Rheumatoid Synovial Subsets of CD3 + T cells defined by PD-1 + and PD-1- Cells Drive Fibroblasts into Distinct Endotypes
Present Address:
SofieRask1
ChristianArnfeldtTerpAndersen2
BenedicteBechAndersen1
MortenTVenø3
YanYan4
MaleneHvid1,5
KimRavnskjaer2✉Emailsrg@biomed.au.dk
BentDeleuran1,6
StinneRavnGreisen1,6
1Department of BiomedicineAarhus UniversityWilhelm Meyers Allé 68000Aarhus CDenmark
2
A
Department Of Biochemistry and Molecular Biology (Bmb), Faculty of ScienceUniversity of Southern DenmarkDenmark3Omiics ApsAarhusDenmark
4
A
EVGenomics Aps5Department of Clinical MedicineAarhus UniversityDenmark
6Department of RheumatologyAarhus University HospitalDenmark
Sofie Rask1, Christian Arnfeldt Terp Andersen2, Benedicte Bech Andersen1, Morten T Venø3, Yan Yan4, Malene Hvid1,5, Kim Ravnskjaer2, Bent Deleuran1,6, Stinne Ravn Greisen1,6
1Department of Biomedicine, Aarhus University, Denmark; 2Department Of Biochemistry and Molecular Biology (Bmb) Faculty of Science University of Southern Denmark, Denmark; 3Omiics Aps, Aarhus, Denmark; 4EVGenomics Aps, 5Department of Clinical Medicine, Aarhus University, Denmark, 6Department of Rheumatology, Aarhus University Hospital, Denmark.
Corresponding author contact: Stinne R. Greisen. Address: Department of Biomedicine, Aarhus University, Wilhelm Meyers Allé 6, 8000 Aarhus C. Email: srg@biomed.au.dk.
Abstract
Introduction
The joint destructive process in rheumatoid arthritis (RA) is highly individual and dominated by complex cellular interactions. In damping the immune activation, the PD-1 pathway is central and highly upregulated in the arthritic joint. Here, we examine how T cells with or without PD-1 influence the function of arthritic mesenchymal cells
Methods
We investigated samples from RA synovial fluid and characterized PD-1 + and PD-1- T cells by flow cytometry, ELISA and RNA sequencing (RNA seq). We performed co-culture experiments to investigate the interaction between PD-1+/- cells, osteoclasts, osteoblasts and fibroblasts. We used flow cytometry, ELISA and RNA seq to characterize the RA and HC fibroblasts after co-culture with PD-1 + and PD-1- T cells.
Results
We found PD-1 + T cells to express other checkpoint receptors and to have a diminished cytokine production upon CD3/CD28 stimulation. The PD-1 + and PD-1- T cells from the RA synovial fluid clustered differently after RNA seq, and the gene signature of the PD-1 + T shared traits with peripheral T helper cells. Fibroblasts co-cultured with PD-1 + T cells expressed genes associated with collagen synthesis, TGFβ production and immune regulation. In contrast, the signature in fibroblasts co-cultured with PD-1- cells was associated with immune activation.
Conclusion
We demonstrate RA synovial fluid CD3 + T cell subsets can drive healthy fibroblasts into distinct pathological endotypes. When evaluating all PD-1 + T cells, these support fibroblast subsets related to the production of collagen and immune regulation.
Keywords
Immune checkpoint receptors
Rheumatoid Arthritis
Fibroblasts
T cell subset
A
Highlights
PD-1 + T cells are present in the rheumatoid synovial joint fluid
PD-1 + synovial fluid T cells have diminished cytokine production upon activation
Healthy fibroblast can be activated to a pathological endotype by synovial T cells
The fibroblast endotype differs depending on the PD-1 expression on the T cell
A
Introduction
Rheumatoid arthritis (RA) is a multifactorial autoimmune disease characterized by joint inflammation, resulting in pain, swelling and ultimately joint destruction1,2.
Programmed Cell Death Protein 1 (PD-1) is an immune checkpoint receptor expressed by activated T cells. Upon interaction with its ligands; Programmed Death Ligand 1 and 2 (PD-L1/2), PD-1 inhibits the adaptive immune response and is, therefore, crucial in maintaining peripheral self-tolerance3,4. Downstream PD-1-signaling leads to reduced activation, cytokine production and proliferation of the T cell5,6. PD-1 is expressed on continuously activated T cells and is often associated with an exhausted T cell profile. Exhausted T cells (Tex) are characterized by an increased expression of inhibitory receptors and loss of effector functions, including cytokine production and effector abilities7.
The importance of the PD-1 pathway in self-tolerance is supported by the development of immune-related adverse events, including development of an RA-like disease in cancer patients treated with anti-PD-1/PD-L1 antibodies8,9. Furthermore, a recent phase two study highlights the effect of using agonistic PD-1 antibodies to decrease disease activity in refractory RA10. These findings support the importance of PD-1 in autoimmunity11. Together with animal- and genetic studies that have linked loss of function in the PD-1/PD-L1 axis with several autoimmune disorders, including inflammatory bowel disease (IBD), Sjögren’s syndrome, systemic lupus erythematosus and RA12.
The PD-1+CD4+CXCR5− subset of peripheral T-helper cells (Tph) is characterized by IL-21 and CXCL13 production and shares effector functions with follicular T-helper cells. They are abundant in RA synovial fluid, and increasing numbers correspond to high disease activity. They exert T-helper cell functions by activating plasma cells and inducing the production of autoantibodies13–15, collectively supporting an increased immune activity.
The interplay between T cells, B cells, and mesenchymal cells is of major importance in RA disease progression. Fibroblasts, osteoclasts, and osteoblasts infiltrate the joint space and the synovium and play an important role in disease progression, joint destruction and joint homeostasis16. Osteoclasts and osteoblasts maintain healthy bone but are also activated upon inflammation, and osteoclasts drive bone erosions in RA17. Immune checkpoint receptors suppress osteoclast activity and may promote osteoblast activity, suggesting an important link between immune regulation and bone homeostasis18. In RA, fibroblast subtypes differ in the production of inflammatory mediators and are localized in the lining or sublining layer of the inflamed synovium19,20. These subtypes infiltrating the synovium have proven to be of importance in the pathogenesis of RA21.
Here, we investigate the relation between synovial fluid PD-1 + and PD-1- T cell and mesenchymal cellular subsets in RA. We report that PD-1 + T cells can drive fibroblasts into an endotype supporting the production of extracellular matrix proteins and immune regulation.
Methods
Patient material and cells
A
RA synovial fluid mononuclear cells (SFMCs) and peripheral blood mononuclear cells (PBMCs) were obtained from patients with chronic RA (n = 9). These patients presented in the outpatient clinic (Aarhus University Hospital, Denmark) with disease flare requiring joint aspiration and glucocorticoid injection. All included patients were seropositive RA patients, with a disease duration of 5 years or more. The diagnosis was based on the ACR/EULAR 2010 classifications criteria for RA22. Healthy control (HC) PBMCs (n = 5) were obtained from buffy coats provided by the Danish blood bank. HC skin fibroblasts were obtained from the skin of healthy donors and used between passages 4–6 (n = 5). In brief, the skin was minced and digested, and fibroblasts were grown on 5 ml Petri dishes in fibroblasts culture medium; DMEM (VWR, cat. no. 392–0407) supplemented with 10% fetal calf serum, 1% penicillin, streptomycin, amphotericin B and glutamine. Collected fibroblasts were stored at -150oC in DMSO freezing medium. The SaOS-2 cell line was used for the mineralization assay23.Characterization of SFMCs
SFMCs were characterized by flow cytometry using the Novocyte Quanteon analyser (Agilent) staining for: CD3 PeCy7 (Biolegend, cat. no: 303118), CD4 BV605 (Biolegend, cat. no:317437), CD8 BV785 (Biolegend, cat. no:344740), TIGIT BV421 (EBioscience, cat. no: 48-9500-42), TIM3 BV510 (Biolegend, cat. no: 345029), LAG3 BV711(Biolegend, cat. no: 369319), PD1 AF488 (Biolegend, cat. no. 329936), CTLA4 PerCP-Cy5.5 (Biolegend, cat. no: 349928), NFAT PE (Biolegend, cat. no: 649606), TOX AF594 (Biolegend, cat. no: 682604), TCF1 APC (Biolegend, cat. no: 655204), L/D nIR (Invitrogen, cat. no. L34976),
Gating was done on live, single cells, and fluorescence minus one (FMO) was used to set the gates (suppl fig S1) Data were analyzed in FlowJo (BD Biosciences).
Sorting and stimulation of cells
SFMCs were thawed and stained using L/D nIR (Invitrogen, cat. no. L34976), anti-CD3 PerCPCy5.5 (Biolegend, cat. no. 317336) and anti-PD-1 AF488 (Biolegend, cat. no. 329936).
Sorting was done on single, live, CD3 + cells and into PD-1 + and PD-1- populations using the 4-laser FACSAria III cell sorter (BD Biosciences) (suppl fig S2). Gating for PD-1 was done excluding the intermediate population, ensuring only to sort true PD-1 + and true PD-1- T cells. RNA bulk sequencing was done on naïve sorted cells. Post sorting, 50,000 cells were seeded, rested overnight in RPMI media and subsequently stimulated for 24 hours using anti-CD3/CD28 beads (Gibco, cat. no. 11131D). After stimulation, cells and supernatants were collected and analyzed or passed to co-cultures. The supernatant was removed, passed for cell culture stimulation, or stored at -20°C until subsequent analysis or culturing.
Supernatant analyses of stimulated cells
V-plex multipanel assay was performed on the supernatants from stimulated PD-1 + and PD-1- cultures, using the proinflammatory kit (cat. K15049D) (n = 7). The multiplex assay was performed in accordance with the manufacturer’s instructions.
Osteoclast cultures and TRAP activity
PBMCs from HC were thawed, and cells allowed to adhere at 37°C for 4 hours. Non-adherent cells were removed, and adherent cells stimulated with M-CSF (Sigma Aldrich, cat. no. 300 − 25) 100 ng/ml for three days. Cells were subsequently restimulated with M-CSF 25 ng/ml and RANKL (Abcam, cat. no. ab157289) 50 ng/ml and 20% supernatant from activated PD-1 + T cells or activated PD-1- T cells (n = 3). Positive control was set as wells only stimulated with RANKL and M-CSF. A commercially available TRAP assay (B-Bridge International, cat. no. AK04) was used to determine osteoclast activity during the 18 days culture. Day 0 was the first day of stimulation.
MTT-assay (Roche, cat. no. 11465007001) was performed on day 3 to determine whether supernatants had any influence on the survival of the adherent cells of the culture. MTT-assay was performed according to manufacturer’s instructions.
Osteoblast cultures and mineralization assay
The osteoblast cell line, SaOS-2 cells (Promocell) were grown in T75 flasks in osteoblast growth media (Promocell, cat. no. C-27001) until confluency was reached. Cells were trypsinized and seeded into a black-walled 96-well plate (Sigma Aldrich, cat. no CLS3603-48EA) in a concentration of 20,000 cells/well in 150 µL growth media. Cells were stimulated with BMP-2 (RnD, cat. no. 355-BM/CF) 50 ng/mL, or 20% supernatant from PD-1 + or PD-1- cells (paired, n = 5). On day 4, media was changed to mineralization media (PromoCell, cat. no. C-27020), and stimulations repeated. Media and stimulations were changed every 3–4 days until day 11, where mineralization was investigated using a mineralization kit (Lonza, cat. no. PA-1503). The assay was conducted according to the manufacturer’s instructions.
RA fibroblast cultures
RA SFMCs (n = 3) were differentiated into fibroblasts-like synovial cells (FLS) as previously described24,25. FLS’s were cultured in T25 flasks (Sarstedt, cat. no. 83.3910) in DMEM low glucose media (VWR, cat. no. 392–0407) until confluency. Cells were trypsinized and seeded in a concentration of 20,000 cells pr. well in a 96-well plate, rested for 24 hours and then stimulated with 20% supernatant from PD-1+/- cultures, IFNγ (10 µg/ml) or no stimulation. Cells were cultured for 48 hours. Supernatant was removed and stored at -20°C for further analyses. MCP-1 production in the supernatants was determined using the MCP-1 ELISA kit (Invitrogen, cat. no. 88-7399-88).
HC fibroblast cultures
Fibroblasts from HC skin in passage 3 were thawed and cultured in a T175 flask (Sarstedt, cat. no. 83.3911.002) in DMEM low glucose media (VWR, cat. no. 392–0407). When confluency was reached cells were trypsinized and seeded into a 96-well plate and stimulated by adding 20% supernatant from PD-1+/- cultures (n = 9), IFNγ (10 µg/ml) (n = 7), PD-1+/- T cells (n = 9), or no stimulation (n = 7). After 48 hours fibroblasts were investigated by flow cytometry. T cells were removed by centrifugation and the supernatant stored at -20°C until further analyses.
RA- and HC fibroblast flow cytometry and MCP-1 ELISA
MCP-1 production was determined in the supernatant from co-cultures using the MCP-1 ELISA kit (Invitrogen, cat. no. 88-7399-88). For flow cytometry staining of the fibroblasts, the panel included: CD90 PeCy7 (BD Biosciences, cat. 561558), ICAM Pacific blue (Biolegend, cat. 353110), VCAM PeCy5 (Biolegend, Cat. 305808), PD-L1 BV650 (BD Biosciences, cat. 328740), FAP AF488 (RnD, cat. FAB3715G), Podoplanin PE (Biolegend, cat. 337004) and L/D nIR (Invitrogen, cat. no. L34976). Gating was done on live, singlets, and CD90+. FMO’s were made for PD-L1, podoplanin ICAM and VCAM (suppl fig S3). Data was analyzed in FlowJo (BD Biosciences).
RNA purification and sequencing of PD-1+ / PD-1- T cells
RNA was purified from non-stimulated, PD-1 + and PD-1- T cells from the synovial joint fluid after sorting and cell lysis (n = 8, paired) using the RNeasy Mini Kit (Qiagen, cat. no. 74104) and the Genomic DNA was removed by using the RNase-Free DNase Set (Qiagen, cat. no. 79254).. The library preparations were performed by the Functional Genomic and Metabolism (FGM) sequencing team at SDU. The RNA was here enriched for mRNA via Poly- A enrichment and quantitatively sequenced by synthesis on the Illumina Novaseq 6000 platform.
RNA-seq data preprocessing
FASTQ files were aligned to the human genome (GRCh38, Ensembl release 101) using STAR (v.2.7.8a). Aligned reads were counted using FeatureCounts (v2.0). Genes with no read count were removed. Ensembl IDs were then converted to Gene symbols using biomaRt (v. 2.52.0). The covariate effects; sex, flow cell, and PD-1+/- groups (PD-1 Levels) were tested using one-way Analysis of Variance (ANOVA) regarding Principal Component (PC) one to six from principle component analysis (PCA) on the 28,408 VST counts. Covariates with a p-value < 0.05 were considered statically significant.
Differential expression analysis and data processing
Differential expression analysis was conducted using DESeq2 (v1.35.0) by fitting counts to a negative binomial model. Differentially expressed genes were identified using a Wald test and alpha-error accumulation was adjusted for using Benjamin-Hochberg implemented in the package. Genes with adjusted p-value < 0.05 and Log2 fold change (Log2FC) ≥ 0.01 and Log2FC ≤ 0.01 were considered statistically significant differentially expressed. Gene counts were normalized by variance stabilizing transformation for visualization and further analysis. Two separate principal component analyses (PCA) were performed on normalized counts for top-500 highly variable genes and DEGs, respectively. One-way analysis of variance (ANOVA) was used to test the transcriptional variance of the co-variates: gender, flow cell and PD-1+/- group.
The initial 28,408 genes were used as background genes for the enrichment analysis. The search for interesting gene ontology (GO) terms was done on the 753 differentially regulated genes with a base mean (BM) expression > 100, Log2FC ≥ 1, and Log2FC ≤ -1. The search was done against the org.Hs.eg.db database using clusterProfiler (v. 3.0.4). The p-value and q-value cutoffs were set to 0.05 and 0.2 for this analysis.
This enrichment focused on the two topics of biological process (BP) and molecular function (MF). BP covered broad and specific GO terms which could be linked to multiple molecular activities, and the MF terms, described activities at a molecular level, which were associated with certain gene products.
RNA purification and bulk sequencing on fibroblasts
RNA was purified from the co-cultured fibroblasts after washing with PBS and cell lysis. RNA-purification was done using the RNeasy Mini Kit (Qiagen, cat. no. 74104) (n = 1 in triplicates), and the Genomic DNA was removed by using the RNase-Free DNase Set (Qiagen, cat. no. 79254).. The RNA integrity number and concentration were measured on Bioanalyzer RNA Nano chip (Agilent, cat.no. 5067 − 1511).
A
The SMARTer Stranded Total RNA Sample Prep Kit - HI Mammalian (Takara, Japan, cat.no. 634876) was used for library preparation with 200ng of the purified RNA as input following the manufacturer`s protocol. The Bioanalyzer High sensitivity DNA analysis kit (Agilent, cat.no. 5067 − 4626) was used to determine the size of the library fragments and the Qubit dsDNA High Sensitivity Kit (Invitrogen, cat. no. Q32851) was used to quantify the library. The libraries were pooled in equal amounts and sequenced on a NovaSeq X Plus sequencing machine using a 10B flow-cell for paired-end 150 cycles (Illumina).Data analysis
Sequencing data were pre-processed by removing the adapter sequence and trimming away low-quality bases with a Phred score below 20 using Trim Galore (0.6.10). Quality control was performed using FastQC (0.12.1), Picard (2.3.1) and Multiqc (1.9) to ensure high-quality data. The filtered RNA-seq data was mapped against the human genome (hg19 / GRCh37) using STAR (2.7.11b) and gene expression was quantified using featureCounts (2.0.0) with gene annotations from Gencode release 37. Differential expression analysis was performed using DESeq2 (1.40.2) in R for gene expression profiles. Plots were done in R. Genes were considered to be significantly differentially expressed with the FDR < 0.05. Enrichment analysis was performed from differentially expressed genes only evaluating coding genes with a log2 fold change between − 1 and 1 or -2 and 2, dependent on the comparison, using the R package ClusterProfiler. The enrichment analysis focused on GO terms covering BP, MF and cellular components (CC).
Data availability
RNA sequencing data on PD-1 + and PD-1- T cells from the synovial joint fluid is available through ArrayExpress; E-MTAB-16009. The RNA sequencing data on HC fibroblasts cultured with PD-1 + and PD-1- T cells or supernatants are also available through ArrayExpress; E-MTAB-15969. The two datasets are marked as associated at ArrayExpress.
Statistical analyses
Data analysis were performed in “R” for sequencing data and GraphPad Prism (Graphpad) for the remaining data. Data fitting a normal distribution were compared using Student’s T-test. When data did not fit a normal distribution, the Wilcoxon signed-rank test was used to compare paired data. A p-value below 0.05 was considered statistically significant. Data are expressed as mean with standard deviation (SD) unless stated otherwise.
Ethics approval and patient and public involvement
Studies were approved by the Danish Data Protection Agency, the Danish Medical Agency, and the Regional Ethics Committee (2012-1329-2), and performed in accordance with the Declaration of Helsinki. All patient identifiers have been removed. Written informed consents were obtained from all participants. HC fibroblasts and cells cannot be identified or traced in accordance with the Danish Regional Ethics Committee no specific approval is needed.
Results
RA synovial fluid PD-1 + T cells express other co-inhibitory receptors
We examined CD3 + CD4+ and CD3 + CD8 + T cells from the RA synovial fluid by flow cytometry. For both CD4 + and CD8 + T cells, we evaluated PD-1 + and PD-1- T cells separately for expression of transcription factors and checkpoint receptors.
PD-1 + cells represented 42%±21% (mean ± SD) of CD4 + T cells and 34%±26% of CD8 + T cells in the RA synovial fluid (SF) of RA patients (P = 0.5). Expression of TIGIT, TIM3 and NFAT were increased in CD4 + PD-1 + T cells (all p < 0.05), suggesting these to be activated, but with either a regulatory- or an exhausted profile (Fig. 1a). A similar yet statistically insignificant tendency was observed for the CD8 + PD-1 + T cells (Fig. 1b). TOX, the transcription factor associated with Tex was equally abundant between PD-1 + and PD-1- T cells and did not vary between CD4 + and CD8 + T cells (Fig. 1a + b).
RA synovial fluid PD-1 + T cells produce less proinflammatory cytokines than PD-1- T cells
Next, we sorted RA synovial fluid CD3 + T cells into a PD-1 + and PD-1- population (suppl fig S2). After activation through CD3/CD28, we examined the production of pro-inflammatory cytokines. Both PD-1 + and PD-1- cells were responsive to stimulation. The production of IFN-γ, TNF-α, and IL-2 in PD-1- cells was up to 20 times greater than in PD-1 + cells. To complete an evaluation of the entire cytokine profile, we log-transformed data and performed a 2-way ANOVA. Overall, cytokine production in PD-1 + T cells was decreased, and the collective cytokine profile in the PD-1 + cells was significantly different compared to the PD-1- cells (p < 0.01, Fig. 1c and suppl fig. S4).
Fig. 1
Characterization of PD-1 + and PD-1- T cells from the RA synovial fluid
a.
(a) Flow cytometric characterization of PD-1 + and PD-1- CD4 + T cells from the RA synovial fluid. PD-1 + T cells co-express other immune checkpoint receptors including TIM-3, TIGIT and LAG-3, but also have a higher expression of NFAT, suggesting previous activation. (b) Flow cytometric characterization of PD-1 + and PD-1- CD8 + T cells from the RA synovial fluid. Both PD-1 + and PD-1- CD8 + T cells have a low expression of other immune checkpoints, and no difference is observed between the PD-1 + and PD-1- populations. Flow data are analyzed by Student’s T test for paired data. Data is presented as mean with SD
b.
(c) Production of pro-inflammatory cytokines in the supernatant from sorted and stimulated PD-1 + and PD-1- CD3 + T cells evaluated by V-plex multiplex. Cytokine production in PD-1 + T cells is significantly reduced compared to PD-1- T cells. Prior to analysis, cytokine data are log transformed to fit the normal distribution and then analyzed by a 2way ANOVA, but are presented in the graph as non-log transformed data. * = p < 0.05.
CD3 + PD-1 + RA synovial fluid T cells express genes associated with cell structure and immune regulation
To further investigate the PD-1 + and PD-1- cells from the inflamed RA joint, we performed bulk RNA seq on the sorted RA synovial fluid CD3 + PD-1 + and CD3 + PD-1- cells. Covariate Analysis on 28,408 genes confirmed the PD-1 levels covariate to drive data variance along PCs 1–6 (p-value < 0.0001) (suppl fig S5a and S5b)
PD-1- and PD-1 + cells separated along principal component 1 (PC1) based on the top-500 highly variable genes (HVG) from our DESeq2 analysis (Fig. 2a). We then evaluated all 753 DEGs and found 432 DEGs enriched in PD-1 + T cells (log2FC ≥ 1, padj. < 0.05) and 321 DEGs enriched in PD-1- T cells (log2FC ≤ -1, padj. < 0.05) (Fig. 2b), with PD-1 + cells expressing higher levels of IL21, CXCL13, CTLA4, IFNG, and IL10. Hierarchical clustering confirmed that patient cells from the two sorted populations separate transcriptionally by their PD-1 levels (Fig. 2c).
To better describe the differentially expressed genes between PD-1 + and PD-1- T cells, a Gene Ontology (GO) analysis was conducted. PD-1+-enriched genes were mainly associated with cell cycle reflected in GO terms, including organelle fission, nuclear division, and chromosome segregation. However, terms like negative regulation of immune system process, B cell activation- and lymphocyte activation- involved in immune response (not shown) were also found here but with adjusted p-values in the range [0.049,0.0006]. For the PD-1–enriched genes, GO terms included regulation of lymphocyte activation, adaptive immune response, and T cell activation (Fig. 2d). When evaluating molecular functions associated with genes enriched in the PD-1- and PD-1 + populations, we observed a similar tendency with the PD-1- enriched genes being associated with T cell activity (Fig. 2e). Conclusively, the RNAseq analysis associated the PD-1 + cells with the cell cycle, suggesting these cells to be in an active and potentially dividing state and immune regulation with a gene expression profile similar to the previously described Tph cells14.
We next sought to investigate how these cells interact with cells of importance for structural maintenance in the synovial joint.
Fig. 2
Differential gene expression analysis of synovial fluid PD-1+/- T cells
(a) PC1 vs PC2 plot from principal component analysis using top-500 highly variable genes. (b) MA-Plot of 753 differentially expressed genes (DEGs) (Log2FC ≥ 1 and Log2FC ≤ -1, base mean (BM) expression > 100), showing selected T cell marker genes. (c) Heatmap of z-score-transformed VST values of 20 DEGs with largest & lowest log2FC. (d) Dot plots of enriched biological processes of 432 DEGs enriched in PD-1 + cells (left, BM > 100, Log2FC ≥ 1) and 321DEGs enriched in PD-1- cells (right, BM > 100, Log2FC ≤ -1). (e) Dot plots of enriched molecular functions of 432 DEGs (left, BM > 100, Log2FC ≥ 1) and 321 DEGs (right, BM > 100, Log2FC ≤ -1). All, DEGs were p.adj. < 0.05, Wald test, Benjamini-Hochberg-corrected.
PD-1 + and PD-1- T cell supernatants do not affect bone homeostatic cells
We first set out to determine if the microenvironment created by PD-1 + cells could influence osteoclast and osteoblast function. We cultured the SaOS-2 osteoblast cell line with PD-1+/- supernatant and investigated the degree of mineralization. When visually evaluating the osteoblast cultures, both PD-1 + and PD-1- cell supernatants resulted in increased mineralization at day 11 (Fig. 3a) When evaluating the mineralization by the quantified mineralization readout, no significant difference was observed between the supernatant from the PD-1 + vs the PD-1- cells (Fig. 3b).
We next investigated if the PD-1+-created microenvironment could influence the generation of osteoclasts. We cultured PBMCs from HC in the presence of RANKL and M-CSF to generate osteoclasts. The supernatant from stimulated PD-1 + and PD-1- cells were added to the cultures (Fig. 3c). Evaluating TRAP in the supernatant as a measure of osteoclast activity, we did not observe any effect of the added supernatants when compared to the RANKL-MCS-F stimulated control (Fig. 3d).
We then continued to investigate the potential interaction between PD-1+/- cells and fibroblasts.
Fig. 3
Stimulation of osteoclasts and osteoblasts
(a) SaOS-2 osteoblasts cultured in mineralization media (Min. media) with PD-1+/- supernatants. Photos by iPhone in 20x light microscope. Visually, more calcification was observed after addition of PD-1 + supernatant. (b) Corresponding quantification of mineralization in osteoblast cultures. No difference was observed. (c) Healthy control osteoclasts cultured with PD-1+/- supernatants. Photos by iPhone in 20x light microscope. (d) All photos are chosen as representatives. Corresponding quantification of TRAP activity in osteoclast cultures. No difference was observed. Data were analysed using the Mann-Whitney U test and expressed as median with IQR.
PD-1- T cells induce a pro-inflammatory response in fibroblasts-like synoviocytes
We first sought to investigate if the microenvironment created by the PD-1 + and PD-1- T cells could influence fibroblast surface marker expression and cytokine production. Fibroblasts-like synoviocytes (FLS) from RA patients were co-cultured with autologous PD-1 + and PD-1- T cell supernatant. MCP-1 production, used as a measure of FLS pro-inflammatory activity, increased significantly when the supernatant from PD-1- cells was added to the FLS culture. PD-1 + supernatant also increased MCP-1 production, however not significantly compared to NT (Fig. 4a). No MCP-1 was detected in the PD-1+/- supernatant (data not shown).
Next, we evaluated the CD90 + FLS endotype by investigating the expression of surface markers (Fig. 4b). PD-1- supernatants increased the FLS ICAM-1 expression similar to IFN-γ stimulation and significantly more when compared to NT and the PD-1 + supernatant. VCAM expression did not differ when comparing stimulations of PD-1 + and PD-1- supernatants. PD-L1 expression induced by PD-1- cell supernatant was comparable to that of IFN-γ stimulation, and higher than from PD-1 + supernatants, despite not reaching the level of significance (p = 0.058). Podoplanin was continuously expressed, suggesting a high baseline activation of the RA FLS.
Healthy control fibroblast can differentiate into an activated endotype in response to co-culture with PD-1- T cells from the RA synovial fluid
We continued to examine whether the microenvironment generated by the RA PD-1+/PD-1- T cells could drive HC fibroblasts isolated from the skin, into an activated endotype. We co-cultured HC fibroblasts with the supernatant collected from stimulated PD-1+/- cells or directly with the stimulated PD-1+/- T cells. Co-culturing RA T cells with HC fibroblasts did not induce a significant degree of cell death when evaluated by MTT assay or live/dead on flow cytometry (suppl fig S6a + b).
As observed in the RA FLS, MCP-1 production was significantly increased by the addition of IFN-γ to the culture. MCP-1 production also increased significantly when adding cells or supernatant from the stimulated PD-1+/- T cells (all p < 0.05). Both PD-1- T cells and supernatant significantly increased MCP-1 production compared PD-1 + T cells and supernatant (Fig. 4c). Evaluating the stimulated HC fibroblasts by flow cytometry, ICAM-1 expression increased in response to IFN-γ stimulation, and again we observed a significant difference in ICAM-1 expression between the PD-1 + and the PD-1- stimulation. VCAM also increased in response to IFN-γ. Here we only observed a difference between the PD-1 + and PD-1- cells. For PD-L1 expression a similar picture emerged. Podoplanin also increased in response to IFN-γ stimulation and the expression differed between PD-1 + and PD-1- supernatant and cells (Fig. 4d). Evaluating the median fluorescence intensity (MFI) we observed a similar picture but with additional significant differences between the NT and the cell/supernatant stimulation (suppl figure S6c).
Co-culturing fibroblasts with PD-1+/- cells or supernatants resulted in fibroblast activation, for both RA FLS and HC fibroblasts, with PD-1- promoting an active endotype evaluated by MCP-1, ICAM-1, podoplanin, and PD-L1.
Fig. 4
Fibroblast- and PD-1+/- co-cultures
(a) RA differentiated synovial-like fibroblasts co-cultured in the presence of the supernatant from autologous RA PD-1 + or PD-1- stimulated T cells. MCP-1 production was measured in the supernatant. More MCP-1 was produced when fibroblasts were cultured with the PD-1- supernatant. (b) RA differentiated synovial-like fibroblasts co-cultured in the presence of the supernatant from autologous PD-1 + or PD-1- stimulated T cells. Flow cytometric analysis of the surface expression of ICAM-1, VCAM, PD-L1 and PDPL evaluated on CD90 + fibroblasts after culture. ICAM-1 differed significantly between PD-1 + and PD-1- supernatant stimulated fibroblasts. Data were analyzed using a paired t-test, and are presented as mean with SD. (c) HC skin fibroblasts co-cultured in the presence of the RA PD-1 + or PD-1- supernatant or cells. MCP-1 production was measured in the supernatant. MCP-1 increased upon culture with cells/supernatant and increased when fibroblasts were cultured with the PD-1- supernatant/cell compared to PD-1 + supernatant/cell. (d) HC skin fibroblasts co-cultured in the presence of the RA PD-1 + or PD-1- supernatant or cells. Surface expression of ICAM-1, VCAM, PD-L1 and PDPL evaluated on CD90 + fibroblasts after co-culture. All surface markers differed significantly between PD-1+/- cell stimulation. Data were analyzed using a Wilcoxen rank test and are presented as median with IQR * = p < 0.05, ** = p < 0.01.
PD-1 + and PD-1- T cells and supernatant induce endotypic differences in fibroblasts
To further understand how the PD-1+/- T cells influenced the HC fibroblast endotypes we used RNA seq to evaluate DEG in HC fibroblasts co-cultured with PD-1+/- cells or exposed to PD-1+/- supernatants.
We used IFN-γ as a positive control for stimulation and compared the PD-1+/- stimulations to the non-treated fibroblast culture. As the fibroblasts were cultured with T cells, these could potentially bind to the fibroblast surface. Therefore, we first investigated for contamination of T cells in the bulk sequencing data. We compared the PD-1+/- T cell data set and the fibroblasts data set and identified 125 DEG to be shared in both data sets. Using Enrichr 26and the CellMarker 2024 database to evaluate the 10 most significant hits we observed a T cell profile among these genes (suppl fig S7a). This led us to exclude these 125 genes as they could represent contamination from T cells in the fibroblast culture.
The stimulated fibroblasts clustered separately on the PCA plot depending on the stimulation (Fig. 5a). Between the PD-1 + and PD-1- cell stimulated fibroblasts, we observed 1031 DEG and 413 DEG between the PD-1 + and PD-1- supernatant stimulated fibroblasts (Fig. 5b). The stimulated fibroblasts differed significantly from the controls; both the non-stimulated blank control and the IFN-γ stimulated fibroblast cultures. Among the top 50 DEG between PD-1 + and PD-1- stimulation was the collagen-related gene COL1A1 as well as TGFB1 suggesting PD-1 + cells and supernatant to drive fibroblasts into an endotype supporting the generation of extracellular matrix (Fig. 5c). When evaluating the genes upregulated in fibroblasts stimulated with PD-1- T cells or PD-1- supernatants, these were dominated by upregulation of HLA-subtypes and CXCL9, suggesting these fibroblasts have a pro-inflammatory signature. The PD-1- cells induced a signature resembling the IFN-γ stimulation, whereas the PD-1 + supernatant differed the least when compared to the non-treated blank fibroblasts (Fig. 5c).
We further investigated enriched gene signatures evaluating KEGG and GO terms. When evaluating KEGG terms for the upregulated genes comparing PD-1 + and PD-1- T cell stimulated fibroblasts we observed pathways associated with viral responses, rheumatoid arthritis, T cell differentiation and cell adhesion molecules (suppl fig S7b). Evaluating biological pathway GO terms for the same comparison, we observed an association to regulation of membrane potential and cell adhesion molecules for the PD-1 + vs PD-1- cell-stimulated fibroblasts. For the supernatant-stimulated fibroblasts, the PD-1+-associated GO terms were BMP signaling pathways, SMAD signaling and negative regulation of cytokines (Fig. 5d).
Collectively, this supports PD-1+/- T cells and supernatants to drive HC fibroblasts into distinct endotypes with unique properties.
Fig. 5
Differential gene expression analysis of fibroblasts after synovial fluid PD-1+/- culture
(a) PC1 vs PC2 plot from principal component analysis of stimulated fibroblasts, excluding the 125 DEG shared between fibroblasts and PD-1+/- T cells. (b) Volcano plot visualizing the DEG between the PD-1+/- supernatant (left) and cells (right). The top20 DEG are indicated. The 125 DEG shared between fibroblasts and PD-1+/- T cells are excluded. (c) Heatmap of z-score-transformed VST values of 50 DEGs with largest & lowest log2FC. The 125 DEG shared between fibroblasts and PD-1+/- T cells are excluded. (d) Dotplots of enriched biological processes GO term evaluating DEG with a Log2FC > 2 or log2FC < 2 from PD-1+/- cell or supernatant stimulated fibroblasts. The 125 DEG shared between fibroblasts and PD-1+/- T cells are excluded. All DEGs were p.adj. < 0.05, Wald test, Benjamini-Hochberg-corrected.
Discussion
A
We confirm that a high percentage of the T cells in the rheumatoid synovial fluid express PD-1. We also show that these cells express other co-inhibitory receptors and that they are less responsive to CD3/CD28 stimulation than T cells not expressing PD-1. This supports their function as being significantly different from naïve or recently activated T cells. Based on RNAseq data, we show that a proportion of the RA CD3 + PD-1 + T cells from the synovial fluid, have a gene expression similar to peripheral T helper cells, with high expression of CXCL13 and IL21 14. High expression of PD-1 is often associated with T cell exhaustion, which has also been shown in RA27. Based on the diminished cytokine production we observed in the PD-1 + T cells, these cells could have undergone a degree of exhaustion. However, considering the transcription factor TOX not being increased28 and the gene profile found in the RNA seq, the majority of the RA PD-1 + T cells do not have a classically exhausted T cell profile. The RA joint is a complex structure with multiple cell types. The interplay between these different cell types all contribute to the outcome of inflammation. Bone destruction is mediated by osteoclasts. The association between osteoclasts and the immune system is well described and inflammatory activity drives osteoclast activity and differentiation both directly through interaction with T cells and via pro-inflammatory cytokines.29 We previously demonstrated the PD-1 pathway to be associated with bone homeostasis30. Therefore, we expected osteoclast formation to be augmented by the PD-1 + supernatant. However, we did not observe any statistical difference between the addition of the PD-1 + and the PD1- supernatant when cultured with osteoclasts. This could be explained by a lack of cellular contact or that the regulation of the inflammatory environment is more complex. The association between the PD-1/PD-L1 axis and bone homeostasis could also influence osteoblasts and support bone formation31. However, we observed no significant difference when using the supernatant from the PD-1 + cells.Fibroblasts are identified as major drivers of disease pathology and inflammation in RA32,33. The fibroblast signature correlates with RA severity34 and fibroblasts are known to activate T cells35. Evaluating only a few selected markers of fibroblast activation, including MCP-1 production we show that pre-activated CD3 + T cells from the RA synovial fluid can activate fibroblasts and drive healthy fibroblasts into a specific pathological endotypes. By sorting the CD3 + T cells based on their PD-1 expression, we observed clear differences between the fibroblast endotypes depending on their stimulation with either PD-1 + or the PD-1- cells/supernatant. Evaluated by flow cytometry and MCP-1 production, we did not observe a significant difference between fibroblasts stimulated with cells or supernatants. We, therefore, suggest that the cytokines produced by the T cell subset are important drivers of the fibroblast endotypes. However, evaluated by RNAseq the fibroblast differentiation is more pronounced when the T cell subset and the fibroblasts are in direct contact. The effect of T cells and T cell cytokines on fibroblast differentiation is confirmed in the arthritic joint35. Supportive of the PD-1 pathways function in relation to tissue supportive functions, we observed the fibroblast subsets induced by the PD-1 + T cell subsets to have a less pro-inflammatory profile and be connected to tissue supportive functions with DEG related to collagen synthesis and TGFβ as well as pathways associated with BMP signaling and IL-6 production36. Despite observing significant differences in fibroblast endotypes after co-culture, our study is limited by only evaluating bulk CD3 + T cells sorted on PD-1 expression. CD8 + T cells with a high expression of PD-1 are often considered to be exhausted37, whereas subsets of CD4 + PD-1 + T cells have been described with distinct functions, including peripheral T helper cells14. Further sorting the T cells into specific subsets would have added further value to the understanding of which T cell subsets drive fibroblasts into distinct endotypes27.
The PD-1 pathway is suggested to play an essential role in fibrosis38. Fibrosis is mediated by TGFβ and chronic inflammation39,40. Fibrosis can be considered an excessive healing process with fibroblasts producing extracellular matrix proteins39,40, supporting immune regulatory pathways to play a role in fibrosis. In the PD-1 + T cell stimulated fibroblasts, we observed not only the collagen-related genes but also TGFB1, further supporting the PD-1 + cells to play an important role in extracellular matrix production, potentially supporting both natural healing and excessive fibrosis.
In conclusion, we demonstrate that RA CD3 + T cells can induce a pathological endotype in healthy fibroblasts. Additionally, this endotype depends on the PD-1 expression on the T cell subset cultured with the fibroblasts. We suggest PD-1 + CD3 + T cells to drive fibroblasts into an endotype supportive of the extracellular matrix.
Figure legends supplementary figures
Supplementary figure S1
Gating strategy for gating on RA synovial fluid cells. Gating is done on lymphocytes, single cells, live cells, CD3 + cells and CD4+/CD8 + cells, and subsequently PD-1 + and PD-1- cells in addition to LAG3, TIGIT, TIM3, NFAT, TCF1 and TOX. Representative plot of all stains and FMOs.
Supplementary figure S2
Representative gating strategy for cell sorting in PD-1 + and PD-1- CD3 + T cell populations. Gating is done on lymphocytes, single cells, live cells and CD3 + PD1+/- cells. Subsequent re-analysis of the sorted populations is presented.
Supplementary figure S3
Gating strategy for fibroblasts. Gating is done on cells, single cells, live cells and CD90 + cells. Representative plot of FMOs and all stains for ICAM, PD-L1, VCAM and PDPL are included.
Supplementary figure S4
Individual plots plots for each cytokine from the multiplex. Individual values are presented and the difference between PD-1 + and PD-1- represented with connecting lines. Data were analyzed by a multiple Wilcoxon signed rank test, presenting both the p value and the adjusted q value.
Supplementary figure S5
(a) Dot plot of covariate effects based on PC 1–6 from PCA of 28,408 VST counts. Dot size and color are proportional to the -log10 p-value, from one-way ANOVA (α = 0.05). (b) MA-Plot of 2,765 DEGs. (Log2FC ≥ 0.01 & Log2FC ≤ -0.01, p.adj. < 0.05, Wald test, Benjamini-Hochberg-corrected).
Supplementary figure S6
(a)
MTT assay evaluating viability in HC fibroblasts cultured alone, with T cell supernatants or with activated T cells. (b) Viability in fibroblasts evaluated by live dead gating on flow plots. Fibroblasts are cultured with IFNγ, PD-1- supernatant or PD-1- cells. (c) Median fluorescence intensity (MFI) for ICAM, VCAM, PD-L1 and PDPL on CD90 + fibroblasts cultured with PD-1 +/- supernatant or cells. Differences are evaluated by Wilcoxon signed rank test.
Supplementary figure S7
(a) Evaluating the 125 shared DEG between PD-1+/- cells and fibroblasts stimulated with PD-1+/- cells. CellEnrichr analysis supporting these DEG to originate from T cell contamination. (b) KEGG pathway analysis on fibroblasts co-cultured with PD-1+/- cells or supernatants.
Declarations
Ethics approval and consent to participate:
Studies were approved by the Danish Data Protection Agency, the Danish Medical Agency, and the Regional Ethics Committee (2012-1329-2). All patients were provided written information and informed written consents were obtained according to the Helsinki declaration. HCs samples cannot be identified or traced.
Consent for publication:
Not applicable
Competing interests
The authors declare thay have no compeeting interests
A
Funding
This study was funded by: The Riisfort foundation, The Lundbeck Foundation (R287-2018-1094), Direktør Jens Aage Sørensen og hustru Edith Ingeborg Sørensens Mindefond, and Fonden til lægevidenskabens fremme.
A
Acknowledgement
We thank the FACS core facility at Aarhus University for help with practical experiments as well as guidance in data analysis. Thanks to lab technician Sulaiman Hussain for sound lab work and for guiding students.
Electronic Supplementary Material
Below is the link to the electronic supplementary material
A
Author Contribution
SR: Performed experiments and data analysis. Initial draft of the manuscripts. CATA: RNAseq experiments and data analysis. Drafting the section of the manuscript. BBA: Performed flow cytometry experiments and data analysis. MV: RNAseq of fibroblasts and data analysis. Conceptual design of fibroblast experiment. YY: RNAseq of fibroblasts and data analysis. Conceptual design of fibroblast experiment. Drafting part of the manuscript. MH: student supervision, project planning and data interpretation. KR: Planning of RNAseq experiments, data interpretation. BD: supervising the project, data interpretations. SRG: project idea and conception, supervision and planning, experiments and data analysis, data interpretation, drafting and refining the manuscript. All authors read, refined and approved the final manuscript.
A
Data Availability
RNA sequencing data on PD-1+ and PD-1- T cells from the synovial joint fluid is available through ArrayExpress; E-MTAB-16009. The RNA sequencing data on HC fibroblasts cultured with PD-1+ and PD-1- T cells or supernatants are also available through ArrayExpress; E-MTAB-15969. The two datasets are marked as associated at ArrayExpress.
References
1.
Scott, D. L., Wolfe, F. & Huizinga, T. W. Rheumatoid arthritis. Lancet 376, 1094–1108. https://doi.org:10.1016/s0140-6736(10)60826-4 (2010).
2.
Wasserman, A. M. Diagnosis and management of rheumatoid arthritis. Am. Fam Physician. 84, 1245–1252 (2011).
3.
Pardoll, D. M. The blockade of immune checkpoints in cancer immunotherapy. Nat. Rev. Cancer. 12, 252–264. https://doi.org:10.1038/nrc3239 (2012).
4.
Fife, B. T. & Bluestone, J. A. Control of peripheral T-cell tolerance and autoimmunity via the CTLA-4 and PD-1 pathways. Immunol. Rev. 224, 166–182. https://doi.org:10.1111/j.1600-065X.2008.00662.x (2008).
5.
Ai, L., Xu, A. & Xu, J. Roles of PD-1/PD-L1 Pathway: Signaling, Cancer, and Beyond. Adv. Exp. Med. Biol. 1248, 33–59. https://doi.org:10.1007/978-981-15-3266-5_3 (2020).
6.
Chen, Y. N. et al. Allosteric inhibition of SHP2 phosphatase inhibits cancers driven by receptor tyrosine kinases. Nature 535, 148–152. https://doi.org:10.1038/nature18621 (2016).
7.
Kurachi, M. CD8(+) T cell exhaustion. Semin Immunopathol. 41, 327–337. https://doi.org:10.1007/s00281-019-00744-5 (2019).
8.
Kumar, P., Saini, S. & Prabhakar, B. S. Cancer immunotherapy with check point inhibitor can cause autoimmune adverse events due to loss of Treg homeostasis. Semin Cancer Biol. 64, 29–35. https://doi.org:10.1016/j.semcancer.2019.01.006 (2020).
9.
Belkhir, R. et al. Rheumatoid arthritis and polymyalgia rheumatica occurring after immune checkpoint inhibitor treatment. Ann. Rheum. Dis. 76, 1747–1750. https://doi.org:10.1136/annrheumdis-2017-211216 (2017).
10.
Tuttle, J. et al. A Phase 2 Trial of Peresolimab for Adults with Rheumatoid Arthritis. N Engl. J. Med. 388, 1853–1862. https://doi.org:10.1056/NEJMoa2209856 (2023).
11.
Abramson, J. & Husebye, E. S. Autoimmune regulator and self-tolerance - molecular and clinical aspects. Immunol. Rev. 271, 127–140. https://doi.org:10.1111/imr.12419 (2016).
12.
Zamani, M. R., Aslani, S., Salmaninejad, A., Javan, M. R. & Rezaei, N. PD-1/PD-L and autoimmunity: A growing relationship. Cell. Immunol. 310, 27–41. https://doi.org:10.1016/j.cellimm.2016.09.009 (2016).
13.
Zhao, L. et al. Circulating CD4(+) FoxP3(-) CXCR5(-) CXCR3(+) PD-1(hi) cells are elevated in active rheumatoid arthritis and reflect the severity of the disease. Int. J. Rheum. Dis. 24, 1032–1039. https://doi.org:10.1111/1756-185x.14170 (2021).
14.
Rao, D. A. et al. Pathologically expanded peripheral T helper cell subset drives B cells in rheumatoid arthritis. Nature 542, 110–114 (2017).
15.
Yoshitomi, H. CXCL13-producing PD-1(hi)CXCR5(-) helper T cells in chronic inflammation. Immunol. Med. 43, 156–160. https://doi.org:10.1080/25785826.2020.1781998 (2020).
16.
Croft, A. P. et al. Distinct fibroblast subsets drive inflammation and damage in arthritis. Nature 570, 246–251. https://doi.org:10.1038/s41586-019-1263-7 (2019).
17.
Kim, J. M., Lin, C., Stavre, Z., Greenblatt, M. B. & Shim, J. H. Osteoblast-Osteoclast Communication and Bone Homeostasis. Cells 9 https://doi.org:10.3390/cells9092073 (2020).
18.
Schett, G. & Osteoimmunology Z. für Rheumatologie 75, 531–533 (2016).
19.
Komatsu, N. & Takayanagi, H. Mechanisms of joint destruction in rheumatoid arthritis - immune cell-fibroblast-bone interactions. Nat. Rev. Rheumatol. 18, 415–429. https://doi.org:10.1038/s41584-022-00793-5 (2022).
20.
Tsaltskan, V. & Firestein, G. S. Targeting fibroblast-like synoviocytes in rheumatoid arthritis. Curr. Opin. Pharmacol. 67, 102304. https://doi.org:10.1016/j.coph.2022.102304 (2022).
21.
Zhang, F. et al. Deconstruction of rheumatoid arthritis synovium defines inflammatory subtypes. Nature 623, 616–624. https://doi.org:10.1038/s41586-023-06708-y (2023).
22.
Kay, J. & Upchurch, K. S. ACR/EULAR 2010 rheumatoid arthritis classification criteria. Rheumatology (Oxford) 51 Suppl 6, vi5-9 (2012). https://doi.org:10.1093/rheumatology/kes279
23.
Pautke, C. et al. Characterization of osteosarcoma cell lines MG-63, Saos-2 and U-2 OS in comparison to human osteoblasts. Anticancer Res. 24, 3743–3748 (2004).
24.
Køster, D. et al. Phenotypic and functional characterization of synovial fluid-derived fibroblast-like synoviocytes in rheumatoid arthritis. Sci. Rep. 11, 22168. https://doi.org:10.1038/s41598-021-01692-7 (2021).
25.
Stebulis, J. A., Rossetti, R. G., Atez, F. J. & Zurier, R. B. Fibroblast-like synovial cells derived from synovial fluid. J. Rheumatol. 32, 301–306 (2005).
26.
Kuleshov, M. V. et al. Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic Acids Res. 44, W90–97. https://doi.org:10.1093/nar/gkw377 (2016).
27.
Baker, K. F. et al. Single-cell insights into immune dysregulation in rheumatoid arthritis flare versus drug-free remission. Nat. Commun. 15, 1063. https://doi.org:10.1038/s41467-024-45213-2 (2024).
28.
TOX is expressed. by exhausted and polyfunctional human effector memory CD8. 1–14 (2020).
29.
Yang, M., Zhu, L. & Osteoimmunology The Crosstalk between T Cells, B Cells, and Osteoclasts in Rheumatoid Arthritis. Int. J. Mol. Sci. 25 https://doi.org:10.3390/ijms25052688 (2024).
30.
Greisen, S. R. et al. The Programmed Death-1 Pathway Counter-Regulates Inflammation-Induced Osteoclast Activity in Clinical and Experimental Settings. Front. Immunol. 13, 773946. https://doi.org:10.3389/fimmu.2022.773946 (2022).
31.
Lee, S. C. et al. Immunomodulatory Effect and Bone Homeostasis Regulation in Osteoblasts Differentiated from hADMSCs via the PD-1/PD-L1 Axis. Cells 11 https://doi.org:10.3390/cells11193152 (2022).
A
32.Bai, Z. et al. Synovial fibroblast gene expression is associated with sensory nerve growth and pain in rheumatoid arthritis. Sci. Transl Med. 16, eadk3506. https://doi.org:10.1126/scitranslmed.adk3506 (2024).
A
33.Orange, D. E. et al. RNA Identification of PRIME Cells Predicting Rheumatoid Arthritis Flares. N Engl. J. Med. 383, 218–228. https://doi.org:10.1056/NEJMoa2004114 (2020).
34.
Hu, X. et al. Deconvolution of synovial myeloid cell subsets across pathotypes and role of COL3A1 + macrophages in rheumatoid arthritis remission. Front. Immunol. 15, 1307748. https://doi.org:10.3389/fimmu.2024.1307748 (2024).
35.
Smith, M. H. et al. Drivers of heterogeneity in synovial fibroblasts in rheumatoid arthritis. Nat. Immunol. 24, 1200–1210. https://doi.org:10.1038/s41590-023-01527-9 (2023).
36.
Wu, M., Wu, S., Chen, W. & Li, Y. P. The roles and regulatory mechanisms of TGF-beta and BMP signaling in bone and cartilage development, homeostasis and disease. Cell. Res. 34, 101–123. https://doi.org:10.1038/s41422-023-00918-9 (2024).
37.
Utzschneider, D. T. et al. Early precursor T cells establish and propagate T cell exhaustion in chronic infection. Nat. Immunol. 21, 1256–1266. https://doi.org:10.1038/s41590-020-0760-z (2020).
38.
Zhao, Y., Qu, Y., Hao, C. & Yao, W. PD-1/PD-L1 axis in organ fibrosis. Front. Immunol. 14, 1145682. https://doi.org:10.3389/fimmu.2023.1145682 (2023).
39.
Zhao, M. et al. Targeting fibrosis, mechanisms and cilinical trials. Signal. Transduct. Target. Ther. 7, 206. https://doi.org:10.1038/s41392-022-01070-3 (2022).
40.
Wynn, T. A. & Ramalingam, T. R. Mechanisms of fibrosis: therapeutic translation for fibrotic disease. Nat. Med. 18, 1028–1040. https://doi.org:10.1038/nm.2807 (2012).