TGF-β-producing tBregs are associated with poor survival rates in PTGS2-high cancers Authors: Kavitha Premkumar, Soundharya Ramu, Mohit Kumar Jolly, Bhavani S. Shankar Published: 2026 ABSTRACT The cues that cause B cells to adopt a regulatory phenotype are unknown. We previously reported on the regulatory B-T axis in mouse fibrosarcoma. This work demonstrates the presence of TGF-β-generating Bregs in the sarcoma microenvironment as well as variable gene expression depending on TGF-β levels. Breg depletion reduced tumor-evoked Bregs and progression of WEHI-164 fibrosarcoma, which was reversed by adoptive transfer of B cells. Tumor-derived PGE2 acting via EP4/STAT3 signaling is a key driver of tBreg formation, ultimately resulting in increased Tregs. Only the TGF-β+ tBregs caused T cell suppression. PGE2 activated STAT3 in a non-canonical manner since pre-treatment of B cells with JNK and NF-κB inhibitors completely restored tBreg-induced T cell suppression. The COX-2 inhibitor NS-398 prevented the generation of tumor-evoked Bregs, resulting in a lower tumor burden, fewer Tregs, and improved T cell responses. TGF-β-producing Bregs were identified in sarcomas and other cancers with elevated PGE2. The B cells from PTGS2hi cancers displayed unique Breg signatures when compared to PTGS2lo cancers. TCGA analysis showed that the Breg signature was higher in patients with high PTGS2 levels, which was also associated with a poor survival rate. In summary, we found that the tumor-derived PGE2 activates non-canonical MAPK/STAT3 in B cells, leading to the formation of TGF-β-generating Breg cells. PTGS2hi Breghi patients were associated with a poor survival rate, while the COX-2 inhibitor NS-398 prevented tBreg generation and reduced tumor burden. This could lead to the development of novel strategies for reducing tBreg production in cancer. Introduction Soft tissue sarcomas are tumors of mesenchymal origin that are heterogeneous in terms of anatomical sites, histology, and molecular features, with multiple subtypes, and can impact individuals across all age groups [1, 2]. Up to 30% of the STS patients who received multimodal therapies later develop a hard-to-treat metastatic condition. When it comes to immunotherapy, only a few types of STS have shown major responses, but the main barriers are heterogeneity and a lack of complete understanding of its immune microenvironment [3]. The Cancer Genome Atlas Research Network analyzed 28 immune signatures across 206 adult STSs and identified positive prognoses associated with NK cell, CD8+ T cell, and dendritic cell scores in 6 different types of sarcoma [4]. Petitprez et al. identified five distinct immune phenotypes in a study on 608 STSs, with A and B being immune-low, D and E as immune-high, and C as the highly vascularized group [5]. The authors conclude that B-cell infiltration with the formation of tertiary lymphoid structures (TLS) was associated with improved survival. However, correlation of immune histochemical assessment of CD20- and CD19-positive cells with prognostic association has yielded both negative and positive correlations with survival [6]. In addition, several signaling pathways, like lactate and TGF-β, inhibit the formation of tertiary lymphoid structures [7, 8]. All of these results suggest that B cells in sarcomas are influenced by the soluble mediators of the tumor microenvironment and need further investigation. B cells are a major component of the immune system known to produce antibodies, secrete immunoregulatory cytokines, and serve as antigen-presenting cells [9]. Besides these functions, a subset of B cells, called regulatory B cells (Breg), have immunosuppressive properties [10]. Even though these Bregs are well investigated in the etiology of autoimmune diseases and infections, emerging data points to their participation in cancer [11–13]. Bregs in the tumor have been shown to have pro-tumor effects by inhibiting T cells and NK cells and enhancing the activity of Tregs and other immunosuppressive cells [14, 15]. Anti-CD20-mediated B cell depletion decreased tumor burden associated with increased Tregs in a methylcholanthrene-induced murine fibrosarcoma model [15]. The existence of similar immunosuppressive B cells is reported in a limited number of preclinical cancer models [16, 17]. A pro-tumoral relationship of B cells in human sarcoma has been reported by Smolle et al., who showed a negative correlation of CD20+ B cells with local recurrence and no influence on overall survival based on studies using 1266 tissue core samples from 188 STS patients [18]. Such pro and anti-tumoral effects of B cells in the STS microenvironment indicate the complexity and heterogeneity of B cell populations. These data also highlight the importance of the spatial distribution of B cells in TME, which could determine the outcome of the immune response rather than the mere presence of these cells. Hence, it is very important to identify the mediators in the tumor microenvironment that influence the B cells of the TME to acquire such varied roles. Recently, we have reported that WEHI-164 fibrosarcoma induced CD19+CD81+CD25hiPD-L1hiCD45Rhi CD85hi MHCIIhiCD27hi and P-STAT3+ Breg cells that secrete IL-10 and TGF-β. These tumor-evoked Bregs (tBregs) suppressed T cell responses and converted CD4+ cells into Treg cells through TGF-β-mediated signaling [19]. In this study, we demonstrate that TGF-β-producing tBregs in sarcoma have enhanced expression of immunosuppressive markers when compared to low TGF-β-expressing B cells. Despite the fact that several studies have found Bregs in cancers, the factors responsible for their generation are not known. Bodogai et al. demonstrated that 5-lipooxygenase metabolites induced tBregs in 4T1 mammary cancer [20]. Other investigators have found that cytokine IL-18, fatty acids, or autophagosomes generated Bregs in different cancers [17, 21]. In this study, we show that the WEHI-164 tumor cells secrete prostaglandins (PG), which induce TGF-β1-producing tBreg cells in BALB/c mice. Both in vitro and in vivo, tBreg induction and immunosuppression were prevented by pharmacologically blocking EP receptor signaling in B cells or inhibiting COX2 in tumor cells. In addition, sarcomas and other cancers with elevated PGE2 had a higher proportion of TGF-β-producing Bregs displaying unique Breg signatures. The proportion of Breg was higher in patients with high PTGS2 levels, which was associated with a poor survival rate. Though PTGER4 is known as a negative regulator of BCR signaling, and elevated levels of PGE2 have been known to sensitize B lymphocytes to antigen-specific tolerogenic signals, how this suppressive effect of PGE2 on B cells will affect anti-tumor immunity and cancer progression was not clear [22, 23]. Through this study, we report the pivotal role of COX2 and tumor-derived PGE2 in inducing immunosuppressive tBreg and Treg cells, leading to increased cancer progression. To the best of our knowledge, this is the first report of PGE2-induced tBregs in cancer. Results Breg cells expressing TGFB1 in sarcoma We previously demonstrated that WEHI-164 fibrosarcoma cells induced the generation of CD19+CD81hiCD86hiCD25+CD45R+STAT3+IL-10+TGF-β+ tumor-evoked Bregs (tBregs) [19]. These tBregs suppressed T cell proliferation and induced Tregs in a TGF-β-dependent manner. SB431542, a pharmacological inhibitor of the TGF-β receptor, restored T cell responses and reduced tumor burden by abrogating the immunosuppressive function of tBregs [19]. However, the existence of such regulatory B cells in sarcomas and other cancers and the specific mechanism of their induction are not well documented. In order to identify the presence of TGF-β+ Bregs in human sarcoma, we downloaded publicly available single-cell RNA sequencing data from 6 osteosarcoma (OS) patients (GSE162454). After quality control for all the samples, the cell annotation showed different immune cell subpopulations, including B cells (Figs. 1A and S1A, supporting information). B cells were associated with clusters 13 and 14 expressing MS4A1 and CD79A (Fig. S1A, supporting information). These two B cell clusters were re-clustered into seven subclusters (Fig. 1B). In these subclusters, we looked for the expression of TGFB1. Most of the clusters expressed TGFB1 (Fig. 1C), and among them, the clusters 0 and 4 expressed the highest level and thereby corresponded to TGF-βhi B cells (Fig. 1D). We observed that the other clusters can be split into TGF-βmid (clusters 5, 6, and 7) and TGF-βlo (clusters 1, 2, and 3) B cells (Fig. 1D). Differentially expressed genes (DEGs) were analyzed between TGF-βhi and TGF-βlo B cells to identify genes specific to TGF-βhi B cells. We observed 853 DEGs between these two groups, out of which 632 were upregulated (log2(FC) > 0.25, FDR < 0.05) and 221 were downregulated (log2(FC) < -0.25, FDR < 0.05). The volcano plot and heatmap show DEGs that are upregulated or downregulated to 1.5-fold or above (log2(FC) ≥ 0.585 or log2(FC) ≤ –0.585) (Figs. 1E and S1B, supporting information).Fig. 1TGF-β-expressing Breg cells in sarcoma.A UMAP projection of cell clustering information, color-coded by clusters. B UMAP plot representation of the B-cell subpopulation, colored and labeled by UMAP sub-cluster annotations. C UMAP representation and D feature plot visualizing the expression of the TGFB1 gene in the B cell subset. E Volcano plot showing DEGs in TGF-βhi and TGF-βlo B cells. F Heatmap presenting the expression of the top 20 upregulated and downregulated genes in the TGF-βhi B cells. G Heatmap depicting the normalized expression scores of the significantly upregulated and downregulated pathways in the TGF-βhi B cells. T cell response in the presence of tBregs generated from WEHI-164 fibrosarcoma. B cells were cultured with 20% TCM (B-TCM) for 72 h and then co-cultured with CFSE-labeled, stimulated Bdep cells along with neutralising anti-TGFβ1 antibody or TGFβRI inhibitor SB431542. H Representative flow histograms of CFSE dye dilution in CD4+ gated T cells from different treatment groups. I Bar diagram of T cell proliferation (n = 3). B- TCM cells were sorted into TGFβ+ and TGFβ- B cells using MACS and co-cultured with CFSE-labeled, stimulated Bdep cells. J Representative flow histograms of CFSE dye dilution in CD4+ gated T cells from different treatment groups. K Bar diagram of T cell proliferation (n = 3). TGF-β-expressing Breg cells in sarcoma. A UMAP projection of cell clustering information, color-coded by clusters. B UMAP plot representation of the B-cell subpopulation, colored and labeled by UMAP sub-cluster annotations. C UMAP representation and D feature plot visualizing the expression of the TGFB1 gene in the B cell subset. E Volcano plot showing DEGs in TGF-βhi and TGF-βlo B cells. F Heatmap presenting the expression of the top 20 upregulated and downregulated genes in the TGF-βhi B cells. G Heatmap depicting the normalized expression scores of the significantly upregulated and downregulated pathways in the TGF-βhi B cells. T cell response in the presence of tBregs generated from WEHI-164 fibrosarcoma. B cells were cultured with 20% TCM (B-TCM) for 72 h and then co-cultured with CFSE-labeled, stimulated Bdep cells along with neutralising anti-TGFβ1 antibody or TGFβRI inhibitor SB431542. H Representative flow histograms of CFSE dye dilution in CD4+ gated T cells from different treatment groups. I Bar diagram of T cell proliferation (n = 3). B- TCM cells were sorted into TGFβ+ and TGFβ- B cells using MACS and co-cultured with CFSE-labeled, stimulated Bdep cells. J Representative flow histograms of CFSE dye dilution in CD4+ gated T cells from different treatment groups. K Bar diagram of T cell proliferation (n = 3). TGF-βhi B cells had high levels of CD19 expression with down-regulation of CD38, JCHAIN, and SDC1 and absence of CD27, indicating that they are naive but not plasma cells. These cells also expressed high levels of immature B cell markers TCL1A, IGHD, SELL, and FCMR. Expression of CD69 and CD83, along with increased chromatin remodeling genes (CHD1, AIRD1A, BAF, HDAC1, HDAC2, and JMJDC1) and ribosome biogenesis genes (MYC, NPM1, NC1, NOP53, DDX17, DDX21, RPS genes, and RPL genes), is a hallmark of cell growth and differentiation and indicates the activated state of TGF-βhi B cells. TGF-βhi B cells highly expressed MHCII- and MHCI-related genes (HLA-DQ, HLA-DP, HLA-DR, HLA-A, HLA-B, and HLA-C) and markers associated with trafficking and chemotaxis (CXCR4, CXCR5, GPR18, CDC42, and ARPC3), suggesting MHC-mediated antigen presentation and interaction with T cells might be one function of these cells. On the other hand, high expression of HLA-E, HLA-F, and HLA-DO indicates that TGF-βhi B cell-T cell interaction could promote tolerogenic T cell differentiation (Fig. 1F). DEG analysis revealed that TGF-βhi B cells expressed high levels of markers associated with immature B cells, activation and proliferation, chemotaxis, and antigen presentation, along with high levels of the immunosuppressive HLA-1b class of MHCs. This suggests a possible interaction of these tolerogenic B cells (or Breg cells) with T cells. To confirm whether these cells exhibited a Breg-like phenotype, we assessed the previously reported Breg-associated markers in the DEGs of TGF-βhi B cells. Most of the Breg-associated genes were highly expressed in TGF-βhi B cells (Fig. 1F). One interesting factor was the upregulated NF-κB pathway genes (NFKB1, NFKB2, REL, TRAF5, and BIRC3). The pathway analysis shows an upregulation of metabolic pathways and biosynthetic processes and a downregulation of various B cell receptor signaling and immune response pathways in TGF-βhi B cells (Figs. 1G and S1C, supporting information). Overall, these data show that TGF-βhi B cells were associated with a gene expression profile related to the regulatory B cell signature. To confirm the role of TGF-β in Breg-mediated suppression, we generated tBregs using WEHI-164 tumor-conditioned medium (B-TCM). We then inhibited TGF-β signaling in a classical Breg-T cell co-culture system using either a neutralizing anti-TGF-β antibody (1D11) (B-TCM αTGFβ) or a TGF-βRI inhibitor SB431542 (B-TCM SB431542) and assessed T cell proliferation. Both neutralizing antibody and receptor inhibitor effectively blocked TGF-β signaling in T cells and reversed tBreg-mediated suppression (Fig. 1H, I). In addition to secreting TGF-β, the tBregs also express surface membrane-bound TGF-β in a LAP-TGFβ form (Fig. S1D, supporting information). To further establish that TGF-βhi B cells are tBregs, we sorted TGF-β+ and TGF-β- cells from tBregs using the LAP-TGFβ1 antibody, followed by MACS (Fig. S1D, supporting information). The sorted cells (TGFβ+ B-TCM and TGF-β- B-TCM) were then co-cultured with stimulated T cells. TGF-β+ B-TCM cells exhibited enhanced T cell suppression as compared to unsorted B-TCM cells, whereas TGF-β- B-TCM cells failed to suppress T cell proliferation (Fig. 1J, K). Collectively, these data demonstrate that TGF-βhi B cells constitute the functional tBreg population responsible for immune suppression in sarcoma. B cell depletion in BALB/c mice reduced Treg and WEHI-164 tumor burden To better understand the mediators of tBreg production, we studied the changes in B and T cells over the time course of tumor progression in WEHI-164 tumor-bearing mice (TBM). The proportion of CD19+ B cells was consistently higher in TBM as compared to non-tumor controls (Fig. 2A). These B cells also expressed the activation marker CD69 in the late phases of tumor progression (Fig. 2B). Similarly, immunosuppressive PD-L1-expressing B cells increased as the tumor grew (Fig. 2C). The observed increase in CD19+CD69+ cells and CD19+PD-L1+ cells in TBM is consistent with our prior findings [19]. In contrast, the percentages of splenic CD4⁺ and CD8⁺ T cells significantly decreased at later time points (Fig. S2A, B, supporting information). To further confirm this decrease, we calculated the absolute number of CD19+ B and CD4⁺ and CD8⁺ T cells per spleen from NT and TBM. The absolute numbers of T cells from NT and TBM were unchanged, and the decrease in the T cells was due to the increased B cell proportion in TBM (Fig. 2D). To determine whether this increase in B cells is related to immune suppression, we performed a previously established T cell suppression assay [19]. As the tumor progressed, B cells progressively became immunosuppressive. By day 10, when tumors became palpable, TBM-B cells significantly suppressed T cell proliferation (Fig. 2E). In addition, as the tumor progressed, TBM displayed a marked increase in the frequency of CD4⁺Foxp3⁺ regulatory T cells (Tregs) (Fig. 2F).Fig. 2B cell depletion reduced Treg and WEHI-164 tumor burden.Splenic immune cell temporal dynamics in WEHI-164 TBM. Percentages of A CD19+ B cells, B CD69-expressing B cells, and C PD-L1-expressing B cells. D absolute numbers of CD19+ B, CD4+ and CD8+ T cells in total spleen cells. E Proliferation in CFSE-labeled CD4+ T cells in a Bdep: tBreg co-culture. F CD4+Foxp3+ Treg cells. Tumor burden and T cell responses in B-depleted TBM. B cells were depleted in mice three days before tumor challenge. B cells purified from NT or TBM were adoptively transferred to B-depleted TBM on days 3, 10, and 17 of tumor challenge. G tumor growth kinetics. H Bar diagram of T cell proliferation from different treatment groups (n = 6). I Representative flow dot plots of CD4+Foxp3+ Treg cells from different treatment groups. J Bar diagram of Treg cells (n = 6). Tumor generation experiments were repeated at least thrice (n = 4–6 at each repeat). B cell depletion reduced Treg and WEHI-164 tumor burden. Splenic immune cell temporal dynamics in WEHI-164 TBM. Percentages of A CD19+ B cells, B CD69-expressing B cells, and C PD-L1-expressing B cells. D absolute numbers of CD19+ B, CD4+ and CD8+ T cells in total spleen cells. E Proliferation in CFSE-labeled CD4+ T cells in a Bdep: tBreg co-culture. F CD4+Foxp3+ Treg cells. Tumor burden and T cell responses in B-depleted TBM. B cells were depleted in mice three days before tumor challenge. B cells purified from NT or TBM were adoptively transferred to B-depleted TBM on days 3, 10, and 17 of tumor challenge. G tumor growth kinetics. H Bar diagram of T cell proliferation from different treatment groups (n = 6). I Representative flow dot plots of CD4+Foxp3+ Treg cells from different treatment groups. J Bar diagram of Treg cells (n = 6). Tumor generation experiments were repeated at least thrice (n = 4–6 at each repeat). To validate the role of tBregs in tumor immunosuppression, we depleted the B cells in mice with anti-CD19 antibody prior to WEHI-164 tumor cell inoculation. Effective B cell depletion was confirmed on the day of tumor challenge, without affecting the frequencies of T cells (Fig. S2C, supporting information). To further validate the role of B cells in immune suppression, purified B cells from either normal mice (B-NT) or WEHI-164 tumor-bearing mice (Breg) were adoptively transferred to B-depleted TBM on days 3, 10, and 17 following tumor cell inoculation. Tumor growth was significantly reduced in B-depleted mice, as compared to TBM (Fig. 2G). Adoptive transfer of both B-NT and Bregs resulted in a significant increase in tumor burden in B-depleted TBM (Fig. 2G). There was no significant change in tumor volume between TBM and B-NT or Breg-adoptively transferred Bdep TBM. On the other hand, the tumor burden of Breg-adoptively transferred Bdep TBM was significantly higher as compared to B-NT-adoptively transferred Bdep TBM (Fig. 2G). The increased tumor burden in B-NT-adoptively transferred Bdep TBM may be due to the conversion of B cells into Bregs in the presence of the tumor, which requires further validation to confirm. On day 20, the percentages of CD4+ T and CD8+ T cells were lower in TBM and improved in B-depleted TBM (Fig. S2D, supporting information). B cell frequency was increased in the TBM but remained low in B-depleted TBM even up to day 20 (Fig. S2E, supporting information). Re-expression of CD19 in the spleens of B-cell-adoptively transferred Bdep TBM confirms the presence of the transferred B cells even at the end of the experiment (Fig. S2E, supporting information). We next validated the effect of B depletion and adoptive transfer on T cell responses. Anti-CD3/CD28-induced proliferation was inhibited in CD4+ T cells from TBM but was significantly restored in B-depleted TBM, indicating a recovery of T cell function in the absence of B cells (Figs. 2H and S2F, supporting information). This increase in T cell proliferation was abrogated in B-NT and Breg-adoptively transferred Bdep TBM (Figs. 2H and S2F, supporting information). The frequency of regulatory T cells was significantly higher in the TBM but not in the B-depleted TBM, and increased to levels similar to those in the TBM in B cell-adoptively transferred B-dep TBM. These data suggest a B-cell-dependent expansion of Tregs during tumor progression (Fig. 2I, J). PGE2-EP2/EP4 signaling induces tBreg-mediated suppression via STAT3 activation We identified the tumor-derived soluble factor as a heat-stable, protease-resistant molecule with a molecular weight of less than 5 kDa (data not shown), and these characteristics are consistent with the lipid signaling mediator, prostaglandins (~350 Da). Prostaglandin E2 (PGE2), a prostanoid produced by the action of cyclooxygenase on arachidonic acid, has been shown to impair host immunity in several cancers [24–27]. Therefore, we measured PGE2 levels in WEHI-164 TCM, as well as 4T1, a previously known PGE2-secreting cell line [28]. WEHI-164 produced very high levels of PGE2 at ~4 ng/mL of TCM, comparable to those secreted by 4T1 (Fig. 3A).Fig. 3PGE2-EP2/EP4 signaling promotes tBreg-mediated suppression via STAT3 activation.A PGE2 in the supernatants of WEHI-164 and 4T1 cultures. PGE2 receptors B EP4 and C EP2 expression in CD4+ and CD19+ cells. D EP4 expression in CD19+ cells in the presence of tumor in vitro and in vivo. T cell responses with tBregs and EP antagonists. B cells were cultured with 20% TCM (B-TCM) along with EP2 and EP4 antagonists for 72 h and then co-cultured with CFSE-labeled, stimulated Bdep cells. E representative flow histograms of CFSE dye dilution in CD4+ gated T cells from different treatment groups. F Bar diagram of T cell proliferation (n = 5). Characterization of B-TCM and B-PGE2 tBregs. Purified B cells were cultured with 20% TCM or 1 μM PGE2 followed by a T cell suppression assay and Treg generation assay. G Secretion of TGFβ1 by B-TCM and B-PGE2 cells. H Representative flow histograms of CFSE dye dilution in CD4+ gated T cells from different treatment groups. I Bar diagram of T cell proliferation (n = 5). J Representative dot plots showing CD4+CD25+Foxp3+ Tregs from different treatment groups (K) Bar diagram of Treg cells (n = 7). Role of STAT3 signaling in tBregs: L P-STAT3 expression in B-TCM and B-PGE2 cells. B cells were pre-treated with the STAT3 inhibitor BP-1-102 and cultured with TCM, followed by a T cell suppression assay. M Representative CFSE histograms of CD4+ gated T cell proliferation during co-culture with tBregs and STAT inhibitor and N Bar diagram representation of T cell proliferation (n = 5). PGE2-EP2/EP4 signaling promotes tBreg-mediated suppression via STAT3 activation. A PGE2 in the supernatants of WEHI-164 and 4T1 cultures. PGE2 receptors B EP4 and C EP2 expression in CD4+ and CD19+ cells. D EP4 expression in CD19+ cells in the presence of tumor in vitro and in vivo. T cell responses with tBregs and EP antagonists. B cells were cultured with 20% TCM (B-TCM) along with EP2 and EP4 antagonists for 72 h and then co-cultured with CFSE-labeled, stimulated Bdep cells. E representative flow histograms of CFSE dye dilution in CD4+ gated T cells from different treatment groups. F Bar diagram of T cell proliferation (n = 5). Characterization of B-TCM and B-PGE2 tBregs. Purified B cells were cultured with 20% TCM or 1 μM PGE2 followed by a T cell suppression assay and Treg generation assay. G Secretion of TGFβ1 by B-TCM and B-PGE2 cells. H Representative flow histograms of CFSE dye dilution in CD4+ gated T cells from different treatment groups. I Bar diagram of T cell proliferation (n = 5). J Representative dot plots showing CD4+CD25+Foxp3+ Tregs from different treatment groups (K) Bar diagram of Treg cells (n = 7). Role of STAT3 signaling in tBregs: L P-STAT3 expression in B-TCM and B-PGE2 cells. B cells were pre-treated with the STAT3 inhibitor BP-1-102 and cultured with TCM, followed by a T cell suppression assay. M Representative CFSE histograms of CD4+ gated T cell proliferation during co-culture with tBregs and STAT inhibitor and N Bar diagram representation of T cell proliferation (n = 5). PGE2 binds to membrane receptors EP1-4 to trigger signaling in target cells. Splenic CD19+ B cells had a significantly higher EP4 expression when compared to CD4+ T cells (Fig. 3B). While over 90% of CD19+ cells expressed the EP4 receptor, both T and B cells had low levels of EP2 expression, with B cells having slightly more EP2 than T cells (Fig. 3C). In addition, EP4 expression in CD19+ B cells increased with either in vitro TCM treatment or in the B cells from tumor-bearing mice (TBM) (Fig. 3D). Whereas EP2 expression in CD19+ B cells was either unchanged or decreased in the presence of the tumor (Fig. S3A, supporting information). To further understand the role of EP receptors, tBregs were generated with TCM (B-TCM) in the presence of EP2 and EP4 neutralizing antibodies or potent EP2 (PF04418948) and EP4 (MK2894) antagonists. Neutralizing antibodies (B-TCMαEP2 and B-TCMαEP4) partially abrogated (Fig. S3B, C, supporting information), whereas both antagonists PF04418948 (B-TCMEP2ant) and MK2894 (B-TCMEP4ant) (Fig. 3E, F) completely reversed the B-TCM-induced suppression of CD4+ cell proliferation. Also, B cells treated with the agonists (PF-04217329, B-EP2ag, and CY10598, B-EP4ag) significantly suppressed T cell proliferation with an effect similar to B-TCM (Fig. S3D, E, supporting information). Collectively, our findings demonstrate that tumor-induced Breg generation is mediated by EP receptor signaling in B cells. To confirm the effect of PGE2, purified B cells were treated with synthetic PGE2 (10 µM) (B-PGE2), and several WEHI-164-induced tBreg markers were examined [19]. In contrast to TCM, PGE2 did not increase the expression of tBreg-associated markers CD69 or PD-L1 (Fig. S3F, G, supporting information). B-PGE2 secreted significantly larger levels of TGF-β, albeit lower than B-TCM (Fig. 3G). B-PGE2 cells secreted IL-6 similar to B-TCM but did not secrete IL-10 in contrast to B-TCM (Fig S3H, supporting information). Similar to B-TCM, B-PGE2 effectively suppressed the proliferation of CD4+ cells, with comparable levels of suppression observed between the two groups (Fig. 3H, I). These cells were also able to generate Treg cells in a CD4+-B cell co-culture system (Fig. 3J, K). Since we previously found increased STAT3 phosphorylation in B-TCM cells [19], we investigated if this was due to PGE2 signaling. Both B-TCM and B-PGE2 elevated STAT3 phosphorylation, as shown by Western blot and flow cytometry (Figs. 3L and S3I, J, supporting information). Decrease of STAT3 phosphorylation with α-EP4 neutralizing antibody in B-TCM confirmed that PGE2 signaling regulated STAT3 phosphorylation (Fig. S3K, supporting information). To validate PGE2-mediated STAT3 phosphorylation in tBregs, we inhibited STAT3 phosphorylation in B-TCM cells using the STAT3 inhibitor BP-1-102 (Fig. S3L, supporting information). BP-1-102 treatment resulted in a complete restoration of B-TCM-induced suppression of responder T cell proliferation (Fig. 3M, N). Collectively, these results indicate that PGE2-induced tBreg generation was facilitated by STAT3 activation. PGE2-induced STAT3 phosphorylation in Bregs is mediated through non-canonical MAPK signaling To further confirm the involvement of PGE2/STAT3 signaling in Breg cell induction, we knocked down EP4 and STAT3 expression in B cells using RNA interference prior to Breg induction (Fig. S4A, B, supporting information). We selected EP4 since it is more abundant in B cells and increases with tumor conditions rather than EP2. Both SiEP4 and SiSTAT3 abrogated tBreg generation and its T cell suppressive function (Fig. 4A, B). Silencing of EP4 also decreased STAT3 phosphorylation, while STAT3 silencing did not affect EP4 expression (Fig. S4A, B, supporting information). These data confirm that STAT3 activation is indeed downstream of PGE2/EP4 signaling. We next sought to identify the downstream pathway leading to STAT3 activation in Breg cells.Fig. 4PGE2 induces STAT-3 phosphorylation in Breg via non-canonical MAPK signaling.Effect of EP4 and STAT3 siRNAs on Bregs. Purified B cells were transfected with si RNAs against EP4, STAT3 and non targeting control (siEP4, siSTAT3 and siNTC) for 24 h. The transfected cells were treated with 20% TCM to generate Breg cells, followed by a T cell suppression assay. A Representative histograms of CD4+ gated T cell proliferation in different treatment groups. B Bar diagram of T cell proliferation (n = 6). Effect of IL-6 signaling on Bregs. B cells were either pre-treated with IL-6 signaling inhibitors Ruxolitinib and Tofacitinib and cultured with 20% TCM or 20 ng/mL recombinant mouse IL-6, followed by a T cell suppression assay. C Representative histograms of CD4+ gated T cell proliferation in different treatment groups. D Bar diagram of T cell proliferation (n = 6). E Western blot images and F Quantification of NF-kB and MAPK protein expression in tBregs treated with TCM and NS-TCM (n = 3). Effect of non-canonical protein inhibitors of Bregs. B-TCM were generated in the presence of inhibitors against ERK, p38, SAPK/JNK, and NFκB (10 μM), followed by a T cell suppression assay. G Representative histogram of CD4+ gated T cell proliferation in different treatment groups. H Bar diagram of T cell proliferation (n = 6). I Western blot of P-STAT3 in tBregs treated with inhibitors and J quantitation of P-STAT3/STAT3 (n = 3). PGE2 induces STAT-3 phosphorylation in Breg via non-canonical MAPK signaling. Effect of EP4 and STAT3 siRNAs on Bregs. Purified B cells were transfected with si RNAs against EP4, STAT3 and non targeting control (siEP4, siSTAT3 and siNTC) for 24 h. The transfected cells were treated with 20% TCM to generate Breg cells, followed by a T cell suppression assay. A Representative histograms of CD4+ gated T cell proliferation in different treatment groups. B Bar diagram of T cell proliferation (n = 6). Effect of IL-6 signaling on Bregs. B cells were either pre-treated with IL-6 signaling inhibitors Ruxolitinib and Tofacitinib and cultured with 20% TCM or 20 ng/mL recombinant mouse IL-6, followed by a T cell suppression assay. C Representative histograms of CD4+ gated T cell proliferation in different treatment groups. D Bar diagram of T cell proliferation (n = 6). E Western blot images and F Quantification of NF-kB and MAPK protein expression in tBregs treated with TCM and NS-TCM (n = 3). Effect of non-canonical protein inhibitors of Bregs. B-TCM were generated in the presence of inhibitors against ERK, p38, SAPK/JNK, and NFκB (10 μM), followed by a T cell suppression assay. G Representative histogram of CD4+ gated T cell proliferation in different treatment groups. H Bar diagram of T cell proliferation (n = 6). I Western blot of P-STAT3 in tBregs treated with inhibitors and J quantitation of P-STAT3/STAT3 (n = 3). IL-6, a classical upstream activator of STAT3, was elevated in both B-TCM and B-PGE2 cells. To mechanistically delineate the role of IL-6 in STAT3 activation, we either blocked IL-6 signaling in Bregs with a JAK1/2 inhibitor, Ruxolitinib, and a JAK1/3 inhibitor, Tofacitinib, or treated B cells with recombinant mouse IL-6. Inhibition of IL-6 signaling did not reverse Breg immunosuppressive function (Fig. 4C, D). Similarly, purified B cells treated with recombinant mouse IL-6 (B-IL-6) did not acquire immunosuppressive properties and failed to suppress T cell proliferation (Fig. 4C, D). These results showed that the tBreg phenotype was independent of IL-6 and may be mediated by non-canonical STAT3 activation. NS-398, a selective cyclooxygenase 2 (COX2) inhibitor, was used to block PGE2 secretion by WEHI-164 cells [29]. This NS-TCM was used to identify the non-canonical MAPK signaling events associated with tBreg generation. NF-kB and MAPK proteins JNK and ERK were hyperphosphorylated in B-TCM and returned to basal levels in B-NSTCM cells (Fig. 4E, F). To identify the individual role of these pathways, B cells were pre-treated with selective pharmacological inhibitors—NF-κB activation inhibitor II (B-NF-κBi), SP600125 (B-JNKi), UO126 (B-ERKi), and SB203580 (B-p38i)—prior to TCM treatment (Fig. S4C, supporting information). The ERK1/2 inhibitor or p-38 inhibitor did not have any effect on the tBreg-induced inhibition of T cell proliferation (Fig. 4G, H). Both JNK and NF-κB inhibitors completely abrogated tBreg-mediated T cell suppression (Fig. 4G, H). Notably, only JNK inhibition, but not NF-κB inhibition, decreased STAT3 activation in B cells (Fig. 4I, J). These data suggest that PGE2/EP4-induced JNK activation mediates STAT3 phosphorylation in Bregs and regulates Breg immunosuppressive function. PGE2/EP4-induced NF-κB activation did not have a direct role in STAT3 activation but regulated Breg immunosuppressive function. Both the transcription factors STAT3 and NF-κB could be independently activated downstream of PGE2/EP4 signaling, and many forms of crosstalk between these two proteins exist, including physical interaction and cooperation at promoters/enhancers [30]. Further studies are therefore required to identify the interplay between JNK, STAT3, and NF-κB in tBreg induction and function in fibrosarcoma. The COX-2 inhibitor NS-398 reduced tumor burden by reversing Breg-mediated T cell suppression Since there was abrogation of B-TCM-induced phosphorylation of the non-canonical STAT3 activators in B-NSTCM, we further evaluated whether NS-398 administration in tumor-bearing mice could lead to functional restoration of B cells. NS-398 was non-cytotoxic to WEHI-164 cells at concentrations up to 100 µM (Fig. S5A, supporting information), and NS-398 induced a substantial reduction in PGE2 release even at 10 μM (Fig. 5A).Fig. 5NS-398 reduced tumor burden by reversing Breg-mediated T cell suppression.A PGE2 levels in TCM and NS-TCM. Effect of NS-TCM on tBreg generation and function. B cells were cultured with 20% TCM or NS-TCM. B TGF-β1 production by B-TCM and B-NSTCM cells. C Representative flow histogram and D Bar diagram of membrane-bound TGF-β expression (n = 3). E P-STAT3 Western blot. F Representative histogram of CD4+ gated T cell proliferation cultured with B-TCM and B-NSTCM. G Bar diagram of T cell proliferation (n = 6). H Treg generation upon co-culture with B-TCM and B-NSTCM and I bar diagram of Treg cells (n = 7). NS398 lowers tumor burden through reduced Breg and Treg. NS-398 (10 mg/kg body weight) was administered in WEHI-164 TBM for 10 days from day 3 of tumor inoculation. On day 20, the mice were sacrificed, and spleens were isolated to assess T and B cells. J tumor volume. K Representative flow histograms and L Bar diagram of TGF-β expressing B cells from TBM and NSTBM (n = 4). M Expression of signaling proteins in B cells purified from TBM and NS-TBM. N Representative histograms and O Bar diagram of CD4+ gated T cell proliferation during co-culture with B cells from TBM and NS-TBM (n = 5). P Dot plots and Q Bar diagram of Treg generation in spleen cells from TBM and NS-TBM (n = 5). Tumor generation experiments were repeated at least thrice (n = 4–6 at each repeat). NS-398 reduced tumor burden by reversing Breg-mediated T cell suppression. A PGE2 levels in TCM and NS-TCM. Effect of NS-TCM on tBreg generation and function. B cells were cultured with 20% TCM or NS-TCM. B TGF-β1 production by B-TCM and B-NSTCM cells. C Representative flow histogram and D Bar diagram of membrane-bound TGF-β expression (n = 3). E P-STAT3 Western blot. F Representative histogram of CD4+ gated T cell proliferation cultured with B-TCM and B-NSTCM. G Bar diagram of T cell proliferation (n = 6). H Treg generation upon co-culture with B-TCM and B-NSTCM and I bar diagram of Treg cells (n = 7). NS398 lowers tumor burden through reduced Breg and Treg. NS-398 (10 mg/kg body weight) was administered in WEHI-164 TBM for 10 days from day 3 of tumor inoculation. On day 20, the mice were sacrificed, and spleens were isolated to assess T and B cells. J tumor volume. K Representative flow histograms and L Bar diagram of TGF-β expressing B cells from TBM and NSTBM (n = 4). M Expression of signaling proteins in B cells purified from TBM and NS-TBM. N Representative histograms and O Bar diagram of CD4+ gated T cell proliferation during co-culture with B cells from TBM and NS-TBM (n = 5). P Dot plots and Q Bar diagram of Treg generation in spleen cells from TBM and NS-TBM (n = 5). Tumor generation experiments were repeated at least thrice (n = 4–6 at each repeat). B-NSTCM produced significantly lower levels of TGF-β (Fig. 5B), IL-6, and IL-10 (Fig. S5B, supporting information). B-NSTCM showed a significant reduction in the percentage of membrane-bound TGF-β (LAP-TGFβ) (Fig. 5C, D). In addition, the CD19+CD81+CD27+ population of B-NSTCM cells decreased significantly when compared to B-TCM cells (Fig. S5C, supporting information). CD86 expression was marginally lower in B-NSTCM cells (Fig. S5C, supporting information). However, other WEHI-164 tBreg markers such as CD40, CD80, IA/IEhi, and PD-L1 were equivalent in B-NSTCM cells compared to B-TCM tBregs (Fig. S5C, supporting information). P-STAT3 levels were significantly reduced in B-NSTCM cells compared to B-TCM tBregs (Figs. 5E and S5D, supporting information). B-NSTCM significantly improved tBreg-mediated suppression of T cell proliferation (Fig. 5F, G). The B-NSTCM cells also failed to generate Treg cells, unlike B-TCM cells (Fig. 5H, I), confirming that tBregs were generated by PGE2 signaling. To confirm that tumor-secreted PGE2 is solely responsible for tBreg generation, we inhibited the other arachidonic acid metabolic intermediates in WEHI-164 cells, resulting in the generation of corresponding TCMs. We suppressed leukotriene generation by inhibiting 5-lipoxygenase with zileuton and thromboxane generation with the thromboxane-A2 inhibitor dipyridamole [31, 32]. Unlike NS-TCM, B cells generated with zileuton- and dipyridamole-treated TCM did not rescue T cell proliferation in a co-culture assay, confirming that PGE2 is the key mediator of immunosuppressive tBreg induction by the WEHI-164 tumor (Fig. S5E, supporting information). Administration of NS-398 significantly decreased tumor growth when compared to TBM (Fig. 5J), along with an increase in CD4+ cells and CD8+ cells (Fig. S5F, supporting information) and no change in CD19+ B cells (Fig. S5F, supporting information). However, NS-398 treatment significantly reduced TGF-β-expressing B cells in TBM (Fig. 5K, L). TBM tBregs exhibited increased phosphorylation of STAT3, NFκB, SAPK/JNK, and ERK1/2 compared to B cells from NS-398-treated TBM (Figs. 5M and S5G, supporting information), supporting our in vitro findings that NS-398 treatment prevents the development of the tBreg phenotype. B cells from vehicle-treated TBM inhibited T cell proliferation, but B cells from NS-398-TBM significantly improved T cell responses ex vivo (Fig. 5N, O). The percentage of Treg cells that increased in vehicle-TBM decreased significantly after NS-398 treatment, showing that the Breg-Treg axis was disrupted (Fig. 5P, Q). To confirm that tumor-derived PGE2 and tBregs do actually cause immunosuppression, we administered NS-398 to B-depleted TBM. NS-398 treatment and B-depletion reduced tumor development individually but not in combination (Fig. S6A, supporting information). The percentages of Treg cells in the NS-Bdep TBM were comparable to those of the individual treatments (Fig. S6B, C, supporting information). The proliferation efficiency of splenic T cells was also evaluated. While NS-398 and B-depletion greatly improved T cell proliferation, combining the two did not further improve T cell responses (Fig. S6D, E, supporting information). These findings show that COX-2 inhibition with NS-398 reduced tumor-derived PGE2. This in turn blocked tBreg generation, disrupted the regulatory B-T axis, relieved tumor-induced immunosuppression, and restored T cell responses, lowering the tumor burden in preclinical models. Since there were no B cells in B-depleted mice, there was no further improvement because of NS-398 treatment. TGFB1-producing Bregs are present in human sarcomas and other cancers with high PGE2 To obtain an insight into PGE2/EP4 signaling-mediated regulatory B cell induction in human cancers, we first evaluated whether such a condition is prevalent in the osteosarcoma (OS) samples using the single-cell transcriptomic data. We first mapped the expression of the PTGS1 and PTGS2 genes that encode for COX1 and COX2 enzymes in individual OS samples (Fig. S7A, supporting information) and grouped OS3 and OS6 as PTGS1/2lo, OS1 and OS2 as PTGS1/2mid, and OS4 and OS5 as PTGS1/2hi samples. EP4 receptor (PTGER4) in B cells from these samples showed increased expression in PTGS1/2hi samples (Fig. S7B, supporting information). Breg-associated gene expression analysis in B cells from these groups showed increased TGFB1 and Breg signature in PTGS1/2hi samples and PTGS1/2mid samples as compared to PTGS1/2lo samples. Most of the Breg genes were highly upregulated in PTGS1/2hi samples. Interestingly, NF-κB pathway genes (NFKB1, NFKB2, REL, and TRAF5) were also upregulated in PTGS1/2hi and PTGS1/2mid B cells, similar to that of TGFBhi Breg cells, indicating that B cells from PTGS1/2-expressing cancers indeed show the Breg phenotype (Fig. 6A).Fig. 6TGF-β-producing Bregs are present in human cancers with high PGE2 levels.Publicly available single-cell transcriptomics data from human cancers (osteosarcoma, LPS, SRC, BRCA, CRC, and OVCA) were analyzed for Breg cells. ScRNAseq samples from OS, LPS, and SRC were categorized into PTGS1/2hi and PTGS1/2lo based on the expression of PTGS1 and PTGS2. B cell subset and expression of common Breg markers were assessed. A A heatmap showing relative (z-scored per row) gene expression levels of Breg genes among individual OS samples arranged in order of PTGS1/2 hi to low. B Breg genes in B cells from PTGS1/2hi and PTGS1/2lo LPS. C Breg genes in B cells from PTGS1/2hi and PTGS1/2lo SRC. B cells were filtered from ScRNA-seq samples of CD45+ TILs from high PTGS2-expressing human BRCA, CRC, and OVCA samples. TGFB1 expression in B cell subclusters of D BRCA, E CRC, and F OVCA. Filtered B cells were grouped into TGF-βlo and TGF-βhi B cells. G bubble plot showing Breg genes in TGF-βhi B cells from BRCA, CRC, and OVCA. TGF-β-producing Bregs are present in human cancers with high PGE2 levels. Publicly available single-cell transcriptomics data from human cancers (osteosarcoma, LPS, SRC, BRCA, CRC, and OVCA) were analyzed for Breg cells. ScRNAseq samples from OS, LPS, and SRC were categorized into PTGS1/2hi and PTGS1/2lo based on the expression of PTGS1 and PTGS2. B cell subset and expression of common Breg markers were assessed. A A heatmap showing relative (z-scored per row) gene expression levels of Breg genes among individual OS samples arranged in order of PTGS1/2 hi to low. B Breg genes in B cells from PTGS1/2hi and PTGS1/2lo LPS. C Breg genes in B cells from PTGS1/2hi and PTGS1/2lo SRC. B cells were filtered from ScRNA-seq samples of CD45+ TILs from high PTGS2-expressing human BRCA, CRC, and OVCA samples. TGFB1 expression in B cell subclusters of D BRCA, E CRC, and F OVCA. Filtered B cells were grouped into TGF-βlo and TGF-βhi B cells. G bubble plot showing Breg genes in TGF-βhi B cells from BRCA, CRC, and OVCA. We then asked whether B cells exhibit a similar Breg phenotype in other human sarcomas with high prostaglandin levels. We utilized publicly available single-cell transcriptomic data of soft tissue sarcoma [33]. We first calculated B cell frequencies in human liposarcoma (LPS, n = 10), undifferentiated pleomorphic sarcoma, and mixofibrosarcoma (SRC, n = 7). Overall, B cell infiltrations were lower in LPS and SRC (Fig. S7C, D, supporting information). We then evaluated the expression of PTGS1/2 in individual samples and grouped them as PTGS1/2hi and PTGS1/2lo based on expression (Fig. S7E, F, supporting information). B cells from each group were then filtered, and Breg markers were assessed. Similar to OS, common Breg-associated markers showed increased expression in PTGS1/2hi LPS and SRC (Fig. 6B, C). Genes like NF-κB pathway genes (NFKB1, NFKB2, and REL), HSP90B1, HMGB1, LTB, CD81, and CD83 were among the commonly upregulated genes in LPS and SRC. TGFB1 was also higher in PTGS1/2hi B cells in LPS and SRC, but the extent of increase was more evident in SRC as compared to LPS (Fig. 6B, C). We next investigated whether this is a sarcoma-specific phenomenon or whether such Bregs can be induced in other human cancer types expressing high levels of PGE2. For this, we utilized single-cell transcriptomic data of tumor-infiltrating CD45+ immune cells from breast cancer (BRCA; n = 5), ovarian cancer (OVCA; n = 5), and colorectal cancer (CRC; n = 5) previously published by Punyawatthananukool et al. [34]. They have identified that these cancer tissues expressed high levels of PGE2 through both COX1 and COX2 via immunohistochemistry [34]. We first enumerated the percentages of immune cells in individual samples (Fig. S7G–I, supporting information). Most of the samples showed B-cell infiltration, and we selected samples with higher B-cell infiltration within each cancer type (BRCA_2, BRCA_4, CRC_1, CRC_3, CRC_4, OVCA_1, OVCA_2, and OVCA_3) for further analysis. B cells from each type were pooled within, subclustered, and checked for the expression of TGFB1. Most of the B cell clusters showed increased TGFB1 in all three cancer types, indicating the presence of TGFB1-producing Breg cells (Fig. 6D–F). We then grouped the B cells as TGF-βlo and TGF-βhi B cells based on the median expression of TGFB1 in each cancer type. The TGF-βhi B cells also showed increased expression of common Breg markers (Fig. 6G). Similar to OS, LPS, and SRC B cells, TGF-βhi B cells from BRCA, CRC, and OVCA showed increased expression of NFκB pathway genes (NFKB1, NFKB2, and REL). These data, along with our experimental results, suggest that tumor-derived PGE2 induces TGF-β-secreting Breg cells, which are regulated at least partially by NF-κB signaling. A high Breg profile is associated with increased immunosuppressive markers and poor survival in PTGS2-high cancer patients A wide range of human cancers have high COX2 levels, which result in high PGE2 levels. Comparison of PTGS2 expression in different cancer types from the TCGA pan-cancer bulk RNA-seq data is given in Fig. S8A, supporting information. To further explore the relevance of the TGFβ-Breg cell subpopulation in COX2-expressing cancers, we deconvoluted the TCGA pan-cancer bulk RNA-seq data using canonical B cell markers MS4A1 and CD79A (log2TPM ≥ 2.58), representing samples with strong B cell infiltration. We then further stratified this B-cell-enriched cohort into PTGS2hi and PTGS2lo according to the optimal survival-based cutoff of PTGS2 expression. Based on the quartile expression of TGFB, we then classified the B cell cohort as “Normal” (TGFB < Q1) and “Breg” (TGFB > Q3). The frequency of the Breg profile was higher in patients with high PTGS2 expression (Fig. 7A). The frequency of patients with “Normal” and “Breg” profiles in different cancer types is shown in Fig. S8B, supporting information.Fig. 7Breg Signature is associated with poor survival in PTGS2-high cancer patients.TCGA-Pancancer samples were classified into PTGS-high and PTGS2-low based on the survival cutoff of PTGS2 expression. The “Breg” signature was defined as strong B cell marker (CD79A) expression along with high TGFB1 expression (>Q3), whereas the “Normal” signature was defined as strong B cell marker (CD79A) expression along with low TGFB1 expression (Q3), whereas the “Normal” signature was defined as strong B cell marker (CD79A) expression along with low TGFB1 expression (