IL-10 from tumoral B cells modulates the diffuse large B-cell lymphoma microenvironment and response to immunotherapy Open Access

Diffuse large B-cell lymphomas (DLBCLs) are the most common type of non-Hodgkin lymphoma (NHL).¹ DLBCLs arise from germinal center (GC) B cells and show marked molecular heterogeneity, with several genetic subgroups^2-4 and 2 main transcriptional subgroups: activated B-cell–like (ABC) and GC B-cell–like (GCB).⁵ The ABC subtype is thought to arise from GC-experienced B cells that avoid terminal plasmacytic differentiation either as plasmablasts or aberrant memory B cells.^6,7 The more aggressive ABC-DLBCL accounts for most of the 30% to 40% of patients experiencing relapsed or refractory responses to R-CHOP immunochemotherapy (rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone).⁸ Identification of factors contributing to malignant DLBCL cell survival might predict immunochemotherapy responses and help design treatment alternatives for higher-risk patients.

Interleukin-10 (IL-10) and IL-10 receptor (IL-10R) have been proposed to play an unfavorable role in DLBCL,^9-12 creating an autostimulatory loop via STAT3 signaling^1,11,13-15 and inducing programmed death-ligand 1 (PD-L1) immune checkpoint expression.¹⁶ IL-10 is a potent growth and differentiation factor for activated B cells,¹⁷ enhancing the viability and expression of major histocompatibility complex (MHC) class II antigens in resting murine B cells.¹⁸ Under inflammatory conditions, frequently found in tumor microenvironments, IL-10 can be upregulated and inhibit the secretion of proinflammatory cytokines, MHC molecules, and costimulatory molecules.^19,20 In normal B cells, STAT3 activation is usually transient, but DLBCL shows constitutive activation of STAT3, promoting tumorigenesis.^14,21,22 Although multiple lines of evidence support the autocrine contribution of IL-10 to lymphomagenesis, the potentially more complex in vivo role of the B-cell IL-10/STAT3 axis in DLBCL progression remains uncertain. It is unclear whether IL-10 secreted by malignant B cells may elicit antitumor responses in DLBCL via paracrine effects, given its pleiotropic roles with both stimulatory and inhibitory functions.^23-25 

To explore in vivo the relevance of lymphoma-derived IL-10, we generated a quintuple transgenic mouse model that we call pBIC^Δ10, after crossing our previous ABC-DLBCL-like pBIC mice,²⁶ a “push-and-block” model driven by NF-κB activation (constitutively active IKK2), genomic instability (p53 loss), and blocked terminal differentiation (Blimp1 deletion), with mice with a floxed Il10 gene,²⁷ for conditional knockout of IL-10 specifically in malignant ABC-DLBCL cells. Analyses of human DLBCL data sets supported the clinical implications of our findings. Collectively, our study provides new functional and mechanistic insights into the role of lymphoma B-cell–derived IL-10 in DLBCL pathogenesis and immunotherapy response.

Control YC and lymphoma pBIC compound mice, previously described,²⁶ carried conditional loxP alleles (p: p53, B: Blimp1, I: IKK2ca.IRES.GFP, or Y: eYFP) and Cγ1-cre (C). pBIC^Δ10 mice were generated by crossing pBIC mice with Il10^F/F mice.²⁷ Both models were followed-up for tumoral onset without chronic immunization and subjected to in vivo immunotherapy combinations (detailed in the supplemental Methods, available on the Blood website). All procedures were approved by the Ethical Committee of Animal Experimentation of the University of Navarra (089c-17, 085-19), following European Directive 2010/63/EU.

Cell viability in fresh primary murine tumors was evaluated by flow cytometry in response to IL-10/JAK1/STAT3 axis inhibitors (anti-IL-10R, ruxolitinib, and pyrimethamine) and anti-CD20 immunotherapy. Murine serum IL-10 levels were determined by OptEIA mouse IL-10 enzyme-linked immunosorbent assay (ELISA) set. See details in the supplemental Methods.

Flow cytometry and chromogenic multiplex immunohistochemistry were performed using population markers and antibodies (supplemental Table 1) and standard protocols as detailed in the supplemental Methods, using CytExpert/FlowJo for cytometry and Leica Aperio CS2/QuPath for image analysis.

RNA was extracted using TRIzol reagent from the population of interest for reverse transcription quantitative polymerase chain reaction (RT-qPCR) and RNA sequencing (RNA-seq; GSE255140 and GSE116290²⁶). For single-cell RNA sequencing (scRNA-seq) and T-cell receptor sequencing (scTCR-seq), Chromium Single-Cell Instrument and Chromium Next GEM Single Cell 5' Reagent Kit v2 (Dual Index) from 10× Genomics were used (GSE253955). Human DLBCL cohorts with transcriptional and survival data were obtained from indicated public repositories. For additional information, data processing, and analysis, see the supplemental Methods.

Statistical analyses were performed using RStudio and GraphPad Prism V.9.0.2 and are described throughout the text and in the supplemental Methods.

Using data from available human DLBCL transcriptional studies, we confirmed that although IL-10 expression is highly variable in DLBCL, it is upregulated compared with normal GC B cells, particularly in the more aggressive ABC or (MYD88^L265P and CD79B mutation-enriched (MCD) subtypes^12-15 (Figure 1A). Because our immunocompetent lymphoma mouse model (pBIC) shows key ABC-DLBCL–like hallmarks,^26,28 we also detected increased expression of Il10 by RNA-seq and RT-qPCR in fluorescence-activated cell sorter (FACS)-sorted B220⁺GFP⁺ lymphoma cells compared with normal GC B cells isolated from sheep red blood cell–immunized control YC mice (Figure 1B). Previous studies in ABC/MCD-DLBCL demonstrated the establishment of a feed-forward prosurvival loop through autocrine IL-10R signaling, relying on JAK1/STAT3 activation (Figure 1C).^13,14,29-31 Consistently, we observed significantly higher levels of intracellular IL-10 and phosphorylated STAT3 (pSTAT3) in CD19⁺GFP⁺ pBIC lymphoma cells than in normal B cells from control mice (Figure 1D). Ex vivo perturbation of this IL-10R/JAK1/STAT3 axis with monoclonal antibodies blocking IL-10R or inhibiting JAK1 (ruxolitinib) and STAT3 (pyrimethamine)^1,32 led to significant killing of pBIC lymphoma cells over adjacent normal B cells (Figure 1E). Furthermore, the known link between IL-10/STAT3 signaling and PD-L1 expression in human DLBCL¹⁶ could also be demonstrated in pBIC tumors, with significant reduction of PD-L1 surface levels in B220⁺GFP⁺ lymphoma cells exposed to JAK1/STAT3 inhibitors (Figure 1F). Altogether, these data support a pro-oncogenic role of autocrine IL-10/STAT3 signaling inside DLBCL cells, promoting cellular viability and PD-L1 immune checkpoint upregulation.

To evaluate in vivo the contribution of autocrine IL-10 to lymphoma progression, we generated a conditional knockout mouse with specific deletion of Il10 in lymphoma cells by crossing Il10^F/F mice²⁷ with our original DLBCL-like pBIC mice (hereafter, pBIC^Δ10). Compared with IL-10–proficient pBIC mice, complete abrogation of IL-10 in GC-triggered lymphomas (supplemental Figure 1) severely impaired overall survival of pBIC^Δ10 mice, whereas a single wild-type Il10 allele (pBIC^Δ10F/+) was haplosufficient to restore the original survival rates (Figure 2A). All pBIC and pBIC^Δ10 mice exhibited splenomegaly at advanced stages (Figure 2B) due to the development of B-cell lymphomas, confirmed by significant expansion of CD19⁺GFP⁺ cells (Figure 2C), which were large (Figure 2D) and failed terminal differentiation into CD138⁺ plasma cells (Figure 2E). Consistent with a more aggressive course, IL-10–deficient pBIC^Δ10 mice evidenced lymphoma progression at earlier stages (Figure 2B-C), which was independent of sex (supplemental Figure 2A). Histopathologic analysis of the tumors confirmed a diffuse growth pattern of large GFP⁺ lymphoma cells that disrupted spleen architecture (Figure 2F; supplemental Figure 3). In addition, RNA-seq profiling of fluorescence-activated cell sorter-sorted lymphoma cells from pBIC/pBIC^Δ10 mice showed downregulation of GCB-linked genes and upregulation of ABC-linked genes,^26,33 supporting the classification of pBIC/pBIC^Δ10 lymphomas as ABC-DLBCL–like (Figure 2G). As expected, these lymphomas were oligoclonal, with downregulation of the normally diverse repertoire of immunoglobulin (Ig) genes expressed in GC B cells (Figure 2H). These results support a host-protective effect of tumor-derived IL-10 in DLBCL progression, likely through direct or indirect paracrine functions, paradoxically overriding the prosurvival effects of the autocrine IL-10/JAK/STAT3 loop.

Given the intriguing antitumoral impact of lymphoma-derived IL-10, we explored the response of both DLBCL models to anti-CD20 immunotherapy (as surrogate for rituximab in patients with DLBCL). Randomized pBIC/pBIC^Δ10 mice at advanced stages of disease were treated intraperitoneally with anti-CD20 (Figure 3A, scheme). Even though IL-10–deficient pBIC^Δ10 lymphomas are more aggressive in the absence of treatment, they surprisingly responded significantly better to anti-CD20 immunotherapy than IL-10–proficient pBIC mice (Figure 3A). Although this is consistent with the association of higher IL-10 levels and worse clinical outcomes with R-CHOP in patients with DLBCL,^9-12 the contradictory aggressiveness of IL-10–deficient murine lymphomas anticipates a complex contribution of tumor-derived IL-10 to DLBCL biology.

Antibody-mediated crosslinking of CD20 molecules induces lymphoma B-cell cytotoxicity via intracellular Ca²⁺ influx,^35,36 requiring the accessory CACNA1C cell surface L-type calcium channels for CD20 stability and rituximab-induced cell death.³⁷ STAT3 directly binds (Figure 3B) and represses CACNA1C expression in pSTAT3^high human DLBCL cell lines.³⁴ Consistently, we observed increased Cacna1c expression in pBIC^Δ10 lymphomas compared with pBIC by RNA-seq (Figure 3C) and after ex vivo treatment with anti-IL-10R or JAK/STAT3 inhibition in primary pBIC lymphoma cells (Figure 3D), supporting the transcriptional repression of L-type CACNA1C calcium channels by the autocrine IL-10/JAK/STAT3 pathway. To further explore the role of this IL-10⁺/pSTAT3⁺/CACNA1C^low axis in the response to anti-CD20, ex vivo treatment of fresh primary pBIC/pBIC^Δ10 tumors with increasing doses of anti-CD20 confirmed increased sensitivity (lower IC50, 50% inhibitory concentration values) of IL-10–deficient pBIC^Δ10 lymphoma cells (Figure 3E). Notably, this was dependent on extracellular Ca²⁺ because chelating calcium with EDTA significantly inhibited anti-CD20–induced cell death in both models (Figure 3F). Altogether, these results suggest that the absence of IL-10/JAK/STAT3 autocrine signaling sensitizes DLBCL cells to antibody-mediated cell killing through increased L-type calcium channels that elicit Ca²⁺ influx after CD20 crosslinking, resulting in better responses to anti-CD20 immunotherapy despite the initially more aggressive behavior of IL-10^–/CACNA1C^high pBIC^Δ10 lymphomas.

To elucidate the complex role of lymphoma-derived IL-10 in vivo, we compared the transcriptomes of pBIC/pBIC^Δ10 lymphomas. IL-10 depletion led to broad changes in gene expression (Figure 4A; supplemental Table 2), with 1779 differentially expressed genes involved in functional categories significantly enriched or depleted in pBIC^Δ10 mice according to gene set enrichment analysis (GSEA) (Figure 4B; supplemental Table 3). Single-sample GSEA (supplemental Table 4) of selected gene signatures revealed a dual role of lymphoma-derived IL-10, enabling immune-associated processes (interferon gamma [IFNG], antigen presentation [AP], cytokine release [CKK], and IL-10 signaling) while repressing others (angiogenesis [ANGIO] and calcium activity [CAL], the later exemplified by the Cacna1c gene; Figure 4C).

To delve deeper into the immunomodulatory role of IL-10 on adjacent immune cells, we performed scRNA-seq of splenocytes from representative pBIC and pBIC^Δ10 tumors (1 from each group), composed of 20% GFP⁺ lymphoma cells and 80% nontumoral GFP^– cells (supplemental Figure 4A). In this exploratory analysis, we identified a total of 11 distinct cell types (Figure 4D) in an integrated data set of 18 343 splenocytes (10 049 from pBIC and 8294 from pBIC^Δ10), verified by the expression of canonical cell type–specific markers (supplemental Figure 4B). The absence of lymphoma-derived IL-10 promoted differences in infiltrating populations (Figure 4E), with relative increase of immune-suppressor cells (regulatory T cells [T_regs] and macrophages) and relative decrease of immune-effector (CD8 and natural killer [NK]/NKT) or antigen-presenting cells (dendritic cells [DCs]). These results support the existence of autocrine and paracrine interconnected gene networks controlled by IL-10, which may ultimately rewire antitumor immune responses.

Remarkably, the accelerated tumoral onset in IL-10–deficient lymphomas was accompanied by a decrease in CD8⁺ T cells and DCs without clear changes in NK/NKT cells, both by scRNA-seq (Figure 4E) and flow cytometry (Figure 5A). Consistently, several chemokines involved in the attraction of CD8 cells (eg, CCL5/CXCL9)³⁸ and DCs (eg, CCL4)^38-41 appeared particularly downregulated in lymphoma cells from pBIC^Δ10 mice, together with reduced expression of MHC-I and -II antigen-presenting genes (Figure 5B). Further inspection of scRNA-seq data revealed 2 subclusters inside DCs (Figure 5C). In both pBIC/pBIC^Δ10 models, DCs were primarily type 2 conventional DCs (cDC2s; Figure 5D) and expressed Cd274/Pd-l1 (Figure 5C), indicating a potential suppressive effect on T cells, in line with previous association of cDC2s with ABC-DLBCL in humans.⁴² In contrast, plasmacytoid DCs (pDCs), which respond to CXCL19/CXCL10 for recruitment,^43,44 are IFN-producing, and may contribute to protective immunity,^39,45 were only found in the pBIC microenvironment (Figure 5D). This reveals a role of lymphoma-derived IL-10 in facilitating antigen presentation and pDC-mediated contribution to tumor control. Indirect support for this was the increased TCR clonal diversity (higher Shannon index, based on scTCR-seq data of the same tumors analyzed with scRNA-seq; supplemental Figure 4A) in IL-10–deficient tumors, coinciding with a reduction in the prevalence of larger clones and TCR clonality (lower Gini index; Figure 5E). Altogether, these results support a role for IL-10 in facilitating CD8 priming and antigen-mediated selection of antitumoral T-cell clones.

Next, we explored the functional state of these CD8⁺ tumor-infiltrating lymphocytes (TILs) in murine DLBCL lymphomas. Combining scRNA-seq and TCR-seq data from pBIC and pBIC^Δ10 CD8⁺ T cells, we estimated the exhaustion state by the average expression of previously published exhaustion and cytotoxic programs⁴⁶ across each T-cell clone (cells with identical TCRα/β-CDR3 clonotypes; supplemental Table 5). CD8⁺ T-cell clones in pBIC^Δ10 had higher exhaustion scores relative to cytotoxicity scores than pBIC (Figure 5F), consistent with a more dysfunctional T-cell compartment in the absence of lymphoma-derived IL-10. Confirmed by flow cytometry, IL-10–deficient pBIC^Δ10 lymphomas exhibited increased proportion of exhausted CD8⁺ T cells, marked by coexpression of inhibitory receptors such as PD-1 and LAG3 (Figure 5G), with especially high expression of LAG3 and PD-1 but lower IFN-γ production (Figure 5H). The hyper-exhausted phenotype of this population of CD8⁺ TILs resembled the effector burned-out phenotype previously described for dysfunctional CD8⁺ TILs, which can predict resistance to immune checkpoint blockade (ICB) in lung cancer.⁴⁷ Moreover, murine pBIC lymphomas evidenced a population of Tcf1⁺Pd1^int/low CD8⁺ cells by scRNA-seq (Figure 5I), confirmed by flow cytometry (Figure 5J), which may derive from an intratumoral niche of stem-like CD8⁺ cells responsive to ICB.^48-50 We observed a decrease in the abundance of these precursor CD8⁺ T cells in pBIC^Δ10 mice (Figure 5I-J), supporting the notion that IL-10R signaling is necessary for the maintenance of stem-like CD8⁺ TILs.⁵¹ Therefore, we reasoned that the shift toward effector burned-out dysfunctional pBIC^Δ10 lymphomas resulted from the simultaneous deterioration of progenitor TCF-1⁺PD-1^int/low CD8⁺ stem-like cells and progressive differentiation of CD8⁺ TILs to terminally exhausted TCF-1^–PD-1^hi cells.

To test whether this unbalance of intratumoral CD8⁺ subsets might contribute to immune escape and resistance to ICB in DLBCL, we evaluated the impact of anti–PD-1 therapy in long-term responses to anti-CD20 immunotherapy in pBIC^Δ10 mice, as we had previously demonstrated in pBIC mice.²⁶ Randomized pBIC/pBIC^Δ10 mice at advanced stages of disease were treated with anti-CD20, anti–PD-1, or their combination. Although T-cell–targeted ICB was ineffective as a monotherapy in both pBIC and pBIC^Δ10 mice (supplemental Figure 5), consistent with our earlier findings,²⁶ a significant response to ICB combination with anti-CD20 was observed only in IL-10–proficient pBIC mice but not in IL-10–deficient pBIC^Δ10 mice (Figure 5K). This suggests that insufficient IL-10 signaling from DLBCL cells fails to sustain the niche of anti–PD-1 responsive stem-like CD8⁺ cells and allows for the chronic activation and burnout of CD8⁺ TILs, ultimately contributing to a less responsive lymphoma microenvironment (LME) and resistance to ICB.

Tumor-associated macrophages (TAMs) have been associated with tumor growth, immune suppression, and resistance to ICB.⁵² TAMs increased in aggressive pBIC^Δ10 tumors, as seen by scRNA-seq (Figure 4E) and flow cytometry (Figure 6A), consistent with IL-10 attenuating the activation, proliferation, and recruitment of TAMs.^53-55 Furthermore, scRNA-seq revealed coexpression of M1/antitumoral and M2/protumoral markers in both pBIC and pBIC^Δ10 TAMs (Figure 6B), recapitulating a noncanonical polarization recently uncovered in human DLBCL.⁵⁶ This suggests that lymphoma-derived IL-10 limits the contribution of macrophages to lymphoma progression.

TAMs are also major producers of proangiogenic factors, contributing to the formation of new blood vessels in B-cell NHL.^57,58 Indeed, transcriptional analysis of pBIC^Δ10 lymphoma cells by RNA-seq revealed an enrichment in the expression of genes associated with angiogenesis compared with pBIC (Figure 4B-C), suggesting that the proangiogenic activity of lymphoma cells could synergize with a larger subset of TAMs toward enhanced tumor vascularization. Because an angiogenesis-related signature (named stromal-2)⁵⁹ has been associated with increased microvessel density in human DLBCL,^60,61 we examined the platelet-endothelial adhesion molecule CD31 (PECAM-1) using chromogenic multiplex immunohistochemistry (Figure 6C). CD31⁺ vascular area was significantly increased in pBIC^Δ10 lymphomas compared with pBIC (Figure 6D), supporting the antiangiogenic effects of IL-10, consistent with previous observations in Burkitt lymphoma⁵⁵ and other tumors.^62-64 Furthermore, pBIC exhibited a recurrent morphology of mosaic vessels with interspersed GFP⁺ lymphoma cells, whereas pBIC^Δ10 tumors showed more continuous CD31 staining and less evidence of focal regions (Figure 6E). Mosaic tumor vessels have been associated with attenuated endothelial markers and abnormal endothelium,^65,66 suggesting stabilization and maturation of vessels in the absence of lymphoma-derived IL-10. This supports the antiangiogenic role of lymphoma-derived IL-10 in DLBCL, at least in part via the expansion of noncanonical TAMs.

We next investigated the contribution of other IL-10–producing immune populations to the maintenance of a tolerant LME in IL-10–deficient DLBCLs. Myeloid-derived suppressor cells (MDSCs) are immune-suppressive immature cells, classified as either granulocytic (Gr-MDSCs) or monocytic MDSCs (Mo-MDSCs).^67,68 Both pBIC and pBIC^Δ10 mice exhibited relatively normal proportions of intratumoral granulocytic MDSCs (Figure 6F), consistent with the small proportion of MDSCs observed in normal mouse spleens (2%-4%) or in peripheral blood from healthy individuals (<2%).^67,69 Only a slight accumulation of Mo-MDSCs was detected in pBIC lymphomas (Figure 6F) or peripheral blood (Figure 6G). Indeed, previous studies have associated IL-10 and increased serum IL-10 levels with Mo-MDSC promotion in B-cell NHL and other cancers.^70-72 However, despite differences in IL-10 production between pBIC and pBIC^Δ10 lymphomas, serum IL-10 was barely detectable and not significantly different (Figure 6H), suggesting that tumor-derived IL-10 may have a rather localized impact in the LME. Altogether, these results suggest a minor pathogenic role of MDSCs in promoting immune suppression in pBIC^Δ10 mice.

To explore other populations that could contribute to the accrued tolerogenic LME of IL-10–deficient lymphomas, we examined the presence of intratumoral CD4⁺FOXP3⁺ T_regs, which have shown otherwise inconsistent associations with DLBCL clinical outcomes.⁷³ In comparison with pBIC lymphomas, T_regs were significantly enriched in IL-10–deficient pBIC^Δ10 tumors, as evidenced by scRNA-seq (Figure 4E) and flow cytometry (Figure 6I). To test the functional relevance of this expanded T_reg compartment in DLBCL progression, pBIC and pBIC^Δ10 mice at advanced stages of disease were treated IV with anti-CD25^NIB (Figure 6J), a T_reg-depleting mouse IgG2a antibody that does not interfere with IL-2 signaling on effector T cells,⁷⁴ which efficiently reduced intratumoral T_regs in vivo (Figure 6I). Notably, anti-CD25^NIB therapy boosted T-cell activation by limiting the proportion of CD8 TILs coexpressing PD-1⁺LAG3⁺ (Figure 6K) and significantly delayed tumoral onset only in the initially more aggressive pBIC^Δ10 tumors (Figure 6L). Altogether, these results suggest that IL-10–deficient lymphomas are particularly dependent on T_regs, acquiring high sensitivity to anti-CD25 immunotherapy.

We then investigated whether the IL-10–associated hallmarks identified in pBIC/ pBIC^Δ10 mice might have some impact on the prognosis and the response to immunochemotherapy in human patients with DLBCL. As a training cohort, we took advantage of the Schmitz et al² data set, which included bulk RNA-seq and survival data of newly diagnosed patients with DLBCL who received R-CHOP–like therapy (dbGaP phs001444.v2.p1; n = 243). We selected the 3 functional and immunophenotypic categories more significantly enriched in the murine LME after Il10 deletion (T_reg [TREG], angiogenesis [ANGIO], and calcium activity [CAL]; supplemental Table 6) and obtained the single-sample GSEA values for each one in every human sample (Figure 7A, top). We then developed a metric called score for IL-10 (SILX) to capture the state of IL-10–associated hallmarks in bulk transcriptomes of human DLBCL samples (supplemental Methods). We ranked all samples according to SILX values (Figure 7A, bottom) and found that the top 25% of the samples, hereafter referred to as SILX^high (n = 61), showed significantly poorer overall survival after R-CHOP immunochemotherapy than the rest of the patients with medium (SILX^med, n = 121) or low values (SILX^low, n = 61; Figure 7B). Multivariate Cox proportional hazard analysis showed that a SILX^high score was a risk factor independent of Eastern Cooperative Oncology Group (ECOG) performance status, transcriptional cell-of-origin (COO),⁵ or mutational LymphGen^2,4 assignments (Figure 7C). Sex was not a contributing factor when included in the multivariate Cox model (supplemental Figure 2B). On the contrary, SILX^low/med patients showed more favorable responses to R-CHOP (Figure 7B). In fact, SILX identified a subset with better outcomes inside the initially worse prognosis DLBCL subtypes (COO/ABC⁵ or mutational/MCD-N1^2,3), as well as a subset with worse outcomes inside the initially good prognosis DLBCL subtypes (COO/GCB⁵ or mutational/ST2-EZB-A53-BN2^2,3; Figure 7D). These findings were highly consistent with our preclinical data, in which IL-10–deficient pBIC^Δ10 mice showed significantly better in vivo responses to anti-CD20 (Figure 3A). Taken together, this suggests that some of the biological vulnerabilities identified in pBIC/pBIC^Δ10 mouse models, captured by the SILX score, add a new layer of information on top of the current molecular DLBCL classifications and could help predict R-CHOP responses.

To add support to these findings, we explored the association between our SILX groups and other stratification signatures previously established. We observed that SILX^low patients were enriched in inflammatory and mesenchymal signatures⁷⁵ and also in lymphoma ecotypes LE7 and LE9⁷⁶ (Figure 7E), which are associated with better responses to R-CHOP. On the contrary, SILX^high samples were overrepresented in the depleted signature⁷⁵ and in lymphoma ecotypes LE1 and LE3⁷⁶ (Figure 7E), associated with poor survival rates to R-CHOP. Furthermore, SILX^low patients resembled the characteristics of pBIC^Δ10 murine tumors according to other stablished microenvironment signatures, such as TILs with increased exhaustion markers, T_regs and macrophages, increased proangiogenic features (including the stromal-2 signature), and increased calcium signaling signatures (Figure 7E, bottom; supplemental Table 6). Importantly, the predictive potential of SILX was further validated in additional pretreatment tumor biopsies from patients with DLBCL with available bulk gene expression and survival data: GSE10846 (n = 232),⁵⁹ GSE98588 (n = 101),³ GSE87371 (n = 116),^77,78 GSE31312 (n = 469),⁷⁹ and GSE117556 (n = 472).⁸⁰ We calculated the SILX score separately in each cohort to avoid batch effects and then merged their survival data. This showed significantly better overall survival rates in SILX^low than SILX^high patients after R-CHOP–like treatment (Figure 7F) and confirmed that SILX is independent of transcriptional COO, identifying patients with significantly worse prognosis within the ABC and GCB subtypes (Figure 7G). These results are fully consistent with the biology described here in our murine models upon conditional genetic ablation of lymphoma-derived IL-10 and offer promising potential for the prediction of the efficacy of novel immunotherapies currently under investigation for patients with DLBCL.

The data presented here provide important insights into the role played by tumor-produced IL-10 in DLBCL biology. Our pBIC lymphoma model recapitulates the IL-10/JAK1/STAT3 autostimulatory prosurvival loop observed in ABC/MCD-DLBCL,^13,14,29-31 in strong agreement with previous observations that anti–IL-10R neutralization or Il10r/Stat3 ablation impairs B-cell proliferation and tumor growth in a cell-intrinsic manner.^12,81 Moreover, the direct correlation between IL-10 and PD-L1 expression in pBIC/pBIC^Δ10 tumors aligns with the recently described immune-suppressive IL-10/STAT3/PD-L1 pathway in DLBCL,¹⁶ supporting immune evasion through the PD-1/PD-L1 axis in IL-10–expressor lymphomas.

Transcriptomic and immunophenotypic analyses of lymphoma-infiltrating immune populations in IL-10–proficient and –deficient murine ABC-DLBCL models uncovered multifaceted roles for lymphoma-derived IL-10: promoting immune-associated processes (immune chemotaxis, antigen presentation, and higher percentages of effector CD8 T cells and dendritic cells) while restraining the pro-oncogenic effects of T_regs or noncanonical M1/M2 macrophages, the major subset of macrophages in human DLBCL.⁵⁶ Therefore, the absence of lymphoma-derived IL-10 results in an immune-suppressed and exhausted microenvironment, prompting aggressive NF-κB–driven pBIC^Δ10 lymphomas that develop more quickly. These findings align with observations in IL-10/IL-10R–deficient pediatric patients, in whom the constitutive absence of functional IL-10 signaling leads to NF-κB dysregulation and local T-cell immunodeficiency, explaining their Mendelian predisposition to B-cell lymphomagenesis.⁸² Altogether, these evidences support a role for IL-10 in maintaining immune equilibrium and tumor suppression.

Additionally, lymphoma-derived IL-10 shows antiangiogenic activity in the stroma, because its absence increased CD31⁺ vascular area and vessels robustness, consistent with the unfavorable angiogenic switch observed in patients with DLBCL.^60,83,84 Our results confirmed another important role for lymphoma-derived IL-10 as a repressor of L-type calcium channels, rendering lymphomas refractory to anti-CD20–dependent cell killing.^37,85 This, together with the increased angiogenic stroma (probably facilitating access of therapeutic agents to tumors⁸⁶) and antibody-dependent phagocytic populations such as TAMs,⁵² could explain the improved survival rates of pBIC^Δ10 mice after anti-CD20 therapy. This paradoxical situation, in which lymphomas exhibit increased aggressiveness without treatment but greater sensitivity to CD20 immunotherapy, highlights the importance of our preclinical models in elucidating the dual impact of IL-10 in antitumoral immunity.

Another important observation with potential translational implications is that therapies involving T-cell–targeted immune checkpoint therapy (anti–PD-1) or T_reg depletion (anti-CD25) were differentially influenced by lymphoma-derived IL-10. Our results show a shift toward burned-out dysfunctional CD8⁺ T cells and expansion of T_regs in the absence of IL-10, which could explain the resistance to anti–PD-1 immunotherapy in pBIC^Δ10 mice. This suggests an antitumoral role of IL-10, acting as an immune modulator that maintains the fitness of CD8 effector TILs, in contrast to the notion that it behaves as a broadly immune-suppressive cytokine. Indeed, IL-10 can promote the expansion and activation of CD8⁺ TILs by enhancing the expression of IFN-γ and MHC-I/II molecules.^19,87 Furthermore, prevention of tumor establishment, growth, and metastasis have been related to IL-10 expression for other cancers,^19,54,88,89 and patients have recently shown promising responses and CD8⁺ T-cell invigoration after treatment with pegylated IL-10.⁹⁰ We show that IL-10 can enhance antigen presentation by improving MHC-I/II gene expression in lymphoma cells and by facilitating chemokine-mediated attraction of cDCs and protective IFN-producing pDCs.^38-41,45 Similarly, CD25^NIB T_reg-depleting therapy⁷⁴ demonstrated superior tumor control in T_reg-infiltrated IL-10–deficient lymphomas, allowing for T-cell activation and antilymphoma immunity.

Given the potential implications of our findings for predicting the response of patients with DLBCL to various therapeutic strategies, we developed a metric that captures characteristic hallmarks of either high or low expression of IL-10 in tumoral B cells (the SILX score). This score stratified several cohorts of human patients with DLBCL (treated with R-CHOP–like regimes) into SILX^high or SILX^med/low groups, showing significant differences in overall survival. In fact, SILX was an independent risk factor, distinct from COO or genetic classifications. Consistent with previous studies associating poor outcomes with high serum IL-10 levels^9,10 and elevated intratumoral IL-10/IL-10R in 30% to 45% of primary DLBCLs,¹¹ SILX^high patients exhibited significantly worse responses to R-CHOP–like treatment. However, because IL-10 sustains the intratumoral subset of anti–PD-1 responsive TCF1⁺PD-1⁺CD8⁺ T cells with stem-like potential^48-50 and IL-10–expressor pBIC mice benefited from combined ICB immunotherapy, our findings suggest that SILX^high patients with DLBCL might be good candidates for combined CD20 targeting and ICB immunotherapy. On the contrary, Mac/T_reg/exhausted TIL signatures, previously defined from transcriptional deconvolution methods in human DLBCL,^75,76 are enriched in SILX^low patients with DLBCL. Because PD-1 blockade did not add significant benefit to anti-CD20 treatment in pBIC^Δ10 mice, we predict that anti–PD-1 or other T-cell–based immunotherapy approaches would show limited efficacy in the dysfunctional LME of SILX^low patients. However, our results suggest that novel combination therapies incorporating T_reg-depleting strategies⁹¹ or even macrophage immune checkpoint inhibitors⁹² might hold promising potential to improve clinical outcomes for SILX^low patients with DLBCL. For clinical applicability, future research should focus on refining SILX normalization, incorporating clinical parameters, and enhancing detection of SILX-integrated biomarkers to improve the stratification of patients with DLBCL for emerging immunotherapy options tailored to the unique immune landscape modulated by lymphoma-derived IL-10.

In summary, we have identified lymphoma-derived IL-10 as a key immunomodulator in DLBCL, acting through interconnected autocrine and paracrine pathways. Our study reveals 2 distinct biological settings and vulnerabilities in IL-10–proficient and –deficient DLBCL microenvironments, recapitulated in genetically engineered pBIC and pBIC^Δ10 mice, respectively. Although IL-10 promotes autocrine lymphoma survival and resistance to anti-CD20 immunotherapy, it also prevents the drift toward a highly immune-suppressive LME, deprived of stem-like reactive CD8 T cells and enriched in T_regs, TAMs, and burned-out exhausted CD8 TILs. Therefore, DLBCL lymphomas low in tumor-derived IL-10 thrive in a T-cell–dysfunctional microenvironment deficient in cytotoxic effector cells but are more sensitive to anti-CD20 immunotherapy because of improved angiogenesis, expanded macrophages, and increased expression of L-type calcium channels. IL-10–associated hallmarks integrated into the SILX score helped in risk stratification and outcome prediction of future immunotherapy interventions in patients with DLBCL. These models help advance our understanding of DLBCL biology and may serve as in vivo platforms for developing and evaluating novel clinical trial strategies.

The authors thank Axel Roers from the Carl Gustav Carus Faculty of Medicine, Institute for Immunology at the Technische Universität Dresden, for access to the mouse line Il10^F/F; and Genentech for the generous gift of in vivo functional grade anti-mouse CD20 monoclonal antibody (mAb; clone 5D2). The authors also acknowledge the excellent assistance provided for bioinformatics by Victor Segura; for cytometry and cell sorting by Diego Alignani and members of CIMA Lab Diagnostics; for immunohistochemistry by the Morphology unit; and for animal husbandry by members of the Animal Facility, all of them at Cima Universidad de Navarra. They also thank the technical staff of the Department of Pathology, Anatomy and Physiology at the University of Navarra for their assistance and support.

This work was supported by grants to S.R. from Proyecto PID2020-112994RB-I00 financed by MICIU/AEI/10.13039/501100011033, and from Proyectos PID2023-15011OB-I00 and SAF2017-82309-R, cofinanced by MICIU/AEI/10.13039/501100011033 and Fondo Europeo de Desarrollo Regional (FEDER) Una manera de hacer Europa, with the collaboration of the Gilead Scholarship Program for biomedical research (GLD19-00082); and by F. Hoffmann-La Roche through NAV4 imCORE project for the study of anti-mouse CD25^NIB mAb. Additional support was provided by Spanish Ministry of Health, Instituto de Salud Carlos III through CIBERONC (CB16/12/00489; S.R. and J.A.M.-C.), CIBEREHD (CB06/04/0006; P.S.), and Fondo de Investigación en Salud (FIS) grant (PI22/00983; J.A.M.-C.), all of them cofinanced by FEDER Una manera de hacer Europa. Training support was received by Ministerio de Universidades (FPU20/03256; S.C.G.) and (FPU22/01581; A.A.-L.), by Plan de Formación y de I+D 2019 from Gobierno de Navarra (34E/2020; J.M.), and by Asociación de Amigos de la Universidad de Navarra (ADA/2023; B.S.B.).

Contribution: M.G.-L., S.C.G., F.J.N., and S.R. were involved in conceptualization; M.G.-L., S.C.G., E.G., I.C.-L., O.B., F.J.N., and S.R. performed formal analysis; S.R. was involved in funding acquisition; M.G.-L., S.C.G., J.M., M.P., S.M.-S., M.A.B., A.A.-L., and B.S.B. performed investigation; M.G.-L., S.C.G., J.M., M.P., S.M.-S., M.A.B., and S.R. contributed to methodology; M.A., C.K., P.U., and P.S. contributed to resources; M.C., J.A.M.-C., J.J.L., and S.R. provided critical feedback on experimental design and data interpretation; M.G.-L., S.C.G., F.J.N., and S.R. were involved in visualization and wrote the original draft; and all authors reviewed and edited the draft.

Conflict-of-interest disclosure: J.A.M.-C. obtained research funding from Roche-Genentech, Bristol Myers Squibb-Celgene, Janssen, Priothera, Palleon, AstraZeneca, and K36 Therapeutics; and is a founder of and holds stock options in MIMO Biosciences. M.C. had a consultancy agreement with BeiGene, Bristol Myers Squibb, Incyte, Janssen, Karyopharm, Kite, Kyowa, Lilly, Roche, and Takeda; and was on the speaker’s bureau for Kite, Kyowa, Roche, and Takeda. M.A., C.K., and P.U. are employed by Roche Innovation Center Zurich and declare ownership of stock and patents with Roche. S.R. obtained research funding from Roche-Genentech. The remaining authors declare no competing financial interests.

Correspondence: Sergio Roa, Department of Biochemistry and Genetics, Universidad de Navarra, 1 Irunlarrea, 31008, Pamplona, Spain; email: sroa@unav.es.