactions during the resolution phase compared to the homeostasis and inflammatory phases. These deregulated cell communication patterns could explain the reduced angiogenic potential of the in vitro endothelial-FLS interaction experiments (Fig. [143]9B, C). Further, we observed that endothelial cells and well as fibroblasts communicated with multiple other cell types in the joints, where the sub lining Fibroblast-2 subtype was the most interactive cell type during the resolution phase, likely driving their pathophysiology (Fig. [144]9D). Finally, we found that during the resolution phase, the endothelial cells received Sema3c ligands from Fibroblast-1, 2, and 3 subtypes as well as the chondrocytes. In turn the endothelial cells communicated with the Fibroblasts through the Tgfb pathway ligands. TGFb-mediated communication was strongest between the endothelial cells and Fibroblast-2 subtypes (Fig. [145]9E). Thus, the distinct cell communication pattern between the FLSs and endothelial cells may play an important role in defining cell behavior during the various pathological states of RA. Fig. 9. Intercellular communication patterns in K/BxN serum transfer-induced arthritis mouse model. [146]Fig. 9 [147]Open in a new tab A UMAP plots to visualize cell clusters in published scRNA-seq from the synovial tissue of K/BxN serum transfer-induced arthritis (GSE 230145). B, C Dot plot of the differentially expressed incoming and outgoing ligand-receptor interactions in the defined cell clusters using CellChat. D Circos plot of cell-cell interaction networks driven by endothelial and fibroblast subsets. E Circos interaction plot of the top incoming and outgoing signaling pathway communications between endothelial cells and fibroblasts. F A schematic showing the persistent aggressive phenotype of fibroblast-like synoviocytes (FLSs) in the resolution phase, which not only promotes inflammatory cytokine production and cell cycle perturbations but also triggers impaired endothelial cell functioning. Discussion Sustained remission without drug treatment has not yet been accomplished for a large proportion of RA patients^[148]53–[149]56. Among the primary contributors underlying these unmet accomplishments are (1) an incomplete understanding of the differences in the cellular and molecular profiles of the systemic and joint resident cells during the clinical remission or quiescent phase compared to that of healthy individuals and (2) the inability of the disease-modifying biological and small molecule drugs to completely reverse the disease-driving features back to a healthy state. In this regard, several recent multi-omics-based studies have started to address these critical gaps in knowledge. In particular, they revealed a role for specific subsets of FLSs, synovial macrophages, and B cells that attain a primed molecular state and maintain altered cell physiology, even when the inflammatory triggers are effectively inhibited by the disease-modifying therapies^[150]20–[151]22,[152]57. In this study, we performed a side-by-side comparison of the cellular and molecular properties of the FLSs derived from healthy, inflammatory, and recovered joints to obtain a better understanding of not only the FLS properties but also their potential effects on endothelial cells which they closely interact with (Fig. [153]9F). We utilized a previously described inducible TNF transgenic mouse that reliably models the cyclic phases of exacerbated and resolved states of RA^[154]22,[155]42. Our data show that the resolution phase FLSs are not merely quiescent cell populations that are epigenetically primed to be in a ready-to-respond state. They not only exhibit distinct transcriptomic and cellular features of an active disease state but also secrete factors that disrupt the physiology of the adjacent tissue-resident cells, such as the endothelial cells. Our data show that the phenotypic characteristics of the inflammatory phase including, synovial hyperplasia, cartilage degeneration, macrophage infiltration, and pannus formation were reversed by stopping the expression of the human TNF transgene. We also found that the invasion of the joint space and the synovial cavity by the activated FLSs was abrogated upon stopping the expression of the human TNF transgene (Fig. [156]1). However, our transcriptomic characterization of the FLSs showed that the molecular changes occurring during the inflammatory phase were incompletely reversed upon abrogating TNF transgene expression (Figs. [157]2 and [158]3). The resolution phase FLSs persistently produced a variety of pro-inflammatory cytokines in the resolution phase (Fig. [159]4). These results suggest that a high degree of inflammatory epigenetic priming occurs in the FLSs in the iTNF-tg mouse model. In this study, we established the resolution phase of RA by abrogating TNF expression for 3 weeks. It could be argued that prolonging the resolution phase longer than 3 weeks might mitigate the pro-inflammatory nature of the resolution phase transcriptome. However, our experimental conditions for the resolution phase consisted of 3 weeks of in vivo doxycycline withdrawal period in addition to an extended period of primary cell culture, for up to 3 passages without any inflammatory stimulus. Thus, it is unlikely that a prolonged withdrawal of inflammatory stimulus might have resulted in the return of the resolution FLSs into a homeostatic state and that the molecular states established in vivo were not completely erased by in vitro culture conditions. Moreover, our comparison of persistent transcriptomic changes in our mouse model with human RA FLS predicted JAK/STAT signaling mediated transcription as an upstream regulator of the resolution phase FLS transcriptome (Fig. [160]2F). Interferon signaling and complement activation pathways were predicted as the persistently activated downstream signaling pathways in the resolution phase FLSs (Fig. [161]2G). Interestingly, STAT family transcription factors, as well as the interferon and complement activation pathways were previously reported to be involved in the establishment of epigenetic and transcriptomic memory in the FLSs and immune cells in cancer and autoimmune diseases. Thus, our data are highly relevant to understanding the epigenetic and transcriptomic properties of the FLSs during the resolution phase of RA^[162]22,[163]58–[164]61. It has been recently reported that specialized pro-resolution mediators which are the metabolites of the omega-3 fatty acids play a vital role in the resolution of inflammation and the re-establishment of homeostasis tissue in diseases including RA^[165]62,[166]63. However, our transcriptomic data suggest that the production of pro-inflammation lipid mediators, such as prostaglandins^[167]64 is increased in the resolution phase because of the upregulation of Ptgs1 and Ptgs2 (encoding for cyclooxygenase 1 and 2 enzymes, Fig. [168]4A). Additionally, the production of pro-resolution lipid mediators is likely reduced in the resolution phase due to the downregulation of the enzyme Arachidonate 15-Lipoxygenase (Alox15, Fig. [169]4A), which catalyzes the production of Maresin a potent pro-resolution mediator^[170]65. Thus, our data also suggest that resolution phase FLSs may not participate in lipid-mediated pro-resolution mechanism. In contrast, we found that the growth arrest specific 6 (Gas6) ligand, a pro-resolution mediator reported to play a role in hematoma resolution and neuroinflammation^[171]66 was persistently upregulated in the FLSs of the inflammatory and resolution phase (Fig. [172]4A, B). These data support the notion that enhancing GAS6 signaling is a potential therapeutic for boosting the healthy resolution of inflammation in the RA joints^[173]67. It has been previously reported that primary FLSs from RA patients are resistant to programmed cell death, which is induced by defective apoptotic stimulus signals which further contributes to the abnormal synovial hyperplasia, pannus formation, and inflammatory cell infiltration, resulting in joint destruction and functional loss^[174]68–[175]71. Our transcriptomic analysis suggests that cell cycle pathways are downregulated in the resolution phase accompanied by increased apoptosis (Fig. [176]5). In addition, we also observed a lack of joint cavity invasion and synovial hyperplasia in the joints during the resolution phase (Fig. [177]1). In addition, transcriptomic data from resolution phase FLSs did not reveal significant change in the expression Fibroblast Activation Protein alpha (Fig. S[178]4), a marker of fibroblast invasiveness^[179]72. Taken together these results suggest that the downregulation of cell cycle pathways increased apoptotic ability coupled with reduced invasiveness of the resolution phase FLSs may contribute to the histopathological recovery of the joints. Importantly, our data demonstrate that the reversal of certain pathological cellular properties does not ensure the reversal of the molecular features that sustain a pro-inflammatory pathology and potentially explain the discrepancies between the establishment of phenotypic or clinical symptomatic remission and the lack of molecular remission^[180]53,[181]73. Angiogenesis is a critical event in the pathogenesis of RA, where the vasculature is constantly reformed to accommodate the growing needs of the hyperplastic synovial tissue^[182]74–[183]78. In physiological conditions, a dynamic balance between pro- and anti-angiogenic factors regulates vascular density in the synovium. In contrast, in the RA synovial microenvironment, an imbalance between the levels of angiogenic mediators and inhibitors leads to increased capillary formation. Therefore, targeting pro or anti-angiogenic factor balance remains a promising RA therapeutic strategy^[184]31. However, translation of angiogenesis targeting drugs from pre-clinical to the clinical setting is challenging^[185]79 likely because of gaps in knowledge regarding the source of the angiogenic factors in the RA joints and the appropriate timing for therapeutic intervention. Our studies suggest that targeting the secretion of angiogenesis factors during the resolution phase FLSs may be beneficial in preventing recurrent flares. Our in vitro results showed resolution phase FLSs secretome contains factors that inhibit angiogenesis (Figs. [186]6 and [187]7). Our data also show that the secretome of resolution phase FLSs downregulated pro-angiogenic genes such as BMP4, SELE, MMP10, and CCL2 in HUVECs (Fig. [188]8). Various studies suggested that BMP4 regulates the formation of blood vessels during embryonic development and adult pathophysiological conditions^[189]80–[190]83. Similarly, the expression of E-selectin, a protein encoded by SELE, is strongly related to the events of angiogenesis^[191]84,[192]85. A recent study suggested that mesenchymal stem/stromal cells (MSCs) that are genetically modified to overexpress E-selectin could increase angiogenesis and enhance ischemic injury healing^[193]86. MMP-10 also can regulate the migration and invasion of tumor cells and the development of endothelial cell tubes^[194]87, whereas, CCL2 was reported to promote endothelial cell proliferation, migration and angiogenesis through the MAPK/ERK1/2/MMP9, PI3K/AKT, Wnt/β‑catenin signaling pathways^[195]88. Interestingly, our analysis of published single-cell RNA-seq data from the KBxN serum transfer RA mouse model supported our in vitro findings that the resolution phase endothelial cells may have impaired angiogenic potential. Cell Communication analysis predicted that the number of ligand-receptor interactions occurring at the endothelial cell surface is the lowest during the resolution phase compared to homeostasis and resolution (Fig. [196]9A, B). This bioinformatic analysis also revealed that endothelial cells and the FLSs may communicate via semaphorin 3 (SEMA3) signaling pathways (Fig. [197]9E). SEMA3 signaling pathway primarily acts as an anti-angiogenic pathway, SEMA3s suppress VEGF-induced angiogenesis by competing with VEGF for binding to Nrp1^[198]89. SEMA3A enhances vascular permeability, decreases endothelial cell proliferation, and promotes apoptosis even in the absence of VEGF, indicating that Sema3A activates its own signaling pathways, which is consistent with its independent role on Nrp1^[199]90. SEMA3A also inhibits integrin activity, which reduces endothelial cell adhesion and migration^[200]91. However, numerous studies indicated semaphorins can either stimulate or inhibit angiogenesis depending on the receptor they interact with, whether it is a transmembrane or secreted protein, and which signaling pathways are triggered. Hence, they are attractive therapeutic targets for regulating angiogenesis in autoimmune diseases^[201]92. Thus, the resolution phase FLSs may primarily disrupt the balance between the pro- and anti-angiogenic activity required for re-establishing synovial homeostasis. In summary, our study reports the dynamic nature of the cellular and molecular changes during the cyclic phases of RA flare pathology. Our results strongly suggest that addressing the detrimental effects of FLSs’ transformation on endothelial cell function is critical for treating flares and achieving drug-free molecular remission. One of the limitations of this study is the utilization of primary FLS as a model. Although we were able to extract clinically relevant conclusions regarding the unique features of the resolution phase of FLSs, future investigations using multi-omics and single-cell platforms may be essential to elucidate the in vivo importance of cell-cell communications (both autocrine and paracrine interactions) within the RA synovium at various cyclic phases of disease progression. Establishing the effect of resolution phase FLSs on immune cells and the various joint resident cells is an additional aspect warranting future investigations. The interaction between the endothelial cells and FLSs is highly dynamic and stage-specific^[202]33. Therefore, it can be expected these interactions would be differentially responsive to the various disease-modifying agents utilized during the clinical care of RA patients. These factors will need to be considered in future study designs. Finally, extensive analysis in various pre-clinical animal model studies will be essential to establish the ability of various epigenetic-modifying drugs to reverse the molecular profile of resolution phase FLSs and to achieve complete clinical and molecular remission in RA and related-autoimmune joint disease patients. Methods Animals All animal procedures were approved by the Institutional Care and Use Committee (IACUC) at Emory University and carried out in compliance with its institutional policies. The animals were maintained under pathogen-free conditions, and all experiments were performed on mice homozygous for both the Tg_rtTA2S transgene and the human TNFα transgene (iTNF-tg)^[203]42. This line was generated by a cross between a first transgenic mouse line carrying the modified reverse tet transactivator rtTA2S-M2 under the hnRNPA2/B1 promoter for ubiquitous expression^[204]93 and a second transgenic line carrying the complete human TNFα coding sequence under the Tet-responsive P-tight promoter. These double-transgenic iTNF-tg mice that ubiquitously express human TNF upon doxycycline administration were previously demonstrated to exhibit characteristics of autoimmune arthritis. It is a distinctive animal model for researching the molecular causes of arthritis, particularly the early stages of the disease development and the tissue remodeling processes that occur when TNFα expression is stopped. Inflammation in the iTNF-tg mouse model starts after one week of doxycycline and continues for as long as doxycycline treatment is maintained. Histological signs of joint inflammation are resolved by 3 weeks of doxycycline treatment^[205]42. In this study, the inflammation phase was induced in female mice by treatment with 1 mg/ml doxycycline and 5% sucrose for 3 weeks. This was followed by an additional 3 weeks of no treatment to establish the resolution phase. Cell culture FLS were prepared from iTNF-tg mice according to an established protocol^[206]43,[207]94. Briefly, forelimbs were separated at the radiocarpal joint while hind limbs were separated at the tibiotalar joint. The phalanges were carefully deskinned, keeping in mind to keep the interphalangeal and metacarpophalangeal joints intact. This was followed by the first Collagenase IV (Sigma-Aldrich, 1 mg/mL for 30 min at 37 °C) treatment to digest the isolated joints for the removal of skin fibroblasts and other surface cells. For the digestion of synovium lining the phalangeal joints, a second digestion using collagenase IV (3 mg/mL for 1 h at 37 °C) was performed. To get rid of the undigested bones and other debris, the digests were filtered using a 50-micron cell strainer. The resultant cell suspension, which contained >90% FLS, was grown for 16 h in DMEM with 10% fetal bovine serum (Corning) and 1% penicillin/streptomycin. Human umbilical vein endothelial cells (HUVECs) were sourced from the American Type Culture Collection (ATCC) (Rockville, MD) and were cultured in EGM (Endothelial Cell Growth Medium) (PromoCell) supplemented with 2.4% EGM supplement mix (PromoCell) and 1% Penicillin/Streptomycin (Sigma). HDMECs were obtained from PromoCell (Heidelberg, Germany) and were cultured in Endothelial Cell Growth Medium MV (PromoCell) supplemented with 2.4% EGM-MV supplement mix (PromoCell) and 1% Penicillin/Streptomycin (Sigma). They were maintained at 37 °C in a humidified 5% (vol/vol) CO[2] incubator and used at passages 3–4. Additionally, the co-culture system of FLS and HDMECs was established to carry out the gene expression analysis and angiogenesis analysis using Transwell inserts. Histology and immunohistochemistry The forepaws from the mice of all three groups were fixed overnight in 4% paraformaldehyde in phosphate-buffered saline, decalcified with Morse’s solution at 4 °C for 5 days (22.50% formic acid and 10% sodium citrate), and embedded in paraffin. Serial sections measuring 4 μm in thickness followed by Toluidine Blue (Sigma) staining. Immunohistochemistry (IHC) was performed using a Vectastain Elite kit (AK-5001, Vector Laboratories, Burlingame, CA) following the manufacturer’s protocol, next the heat-mediated antigen retrieval by 10 mM Citrate buffer pH 6.0. To reduce further nonspecific background staining, slides were incubated with 3% normal goat serum for 20 min. All slides then were incubated at 4 °C overnight with primary antibodies anti-CD31 (1:200, 77699, Cell Signaling), and anti F4/80 (1:200, 70076, Cell Signaling). Negative controls were produced by eliminating the primary antibodies from the diluents. Sections then were incubated with Vectastain ABC-AP reagent (Vector Laboratories, Burlingame, CA) for 30 min at room temperature, followed by incubation with ImmPACT Vector Red Substrate for alkaline phosphatase (SK-5105, Vector Laboratories, Burlingame, CA) for 30 min at room temperature. Fast Green FCF (F99-10, Fisher Chemical, Fisher Scientific Company, NJ) was used as the counterstain. Flow cytometric analysis FLSs were resuspended at approximately 1 × 10^5 cells/ml with PBS containing 2% FBS and stained with fluorochrome-conjugated monoclonal antibodies against Alexa Flour 700-conjugated CD45 (1:500; 103127; BioLegend), allophycocyanin (APC)-conjugated CD31 (1:1000; 102409; BioLegend), and R-phycoerythrin (PE)-cyanine 7 conjugated PDGFRA/CD140 (1:200; 25-1401-82; eBioscience). Cells were labeled with antibodies for 30 min at 4 °C in the dark, then washed and resuspended in PBS containing 2% FBS. Cells were stained with Zombie Violet fixable viability kit (423113; BioLegend) to distinguish live and dead cells according to manufacturer’s protocol. Then, cells were flow-cytometrically acquired and analyzed by BD FACSymphony A3 analyser (BD Biosciences, San Jose, CA) equipped with FACSDIVA™ software. FlowJo software (v.10.10.0) was used to analyze the data. Bulk RNA-seq and data analysis FLSs were cultured in 6-well plates after the density of 2 × 10^5 cells/well. The total RNA was extracted and purified using Direct-zol RNA MicroPrep (Zymo Research) following the manufacturers’ protocol. RNA quality and quantity were assessed using a 2100 Bioanalyzer (Agilent Technologies). Only samples with an RNA integrity number >8 were used for sequencing. Four biological replicates were used per each condition. Libraries were generated from 250 ng RNA using TruSeq Stranded Total RNA Sample Prep Kit (Illumina) using the Poly A enrichment method. Sequencing was carried out using the NovaSeq PE 6000 system (Novogene UC Davis Sequencing Center, Novogene Corporation Inc). Raw data were exported in FASTQ (fq) format, and quality checks were made for the distribution of GC content and error rates. Data filtering was also done to get rid of adapter reads and low-quality reads. FASTQ files from these samples were uploaded to DNASTAR Lasergene (version 17.3.0.57) and ArrayStar software and were analyzed as described below^[208]44,[209]95. Paired-end reads were mapped to the GRCm38 mice genome assembly. The differential gene expression (DEG) analysis was performed using the DESeq2 method and pairwise gene expression levels were calculated using RPKM (read per kilobase of transcript sequence per million base pairs sequenced) value by performing Student’s t-test for genes at 95% confidence. FC (fold change) in gene expression was performed on filtered datasets using normalized signal values. DEGs were identified using the BioJupies web tool. Log FC represented the fold change of gene expression, and p < 0.05 and log2 FC > 2 were set for statistically significant DEGs. Multiple correction testing was performed using a false discovery rate (FDR). Pathway analysis was performed utilizing the Reactome Pathway Database available on the Enrichr pathway enrichment analysis tool^[210]96. Data was visualized using heatmaps, volcano plots, and principal component analysis using GraphPad Prism version 9.4.1, and BioJupies^[211]97 Clustergrammer^[212]98, and ChEA3^[213]99 web tools. Single-cell RNAseq data analysis Published scRNA-seq from the synovial tissue of K/BxN serum transfer-induced arthritis (GSE 230145^[214]51). Data from live CD45 negative synovial cells from the hind limbs of the mice (n = 3) on day 0 = control (homeostasis), day 7-9 = peak (inflammation) and day 22 = resolved phase were analyzed in the Seurat pipeline^[215]100,[216]101 (Version 5.1.0). UMAP plots were used to visualize cell clusters that were defined based on previously reported marker genes. The cluster identities were transferred into CellChat^[217]52 (version 1.6.1), a cell-cell communication prediction tool based on the differentially expressed ligand-receptor pairs in the defined cell clusters. Dot plots and Circos plots were utilized for data visualization. Cell viability and cell cycle analysis FLSs and HUVECs were cultured in a 12-well plate after the density adjustment at 1 × 10^5 cells/mL. The percentage of cells in the G0/G1, S/G2/M phases, and sub-G0/G1 phases were determined at 24, 48, 72, and 96 h as described previously^[218]44. Briefly, the cells were permeabilized with 70% ethanol, and their nuclei were stained with DAPI (4 nM, Nexcelom Bioscience) and examined by measuring the amount of DNA per single cell in a multifluorescent channel analysis (blue 377/477) on the Celigo Imaging Cytometer, according to Nexcelom Bioscience protocols (Assay ID: Celigo_02_0001). The percentage of apoptotic cells was measured by gating for intact cells containing less than 2X the amount of DNA calculated at the interface of DAPI integrated intensity on the Celigo Imaging Cytometer. Tube formation assay Briefly, Matrigel Basement Membrane Matrix (Corning) was coated to the precooled 96-well and 24-well tissue culture plate for HUVECs and HDMECs respectively. Pipette tips and Matrigel solution were kept cold throughout to avoid solidification. The plate was then incubated at 37 °C for 30 min to allow the matrix solution to solidify. For HUVECs tube formation assay, a total of 1.25 × 10^4 cells along with 10% of FLSs conditioning medium in a final volume of 100 μL of EGM-supplemented medium were seeded onto the surface of each well containing the polymerized matrix for 3–5 h at 37 °C. For HDMECs tube formation assay, a total of 5 × 10^4 FLSs were seeded in transwells of 0.4 μm pore size and co-cultured with an equivalent number of HDMECs in the EGM-MV medium in 24 well culture plate onto the surface of each well containing the polymerized matrix for 3–5 h at 37 °C. The HDMECs were treated with calcein AM at a final concentration of 2 µg/ml prior to the visualization of tubes for 30 min in the dark at 37 °C and 5% CO[2]. Tube formation was inspected under an inverted microscope (Olympus CKX53) and images were acquired using cellSens imaging software. The quantification of cellular networks and organization of HDMECs cultured in matrigel was done using the Angiogenesis analyzer for ImageJ software. Three biological replicates were used per each condition. Quantitative reverse transcription-PCR The total RNA was extracted from direct cultures of HUVECs with 20% supernatants of FLSs as well as indirect co-cultures of HDMECs with FLSs using Transwell inserts using Direct-zol RNA MicroPrep (Zymo Research). The cDNA was synthesized from 1 μg of RNA using qScript™ cDNA SuperMix 5× (QuantaBio). Quantitative real-time polymerase chain reaction qRT-PCR was performed using PerfeCTa SYBR® Green FastMix 2× (QuantaBio) on the Rotor-Gene Q Real-Time PCR cycler (QIAGEN). Fold change (FC) was determined through the 2^−ΔΔCt method and normalization of the amount of expressed mRNA was done by using human TBP or mouse Tbp as the internal housekeeping gene. Primer sequences are detailed in Supplementary Table [219]1 and [220]2. Statistics and reproducibility Data are expressed as mean ± standard deviation of the mean of at least three independent experiments. The results were analyzed using the software GraphPad Prism version 9.4.1. Statistical comparisons between two groups (TNF treated vs TNF untreated) were performed using a two-tailed Student’s t-test for comparing two groups. One-way analysis of variance (ANOVA) followed by Dunnett’s multiple comparisons test was performed for more than two groups. Two-way ANOVA followed by Tukey’s multiple comparison test was performed for more than two independent variables and multiple group analyses. p-value less than 0.05 was considered statistically significant. The source data for the results and figures are described (Supplementary Data [221]2). Reporting summary Further information on research design is available in the [222]Nature Portfolio Reporting Summary linked to this article. Supplementary information [223]Supplementary Information^ (1.7MB, pdf) [224]42003_2025_8250_MOESM2_ESM.pdf^ (73.6KB, pdf) Description of Additional Supplementary Files [225]Supplementary Data 1^ (139.2KB, xlsx) [226]Supplementary Data 2^ (893.8KB, xlsx) [227]Reporting Summary^ (2.4MB, pdf) Acknowledgements