The TME contains activators of the IRE1/XBP1s axis that are detected by immune cells (25, 26, 31, 32). To identify relevant cell types activating IRE1 RNase in tumors, we analyzed the immune composition of B16 melanoma tumors of ERAI mice, a mouse strain that reports IRE1 RNase activity through expression of Venus Fluorescent Protein (VenusFP) fused with the sequence of Xbp1s (33) [validated in (21, 22, 31)]. Multi-color flow cytometry data (17-color) from ERAI tumors is depicted on a t-distributed stochastic neighbor embedding (t-SNE) map, identifying 15 cell clusters including CD8⁺ T cells, CD4⁺ T cells, monocyte-derived cells (MdCs), NK cells, NKT cells, B cells, neutrophils, cDC1s plus two undefined clusters (Cluster 11: CD4⁺ CD11c⁺ CD26⁺, Cluster 14: CD3⁺ CD4⁺ CD11b^int F4/80⁺ MHC-II^high CD11c^int CD26^high) (Figure 1A, Supplementary Figures 1A, B). Analysis per cluster identified tumor cDC1s as the population with highest VenusFP fluorescence intensity (Figures 1B, C, Supplementary Figure 1C, cluster 15). cDC2s, MdCs, TAMs, neutrophils, NK cells and cells from cluster 11 also showed noticeable VenusFP levels, albeit lower than cDC1s; whereas CD4⁺ T cells, CD8⁺ T cells and B cells showed little or no VenusFP expression (Figures 1B, C). Manual gating analysis confirmed these findings (Figure 1D, Supplementary Figure 1D, see gating strategy in Supplementary Figure 2A), and similar results were observed in cDC1s infiltrating MC38 tumors (Supplementary Figure 1E). Data obtained with ERAI mice were further confirmed by PCR in cDC1s isolated from non-reporter animals bearing B16 tumors, which showed elevated Xbp1 spliced/unspliced ratio (See XBP1^WT, Figure 2C). Altogether, these data indicate that cDC1s display prominent activation of the IRE1 RNase/XBP1s axis in tumors.

We next interrogated whether the high IRE1 RNase activity observed in tumor cDC1s is a lineage-intrinsic signature or if it is a feature regulated by the tumor microenvironment. We compared IRE1 RNase activity in cDC1s from tumor, tumor draining lymph node (TdLN) and inguinal lymph nodes (LN) from tumor-free mice (gating strategy in Supplementary Figure 2B). VenusFP signal was elevated in all cDC1s analyzed but was not further increased in tumor cDC1s or in migratory cDC1s from TdLNs (Figure 1E). Conversely, a slight decrease in VenusFP signal was noted in tumor cDC1s when compared to TdLN-resident cDC1s from tumor-bearing or tumor-free mice (Figure 1E). Similar findings were observed in cDC1s from animals bearing MC38 tumors, suggesting that this may be a shared feature across tumor models (Supplementary Figure 1F). Of note, these observations were not replicated in monocytes or in macrophages from B16 tumors, which express higher VenusFP fluorescence in the tumor site compared to the spleen, indicating that the TME of melanoma is competent to elicit IRE1 activation in myeloid cells (Figure 1F), in line with previous findings (27). Thus, although tumor cDCs display elevated IRE1 RNase activity, this feature corresponds to a stable lineage-intrinsic trait not further enhanced by the TME.

To gain insights in the role of the IRE1/XBP1s axis in tumor DC biology, we generated double conditional knock-out mice lacking the IRE1 RNase domain and XBP1s in CD11c-expressing cells by crossing the Itgax-Cre mice line with Xbp1^fl/fl and Ern1^fl/fl floxed mice lines (35–37) (referred to as “XBP1^ΔDCIRE1^truncDC mice”); and single conditional knock-out mice (referred to as “XBP1^ΔDC mice”) lacking XBP1s in CD11c-expressing cells (see methods for details). The rationale for studying single deletion of XBP1s in DCs is because IRE1 can promote multiple effects via its RNase domain, independent of the XBP1s transcriptional activity and involving degradation of mRNAs/miRNAs through the RIDD branch (16–19), which is observed in steady state cDC1s (21, 22). Also, the XBP1^ΔDC mice line has been previously studied in other tumor models, showing delay in ovarian cancer progression (26). XBP1^ΔDC and XBP1^ΔDCIRE1^truncDC mice are compared to control animals (Xbp1^fl/fl and Xbp1^fl/flErn1^fl/fl littermates with no expression of Cre, respectively). The frequency of tumor cDCs was unaltered in XBP1^ΔDC and XBP1^ΔDCIRE1^truncDC mice compared to control littermates upon implantation of B16F10 tumor cells (Figures 2A, B). We measured IRE1 RNase activity in tumor cDC1s from XBP1^ΔDC and XBP1^ΔDCIRE1^truncDC mice by analyzing the expression of Xbp1 spliced and unspliced mRNA. Although DCs from XBP1^ΔDC mice are unable to synthesize XBP1s protein, these cells still generate mRNA containing the sites for cleavage by IRE1, allowing assessment of IRE1 RNase activity by PCR (Scheme in Supplementary Figure 3A). Whereas tumor cDC1s from control mice displayed high levels of Xbp1 spliced mRNA (Figure 2C), tumor cDC1s isolated from XBP1^ΔDCIRE1^truncDC mice were unable to induce Xbp1 splicing due to the lack of a functional IRE1 RNase domain (Figure 2C). In contrast, tumor cDC1s from XBP1^ΔDC mice showed signs of IRE1 hyperactivation, as indicated by higher expression of the Xbp1 spliced over the unspliced form (Figure 2C).

Considering that XBP1s coordinates expression of immunosuppressive genes in TAMs in melanoma (27, 28), we determined the transcriptomic signature of cDC1s isolated from B16 melanoma tumors of control, XBP1^ΔDC or XBP1^ΔDCIRE1^truncDC mice by bulk RNA sequencing (RNA-seq). Surprisingly, only 51 differentially expressed genes (DEG) were identified among XBP1^ΔDC or XBP1^ΔDCIRE1^truncDC tumor cDC1s (Figure 2D, Supplementary Table 1), corresponding mainly to ER proteins and constituents of the response to misfolded proteins (Supplementary Figure 3B). Gene Set Enrichment Analysis (GSEA) confirmed that canonical XBP1-target genes (34) were downregulated in both XBP1s-deficient and IRE1/XBP1-deficient tumor cDC1s (Figure 2E, Supplementary Figure 3C). These include genes coding for protein disulfide isomerases, chaperones, glycosylation proteins, proteins involved in transport to the ER and from the ER to Golgi. Furthermore, XBP1 single deficient but not IRE1 RNase/XBP1 double deficient cDC1s showed also a decrease in the levels of the RIDD-targets (34) (Figure 2E, Supplementary Figure 3C) Bloc1s1, St3gal5, Itgb2, Rpn1, Erp44, Mlec, Rnf130, Abca2, Stim2, Gm2a, Tmem238, and Paqr7 (16, 21, 34). These data suggest that loss of XBP1s elicits RIDD in tumor cDC1s.

Notably, expression of XBP1s-driven triglyceride biosynthesis genes reported to curtail the function of MoDC/cDC2 in ovarian cancer models (26) was unaltered in XBP1^ΔDC or XBP1^ΔDCIRE1^truncDC mice tumor cDC1s, (Figure 2F, Supplementary Figure 3D). These results demonstrate that cDC1s isolated from B16 tumors maintain a stable transcriptomic signature and that deletion of the entire IRE1/XBP1s axis in tumor cDC1s only alters a discrete subset of genes related with protein homeostasis without affecting expression of pro-tumorigenic genes.

To evaluate if the loss of the IRE1/XBP1s axis in DCs alters the course of tumor growth, we studied immunogenic B78/B16 lines expressing the model antigen ovalbumin. We studied the immunogenic B78ChOVA subcutaneous tumor model that is a B16 variant expressing OVA and mCherry fluorescent protein (2) and that elicits cDC1-mediated antitumor CD8⁺ T cell responses (2, 3). After subcutaneous B78ChOVA implantation, tumor growth in XBP1^ΔDCIRE1^truncDC animals was comparable to that observed in control animals (Figure 3A). Also, similar frequencies of tumor infiltrating CD8⁺ T cells were found in XBP1^ΔDCIRE1^truncDC and control animals (Figure 3B) and no differences were observed in IFN-γ-producing CD8⁺T cells, TNF-producing CD8⁺T cells, IL-2-producing CD8⁺T cells between tumors of XBP1^ΔDCIRE1^truncDC mice and control animals (Figure 3C). Similar results were observed for CD4⁺ T cells (Supplementary Figures 4A, B). We extended our analysis to measure parameters of CD8⁺ T cell exhaustion (38–40) and found similar proportions of ‘exhausted’ TIM-3⁺CD8⁺ T cells and ‘precursor exhausted’ TCF-1⁺CD8⁺ T cells in tumors from XBP1^ΔDCIRE1^truncDC mice compared to controls (Figure 3D). These data show that deletion of the IRE1 RNase/XBP1s branch of the UPR in DCs does not alter the course of melanoma tumor growth or the generation of effector/exhausted T cells at the tumor site.

We investigated if XBP1^ΔDC mice display alterations in tumor growth. Animals bearing single deficiency of XBP1s in DCs showed a modest acceleration of B78ChOVA growth and larger tumor size than tumors from XBP1^WT mice on day 12 post implantation (Figure 3E). Tumor growth kinetics of subcutaneous MC38 tumors from XBP1^ΔDC and XBP1^WT mice did not reach statistical significance (Supplementary Figure 5A).

XBP1^WT and XBP1^ΔDC mice show comparable B78ChOVA tumor immune cell composition and T cell infiltration (Figure 3F Supplementary Figure 6A, B). However, B78ChOVA tumors from XBP1^ΔDC mice contained lower frequencies of IFN-γ-producing and TNF-producing CD8⁺ T cells, decreased frequencies of IFN-γ⁺TNF⁺ CD8⁺ T cells and IFN-γ⁺TNF⁺IL-2⁺ CD8⁺ T cells (Figures 3F, G). Some of these observations were also extended to CD4⁺ T cells, which showed decreased frequencies of IFN-γ⁺ and IFN-γ⁺TNF⁺ CD4⁺ T cells in B78ChOVA tumors and in MC38 tumors (Supplementary Figure 6C, Supplementary Figure 5B, C). B78ChOVA tumors from XBP1^ΔDC mice also show reduced percentages of precursor exhausted TCF-1⁺CD8⁺ T cells compared to control animals (Figure 3H, Supplementary Figure 6D).

Interestingly, the reduced frequencies of effector T cells in B78ChOVA tumors from XBP1^ΔDC mice were not associated with defective cross-presentation of a model antigen in melanoma, as XBP1^WT and XBP1^ΔDC mice lines contained similar frequencies of endogenous OVA-specific CD8⁺ T cells in TdLN and tumors (Supplementary Figure 6E, F). A similar response was obtained when tracking proliferation/early activation of CD8⁺ T cells isolated from pmel mice, which possess transgenic CD8⁺ T cells bearing a TCR selective for the melanoma-associated antigen gp100 (41) (Figure 3I). Also, cDC1, cDC2 and TAMs from melanoma tumors of XBP1^ΔDC mice expressed normal levels of PD-L1 (Figure 3J), an immunoregulatory molecule target of checkpoint blockade therapy which expression is attributed to the IRE1/XBP1s axis in tumor macrophages (28). In addition, in vitro generated bone marrow cDC1s from XBP1^WT and XBP1^ΔDC mice were equally able to produce IL-12 upon stimulation with B78ChOVA lysates (Supplementary Figure 7). We conclude that XBP1 deletion in tumor-associated DCs tunes effector T cell responses independent of cross-presentation or canonical DC activation. Altogether, these data show that double ablation of the IRE1/XBP1 axis in DCs does not alter the course of tumor growth, whereas single XBP1 deficiency in DCs results in mild changes in tumor growth and T cell effector/exhausted profiles at the tumor site. The discrepancies between double versus single knock-out models may be attributed to RIDD counteractivation in XBP1-deficient tumor cDC1s (Figures 2D, E), which is a phenotype reported in XBP1-knock-out cDC1s in steady state (21, 22). In fact, tumor cDC1s, cDC2s and TAMs from XBP1^ΔDC mice display signs of RIDD at protein level, as surface expression of the integrin CD11c, an obligate dimeric partner of the RIDD substrate Itgb2 (coding the integrin CD18) (21) is reduced in tumor cDC1s from XBP1^ΔDC mice (Supplementary Figure 6G).

Finally, we interrogated if selective IRE1/XBP1s deletion in the cDC1 compartment could alter the course of tumor growth and T cell responses. To specifically target cDC1 compartment for genetic deletion of IRE1 RNase and/or XBP1s, we crossed the Xcr1-Cre transgenic mice line (42) with Ern1^fl/fl x XBP1^fl/fl mice (referred to as ‘XBP1^ΔcDC1IRE1^trunc-cDC1 mice’), or with XBP1^fl/fl mice (‘XBP1^ΔcDC1 mice’). XBP1^ΔcDC1IRE1^trunc-cDC1 and XBP1^ΔcDC1 mice showed normal tumor cDC infiltration in MC38 (Supplementary Figures 8A, B) and tumor cDC1s efficiently recombined loxP-flanked sites (Supplementary Figure 8C, D).

We next carried out tumor kinetic studies. Analysis revealed that XBP1^ΔcDC1IRE1^trunc-cDC1 mice showed normal B16ChOVA melanoma tumor growth (Figure 4A), and normal CD8⁺/CD4⁺ T cell tumor infiltration and function compared to control littermates, and a modest decrease in TNF⁺ CD8⁺ T cells (Figure 4B, Supplementary Figure 8E). Similar results were obtained in MC38 tumor models (Figures 4C, D, Supplementary Figure 8E). In agreement with previous results, these data indicate that loss of IRE1 RNase/XBP1s in cDC1s does not impact the outcome of tumor growth or antitumor T cell function.

In addition, analysis of mice carrying single deletion of XBP1s in cDC1s (XBP1^ΔcDC1) also showed comparable B16ChOVA tumor size with control counterparts (Figure 4E), along with normal effector/exhausted T cell responses (Figure 4F, Supplementary Figure 8F). Similar results were also observed in MC38 tumors (Figures 4G, H, Supplementary Figure 8F). As seen with XBP1^ΔDC mice, tumor cDC1s from XBP1^ΔcDC1 mice also showed signs of RIDD, as indicated by an increase in XBP1s splicing ratio (Supplementary Figure 8G) and loss of CD11c surface expression (Supplementary Figure 8H), albeit to a lower extent than observed with Itgax-Cre mice line (Supplementary Figure 6G). Altogether, these data suggest that alterations in the IRE1/XBP1s axis in the tumor cDC1 compartment are not sufficient to shift the balance of antitumor T cell responses.

No statistical methods were used to predetermine sample size. The experiments were not randomized, and the investigators were not blinded to allocation during experiments and outcome assessment. Statistical analysis was conducted using GraphPad Prism software (v9.1.2). Results are presented as mean ± SEM. Two groups were compared using two tailed t-test for normal distributed data (Shapiro-Wilk test) or using a non-parametric two-tailed Mann-Whitney test as indicated in figure legends. Multiple groups were compared using one-way ANOVA with Tukey post-test. A p-value < 0.05 was considered statistically significant.

All animal procedures were approved and performed in accordance with institutional guidelines for animal care of the Fundación Ciencia y Vida, the Faculty of Medicine, University of Chile and the VIB site Ghent-Ghent University Faculty of Sciences and were approved by the local ethics committee.