Transcription factor SPI1 drives immunosuppressive CD45+ erythroid progenitor cells to ameliorate rheumatoid arthritis

Abstract

Objective

Anaemia is common in rheumatoid arthritis (RA), but the role of erythroid-lineage cells is unclear. We investigated the function of CD45⁺ erythroid progenitor cells (CD45⁺ EPCs) in RA.

Methods

We analysed CD45⁺ EPC frequency in patients with RA and mice with collagen-induced arthritis (CIA). Transcriptomics, functional studies and mechanistic assays (Transwell and dual-luciferase reporter assays, chromatin immunoprecipitation followed by quantitative PCR) were used. Therapeutic potential was tested in RA synovial organoids and via splenectomy/adoptive transfer in mice with CIA.

Results

CD45⁺ EPCs were expanded in RA circulation and CIA mouse spleens, correlating positively with disease activity and negatively with haemoglobin. They displayed an immunosuppressive transcriptome, enriched for transforming growth factor (TGF)-β and chemokine signalling. RA-derived CD45⁺ EPCs showed enhanced proliferation and TGF-β/reactive oxygen species production. High C-C Motif Chemokine Receptor 2 (CCR2) expression made them susceptible to recruitment by macrophage-derived C-C Motif Chemokine Ligand 2 (CCL2). In RA synovial organoids, CD45⁺ EPCs suppressed growth and inflammation via TGF-β, while organoid-conditioned media promoted their migration via CCL2. Recruited CD45⁺ EPCs suppressed M1 and promoted M2-like macrophage polarisation. The transcription factor SPI1 was upregulated in RA CD45⁺ EPCs, bound the TGFB1 promoter and drove TGF-β production. In vivo, splenectomy worsened CIA, whereas adoptive transfer of CD45⁺ EPCs ameliorated arthritis.

Conclusion

We identify CD45⁺ EPCs as a novel, SPI1-driven immunosuppressive population in RA. Recruited via the CCR2-CCL2 axis, they attenuate inflammation by modulating macrophages through SPI1/TGF-β signalling, revealing a new immunoregulatory axis and potential therapeutic targets.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Although anaemia is a common comorbidity in rheumatoid arthritis (RA) and CD45⁺ erythroid progenitor cells (CD45⁺ EPCs) with immunomodulatory potential have been identified in some diseases, their specific role and regulatory mechanism in RA pathogenesis remain largely unexplored.

WHAT THIS STUDY ADDS

• CD45⁺ EPCs are expanded in patients with RA and function as a novel immunosuppressive population.

• Recruited via the CCR2-CCL2 axis, CD45⁺ EPCs attenuate arthritis by skewing macrophages towards an M2-like phenotype through transforming growth factor (TGF)-β.

• The transcription factor SPI1 drives this immunoregulatory function by directly binding the TGFB1 promoter to enhance TGF-β production in CD45⁺ EPCs.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Targeting the SPI1-TGF-β axis in CD45⁺ EPCs or harnessing their immunomodulatory function could represent a novel therapeutic strategy for RA.

Introduction

Rheumatoid arthritis (RA) is a systemic autoimmune disease with a global prevalence of approximately 1%, characterised primarily by chronic synovitis, progressive bone and cartilage erosion and various immune system dysfunctions.¹,³ Despite extensive research, the comprehensive pathogenesis of RA remains incompletely elucidated.²,⁴ Anaemia happens in 64% of patients with RA, the severity of which has been observed to correlate with joint symptom activity.² However, the precise pathophysiological role and impact of anaemia in RA development and progression are not well defined.⁵

The pathology of RA is marked by a significant imbalance in immune cell subsets, with well-established roles for macrophages, fibroblasts and T cells in driving inflammation and tissue damage.^{2 4 6} Beyond these classical players, a population of CD45⁺ erythroid progenitor cells (CD45⁺ EPCs) has recently gained attention.⁷ These cells expand in conditions associated with anaemia, such as cancer,^{8 9} chronic hepatitis B infection,¹⁰ autoimmune diseases¹¹ and pregnancy,¹² and are reported to possess potent immunomodulatory capabilities, through the secretion of factors like TGF-β, interleukin (IL)-10, reactive oxygen species (ROS) and arginase, CD45⁺ EPCs can suppress T-cell proliferation and function.^{11 13 14} Notably, the frequency of CD45⁺ EPCs correlates positively with the severity of anaemia. While their potential interactions with macrophages have been noted in former research,^{8 15} the role of CD45⁺ EPCs and their interactions with macrophages in RA is largely unexplored.

The transcription factor SPI1 (also known as PU.1), a member of the E26 transformation-specific (ETS) family, is a critical regulator of haematopoiesis and the development and differentiation of multiple immune cell lineages, including T cells and dendritic cells.¹⁶,¹⁹ Dysregulation of SPI1 has been implicated in various autoimmune diseases, yet its specific function and mechanisms within the cellular landscape of RA are still unclear.^{20 21}

Given that SPI1 is a master transcriptional regulator of immune cell function,¹⁶,¹⁹ and considering the immunomodulatory nature of CD45⁺ EPCs,^{7 11 13} we hypothesise that SPI1 may influence RA pathogenesis by modulating the function of CD45⁺ EPCs, particularly their production of TGF-β.²²,²⁶ This study aims to investigate the potential role of CD45⁺ EPCs in RA progression and to explore the hypothesis that SPI1 exerts a protective effect by enhancing their TGF-β-mediated immunoregulatory functions.

Methods

Patients and healthy donors

The diagnosis of RA followed the 2010 clinical and radiological criteria outlined by the American College of Rheumatology and EULAR. Between June 2023 and September 2025, 50 patients with RA were enrolled in this study. The patients underwent treatment for RA at the Third Affiliated Hospital of Sun Yat-sen University, Guangzhou, China. A total of 31 healthy adult donors were included as controls. All patients and healthy controls were screened for serum HIV antibodies, hepatitis B surface antigen, hepatitis C virus antibodies, hepatitis D virus antigens and hepatitis D virus antibodies. Patients and healthy controls who were positive for HIV, hepatitis infection or other acute infections (including pneumonia and urinary tract infections); who were pregnant or who had a fever were excluded from this study. Demographic and clinical characteristics of the study populations are shown in online supplemental table 1.

Mice

Male DBA/1JGpt mice, aged 8 weeks, were purchased from the Guangdong Medical Laboratory Animal Center (Guangzhou, China). All mouse experiments were conducted in pathogen-free facilities.

Induction of collagen-induced arthritis

On day 0, mice (n=60, weight: 20.23±0.63 g) were anaesthetised via intraperitoneal injection of 1% pentobarbital sodium (160 μL per 25 g body weight). Then, the fur on the back was shaved. For initial immunisation, each mouse received an intradermal injection of 0.1 mL of the Bovine Type II Collagen (CII, Chondrex, WA, USA)/Freund's Complete Adjuvant (CFA, Thermo Scientific, Shanghai, China) emulsion at multiple sites (approximately 3–5 sites) on the shaved back. On day 21, a booster immunisation was administered via an intradermal injection of 0.10 mL of the emulsion at the base of the tail.

Real-time quantitative reverse transcription PCR

RNA was extracted using the Multisource Total RNA Miniprep Kit (Axygen, Union City, California, USA). The purity and concentration of RNA were determined by measuring absorbance at A260/280 nm using NanoDrop 2000 (Thermo Fisher Scientific). Quantitative reverse transcription PCR (qRT-PCR) was performed using commercially available primers (online supplemental table 2) and SYBR Premix Ex Taq II (Code, DRR081; Takara Biotechnology, Dalian, China).

Fluorescence for each cycle was quantitatively analysed using the ABI Prism 7000 sequence detection system (Life Technologies, Carlsbad, California, USA). The results are reported as relative expression, normalised with Actin Beta (ACTB) as a housekeeping gene endogenous control and expressed in arbitrary units.

CD45⁺ EPC transfer experiment

CD45⁺ EPCs were isolated by FACS from spleen of mouse with CIA. Then CD45⁺ EPCs were stained with 5-(and -6)-carboxyfluorescein diacetate succinimidyl ester (CFSE, 2.5 µM, Tonbo Biosciences, San Diego, California, USA) for 10 min at 37°C. Next, 1.5×10⁶ CD45⁺ EPCs were transferred into mice with CIA via caudal vein injection (n=9, weight: 20.23±0.64). After 12 hours, three mice were sacrificed, and CFSE⁺ CD45⁺ EPCs were assessed in the blood and spleen. Other mice were sacrificed after 15 days for analysis of joint inflammation.

Bulk RNA-sequencing

CD45⁺ EPCs sorted from peripheral blood mononuclear cells (PBMCs) of patients with RA or healthy donors were analysed through high-throughput transcriptome sequencing at Beijing Genomics Institute (BGI, Shenzhen, China) to identify aberrantly and specifically expressed genes. Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathway analysis was performed to estimate the function of dysregulated genes. Sequencing and analysis were performed according to BGI’s standard RNA-sequencing (RNA-seq) protocol.

Single cell RNA-seq

CD45⁺ EPCs sorted from PBMCs of patients with RA were analysed using DNBelab C4 Single-Cell RNA-seq at BGI. Sequencing and analysis were performed according to BGI’s standard RNA‑Seq protocol.

CD45⁺ EPC migration analysis

Cell migration was assessed using a Transwell system. The lower chamber contained complete medium (control), Tohoku Hospital Pediatrics-1 (THP-1)-derived macrophages, macrophages with anti-CCL2 neutralising antibody (BioLegend, San Diego, CA, USA), human RA fibroblast-like synoviocytes cell line (MH7-A) or Human Umbilical Vein Endothelial Cells (HUVEC). CD45⁺ EPCs were seeded in the upper chamber. After 48 hours of incubation, migrated cells in the lower chamber were quantified by cell counting. Conditioned medium was collected and stored at −80°C for further analysis.

Intracellular staining for cytokines and transcription factors

For intracellular cytokine staining, freshly isolated mononuclear cells were stimulated with Cell Stimulation Cocktail (eBioScience, California, USA) for 4 hours at 37°C. For transcription factor analysis, unstimulated cells were used. Cells were first stained with surface marker antibodies, then fixed and permeabilised using the Foxp3/Transcription Factor Staining Buffer Set (eBioScience) or the Intracellular Fixation and Permeabilisation Buffer Set (eBioScience) at 4°C overnight. After phosphate-buffered saline washing, cells were stained with antibodies against intracellular cytokines or transcription factors. All samples were analysed by flow cytometry.

Statistical analysis

Variables in different groups were compared using the χ² test (or Fisher’s exact test, if indicated) and the t-test or non-parametric Mann-Whitney U test. For in vitro experiments, statistical analyses were conducted using unpaired t-tests. Correlations between different parameters were analysed using the Spearman’s rank test or logistic regression. The criterion for statistical significance was set at α=0.05, and all p values were based on two-sided tests. Statistical tests were performed using GraphPad Prism V.9.0, R statistical software V.3.6.0 and SPSS Statistics V.25.0. Descriptions of the other methods are provided in the online supplemental material. The antibodies and reagents used are listed in online supplemental table 3. The gating strategy for cytometry is shown in online supplemental figures 3-7.

Results

Expansion of CD45⁺ EPCs in patients with RA and mice with arthritis

We first assessed the frequency of CD45⁺ EPCs in PBMCs from patients with RA and healthy controls. Flow cytometric analysis revealed a significantly higher proportion of CD45⁺ EPCs in patients with RA compared with normal controls (figure 1A,B). To understand the clinical relevance of this expansion, we performed correlation analyses. The proportion of CD45⁺ EPCs showed a significant positive correlation with the Disease Activity Score 28 (DAS-28) (r=0.5742, p=0.003; figure 1C) and a significant negative correlation with haemoglobin (HGB) concentration (r=0.6166, p=0.0001; figure 1D).

Figure 1. Expansion of CD45⁺ EPCs in patients with RA and mice with arthritis. (A) Gating strategy for flow cytometric analysis of human CD45⁺ EPCs (CD45⁺CD71⁺CD235a⁺). (B) Flow cytometric analysis of the frequency of CD45⁺ EPCs in PBMCs from HCs and patients with RA. (C) Correlation analysis between the proportion of CD45⁺ EPCs and DAS28. (D) Correlation analysis between the proportion of CD45⁺ EPCs and HGB concentration in patients with RA. (E) Representative immunofluorescence images showing the presence of CD45⁺ EPCs (white arrows) in synovial tissues from patients with RA. Nuclei are stained with 4',6-diamidino-2-phenylindole (DAPI, blue). Scale bar, 100 μm. (F) Gating strategy for flow cytometric analysis of mouse CD45⁺ EPCs (CD45⁺CD71⁺TER119⁺). (G) Flow cytometric analysis of the frequency of CD45⁺ EPCs (CD45⁺CD71⁺TER119⁺) in splenocytes from healthy mice and mice with CIA. (H) Representative Wright-Giemsa staining of sorted human (left) and murine (right) CD45⁺ EPCs. Scale bar, 5 μm. Data are presented as mean±SEM. **p<0.01, ***p<0.001. CIA, collagen-induced arthritis; DAS28, Disease Activity Score in 28 joints; EPC, erythroid progenitor cell; FSC, forward scatter; HC, healthy control; HGB, haemoglobin; PBMCs, peripheral blood mononuclear cells; RA, rheumatoid arthritis; SSC, side scatter.

We next examined the synovium, the primary site of inflammation in RA. Immunofluorescence staining of synovial tissues from patients with RA confirmed the presence of CD45⁺ EPCs; however, their infiltration was notably scarce (figure 1E), suggesting a potential inverse relationship between local CD45⁺EPC abundance and joint inflammation.

To corroborate these findings in a controlled experimental setting, we employed the CIA mouse model.²⁷ Consistent with the human data, the proportion of CD45⁺ EPCs was significantly elevated in the spleens of mice with CIA compared with healthy controls (figure 1F and G). To phenotypically characterise these cells, CD45⁺ EPCs were sorted from human PBMCs and murine spleens and subjected to Wright-Giemsa staining. Morphological analysis identified these cells as erythroid precursors at stages ranging from polychromatic to orthochromatic erythroblasts (figure 1H).

CD45⁺ EPCs ameliorate arthritis in a CIA mouse model

To evaluate the in vivo function of CD45⁺ EPCs in RA, we first performed splenectomy in a CIA model.^{27 28} Splenectomy significantly reduced the proportion of CD45⁺ EPCs in PBMCs of mice with CIA (figure 2A). Compared with sham-operated controls, splenectomised mice exhibited more severe arthritis, as indicated by significantly higher clinical arthritis scores (figure 2B) and increased ankle joint swelling (figure 2C). Histopathological analyses further confirmed aggravated joint pathology in splenectomised mice, including enhanced inflammatory cell infiltration shown by H&E staining (figure 2D), greater cartilage destruction revealed by Safranin O–Fast Green staining (figure 2E) and increased osteoclast formation demonstrated by tartrate-resistant acid phosphatase staining (figure 2F).

Figure 2. CD45⁺ EPCs ameliorate arthritis in a CIA mouse model. (A) Flow cytometric analysis of CD45⁺ EPC proportion in PBMCs from splenectomised (SpleenX) or sham-operated mice with CIA (n=6). (B, C) Splenectomised mice with CIA developed more severe arthritis, as shown by higher clinical scores (B) and increased ankle swelling (C). (D–F) Histopathological analysis of ankle joints from SpleenX (n=6) or sham-operated mice with CIA (n=6). H&E staining for inflammation (D), Safranin O–Fast Green for cartilage (E) and TRAP for osteoclasts (F). Scale bar, 500 μm. (G) Homing of adoptively transferred CFSE-labelled CD45⁺ EPCs to the spleen, detected by flow cytometry 1 day post-transfer (n=3). (H, I) Mice with CIA receiving CD45⁺ EPC transfer showed lower clinical scores (H) and reduced ankle swelling (I).(J–L) Histopathological analysis of ankle joints from mice with CIA receiving CD45⁺ EPC transfer (n=6) or PBS (n=6). H&E staining for inflammation (J), Safranin O–Fast Green for cartilage (K) and TRAP for osteoclasts (L). Scale bar, 500 μm. Data are presented as mean±SEM. *p<0.05. CFSE, carboxyfluorescein diacetate succinimidyl ester; CIA, collagen-induced arthritis; EPC, erythroid progenitor cell; PBMCs, peripheral blood mononuclear cells; PBS, phosphate-buffered saline; TRAP, tartrate-resistant acid phosphatase.

To directly assess the therapeutic potential of CD45⁺ EPCs, we adoptively transferred CFSE-labelled CD45⁺ EPCs isolated from spleens of mice with CIA into recipient mice via tail vein injection after the second immunisation. Analysis of mice 1 day post-transfer confirmed the homing of CFSE⁺ CD45⁺ EPCs to the spleen via peripheral circulation (figure 2G). Mice receiving CD45⁺ EPC transfer showed significant alleviation of arthritis, as evidenced by lower clinical scores (figure 2H), reduced ankle swelling (figure 2I), diminished inflammatory infiltration (figure 2J), attenuated cartilage damage (figure 2K) and decreased osteoclast numbers (figure 2L).

To further investigate the therapeutic potential of CD45⁺ EPCs in a splenectomised setting, we performed adoptive transfer of these cells into mice with CIA that had undergone splenectomy (online supplemental figure S1A). Mice receiving CD45⁺ EPC transfer exhibited a significant alleviation of arthritis severity compared with control splenectomised mice with CIA. This was evidenced by lower clinical arthritis scores over time (online supplemental figure S1B), reduced ankle joint swelling (online supplemental figure S1C) and improved histopathological outcomes. Histological analysis revealed markedly diminished inflammatory cell infiltration (online supplemental figure S1D), attenuated cartilage damage (online supplemental figure S1E) and a decrease in osteoclast numbers (online supplemental figure S1F) in the treatment group. These results support the immunosuppressive role of CD45⁺ EPCs.¹⁵

Bulk RNA-seq reveals an immunosuppressive gene signature in RA-derived CD45⁺ EPCs

To investigate the functional properties of CD45⁺ EPCs in RA, we performed bulk RNA-seq on these cells sorted from the PBMCs of patients with RA and healthy controls. Comparative analysis identified 422 differentially expressed genes (DEGs) between the two groups, with 221 genes upregulated and 201 genes downregulated in RA-derived CD45⁺ EPCs (figure 3A–C).

Figure 3. Bulk RNA-seq reveals an immunosuppressive gene signature in RA-derived CD45⁺ EPCs. (A) Heatmap showing the expression patterns of 422 DEGs between CD45⁺ EPCs from patients with RA and HCs. (B) Bar plot illustrating 221 upregulated genes and 201 downregulated genes in RA versus HC CD45⁺ EPCs. (C) Volcano plot displaying the DEGs. (D) KEGG pathway enrichment analysis of DEGs. (E) GO pathway enrichment analysis of DEGs. (F) GSEA plots of RA versus HC CD45⁺ EPCs. Data are representative of at least three independent experiments. DEG, differentially expressed genes; GO, Gene Ontology; GSEA, gene set enrichment analysis; HC, healthy control; IL, interleukin; KEGG, Kyoto Encyclopaedia of Genes and Genomes; NES, Normalised Enrichment Score; RA, rheumatoid arthritis; RNA-seq, RNA-sequencing; TGF, transforming growth factor; TNF, tumour necrosis factor.

To elucidate the potential biological pathways involved, we performed KEGG and Gene Ontology enrichment analyses on the DEGs. These analyses revealed significant enrichment for pathways related to RA, cytokine-cytokine receptor interaction and chemokine signalling pathway (figure 3D and E). Furthermore, gene set enrichment analysis demonstrated significant positive enrichment of gene sets associated with inflammatory response, TGF-β signalling and ROS metabolism (figure 3F).

Single-cell RNA-seq reveals a proliferating, immunosuppressive subset of CD45⁺ EPCs in RA

To dissect the heterogeneity of CD45⁺ EPCs in RA, we performed single-cell RNA-seq on these cells sorted from PBMCs of patients with RA. Unsupervised clustering analysis identified four distinct subpopulations (clusters 0–3), demonstrating previously unappreciated cellular diversity within the CD45⁺ EPC pool (figure 4A). Marker gene expression analysis defined unique transcriptional profiles for each cluster (figure 4B).

Figure 4. Single-cell RNA sequencing reveals a proliferating and immunosuppressive CD45⁺ EPC subset in RA. (A) UMAP projection of CD45⁺ EPCs from PBMCs of patients with RA, coloured by four distinct clusters (0–3). (B) Heatmap of top marker genes defining the identity of each cluster. (C) UMAP projection highlighting CD45⁺ EPCs in G2/M phase (red). (D–G) Pseudotime trajectory analysis. UMAP coloured by different state (D). UMAP coloured by inferred pseudotime (E). Expression patterns of TGFB1 (F) and SPI1 (G) overlaid on the pseudotime trajectory. (H–L) Dot plots showing high expression of immunomodulatory genes (TGFB1, GPX4, CCR2, SPI1) in cluster. (M–O) Dot plot showing minimal or absent expression of pro-inflammatory cytokines (TNFA, IL1B, IL6) across all clusters. Data are representative of cells sorted from multiple patients with RA. CCR2, C-C Motif Chemokine Receptor 2; EPC, erythroid progenitor cell; GPX4, Glutathione Peroxidase 4; IL, interleukin; PBMCs, peripheral blood mononuclear cells; RA, rheumatoid arthritis; SPI1, Spleen Focus Forming Virus (SFFV) Proviral Integration Oncogene; TGF, transforming growth factor; TNFA, tumour necrosis factor Alpha; UMAP, Uniform Manifold Approximation and Projection.

Analysis of cell cycle phase assignment revealed that 188 cells were in G2/M phase, providing direct evidence that a proportion of CD45⁺ EPCs in patients with RA are actively proliferating (figure 4C).²⁹

We next applied pseudotime trajectory analysis to reconstruct the developmental relationship between these clusters. This analysis revealed a continuous differentiation trajectory from a putative resting state to an activated, immunosuppressive state. Notably, the expression of key immunosuppressive mediators, including TGFB1, and the transcription factor SPI1 increased progressively along this pseudotime axis (figure 4D–G).

Further characterisation of the four clusters identified cluster 1 as the most immunomodulatory, exhibiting the highest expression levels of TGFB1, GPX4, CCR2 and SPI1 (figure 4H–L). Importantly, all four clusters showed low or undetectable expression of classic pro-inflammatory cytokines such as TNFA, IL1B and IL6, reinforcing a non-inflammatory phenotype (figure 4M–O).

The proliferative and immunosuppressive phenotype of CD45⁺ EPCs in RA

To functionally characterise CD45⁺ EPCs in RA, we first assessed their viability and proliferative capacity. Using flow cytometry with an apoptosis assay kit, we found that the viability of CD45⁺ EPCs was significantly higher in patients with RA compared with healthy donors (figure 5A). Furthermore, 5-ethynyl-2’-deoxyuridine (EdU) incorporation assays revealed a markedly increased proportion of proliferating CD45⁺ EPCs in patients with RA (figure 5B), indicating enhanced cellular expansion in vivo.

Figure 5. CD45⁺ EPCs from patients with RA exhibit enhanced proliferation and an immunosuppressive phenotype. (A) Flow cytometry analysis for viability of CD45⁺ EPCs assessed by Annexin V/7AAD staining in patients with RA (n=6) and HCs (n=6). (B) Flow cytometry analysis for proliferation of CD45⁺ EPCs measured by EdU incorporation assay (n=6). (C, D) mRNA expression levels of TGFB1 (C) and IL10 (D) in CD45⁺ EPCs from HCs and patients with RA (n=3), as determined by quantitative PCR. (E) Protein levels of TGF-β and IL-10 in CD45⁺ EPCs analysed by intracellular cytokine staining through flow cytometry (n=6). (F, G) Levels of TGF-β and IL-10 in conditioned medium of CD45⁺ EPCs were measured by ELISA (n=4). (H) Flow cytometry analysis for intracellular ROS levels in CD45⁺ EPCs from HCs (n=6) and patients with RA (n=6) by DCFH-DA staining (n=6). Data are presented as mean±SEM. *p<0.05; **p<0.01. 7AAD, 7-aminoactinomycin D; DCFH-DA, 2′,7′-dichlorodihydrofluorescein diacetate; EPC, erythroid progenitor cell; HC, healthy control; IL, interleukin; MFI, Mean Fluorescence Intensity; mRNA, messenger RNA; NS, not significant; RA, rheumatoid arthritis; ROS, reactive oxygen species; TGF, transforming growth factor.

We next investigated the expression of immunomodulatory molecules. qPCR analysis demonstrated that CD45⁺ EPCs from patients with RA expressed significantly higher levels of TGFB1 mRNA compared with those from healthy donors. In contrast, the expression level of IL10 mRNA showed no significant difference between the two groups (figure 5C and D). This specific upregulation of TGFB1 was confirmed at the protein level by flow cytometry intracellular cytokine staining and ELISA, which again showed no difference in IL-10 production (figure 5E–G).

Finally, given the role of ROS as an immunosuppressive mechanism,¹³ we measured ROS levels in CD45⁺ EPCs. Flow cytometric analysis showed a significant increase in ROS production in CD45⁺ EPCs from patients with RA relative to healthy controls (figure 5H).

CD45⁺ EPCs are recruited via the CCR2-CCL2 axis and inhibited the M1-like macrophage phenotype

Building on our transcriptomic data, which highlighted enrichment of chemokine signalling pathways in RA CD45⁺ EPCs (figure 3D), we focused on the chemokine receptor CCR2, which was significantly upregulated at the transcriptional level, as shown in the associated heatmap and volcano plot (online supplemental figure S1C–D). This finding was validated by both qPCR and flow cytometry, confirming higher CCR2 expression on CD45⁺ EPCs from patients with RA compared with healthy donors (figure 6A).

Figure 6. CD45⁺ EPCs are recruited via the CCR2-CCL2 axis and inhibit the M1-like macrophage phenotype. (A, B) CCR2 expression is elevated on CD45⁺ EPCs from patients with RA compared with HCs, as shown by qPCR (A) and flow cytometry (B). (C) Transwell migration assay of CD45⁺ EPCs induced by macrophages (MΦ), fibroblast cells, macrophages+CCL2-neutralising antibody blockage, Human Umbilical Vein Endothelial Cells (HUVEC). (D) qPCR of TNF, IL1B, IL-6, CD80, CD86 in macrophages co-cultured with PBMCs (Con) or CD45⁺ EPCs in a Transwell system. (E) ELISA of TNF-α, IL-1β and IL-6 from macrophages co-cultured with PBMCs (Con) or CD45⁺ EPCs in a Transwell system. (F–H) Flow cytometric analysis of M1 macrophage markers, CD80 (F), CD86 (G), ROS (H) in macrophages co-cultured with PBMCs (Con) or CD45⁺ EPCs. (I) qPCR of IL10, TGFB1, ARG1, CD206, CD163 in macrophages co-cultured with PBMCs (Con) or CD45⁺ EPCs in a Transwell system. (J) ELISA of IL-10 and TGF-β from macrophages co-cultured with PBMCs (Con) or CD45⁺ EPCs in a Transwell system. (K) Flow cytometric analysis of CD163 and CD206 in macrophages co-cultured with PBMCs (Con) or CD45⁺ EPCs. Data are presented as mean±SEM. *p<0.05; **p<0.01, ***p<0.001, ****p<0.0001. CCL, C-C Motif Chemokine Ligand; CCR, C-C Motif Chemokine Receptor; Con, control; EPC, erythroid progenitor cell; HC, healthy control; IL, interleukin; MFI, Mean Fluorescence Intensity; ns, not significant; PBMCs, peripheral blood mononuclear cells; qPCR, quantitative PCR; RA, rheumatoid arthritis; ROS, reactive oxygen species; TGF, transforming growth factor; TNF, tumour necrosis factor.

Since the primary ligand for CCR2 is CCL2 (also known as monocyte chemoattractant protein-1, MCP-1), a chemokine predominantly secreted by macrophages,³⁰,³⁴ we hypothesised a recruitment mechanism. Using a Transwell migration assay, we demonstrated that macrophages, but not fibroblasts or HUVEC, robustly attracted CD45⁺ EPCs. This chemotactic effect was effectively blocked by a CCL2-neutralising antibody, establishing the CCR2-CCL2 axis as the key mechanism for macrophage-mediated recruitment of CD45⁺ EPCs (figure 6B).

To investigate the functional impact of CD45⁺ EPCs on macrophages, we employed a Transwell co-culture system that prevents direct cell-cell contact. Co-culture with CD45⁺ EPCs significantly suppressed the production of the pro-inflammatory cytokines TNF and IL-1β by macrophages, while leaving IL-6 levels unchanged, as measured by ELISA (figure 6C). Furthermore, flow cytometric analysis revealed that CD45⁺ EPCs inhibited macrophage M1 polarisation, evidenced by reduced expression of TNFA, IL1B, CD80 and CD86, decreased ROS production (figure 6D–H). CD45⁺ EPCs concurrently promoted an M2-like phenotype, indicated by increased expression of IL10, TGFB1, CD163 and CD206 (figure 6I–L).⁶

CD45⁺ EPCs attenuate rheumatoid arthritis organoid inflammation through a TGF-β-dependent mechanism and are recruited by CCL2

To model the joint microenvironment in vitro,^{35 36} we established synovial organoids derived from patients with RA and performed co-culture experiments. Phase-contrast microscopy revealed that CD45⁺ EPCs significantly suppressed the growth of RA organoids over a 7-day period. This suppressive effect was effectively reversed by the addition of a TGF-β neutralising antibody (figure 7A), indicating a TGF-β-dependent mechanism.^{23 24 26}

Figure 7. CD45⁺ erythroid progenitor cells attenuate RA organoid inflammation through a TGF-β-dependent mechanism and are recruited by CCL2. (A) Representative phase-contrast microscopy images showing the morphology of RA synovial organoids over a 7-day culture period under the indicated conditions: Con, co-culture with CD45⁺ EPCs and co-culture with CD45⁺ EPCs in the presence of a TGF-β neutralising antibody. Scale bar, 100 μm. Co-culture with CD45⁺ EPCs significantly suppressed organoid growth, an effect reversed by TGF-β blockade. (B–G) Cytokine quantification in conditioned media and chemotaxis assay. Co-culture with CD45⁺ EPCs significantly increased the secretion of the anti-inflammatory cytokines (B) TGF-β and (C) IL-10, while reducing the levels of the pro-inflammatory cytokines (D) IL-1β and (E) TNF-α. These effects were partially reversed by neutralising TGF-β. (F) Secretion of the chemokine CCL2 was elevated in the co-culture system. (G) Conditioned media from RA organoids promoted the migration of CD45⁺ EPCs in a transwell assay, which was substantially inhibited by an anti-CCL2 neutralising antibody. Data are presented as mean±SD. Statistical significance was determined by one-way ANOVA with Tukey’s post hoc test (B–E, G) or Student’s t-test (F). *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. CCL2, C-C Motif Chemokine Ligand 2; Con, control; EPC, erythroid progenitor cell; IL, interleukin; RA, rheumatoid arthritis; TGF, transforming growth factor; TNF, tumour necrosis factor.

Cytokine profiling of the conditioned media demonstrated that co-culture with CD45⁺ EPCs markedly increased the secretion of the anti-inflammatory cytokines TGF-β and IL-10 (figure 7B and C). Conversely, the levels of the key pro-inflammatory cytokines TNF-α and IL-1β were significantly reduced (figure 7D and E). Neutralisation of TGF-β partially reversed these cytokine shifts, attenuating the increase in IL-10 and restoring the secretion of TNF-α and IL-1β, confirming that CD45⁺ EPCs drive the microenvironment towards an anti-inflammatory state via TGF-β.

Furthermore, we identified a potential recruitment loop between CD45⁺ EPCs and RA organoids. The co-culture system led to a significant increase in the secretion of the chemokine CCL2 (figure 7F). Conditioned media from RA organoids potently induced the migration of CD45⁺ EPCs, and this migratory response was substantially inhibited by a neutralising anti-CCL2 antibody (figure 7G). These data suggest that the inflamed joint microenvironment may recruit CD45⁺ EPCs via CCL2,^{30 31 33 34} and the recruited cells subsequently exert local immunosuppressive functions through TGF-β, forming a potential negative feedback regulatory circuit.

The transcription factor SPI1 directly promotes TGF-β expression in CD45⁺ EPCs

Our initial transcriptomic analysis of TGF-β-related transcription factors revealed a marked upregulation of SPI1 in RA CD45⁺ EPCs (online supplemental figure S2A and B). We validated this finding, confirming that SPI1 expression was significantly elevated at both the mRNA and protein levels in RA CD45⁺ EPCs compared with controls (figure 8A,B).

Figure 8. SPI1 directly promotes TGF-β expression in CD45⁺ EPCs. (A, B) SPI1 expression is significantly upregulated in RA CD45⁺ EPCs compared with controls, as shown by mRNA levels (qPCR, A) and protein levels (flow cytometry, B). (C, D) Overexpression of SPI1 in CD45⁺ EPCs enhances TGF-β expression, measured by qPCR (C) and flow cytometry (D). (E, F) Knocking down of SPI1 in CD45⁺ EPCs decreases TGF-β expression, measured by qPCR (E) and flow cytometry (F). (G–J) Bioinformatic analysis predicts SPI1 binding motifs in the promoter regions of human and mouse TGFB1 genes (JASPAR database; G–J). (K, L) SPI1 is identified as a potential master transcriptional regulator of TGFB1 in EPCs (Cistrome DB analysis; K, L). (M) Luciferase reporter assay demonstrates direct binding of SPI1 to TGF-β promoter region in CD45⁺ EPCs (n=3). (N) ChIP-qPCR confirms direct binding of SPI1 protein to TGFB1 promoter region in CD45⁺ EPCs (n=3). Data are presented as mean±SEM. *p<0.05; **p<0.01, ***p<0.001, ****p<0.0001. ChIP, chromatin immunoprecipitation; EPCs, erythroid progenitor cells; HC, healthy control; H. sapiens, Homo sapiens; M. musculus, Mus musculus; qPCR, quantitative PCR; RA, rheumatoid arthritis; SSC, side scatter; TGF/Tfg, transforming growth factor; MFI, Mean Fluorescence Intensity; OE, overexpression; NC, Negative Control; pc-DNA, plasmid coding DNA.

To investigate the functional consequence of SPI1 upregulation,^{16 17 20 21} we overexpressed SPI1 in CD45⁺ EPCs using an adenoviral vector. Overexpression of SPI1 led to a significant increase in TGF-β expression, as measured by both qPCR and flow cytometry (figure 8C and D), while knocking down SPI1 by small interfering RNA decreased TGF-β expression (figure 8E and F), indicating that SPI1 is sufficient to enhance TGF-β production.²⁵

To explore the underlying mechanism, we performed bioinformatic analyses. Interrogation of the JASPAR database predicted conserved binding motifs for SPI1 within the promoter regions of both the human and mouse TGFB1 genes (figure 8G–J). Consistent with this, analysis of the Cistrome DB database suggested that SPI1 is a potential master transcriptional regulator of TGFB1 in EPCs (figure 8K and L).

Finally, to definitively establish direct transcriptional regulation, we conducted a luciferase reporter assay and chromatin immunoprecipitation followed by qPCR (ChIP-qPCR). The luciferase assay demonstrated that SPI1 overexpression could activate transcription from the TGFB1 promoter (figure 8M). Crucially, ChIP-qPCR confirmed the direct physical binding of SPI1 to the endogenous TGFB1 promoter region in CD45⁺ EPCs (figure 8N).

Discussion

The transcription factor SPI1 (PU.1), initially recognised for its fundamental role in haematopoiesis and immune cell development, has been extensively studied in haematological malignancies¹⁶,¹⁹ but remains poorly defined in autoimmune diseases like RA.^{20 21} Our study unveils a novel and pivotal role for SPI1 within CD45⁺ EPCs, identifying a previously unrecognised immunomodulatory axis in RA pathogenesis. We demonstrate that CD45⁺ EPCs are expanded in the peripheral blood of patients with RA and spleens of mice with CIA, correlating positively with disease activity and negatively with haemoglobin levels, indicating a compensatory response to inflammation.^{5 14 37} These cells exhibit an immunosuppressive transcriptome enriched in TGF-β signalling,^{23 24 26} chemotaxis³¹ and ROS metabolism,¹³ with functional assays confirming enhanced proliferation and selective production of TGF-β and ROS, but not IL-10.

A key mechanistic insight is the delineation of the CCR2-CCL2 axis as a critical recruitment signal for CD45⁺ EPCs.³⁰,³⁴ High CCR2 expression enables their migration towards macrophage-derived CCL2 in the synovial microenvironment, where they skew macrophage polarisation from pro-inflammatory M1 to anti-inflammatory M2-like phenotypes,⁶ suppressing TNFα/IL-1β production and enhancing CD163/CD206 expression. Central to this process, SPI1 is upregulated in RA CD45⁺ EPCs and directly binds the TGFB1 promoter, driving TGF-β production—a finding validated by gain-of-function and loss-of-function experiments, including SPI1 knockdown that impaired TGF-β expression and immunoregulatory capacity.

While splenectomy has been reported to ameliorate RA in clinical settings like Felty syndrome²⁸ by removing a source of inflammatory cells, in mice with CIA, splenectomy exacerbated arthritis by depleting the splenic reservoir of protective CD45⁺ EPCs. This discrepancy highlights species-specific differences in extramedullary haematopoiesis and the systemic versus local dynamics of CD45⁺ EPCs. In humans, systemic expansion may represent a compensatory anti-inflammatory response, but impaired homing to synovium due to dysregulated chemokine axes allows local inflammation to persist. Conversely, in mice, the spleen serves as a key niche for CD45⁺ EPCs,^{15 36} and their adoptive transfer postsplenectomy rescued arthritis severity, underscoring their homeostatic role.

To bridge mechanistic findings to human relevance, we employed a three-dimensional RA synovial organoid model,³⁵ which recapitulates key cellular components of the inflamed joint. Co-culture with CD45⁺ EPCs significantly suppressed organoid growth and pro-inflammatory cytokine (TNF-α, IL-1β) secretion while enhancing anti-inflammatory TGF-β and IL-10 production—effects reversed by TGF-β neutralisation, confirming TGF-β dependence. Additionally, organoid-conditioned media promoted CD45⁺ EPC migration via CCL2, which was inhibited by anti-CCL2 antibody, establishing a recruitment loop. These data validate the SPI1-TGF-β axis and CCR2-CCL2 recruitment in a human context, emphasising the translational potential.

Despite these advances, limitations remain: the initial signals triggering SPI1 upregulation in CD45⁺ EPCs within inflammatory milieus are unclear, and the relative contributions of TGF-β versus ROS to immunosuppression warrant further elucidation. In conclusion, our study redefines CD45⁺ EPCs as active, SPI1-driven modulators in RA, where inflammation-driven expansion and CCR2-CCL2-mediated recruitment enable TGF-β-dependent suppression of macrophage activation. Targeting this axis may offer novel therapeutic strategies, and future work should focus on upstream regulators of SPI1 and broader immunomodulatory effects in RA and other inflammatory diseases.²¹

Conclusion

In summary, we provide clinical and experimental evidence that CD45⁺ EPCs are expanded in RA and identified as a novel immunosuppressive population. Furthermore, these cells are recruited to inflammatory sites via the CCR2-CCL2 axis and subsequently ameliorate arthritis by polarising macrophages towards an M2-like phenotype through TGF-β. Central to this process, the transcription factor SPI1 is upregulated in RA CD45⁺ EPCs and drives TGF-β production by directly binding to its promoter. Above all, we propose that targeting the SPI1-TGF-β axis in CD45⁺ EPCs represents a novel immunoregulatory therapeutic strategy for RA.

Data availability statement

Data are available on reasonable request.