Erythropoietin receptor on cDC1s dictates immune tolerance

Abstract

Type 1 conventional dendritic cells (cDC1s) are unique in their efferocytosis¹ and cross-presenting abilities², resulting in antigen (Ag)-specific T cell immunity³ or tolerance^4–8. However, the mechanisms underlying cDC1 tolerogenic function remain largely unknown. Here, we show that the erythropoietin receptor (EpoR) acts as a critical switch that determines the tolerogenic function of cDC1s and the threshold of Ag-specific T cell responses. In total lymphoid (TLI)-induced allograft tolerance^9,10, cDC1s upregulate EpoR expression, and conditional knockout of EpoR in cDC1s diminishes Ag-specific FOXP3⁺ T_reg induction and expansion, resulting in allograft rejection. Mechanistically, EpoR promotes efferocytosis-induced tolerogenic maturation^7,11 of splenic cDC1s towards late-stage CCR7⁺ cDC1s characterized by elevated integrin β8 gene¹² (Itgb8) expression, and conditional knockout of Itgb8 in cDC1s impairs TLI-induced tolerance. Migratory cDC1s in peripheral lymph nodes (pLNs) preferentially express EpoR and their FOXP3⁺ T_reg inducing capacity is enhanced by EPO. Reciprocally, loss of EpoR enables immunogenic maturation of both pLN migratory and splenic CCR7⁺ cDC1s by upregulating genes involved in MHC class II-mediated Ag presentation, cross-presentation, and costimulation. EpoR deficiency in cDC1s reduces tumor growth by enhancing anti-tumor T-cell immunity, particularly increasing the generation of precursor exhausted tumor Ag-specific CD8⁺ T cells (T_pex)¹³ in tumor draining LNs and supporting their maintenance within tumors, while concurrently reducing intratumoral T_regs. Targeting EpoR on cDC1s to induce or inhibit T cell immune tolerance could pave the way for treating a variety of diseases.

Immune tolerance¹⁴, a state of indifference or non-reactivity towards a substance that would normally be expected to excite an immunological response, is beneficial in transplantation and autoimmune diseases but detrimental in cancer¹⁵. cDC1s are prototypical Ag-presenting cells specialized in acquiring cell-associated Ags by taking up apoptotic cells via a process called efferocytosis^1,16 or necrotic cells¹⁷ to induce corresponding cell-associated Ag-specific T cell immune responses^18–20. By virtue of their unique capacity to cross-present cell-associated Ags to CD8⁺ T cells², cDC1s are required for immunity against tumors and viral infections²¹. In the cancer-immunity cycle²², cDC1s are vital for cancer immune surveillance by priming tumor Ag-specific CD8⁺ T cells leading to the generation of precursor exhausted T cells (T_pex) in tumor-draining lymph nodes (tdLNs) and recruiting and restimulating immune effector cells in the tumor microenvironment (TME)^23–30. The functional state of cDC1s coupled with Ag uptake and subsequent Ag presentation dictates both the direction, i.e., immunogenic versus tolerogenic, and intensity of an Ag-specific immune response³¹. The acquisition of dead cells by cDC1s not only initiates both CD4⁺ and CD8⁺ T cell priming and activation¹⁹ but can also result in tolerogenic programming and maturation of cDC1s^5,7,32, leading to the induction of Ag-specific CD4⁺FOXP3⁺ T_regs^4,5 and deletion of Ag-specific CD8⁺ T cells³³. cDC1s contribute to homeostatic tolerance⁵, tolerance to dietary Ags^34,35, and tolerance in autoimmunity^6,36 and transplantation³⁷. However, the mechanisms that determine how cDC1s become tolerogenic for T cell adaptive immunity remains unknown.

Tolerogenic cDC1s upregulate EpoR after TLI

To investigate the mechanisms of cDC1-mediated tolerance, we used a protocol combining total lymphoid irradiation (TLI), anti-thymocyte serum (ATS), and allogeneic donor bone marrow (BM) infusion³⁸, which induces mixed chimerism and donor-specific tolerance in kidney transplant patients, enabling the patients to be weaned off immune-suppressive drugs without graft rejection^9,10. In mice, this approach reliably induces lifelong mixed chimerism and tolerance to fully MHC-mismatched organs³⁹. cDC1s are indispensable for immune tolerance³⁸, as Batf3^−/− mice lacking cDC1s failed to establish BM chimerism (Fig. 1a,b) or maintain allo-heart grafts (Fig. 1a,c). After TLI/ATS, total splenic cell numbers decreased by ~60% (Extended Data Fig. 1a), but cDC1s increased over twofold among cDCs (Extended Data Fig. 1b,c), with >97% expressing XCR1⁴⁰ (Extended Data Fig. 1d). cDC1s from TLI/ATS–treated mice showed increased Ki67, indicative of proliferation (Extended Data Fig. 1e). Their identity was confirmed by high expression of Batf3⁴¹, IRF8⁴² (Extended Data Fig. 1f,g), Zbtb46-GFP⁴³ (Extended Data Fig. 1h) and low expression of MafB-mCherry⁴⁴ (Extended Data Fig. 1i–k). RNA–seq and Gene Ontology (GO) analysis of XCR1⁺CD8α⁺ cDC1s after TLI/ATS revealed upregulation of genes involved in phagocytosis, apoptotic cell clearance, erythropoiesis, and myeloid cell migration (Fig. 1d), aligning with increased apoptotic lymphocytes (Extended Data Fig. 2a,b), serum EPO (Extended Data Fig. 2c) and splenic erythropoiesis (Extended Data Fig. 2d,e). Epor was upregulated (log₂FC = 3.45; P = 0.000164; Fig. 1e) in cDC1s following TLI/ATS, and qPCR confirmed elevated expression of Epor, Axl, Mertk, and Cd5l (Fig. 1f).

Fig. 1 |. Efferocytotic tolerance-inducing cDC1s upregulate Epor following TLI.

a, Schematic of TLI/ATS treatment, allo-BM infusion, and heart transplantation in C57BL/6 and Batf3^−/− mice. b, Donor-type (H2Kᵈ⁺) leukocytes (B220⁺, TCRβ⁺, Ly6G⁺, CD64⁺) in peripheral blood 28 days post-BM infusion. WT (n=10) vs. Batf3^−/− (n=7). c, Heart allograft survival in C57BL/6 (n=10) vs. Batf3^−/− (n=7). d-e, RNA-seq of splenic cDC1s from UNT or TLI/ATS-treated mice. d, Enriched GO biological processes. e, Volcano plot of differentially expressed genes. f, qPCR of selected genes (UNT, n=8; TLI/ATS, n=6). g-i, Flow cytometry analysis of Epor-tdTomato (tdT) on cDC1s/cDC2s from UNT, TLI, and TLI/ATS-treated mice; summary of cDC1 frequency (h) and EpoR-tdT⁺ cells (i), (n=5/group). j, tdT expression in Epor⁺ cDC1s: histogram (left) and MFI (right) (UNT, n=5; TLI/ATS, n=4). k-l, RNA-seq of splenic Epor-tdT⁺ vs. EpoR-tdT⁻ cDC1s post-TLI/ATS (n=2/group, pooled from 15 mice). k, Volcano plot of upregulated genes. l, Heatmap of genes enriched in EpoR-tdT⁺ cDC1s. Data are pooled from 2 independent experiments (b). Data are shown from one experiment, representative of at least three independent experiments with similar results (c,g,h,i,j). Statistical analysis was performed using unpaired two-tailed Student’s t-test (b,f,j), one-way ANOVA (h) or two-way ANOVA with Tukey’s multiple-comparison test (i), or Kaplan-Meier survival analysis with Mantel-Cox test (c). P-values were calculated using hypergeometric tests with Benjamini-Hochberg correction (d) or two-sided generalized linear model (GLM) likelihood ratio tests with Benjamini-Hochberg correction (e,k). Data are mean ± s.e.m. (b,f,h,i,j). The diagram in a was created with BioRender by Xiangyue Zhang (2025).

To confirm EpoR expression on cDC1s, we utilized Epor-tdTomato-Cre mice⁴⁵. After TLI/ATS, cDC1s increased threefold (Fig. 1g,h), similar to wild-type (WT) C57BL/6 mice (Extended Data Fig. 1c), and Epor-tdT expression specifically increased markedly on cDC1s (Fig. 1g,i,j). Five days of treatment with recombinant EPO (Extended Data Fig. 2f) expanded cDC1s, increased their Epor-tdT expression (Extended Data Fig. 2g,h), and elevated erythroid progenitors (Extended Data Fig.2i) and red pulp macrophages (RPMΦs) (Extended Data Fig. 2j), validating Epor-tdT as a functional EpoR marker. After TLI/ATS, Epor-tdT was enriched on CCR7⁺ mature^5,7 cDC1s (Extended Data Fig. 2k), and was accompanied by increased CD103⁷ expression (Fig. 1g), which is associated with high engulfment capacity. Gene set enrichment analysis (GSEA) revealed enrichment of metabolic and mTOR gene sets (Extended Data Fig. 3a–c), consistent with enhanced EPO-EpoR signaling⁴⁶ in cDC1s, as evidenced by increased phosphorylation of Akt, Erk, Stat5, S6, and 4E-BP1 (Extended Data Fig. 3d), with minimal activation in cDC2s (Extended Data Fig. 3e).

Fig. 3 |. scRNA-seq analysis reveals that TLI/ATS promotes EpoR-dependent, efferocytosis-triggered tolerogenic maturation of splenic cDC1s.

a, UMAP of splenic cDC1s colored by subtype with dot plot of marker genes. b, cDC1 subtype proportions in UNT vs. TLI/ATS-treated Epor^flox/flox; box highlights increased mature cDC1s after TLI/ATS. c, UMAP by sample identity with violin plots of Itgae, Lgals3, Apol7c. d, Log₂ fold-change of cDC1 subtypes post-TLI/ATS in Epor^flox/flox and Epor^ΔXcr1 mice; box shows EpoR-dependent differences in mature cDC1s. e, Flow cytometry showing cDC1% in splenic Lin⁻SiglecH⁻PDCA-1⁻CD11c^highMHCII^high cDCs (Epor^flox/flox: UNT, n=5; TLI/ATS, n=6; Epor^ΔXcr1: UNT, n=5; TLI/ATS, n=6). f, UMAP with violin plots of Cd83, Rel, Dnase1l3 in cDC1s from Epor^flox/flox and Epor^ΔXcr1 mice post-TLI/ATS. g, Bar charts of cDC1 subtype proportions and h, UMAP and violin plots of Dnase1l3, Xcr1, Cd274 in Epor^flox/flox and Epor^ΔXcr1 mice at baseline (UNT). i, qPCR of selected genes in CCR7⁺XCR1⁺SIRPα⁻ cDC1s from Epor^flox/flox (n=5/condition) and Epor^ΔXcr1 mice (n=5/condition) (UNT or TLI/ATS). j, Itgb8^ΔXcr1 vs. control recipients UNT or treated with TLI/ATS, and infused with BALB/c BM; frequencies of recipient H-2K^b+TCRβ⁺CD4⁺ and FOXP3⁺ T_regs in CD4⁺ T cells were analyzed on day 0 (UNT, n=5; n=5 and TLI/ATS, n=5;n=5) or 14 days post-BM infusion (UNT, n=5; n=5 and TLI/ATS, n=5; n=5). k,l, Itgb8^ΔXcr1 (n=8) vs. controls (n=5) and infused with 2W1S-BALB/c BM. Donor-type leukocyte percentages (k) and the frequency and absolute number of FOXP3⁺2W1S-tetramer⁺CD44⁺CD4⁺ T_regs in spleens (l) were analyzed 14 days later. Data are representative of two (e,i) or one (j,k,l) independent experiments. Statistical analysis was performed using unpaired two-tailed Student’s t-test (i,k,l), or two-way ANOVA followed by Tukey’s multiple-comparison test (e,j), or propeller test, two-sided, no multiple-comparison correction (b.g), or Wilcoxon rank sum test, two-sided, with Bonferroni correction (c,f,h). Data are mean ± s.e.m. (e,i,g,k,l)

RNA-seq on fluorescence-activated cell sorting (FACS)-sorted Epor-tdT⁺ and Epor-tdT⁻ cDC1s following TLI/ATS again showed that EpoR was highly upregulated in EpoR-tdT⁺ cDC1s (Fig. 1k,l) along with a panel of genes contributing to immune regulation (Siglece, Vsig4, Tgfb2, Arg1, Aldh1a1), efferocytosis (Axl, Gas6, Pros1, Lrp1, Timd4), lipid metabolism (Cd5l, Fabp4, Lpl, Ptgs1, Pparg, Nr1h3, Apoe, Abca1), and iron metabolism (Hmox1^47,48, Slc40a1) (Fig. 1k,l). Thus, TLI/ATS markedly increased both the cycling of cDC1s and their EpoR expression. Combined with local accumulation of apoptotic cells and elevated EPO, these findings point to the possibility that EPO-EpoR signaling is involved in the tolerogenic role of cDC1s in the context of efferocytosis⁴⁹.

cDC1-EpoR is required for TLI-induced tolerance

To further investigate the tolerogenic role of EpoR on cDC1s, we next generated Epor^ΔXcr1 (Xcr1^Cre/+Epor^{flox/flox6,50}; H-2K^b+) mice and found that Akt-mTOR signaling activation following TLI/ATS was abrogated specifically in cDC1s but not in cDC2s (Extended Data Fig. 3f,g), confirming efficient EpoR deletion. Ex vivo EPO stimulation further validated the cDC1-specific responsiveness of EpoR to EPO, as shown by enhanced phosphorylation of Akt, Erk, Stat5, S6, and 4E-BP1 in cDC1s isolated from EpoR intact TLI/ATS-treated Epor^flox/flox mice (Extended Data Fig. 3h). As in Batf3^−/− mice (Fig. 1b,c), loss of EpoR in cDC1s abrogated TLI/ATS-induced BM chimerism (Fig. 2a,b) and led to allograft rejection within 2 weeks (Fig. 2a,c), indicating that cDC1-EpoR signaling is required for TLI/ATS-induced tolerance.

Fig. 2 |. Absence of EpoR on cDC1s abrogates T_reg-mediated alloantigen-specific tolerance following TLI/ATS, resulting in allograft rejection.

a, Schematic of TLI/ATS treatment, allo-BM infusion, and heart transplantation in Epor^flox/flox and Epor^ΔXcr1 mice. b, Donor-type (H2Kᵈ⁺) leukocytes (B220⁺, TCRβ⁺, Ly6G⁺, CD64⁺) in peripheral blood 28 days post-BM infusion. Epor^flox/flox (n=10) vs. Epor^ΔXcr1 (n=10). c, Heart allograft survival in Epor^flox/flox (n=10) vs. Epor^ΔXcr1 mice (n=8). d,e, Foxp3-DTR mice conditioned with TLI/ATS; Group A received DT from day 1 to day 14 post-BM infusion (n=8; n=8; n=8), Group B received DT from day 15 after confirmation of BM chimerism (n=8; n=5; n=8). BM chimerism was assessed on the indicated days. e, Summary of BM chimerism on the indicated days post-BM infusion. f, Epor-tdT⁺ or Epor-tdT⁻ cDC1s cocultured with CTV-labeled naïve OT-II cells and EPO for 5 days or with PBS (w/o); FOXP3 expression on OT-II cells was assessed by flow cytometry (Epor-tdT⁺ cDC1s: +PBS, n = 5; +EPO, n = 6; Epor-tdT⁻ cDC1s: +PBS, n = 4; +EPO, n = 6). g, C57BL/6, Batf3 ^−/−, or Epor^ΔXcr1 recipients treated with TLI/ATS and infused with BALB/c BM; frequencies of recipient H-2K^b+ TCRβ⁺CD4⁺ and FOXP3⁺ CD4⁺ T cells analyzed on day 0 (UNT, n=5; n=5; n=5 and TLI/ATS, n=6; n=5; n=5) or 14 days post BM infusion (UNT, n=5; n=5; n=5 and TLI/ATS, n=5; n=5; n=5). h, Epor^flox/flox and Epor^ΔXcr1 recipients given TLI/ATS or UNT; 14 days post-2W1S-BALB/c donor BM infusion, 2W1S-tetramer⁺CD44⁺H-2K^b+TCRβ⁺CD4⁺ T cells from the spleens were analyzed for FOXP3 expression (Epor^flox/flox, n=7; Epor^ΔXcr1, n=7). Data are pooled from 2 independent experiments (b) or shown from one experiment, representative of at least two independent experiments with similar results (c,e,f,g,h). Statistical analysis was performed using unpaired two-tailed Student’s t-test (b,h), or two-way ANOVA with Tukey’s multiple-comparison test (e,f,g), or Kaplan-Meier survival analysis with Mantel-Cox test (c). Data are mean ± s.e.m. (b,e,f,g,h,). The diagrams in a,d,g,h were created with BioRender by Xiangyue Zhang (2025).

Given the critical role of FOXP3⁺ T_regs in implementing and maintaining immune tolerance⁵¹, and the fact that stable BM chimerism is crucial for tolerance and long-term graft survival⁵², we hypothesized that EpoR⁺ cDC1s mediate their tolerogenic effects by inducing T_regs in response to allo-BM infusion. To assess this hypothesis, we depleted FOXP3⁺ T_regs in Foxp3-DTR mice with diphtheria toxin (DT) during days 0–14 (Group A, red) or 15–28 (Group B, blue) (Fig. 2d) following TLI/ATS and allo-BM infusion. DT reduced FOXP3⁺ T_regs by >90% (Extended Data Fig. 4a). Depletion on days 0–14 abolished BM chimerism (Fig. 2e, red), whereas depletion on days 15–28 prevented its further increase and caused a decline by day 28 (Fig. 2e, blue). These data demonstrate that T_regs are required for both induction and maintenance of immune tolerance following TLI/ATS treatment. To test whether EpoR signaling in cDC1s can drive T_reg induction in vitro, we first co-cultured cell trace violet (CTV)-labeled naïve OT-II T cells with Epor-tdT⁺ or Epor-tdT⁻ cDC1s together with apoptotic Act-mOVA⁵³ thymocytes. EpoR⁺ cDC1s generated more Ag-specific FOXP3⁺ T_regs than EpoR⁻ cDC1s, which was further enhanced by EPO (Fig. 2f). Next, to determine if cDC1 EpoR expression is required for T_reg induction in vivo, we measured FOXP3⁺ T_regs during TLI/ATS tolerance induction in WT, Batf3^−/−, and Epor^ΔXcr1 mice vs. untreated (UNT) controls. After TLI/ATS, on day 0, before BM infusion, CD4⁺ T-cell and FOXP3⁺ T_reg frequencies were similar across strains (Fig. 2g, top; Extended Data Fig. 4b,d). At 14 days after allo-BM infusion, conventional CD4⁺ T cells rose from 1–2% to >10% in Batf3^−/− and Epor^ΔXcr1 mice, whereas they remained 1–2% in WT mice (Fig. 2g; Extended Data Fig. 4c,e). T_reg frequencies increased in WT mice but markedly decreased in Batf3^−/− and Epor^ΔXcr1 mice (Fig. 2g, bottom; Extended Data Fig. 4c,e). Additionally, in MHCII^ΔXcr1 (Xcr1^Cre/+H2-Ab1^flox/flox54) recipients (Extended Data Fig. 4f), BM chimerism establishment (Extended Data Fig. 4g) and FOXP3⁺ T_reg expansion failed, while conventional CD4⁺ T cells increased (Extended Data Fig. 4h,i), indicating that EpoR signaling in cDC1s drives tolerance via MHCII-mediated T_reg induction.

Fig. 4 |. EPO-activated cDC1 EpoR supports Ag-specific FOXP3⁺ T_reg induction in pLN and restrains the immunogenic maturation of CCR7⁺ cDC1s.

a,b, Epor-tdT expression on cDCs in pLNs (n=5) and mLNs (n=5) of Zbtb46^GFP/+Epor^tdT/+ mice (a) and statistical graph comparison (b). c, Epor-tdT⁺ cells were identified in migratory XCR1⁺ cDC1s or XCR1⁻ cDC2s (upper) and in migratory vs. resident cDCs (lower) in pLN Zbtb46-GFP⁺CD11c⁺ cDCs (n=5). d, Epor-tdT expression in pLNs and e, mammary fat pad cDC1s of Epor^tdT/+mice (n=8). f, Efferocytosis of PKH67-labeled CD45.1⁺ apoptotic thymocytes by migratory or resident cDCs in the dLN of CD45.2⁺Epor^tdT/tdT mice (n=7) 12 h after injection of the apoptotic cells into the 3^rd mammary fat pad. g, Effect of EPO on CD45.1⁺ OT-II T_reg induction after Act-mOVA thymocyte injection into the 3^rd mammary fat pad (Epor^flox/flox: +PBS, n=5; +EPO, n=5; Epor^ΔXcr1: +PBS, n=5; +EPO n=5). h, Frequency and absolute number and i, XCR1 MFI of migratory cDC1s per pLN (Epor^flox/flox, n= 15; Epor^ΔXcr1, n=15). j,k, scRNA-seq of pLN migratory cDC1s from Epor^flox/flox and Epor^ΔXcr1; heatmap of DEGs and UMAP colored by cluster identity, gene expression. The violin plots represent module score of CD4⁺ T helper licensing gene signature (259 genes) in migratory cDC1s from Epor^flox/flox (n = 6,890 cells) and Epor^ΔXcr1 (n = 7,225 cells) pLNs. The boxes inside the violin plots show the median (centre line) and the interquartile range (25% to 75%; box limits). l, MFI of indicated molecules on pLN migratory cDC1s from Epor^flox/flox (n=6) and Epor^ΔXcr1 mice (n=6). m, Heatmap of top shared DEGs in pLN migratory and splenic CCR7⁺ cDC1s from bulk RNA-seq (Epor^flox/flox vs. Epor^ΔXcr1). Data are shown from one experiment, representative of at least three independent experiments with similar results (a-h, l). Statistical analysis was performed by using unpaired two-tailed Student’s t-test (b,c,e,f,h,i,l), or two-way ANOVA with Tukey’s multiple-comparison test (g), or uncorrected Wilcoxon rank-sum test, one-sided (k). Data are mean ± s.e.m. (b,c,e,f,h,i,l). The diagrams in b,c,d,e,f,g were created with BioRender by Xiangyue Zhang (2025).

TLI/ATS reprograms the immune system within a strict time window to recognize allo-BM as self, enabling durable chimerism and donor Ag-specific tolerance^9,55. To confirm the Ag-specificity of the T_regs induced by EpoR⁺ cDC1s following allo-BM infusion, we infused the recipients with 2W1S-BALB/c⁵⁶ BM (Fig. 2h); the high precursor frequency of naïve CD4⁺ T cells that recognize MHC class II I-A^b 2W1S_55–68⁵⁶ allows for identification of endogenous donor-specific recipient FOXP3⁺ T_regs using MHCII tetramer staining. In UNT mice, 14 days after 2W1S-BALB/c allo-BM infusion, only 2–3% of CD44⁺2W1S-tetramer⁺CD4⁺ T cells were FOXP3⁺, in either Epor^flox/flox or Epor^ΔXcr1 recipients (Fig. 2h), indicating a strong alloreactive response of UNT recipient CD4⁺ T cells upon allo-Ag encounter. In contrast, TLI/ATS-conditioned Epor^ΔXcr1 recipients showed 15–22% FOXP3⁺ T_regs in 2W1S-tetreamer⁺CD4⁺ T cells, compared with > 55% in Epor^flox/flox recipients (Fig. 2h).

To investigate whether cDC1-specific EpoR expression drives T_reg differentiation from FOXP3 negative naïve precursors, we compared FOXP3 expression among CD4⁺ T cells with I-A^b:2W1S specificity from FOXP3^DTRCD45.1⁺⁵⁷ donor CD4⁺ T cells after adoptive transfer into FOXP3^WT CD45.2⁺ Epor^ΔXcr1 or Epor^flox/flox recipient mice which were then treated with DT to eliminate natural T_regs or not. Although there were comparable FOXP3⁺ T_reg percentages between FOXP3^DTR CD45.1⁺ transferred CD4⁺ T cells and FOXP3^WTCD45.2⁺ endogenous CD4⁺ T cells in either Epor^ΔXcr1 or Epor^flox/flox recipients without DT treatment, we observed reduced FOXP3⁺ T_regs in both populations in Epor^ΔXcr1 recipients but not Epor^flox/flox recipients (Extended Data Fig. 4j). In DT-treated Epor^flox/flox recipients, FOXP3^DTRCD45.1⁺CD4⁺ T cells differentiated into T_regs at lower levels than endogenous FOXP3^WTCD45.2⁺CD4⁺ T cells (Extended Data Fig. 4j). Of note, in DT-treated Epor^ΔXcr1 recipients, T_reg differentiation in transferred cells was reduced by more than 50% relative to the already diminished FOXP3 expression seen in both endogenous FOXP3^WTCD45.2⁺ CD4⁺ T cells and transferred FOXP3^DTRCD45.1⁺CD4⁺ T cells from untreated Epor^ΔXcr1 recipients, when compared with Epor^flox/flox controls (Extended Data Fig. 4j). These findings demonstrate the necessity of cDC1-specific EpoR expression for T_reg induction as well as expansion of preexisting T_regs.

EpoR promotes tolerogenic maturation of cDC1s

Since splenic cDC1s undergo efferocytosis-induced tolerogenic maturation in the homeostatic state^7,11, we hypothesized that EpoR signaling promotes this program following TLI/ATS. To address this possibility, we used the 10x Genomics platform to perform scRNA-seq on splenic cDC1s from Epor^flox/flox and Epor^ΔXcr1 mice, under UNT vs. TLI/ATS conditions (Fig. 3a–d,f,g,h). Samples were processed in parallel with cell hashing, yielding a dataset of 30,938 cells. Aligning with previous reports⁷, in both UNT and TLI/ATS mice, unsupervised clustering and differential gene expression (DEG) analysis identified pre-cDC1s along with proliferating, early- and late-immature, early-mature, and Ccr7⁺ late-mature cDC1s (Fig. 3a) that are tolerogenic in the homeostatic state¹¹. Distinct gene expression signatures were linked with TLI/ATS (e.g., Txn1, Xcr1, Atp5k) and Xcr1^Cre6-driven EpoR conditional deletion (e.g., heat-shock protein genes), shared across multiple cDC1 subtypes (Fig. 3a–c; Extended Data Fig. 5a,b). Notably, TLI/ATS treatment increased early and Ccr7⁺ late-mature cDC1s (Fig. 3b, box) and upregulated efferocytosis-related genes (e.g., Itgae⁷, Lgals3⁵⁸), mirroring bulk RNA-seq analysis (Fig. 1e), while reducing Apol7c⁵⁹ expression (Fig. 3c), indicating enhanced tolerogenic maturation with reduced cross-presentation capacity of cDC1s in Epor^flox/flox mice. By contrast, Epor^ΔXcr1 mice displayed attenuated shifts: Ccr7⁺ late-mature cells decreased 1.5-fold vs. a 1.5-fold increase in Epor^flox/flox controls, and early-mature cells rose only 1.9-fold vs. 3.4-fold (Fig. 3d). Flow cytometry showed a slightly lower baseline proportion of splenic cDC1s in UNT Epor^ΔXcr1 mice compared with Epor^flox/flox controls (Fig. 3e; Extended Data Fig. 5c). Following TLI/ATS, in Epor^ΔXcr1 mice, cDC1 expansion was reduced (2.6-fold vs. 1.7-fold) and absolute cDC1 numbers were only one-third of those in Epor^flox/flox controls (Fig. 3e; Extended Data Fig.5c). EpoR deletion also downregulated tolerance-associated genes (Cd83⁶⁰, Rel, Dnase1l3⁶¹; Fig. 3f), highlighting the essential role of EpoR in cDC1 expansion and tolerogenic maturation after TLI/ATS.

To compare the transcriptional profiles of EpoR⁺ and EpoR⁻ cDC1s, we performed scRNA-seq on sorted Epor-tdT⁺ and Epor-tdT⁻ cDC1s from TLI/ATS-treated Epor-tdTomato-Cre⁶² mice, yielding 24,761 cells (Extended Data Fig. 5d–h). These data revealed that EpoR⁺ cDC1s spanned the continuum of tolerogenic maturation (Extended Data Fig. 5d,e) and thus did not constitute a specific subpopulation of cDC1s. Further, DEG analysis revealed unique sets of genes associated with Epor-tdT⁺ or Epor-tdT⁻ cDC1s (Extended Data Fig. 5f). Notably, Epor-tdT⁻ cDC1s exhibited increased expression of genes associated with Epor^ΔXcr1 mice following TLI/ATS, whereas Epor-tdT⁺ cDC1s expressed elevated levels of genes associated with Epor^flox/flox mice conditioned with TLI/ATS (Extended Data Fig. 5g). Similarly, Epor-tdT⁺ cDC1s were proportionally biased towards more mature cDC1 subtypes (i.e., immature, early mature and Ccr7⁺ late mature), while EpoR⁻ cDC1 proportions correlated with more immature states (Extended Data Fig. 5h). Complementing these results, we observed that in UNT Epor^ΔXcr1 mice there was a reduction in the proportion of Ccr7⁺ late mature cDC1s accompanied by an increase in immature cDC1s (Fig. 3g). EpoR deletion reduced Dnase1l3⁶¹ and Xcr1 expression levels across all splenic cDC1s (Fig. 3h, top), as well as decreased expression of the coinhibitory receptor Cd274 in both early and Ccr7⁺ late mature cDC1s (Fig. 3h, bottom). Collectively, these results illustrate that EpoR⁺ cDC1s do not represent a unique cDC1 subtype and instead reflect a unique transcriptional program associated with cDC1 tolerogenic maturation.

cDC1-integrin β8 is critical for TLI-induced tolerance

CCR7⁺ late-mature cDC1s showed higher expression of T_reg-inducing and -maintaining genes, i.e., Itgb8¹², Scube3, Tgfb1, Ccl22, Aldh1a2⁶³ compared with CCR7⁻ cDC1s^7,11. qPCR confirmed that these genes were significantly upregulated in CCR7⁺ cDC1s after TLI/ATS in an EpoR-dependent manner (Fig. 3i), and their expression levels in CCR7⁺ cDC1s in the homeostatic state were also reduced in the absence of EpoR (Fig. 3i). Ex vivo coculture of CCR7⁺ cDC1s obtained after i.v. injection of apoptotic Act-mOVA thymocytes confirmed the involvement of TGFβ in the enhanced capacity of EpoR expressing CCR7⁺ cDC1s to induce Ag-specific FOXP3⁺ T_regs (Extended Data Fig. 5i). αvβ8 contributes to peripheral T_reg differentiation due to its ability to activate latent TGFβ¹². To test whether EpoR⁺ cDC1-mediated induction of allo-BM-specific T_regs depends on integrin β8, we generated Itgb8^ΔXcr1 mice and infused them or littermate controls with allo-BM. 14 days after allo-BM infusion, T_reg frequencies increased in control mice but markedly decreased in Itgb8^ΔXcr1 mice (Fig. 3j and Extended Data Fig. 5j,k,l,m), similar to our observations in EpoR^ΔXcr1 mice (Fig. 2g and Extended Data Fig. 4b,c,d,e). Moreover, Itgb8^ΔXcr1 recipients exhibited impaired BM chimerism (Fig. 3k), albeit to a lesser extent than Epor^ΔXcr1 mice (Fig. 2b), and showed a lower proportion and cell number of FOXP3⁺ T_regs among CD44⁺2W1S-tetramer⁺ CD4⁺ T cells (Fig. 3l). Aldh1a2 encodes retinaldehyde dehydrogenase 2 (RALDH2)⁶⁴, which catalyzes the production of retinoic acid to support T_reg induction⁶³. We next generated mixed BM chimeras by reconstituting CD45.1⁺ mice with a 1:1 ratio of Aldh1a2^ΔCD11c: Batf3^−/− BM cells, in which only cDC1s were deficient in Aldh1a2 expression, or with 1:1 ratio of Aldh1a2^flox/flox: Batf3 ^−/− mixed BM cells as controls. Unlike Epor^ΔXcr1 and Itgb8^ΔXcr1 mice, there was no difference in either BM chimerism or 2W1S-specific FOXP3⁺ T_regs between these mice (Extended Data Fig. 5n,o). These findings verify the critical role of EpoR in facilitating efferocytosis-triggered tolerogenic maturation of cDC1s towards late mature stage CCR7⁺ cDC1s and demonstrate that integrin β8, but not Aldh1a2, is a critical tolerogenic downstream mediator under EpoR control in TLI/ATS-induced tolerance.

cDC1-EpoR limits CD8⁺ and CD4⁺ T cell priming

cDC1s specialize in cross-presenting exogenous cell-associated Ags to CD8⁺ T cells² and are also required for CD4⁺ T-cell priming¹⁹. Although Epor^ΔXcr1 mice had a slightly lower percentage of cDC1s than Epor^flox/flox controls (Fig. 3e), their cDC1s expressed significantly higher levels of CD40, CD80, MHCI, DEC-205, and the antiapoptotic CD40-dependent protein Bcl-xL⁶⁵, while showing reduced expression of PD-L1 (Extended Data Fig. 6a). In contrast, cDC2s displayed no differences in the expression of these markers between Epor^ΔXcr1 and Epor^flox/flox mice (Extended Data Fig. 6b,c). Flow cytometry confirmed the scRNA-seq finding of reduced CCR7⁺ late-mature cDC1s in Epor^ΔXcr1 mice (Fig. 3g and Extended Data Fig. 6d). Notably, while CCR7⁺ cDC1s normally express higher CD40, CD80, and PD-L1 than CCR7⁻ cDC1s, both subsets in Epor^ΔXcr1 mice displayed increased CD40 and CD80 but reduced PD-L1 compared with Epor^flox/flox controls (Extended Data Fig. 6e), consistent with the scRNA-seq results (Fig. 3h). While the frequencies of splenic conventional CD4⁺ and CD8⁺ T cells were unchanged (Extended Data Fig. 7a,d), Epor^ΔXcr1 mice had reduced FOXP3⁺CD25⁺ T_regs (Extended Data Fig. 7b) and increased CD44^highCD62L^low effector CD4⁺ and CD8⁺ T cells (Extended Data Fig. 7c, e).

Next, to examine the effect of cDC1-specific EpoR deficiency on cross-priming and priming of cell-associated Ags in vivo, we transferred CTV-labeled naïve OT-I or OT-II T cells into Epor^ΔXcr1 or Epor^flox/flox mice immunized with apoptotic Act-mOVA thymocytes. Notably, Epor^ΔXcr1 mice showed enhanced priming of both Ag-specific CD8⁺ (Extended Data Fig. 6f) and CD4⁺ T cells (Extended Data Fig. 6g). Aligning with a previous report¹⁹, OT-II priming and proliferation required cDC1s (Extended Data Fig. 7f), and exogenous EPO enhanced FOXP3 expression in OT-II cells in a manner dependent on cDC1-EpoR expression (Extended Data Fig. 7g). Together, these findings show that even at low homeostatic EPO levels, cDC1-EpoR limits both CD8⁺ T cell cross-priming and CD4⁺ T cell priming in response to cell-associated Ags.

T_reg-induction by pLN cDC1s is promoted by EPO

cDC1s are widely distributed in peripheral tissues and LNs where they comprise both LN-resident and migratory subsets^4,66. In peripheral tissues, cDC1s act as sentinels of the immune system, continuously migrating to the draining LNs (dLNs) to initiate T cell adaptive immunity via afferent lymphatics after Ag uptake and CCR7 upregulation^66,67. To assess steady-state EpoR expression in LN cDC1s, we examined peripheral (pLNs) and mesenteric LNs (mLNs) from Zbtb46^GFP/+ Epor^tdTomato/+ dual-reporter mice by defining cDCs as Zbtb46-GFP⁺CD11c⁺ (Fig. 4a,b). About 7% of pLN cDCs expressed EpoR, nearly all of which were XCR1⁺CD11c^intMHCII^high, i.e., migratory, while such cells were almost absent in mLNs (Fig. 4a,b). EpoR expression was much greater on cDC1s than cDC2s, indicating preferential expression on migratory cDC1s in pLNs (Fig. 4a–c).

CCR7 is required for the migration of cDC1s to the dLNs, where they induce Ag-specific CD4⁺ FOXP3⁺ T_regs in the steady state^4,68. EpoR⁺ migratory cDC1s, while prominent in the pLNs of Epor^tdTomato/+ mice (Fig. 4d), were rare in the pLNs of Ccr7^−/−Epor^tdTomato/+ mice and Batf3^−/− Epor^tdTomato/+ mice (Extended Data Fig. 8a–b), indicating that Ccr7 and Batf3 are required for the presence of EpoR⁺ migratory cDC1s. Prominent EpoR expression in migratory cDC1s was consistently observed across all pLNs examined, independent of their drainage site (Extended Data Fig.8c). Thus, although migratory cDC1s may retain tissue-specific imprints^69,70, Epor-tdT expression in pLN migratory cDC1s remains conserved under homeostatic conditions and is not impacted by pLN environment or location. Indeed, Epor-tdT expression was detected in migratory cDC1s from diverse tissues examined, including brain, skin, and lung (Extended Data Fig. 8d), indicating that cDC1s acquire EpoR expression prior to migration to draining LNs. Thus, EpoR⁺ cDC1s observed in pLNs are attributed to the migration of peripheral EpoR⁺ cDC1s to the dLNs.

Migratory Epor-tdT⁺ cDC1s in pLNs expressed higher levels of inhibitory molecules PD-L1, AXL⁷¹, TIM-3⁷², and CD131 (Extended Data Fig. 8e), suggestive of their tolerogenic potential. We then compared their ability to induce Ag-specific T_regs using apoptotic Act-mOVA thymocytes or DEC205-OVA⁸ that specifically targets cDC1s. Although Epor-tdT⁺ migratory cDC1s were superior to Epor-tdT⁻ cDC1s at inducing Ag-specific T_regs against both sources of Ag (Extended Data Fig. 9a,b), they were more efficient at inducing Ag-specific T_regs to cell-associated Ags (Extended Data Fig. 9b). T_reg induction by Epor-tdT⁻ cDC1s was enhanced in the presence of exogenous EPO, which is consistent with efferocytosis-induced EpoR upregulation (Extended Data Fig. 9b,c). Exogenous EPO administration also increased the Ag-specific T_reg induction capacity of pLN migratory cDC1s, while this effect disappeared when the migratory pLN cDC1s were replaced by Epor^ΔXcr1 pLN migratory cDC1s (Extended Data Fig. 9d).

To determine whether cDC1 EpoR is required for migratory cDC1-mediated FOXP3⁺ T_reg induction, we injected apoptotic cells into the mammary fat pad (MFP) and tracked local EpoR⁺ cDC1s. Over 70% of XCR1⁺Zbtb46⁺CD11c⁺ cDC1s in the MFP expressed EpoR and CD103 (Fig. 4e). In response to injection of PKH67-labeled CD45.1⁺ apoptotic cells, migratory cDCs in the draining inguinal LN showed a stronger PKH67 signal than resident cDCs and all PKH67⁺ migratory cDCs were EpoR⁺ (Fig. 4f). cDC1s engulfed more apoptotic cells than cDC2s, as evidenced by higher PKH67⁺ frequencies and signal intensity (Extended Data Fig. 9e). Furthermore, the induction of OT-II FOXP3⁺ T_regs in Epor^flox/flox control mice injected with Act-mOVA apoptotic cells was further enhanced by exogenous EPO, whereas both effects were abrogated in Epor^ΔXcr1 mice (Fig. 4g).

To validate the role of cDC1 EpoR in inducing FOXP3⁺ T_regs to endogenous cell-associated Ags, CD45.1⁺CD45.2⁺ Act-mOVA mice were reconstituted with BM from either Epor^flox/flox or Epor^ΔXcr1 donors (Extended Data Fig. 9f). In this model, membrane-bound OVA is expressed ubiquitously⁵³, and MHC class II-mediated Ag presentation depends entirely on donor hematopoietic-derived antigen-presenting cells (APCs). Naïve OT-II cells were adoptively transferred (day 0) into the chimera mice, and EPO was administered on days −2 to 2. Prominent expression of FOXP3 was observed in OT-II cells on day 9 in inguinal LNs from Epor^flox/flox BM-reconstituted Act-mOVA mice, whereas this induction was markedly impaired in mice reconstituted with Epor^ΔXcr1 BM (Extended Data Fig.9f). These results confirm the importance of EPO-activated cDC1 EpoR in mediating peripheral T_reg induction to endogenously derived cell-associated Ags upon systemic EPO administration.

Loss of EpoR results in immunogenic cDC1s

Consistent with the reduced frequency and XCR1 expression of splenic CCR7⁺ cDC1s in Epor^ΔXcr1 mice (Extended Data Fig. 6d and Figs. 3e,g), these mice also had fewer migratory cDC1s (Fig. 4h) with lower XCR1 expression (Fig. 4i) in pLNs. Thus, EpoR similarly regulates the homeostatic maturation of migratory cDC1s. scRNA-seq of FACS-sorted pLN migratory cDC1s from Epor^ΔXcr1 and Epor^flox/flox mice revealed four shared clusters (Fig. 4j,k). Cluster I, which was enriched for antigen-presenting genes (H2-Ab1, Rab43⁷³, Cd74⁷⁴) and Il12b was overrepresented in Epor^ΔXcr1 mice (Fig. 4j,k). However, clusters II and III characterized by high Itgb8 expression were significantly underrepresented in Epor^ΔXcr1 samples (Fig. 4j,k), a change potentially compounded by the overall decreased frequency of migratory cDC1s in Epor^ΔXcr1 mice (Fig. 4h). Consistent with these findings, the proportion of Cluster II enriched in immunoregulatory genes such as Mt1, Mt2, Clu, and Mfge8⁷⁵ was reduced. Conversely, EpoR-deficient pLN cDC1s displayed greater enrichment of the “CD4⁺ T-helper licensing” signature⁷⁶ (259 genes) (Fig. 4k; Supplementary Table 1), indicating that EpoR loss itself can mimic CD4⁺ Th-induced transcriptional programming⁷⁶. As revealed by flow cytometry, CD40 and CD86 were upregulated in Epor^ΔXcr1 cDC1s, while PD-L1 remained unchanged (Fig. 4l), which differs from splenic CCR7⁺ cDC1s (Fig. 3h and Extended Data Fig. 6e). Further bulk RNA-seq confirmed that EpoR regulates shared gene programs in pLN migratory and splenic CCR7⁺ cDC1s, upregulating 319 and downregulating 358 genes (Supplementary Table 2). EpoR supported the expression of key immune-regulatory genes such as Apoe⁷⁷ and Tnfaip3 (A20⁷⁸), while the loss of EpoR led to increased expression of genes involved in MHCII-mediated Ag presentation (H2-Ab1, Cd74-invariant chain⁷⁴, Ciita⁷⁹), cross-presentation (Wdfy4⁸⁰, Rab43⁷³), cytotoxic T cell responses (Il15ra⁸¹), costimulation (Cd86), and immunogenic maturation (Il1b), as well as toll-like receptor and type I interferon signaling (Tlr9, Myd88, Irf7, Ifi205) and Tnfrsf1a (TNFR1) (Fig. 4m). These results indicate that loss of EpoR enables the immunogenic maturational programming of CCR7⁺ cDC1s at both anatomical sites. Accordingly, similar to the T cell immune profile in the spleen, FOXP3⁺CD25⁺ T_regs were reduced, while CD44^highCD62L^low effector CD4⁺ and CD8⁺ T cells were increased in pLNs (Extended Data Fig. 10a–e).

Loss of EpoR on cDC1s promotes anti-tumor immunity

Interactions between cDC1s and T cells are critical throughout the cancer-immunity cycle²², not only in tdLNs^29,30 for priming naïve T cells but also in the TME²⁴, where cDC1s play a unique role in determining tumor Ag-specific CD8⁺ T cell fate by recruiting T cells, secreting cytokines and presenting tumor Ags to enhance cytotoxic T cell effector function⁸². cDC1s serve as an autonomous platform for both CD4⁺ and CD8⁺ T cell priming, directly orchestrating their crosstalk, i.e., cDC1 “licensing” in the TME for optimal anti-tumor immunity^19,65,76. Given that EpoR signaling in cDC1s promotes FOXP3⁺ T_reg induction and suppresses CD8⁺ T cell cross-priming, we investigated the impact of cDC1–EpoR on anti–tumor immunity.

To examine EpoR expression on tumor–infiltrating cDC1s, we used Zbtb46^GFP/+Epor^tdTomato/+ mice, defining tumor cDCs as CD45⁺Zbtb46–GFP⁺CD11c⁺ (Extended Data Fig. 11a–d). Epor–tdT⁺ XCR1⁺CD103⁺ cDC1s were detected in multiple tumor models (Extended Data Fig. 11a–d). EpoR was preferentially expressed by CCR7⁺Zbtb46-GFP⁺XCR1⁺CD103⁺ cDC1s rather than CCR7^- cDCs (Fig. 5a and Extended Data Fig. 11e). Of note, CCR7⁺Epor-tdT⁺ cDC1s displayed a maturation-associated regulatory phenotype based on their significantly higher levels of CD40, CD80, CD86, MHCI, and PD-L1 than other cDC populations (Extended Data Fig. 11f), indicating that they might have undergone tolerogenic maturation^5,71. In tdLNs, Epor-tdT was similarly restricted to migratory cDC1s (Extended Data Fig. 11g). Serum EPO positively correlated with tumor growth (Extended Data Fig. 11h), likely reflecting its impact on EpoR⁺ cDC1s to facilitate uptake and processing of apoptotic tumor cells. Utilizing ZsGreen-labeled B16F10-OVA⁷² tumor (Extended Data Fig. 11i), we found that in both tdLN migratory cDC1s (Extended Data Fig. 11j) and tumor-infiltrating cDC1s (Extended Data Fig. 11k), ZsGreen^high cDC1s were also Epor-tdT⁺ (Extended Data Fig. 11j,k), indicating that EpoR⁺ cDC1s engulf tumor debris which could facilitate presentation of tumor-Ags to promote tumor-specific tolerance. Supporting this hypothesis, we observed reduced growth of MC38-OVA^dim72 and B16F10-OVA⁷² in Epor^ΔXcr1 mice compared to Epor^flox/flox controls (Fig. 5b and Extended Data Fig. 12a–b).

Fig. 5 |. EpoR expression on cDC1s hinders Ag-specific anti-tumor T cell immunity and the loss of EpoR in cDC1s leads to tumor reduction.

a, B16F10-OVA tumor cells were implanted s.c. into the flanks of Zbtb46^GFP/+Epor^tdT/+ mice, and 10 days later, Epor-tdT expression on tumor-infiltrating cDC subsets (CCR7⁺ vs. CCR7⁻ populations) was determined. b, MC38-OVA^dim tumor size and weight on day 14 following implantation. Epor^flox/flox (n=8) vs. Epor^ΔXcr1 (n=7). c, CTV-labeled naïve CD45.1⁺ OT-I cells were i.v. transferred 4 days after B16F10-OVA implantation; tdLNs were analyzed 6 days later for OT-I proliferation. Representative flow plots are shown for each marker expressed on OT-I cells vs. CTV dilution. Proliferating CD44⁺CTV^lowOT-I cells were quantified. Epor^flox/flox (n=5) vs. Epor^ΔXcr1 (n=7). d-j, B16F10-OVA tumors implanted in Epor^flox/flox (n=6) or Epor^ΔXcr1 (n=6) mice; TILs analyzed on day 12: d, Frequencies of CD45⁺ TILs, CD8⁺ and CD4⁺ T cells. e, OVA_257–264-dextramer⁺ CD8⁺ T cells. f,g, Representative flow plots and quantification of CD8⁺ T cells expressing TIM-3 and PD-1 (f) and TCF1⁺TIM-3⁻CD8⁺ T cells (g). h, Representative histograms and quantification of perforin, granzyme-B, IFNγ and TNFα expression in CD8⁺ T cells. i, Frequencies and absolute cell numbers of FOXP3⁺ T_regs in CD4⁺ T cells with representative flow contour. j, Frequencies of T-bet⁺CXCR3⁺ T_regs in CD4⁺FOXP3⁺ T_regs with representative flow contour. k, B16F10-OVA tumor growth in Epor^flox/flox vs. Epor^ΔXcr1 treated with anti-PD-1 (n=6 and n=6) or an IgG2a isotype control (n=6 and n=6). Data are shown from one experiment, representative of at least three independent experiments with similar results (a-k). Statistical analysis was performed using unpaired two-tailed Student’s t-test (b,c,d,e,g,h, i,j), or two-way ANOVA followed by Tukey’s multiple-comparison test (f), or Šídák’s multiple-comparison test (k). Data are mean ± s.e.m. (b-k). The diagram in c was created with BioRender by Xiangyue Zhang (2025).

Tumor-specific CD8⁺ T cell activation takes place in two phases: initial activation in tdLNs to generate TCF1⁺PD-1⁺SLAMF6⁺ precursor exhausted T cells (T_pex) and subsequent acquisition of effector programs by CD8⁺ T cells within the tumors^83,84. cDC1s maintain a reservoir of tumor-Ag specific T_pex in tdLNs²³, and intratumoral cDC1-CD8⁺ T cell clusters⁸⁵, which constitute niches for TCF1⁺ T_pex stimulation and play a critical role in promoting tumor Ag-specific CD8⁺ T cell expansion and effector function²⁷. Moreover, the therapeutic response to anti-PD-1 is proportional to the abundance of T_pex^86,87 and APC niches in tumors⁸⁵, and the functionality of tdLNs⁸⁸ is critical. Therefore, we reasoned that, in addition to inducing FOXP3⁺ T_regs, EpoR signaling in cDC1s may suppress anti-tumor immunity by limiting T_pex generation in tdLNs and effector CD8⁺ T cell function at tumor sites.

In tdLNs, tumor Ag–specific CD8⁺ T–cell priming was enhanced in Epor^ΔXcr1 mice, as indicated by greater proliferation of transferred naïve OT–I cells with a T_pex phenotype, i.e., high SLAMF6 and low TIM3 expression (Fig. 5c). Loss of EpoR increased CD40 that is crucial for “cDC1 licensing”^2,65,89 on tumor Ag–carrying migratory cDC1s in tdLNs and CD80/CD86 on tumor cDC1s (Extended Data Fig. 12c–e). Accordingly, Epor^ΔXcr1 mice had more CD45⁺ TILs, a higher percentage of CD8⁺ T cells, and expanded SIINFEKL–H–2Kb⁺CD8⁺ T cells (Fig. 5d,e; Extended Data Fig. 12f,g). Exhausted PD–1⁺TIM–3⁺CD8⁺ T cells decreased (Fig. 5f; Extended Data Fig. 12h), while TCF1⁺TIM–3⁻ T_pex (Fig. 5g; Extended Data Fig. 12i) and the expression of effector molecules (perforin, GzmB, IFNγ, TNFα) increased (Fig. 5h; Extended Data Fig. 12j). Conventional CD4⁺ tumor-infiltrating T cells increased, whereas FOXP3⁺ T_regs and T–bet⁺ CXCR3⁺ T_regs^90–92 decreased (Fig. 5d,i,j; Extended Data Fig. 12f,k,l). Loss of EpoR also enhanced anti–PD–1 efficacy in B16F10–OVA tumors (Fig. 5k). Thus, removal of EpoR signaling in cDC1s promotes anti-tumor T cell immunity, restrains tumor growth and enhances the efficacy of immune checkpoint blockade.

Discussion

Our findings reveal that EPO-EpoR signaling in cDC1s serves as a conserved mechanism that promotes cDC1 tolerogenic maturation^5,7 and regulates cell-associated Ag-specific T cell tolerance, thus establishing a new paradigm for the tolerogenic function of cDC1s. EpoR signaling in MΦs promotes the clearance of dying cells and fosters immune tolerance⁵⁰, and EPO has been reported to suppress splenic DC function by enhancing T_reg expansion⁹³. However, our study highlights cDC1-EpoR-dependent tolerogenic maturation and function in shaping both cell-associated Ag-specific CD4⁺ and CD8⁺ T cell adaptive immunity. Removal of EpoR on cDC1s enhances tumor Ag-specific CD8⁺ T cell priming and profoundly affects the anti-tumor CD8⁺ T cell immune responses in both tdLNs and tumor sites by promoting T_pex generation in tdLNs and their maintenance in TME, and effector CD8⁺ T cell functionality. Especially noteworthy in mice lacking EpoR on cDC1s is the reduction of T-bet⁺CXCR3⁺ T_regs that were shown recently to inhibit costimulatory molecule expression on cDC1s^90–92, thereby restraining cDC1-mediated anti-tumor CD8⁺ T cell immunity. Such bidirectional crosstalk between EpoR⁺ cDC1s and T-bet⁺CXCR3⁺ FOXP3⁺ T_regs likely represents a critical mechanism by which cDC1 EpoR signaling impedes anti-tumor CD8⁺ T cell immunity. Therefore, in patients with cancer, blockade or removal of EpoR from cDC1s, alone or in combination with anti-PD-1, would be expected to diminish tumor growth and spread.

Upon efferocytosis of tumor-associated Ags, cDCs upregulate CCR7^71,94 and become mregDCs⁷¹, a cDC maturational state that can be either tolerogenic or immunogenic^5,71,95,96. cDC1s undergo tolerogenic maturation following efferocytosis¹¹, a process that is dependent on the expression of EpoR and markedly enhanced by TLI/ATS. EpoR signaling promotes tolerogenic maturation of cDC1s towards a CCR7⁺ late mature stage^7,11 with elevated expression of Itgb8, a crucial downstream effector of EpoR in cDC1-induced tolerance. Our findings highlight the conserved role of EpoR as a molecular switch in facilitating the tolerogenic maturation while restraining the immunogenic maturation of both pLN migratory and splenic CCR7⁺ cDC1s. EpoR-deficient pLN migratory cDC1s were enriched for the “CD4⁺ T-helper licensing” gene signature⁷⁶, indicating that EpoR regulates cDC1 functional programming not only during maturation but also in the three-cell CD4⁺ T-cDC1-CD8⁺ T interaction⁹⁷, thereby regulating cDC1-mediated CD4⁺ T-helper licensing essential for effective CD8⁺ T-cell responses⁹⁷. Both the frequency of EpoR-expressing cDC1s and the intensity of EpoR expression on cDC1s vary with their efferocytotic activity as well as with EPO exposure. Accordingly, the activity of these cells can be manipulated, providing a compelling rationale for developing immunotherapies that target EpoR on cDC1s, including agonists to induce tolerance in transplantation or autoimmune disease and antagonists to break tolerance and promote immunity to infection and tumors.

Methods

Mice

The following mice were obtained from The Jackson Laboratory (Bar Harbor, ME): Adult 8- to 10-week-old male wild-type BALB/cJ (H-2K^d) (Jackson, 000651) and C57BL/6J (H-2K^b) (Jackson, 000664), B6.129S(C)-Batf3tm1Kmm/J (Batf3^−/−) (Jackson, 013755), B6.129P2(C)-Ccr7tm1Rfor/J (Ccr7-) (Jackson, 006621), Zbtb46 tm1.1Kmm/J (Zbtb46^gfp) (Jackson, 027618), B6N(129S4)-Mafbtm1.1 (cre) Kmm/J (MafB-mCherry-Cre) (Jackson, 029664), C57BL/6-Tg (CAG-OVA) 916Jen/J (Act-mOVA) (Jackson, 005145), C57BL/6-Tg(TcraTcrb)1100Mjb/J (OT-I) (Jackson, 003831), C57BL/6-Tg (TcraTcrb) 425Cbn/J (OT-II) (Jackson, 004194), B6.129(Cg)-Foxp3tm3(HBEGF/GFP)Ayr/J (Foxp3^DTR) (Jackson, 016958), CByJ.SJL(B6)-Ptprca/J (CD45.1) (Jackson, 006584) and H2-Ab1^fl (B6.129X1-H2-Ab1^tm1Koni/J) (Jackson, 013181). FOXP3^DTR/DTR98 were cross bred with CD45.1 to generate CD45.1/CD45.1 FOXP3^DTR/DTR. OT-I or OT-II mice were cross bred with CD45.1 to generate CD45.1/CD45.1 OT-I or OT-II mice. Epor^flox/flox mice⁵⁰ (provided by Hong Wu, University of California, Los Angeles and Peking University), Epor-TdTomato-Cre mice were generated as previously described⁴⁵, Xcr1^Cre-mTFP1 mice⁶ (provided by Bernard Malissen, Centre d’ Immunologie de Marseille-Luminy, Marseille, France), which were generated with JM8.F6 ES cells and were originally on a C57BL6/N background. They were then backcrossed for more than eight generations onto C57BL6/J mice, resulting in a pure C57BL6/J background before breeding with flox/flox mice. Epor^flox/flox mice were generated on an Sv129/C57BL/6 background and were backcrossed onto the C57BL6/J strain for more than eight generations before crossed with Xcr1^Cre-mTFP1 to generate cDC1-specific Epor genetically deleted (Epor^ΔXcr1) mice. Gender-matched littermates of Epor^ΔXcr1 and Epor^flox/flox were utilized for each experiment. Epor^ΔXcr1 did not develop anemia, maintained normal levels of RBCs (7–10 million per microliter), hematocrit (40–50%), hemoglobin (12–15 g/dL) and reticulocytes (1–6%) in peripheral blood and displayed no differences in these parameters in comparison with Epor^flox/flox mice. Itgb8^flox/flox12, Itgb8^ΔXcr199, Aldh1a2^flox/flox and Aldh1a2^ΔCD11c100, 2W1S_52–68-expressing BALB/c (H-2K^d)⁵⁶ have been previously described. Epor^{tdTomato/tdTomato} mice were bred with Zbtb46^GFP/GFP to generate dual-color reporter Zbtb46^GFP/+Epor^tdTomato/+. Epor^{tdTomato/tdTomato} mice were bred with Ccr7^−/− or Batf3^−/− mice to generate Ccr7^−/−Epor^tdTomato/+ or Batf3^−/−Epor^tdTomato/+ mice. BM cells from BALB/cJ (H-2K^d) or 2W1S_52–68-expressing BALB/c (H-2K^d) mice were used for determining BM chimerism following combined allogeneic heart and BM transplantation. Newborn BALB/cJ (H-2K^d) mice as allogeneic heart donors were obtained from Charles River Laboratories. Unless otherwise specified, experiments were performed with mice between 6 and 10 weeks of age. No differences were observed between male and female mice in any assays performed, and so mice of both genders were used interchangeably throughout the study. Within individual experiments, mice used were age- and sex-matched littermates whenever possible. Mice were housed in animal facilities accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC). All experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at Stanford University (protocols APLAC-28636 and APLAC-17466) and conducted in accordance with Stanford University’s animal care guidelines.

BM transplantation, rabbit ATS, and TLI

BM harvest and transplantation procedures were performed as previously described³⁸. C57BL/6J background recipients were injected intraperitoneally (i.p.) with 0.05 mL of rabbit anti-thymocyte serum (ATS) (AIA3940T/20, Accurate Chemical and Scientific Inc) in 0.5 mL of saline on days 0, 2, 6, 8, and 10. Using a 250-Kv x-ray machine, TLI was delivered to the LNs above and below the diaphragm, thymus, and spleen with lead shielding of the skull, limbs, pelvis, and tail. 240 cGy was administered 5 times per week for two weeks. The last dose of TLI was administered to recipient mice 24 hours before the infusion of allo-BM cells from BALB/cJ or 2W1S_52–68-expressing BALB/c (H-2K^d+) mice. On the next day following the last dose of TLI/ATS, 30 × 10⁶ BALB/c donor BM cells were injected i.v. To deplete FOXP3⁺ T_regs in adoptively transferred CD45.1⁺FOXP3^DTR/DTR CD4⁺ T cells in Epor^flox/flox and Epor^ΔXcr1 mice prior to TLI/ATS treatment, recipient mice were adoptively transferred with 30 million CD45.1/CD45.1 FOXP3^DTR/DTRCD4⁺ T cells that were purified by MACS with CD4⁺ T Cell Isolation Kit (130–104-454, Miltenyi Biotec). Following the transfer, the mice were injected intraperitoneally with purified diphtheria toxin (DT) (D0564, Sigma-Aldrich) at a dosage of 0.5 μg per day for two consecutive days or with phosphate-buffered saline (PBS) as control.

BM chimerism analysis, heart transplantation and monitoring for graft survival

Analysis of chimerism in the blood was performed by flow cytometry using multi-color staining of total white blood cells or cell subsets with anti-H-2K^d mAb as described³⁸. Anti-MHCI (2K^d), anti-Ly6G (granulocytes), anti-TCRβ (T cells), anti-CD64 (macrophages), anti-B220 (B cells) were used to identify immune cell types. Neonatal BALB/c heart grafts were transplanted into a pouch in the ear pinna of C57BL/6, Epor^flox/flox, Batf3^−/− , Epor^ΔXcr1 hosts at least 21 days after BM infusion as described previously³⁸. Grafted heart survival was assessed by daily palpation, and rejection determined by cessation of heartbeat. Heart grafts that failed within 72h were excluded from the experimental groups as “technical failures”.

BM chimeras

Bone marrow (BM) chimeras were generated by retro-orbitally injecting 4 × 10⁶ total donor BM cells into lethally irradiated 8-week-old recipient mice (two doses of 5.5 Gy administered 6 hours apart). Recipients were supplemented for 3 weeks with UNIPRIM^®℞ Trimethoprim and Sulfadiazine supplied by Stanford Veterinary Service Center (VSC). Mice were allowed 8 weeks for reconstitution before experimental use. Successful reconstitution (minimum 90%) was assessed by flow cytometry analysis of peripheral blood.

Flow cytometry

For surface staining, cells were preincubated with anti-Fc receptor antibody (BE0307, Bio X Cell) and stained with appropriate antibodies in PBS containing 5 mM EDTA and 2% FBS at 4°C for 25 min. Viability was assessed by staining with 4′,6-diamidino-2-phenylindole (D1306, Thermo Fisher Scientific) or Fixable LIVE/DEAD Blue (L23105, Thermo Fisher Scientific) or Aqua (L34957, Thermo Fisher Scientific) Cell Stain. For intracellular cytokine detection, cells were stimulated for 4 to 5 h with phorbol 12-myristate 13-acetate (PMA) and ionomycin in the presence of monensin, eBioscience^™ Cell Stimulation Cocktail (plus protein transport inhibitors) (00–4975-93, Thermo Fisher Scientific) before staining according to the manufacturer’s instructions. For intracellular cytokine staining, cells were stained with antibodies against surface markers and then fixed with 2% (w/v) paraformaldehyde for 12 min at 25°C and permeabilized using eBioscience^™ Permeabilization Buffer (00–8333-56, Thermo Fisher Scientific). The fixed and permeabilized cells were subsequently stained with anti-IFNγ-BUV737 (XMG1.2, 612769, BD Biosciences, 1:100) or anti-TNFα-BV605 (MP6-XT22, 506329, BioLegend, 1:100) antibody for 60 min at 4°C. For intranuclear staining, cells were stained with antibodies against specified surface markers, and fixation-permeabilization was performed using the eBioscience^™ Foxp3 / Transcription Factor Staining Buffer Set (00–5523-00, Thermo Fisher Scientific) according to the manufacturer’s instructions. Flow cytometry was performed on a LSRFortessa X-20 or FACSymphony^™ A5 Cell Analyzer (BD Biosciences) with BD FACSDiva (v8), and data were analyzed with FlowJo (v10.10.0, BD Biosciences). Doublets and dead cells were excluded from analyses. Biotin-conjugated antibodies were detected using streptavidin-conjugated Brilliant Violet^™ 421 (405225, BioLegend, 1:400). For detection of phosphorylated proteins, cells were stimulated and immediately fixed with Phosflow Lyse/Fix buffer (558049, BD Biosciences), followed by permeabilization with Phosflow Perm buffer III (558050, BD Biosciences), and staining with antibodies to phosphor-signaling molecules. Tumor antigen-specific T cells were determined by H-2K^b/OVA_257–264 PE dextramer (JD02163-PE, 1:100) staining following the manufacturer’s protocol (Immudex). Cell counting was performed by using 123count eBeads^™ Counting Beads (01–1234-42, Invitrogen). The following anti-mouse antibodies (Target, fluorophore, clone, catalog number, and manufacturer; all antibodies were used at a 1:200 dilution unless otherwise noted) were used: CD11c-PE/Cy7 (N418, 117318, BioLegend), CD11c-BV711 (N418, 117349, BioLegend), MHCII (I-A/I-E)-APC (M5/114.15.2, 107614, BioLegend), MHCII (I-A/I-E)-APC/Cy7 (M5/114.15.2, 107628, BioLegend), MHCII (I-A/I-E)-BV510 (M5/114.15.2, 107636, BioLegend), CD8α-BV785 (53–6.7, 100750, BioLegend), CD8α-BV421 (53–6.7, 100738, BioLegend), CD8β-PE/Cy7 (YTS156.7.7, 126616, BioLegend), XCR1-PerCP/Cy5.5 (ZET, 148208, BioLegend), XCR1-BV785 (ZET, 148225, BioLegend), CD172a (SIRPα)-FITC (P84, 144006, BioLegend), CD172a (SIRPα)-BUV395 (P84, 740282, BD Biosciences), CD172a (SIRPα)-BV421 (P84, 740071, BD Biosciences), CD172a (SIRPα)-BUV661 (P84, 741593, BD Biosciences), CD103-BV421 (2E7, 121422, BioLegend), B220/CD45R-FITC (RA3–6B2, 103206 BioLegend), B220/CD45R-APC (RA3–6B2, 103212, BioLegend), CD19-APC (6D5, 115512, BioLegend), CD19-FITC (1D3/CD19, 152404, BioLegend), CD19-PE/Cy7 (6D5, 115520, BioLegend), SiglecH-BV605 (440c, 747673, BD Biosciences), SiglecH-APC (551, 129612, BioLegend), PDCA-1 (CD317, BST2)-BV711 (927, 127039, BioLegend), PDCA-1 (CD317, BST2)-APC (927, 127016, BioLegend), CD11b-FITC (M1/70, 101206, BioLegend), CD11b-BUV737 (M1/70, 741722, BD Biosciences), Ki67-BV605 (SolA15, 406–5698-82, eBioscience), IRF8-PE (V3GYWCH, 12–9852-82, eBioscience), TER119-APC (TER-119, 116212, BioLegend), TER119-FITC (TER-119, 116206, BioLegend), CD71-PerCP/Cy5.5 (RI7217, 113816, BioLegend), TCRβ-PE/Cy7 (H57–597, 109222, BioLegend), TCRβ-BV421 (H57–597, 109229, BioLegend), TCRβ-PE/Cy5 (H57–597, 109210, BioLegend), CD64-PE (X54–5/7.1, 139304, BioLegend), CD64-BV711 (X54–5/7.1, 139311, BioLegend), Ly6G-PE/Cy7 (1A8, 127618, BioLegend), Ly6C-BV421(AL-21, 562727, BD Biosciences), Ly6C-PerCP/Cy5.5 (HK1.4, 128012, BioLegend), F4/80-BUV395 (T45–2342, 565614, BD Biosciences), F4/80-BV711 (T45–2342, 565612, BD Biosciences), NK1.1-BV711 (PK136, 108745, BioLegend), NK1.1-FITC (PK136, 108706, BioLegend), NK1.1-APC (PK136, 108710, BioLegend), CD49b-APC (DX5, 108910, BioLegend), Siglec-F (CD170)-APC (S17007L, 155508, BioLegend), H-2K^d-PerCP-eFluor^™ 710 (SF1–1.1.1, 50–245-930, eBioscience), H-2K^b-PE (AF6–88.5, 561072, BD Biosciences), CD3ε-PE/Cy7 (500A2, 152314, BioLegend), CD3ε-APC (500A2, 152306, BioLegend), CD4-BUV737 (RM4–5, 612844, BD Biosciences), CD25-BUV395 (PC61, 564022, BD Biosciences), CD44-APC-R700 (IM7, 565480, BD Biosciences), CD62L-BV711 (MEL-14, 104445, BioLegend), CD326 (Ep-CAM)-PE/Cy7 (G8.8, 118216, BioLegend), CD40-APC (3/23, 558695, BD Biosciences), CD80-BV421 (16–10A1, 562611, BD Biosciences), CD86-BV785 (GL-1, 105043, BioLegend), CD274 (PD-L1)-BV421 (10F.9G2, 124315, BioLegend), CD205 (DEC-205) (V18–949, 566376, BD Biosciences), Axl-APC (MAXL8DS, 17–1084-82, eBioscience), CD131-BV421 (JORO50, 740050, BD Biosciences), CCR7-Biotin (4B12, 13–1971-82, eBioscience, 1:100), CD24-BV615 (30-F1, 752769, BD Biosciences), CD40-BV750 (3/23, 746970, BD Biosciences), CD80-BUV563 (16–10A1, 741272, BD Biosciences), CD86-BV510 (PO3, 745059, BD Biosciences), MHCII (I-A/I-E)-Alexa Fluor 700 (M5/114.15.2, 107622, BioLegend), CD274 (PD-L1)-BV605 (10F.9G2, 124321, BioLegend), CXCR3 (CD183)-PE (CXCR3–173, 126506, BioLegend), CD45.1-BV785 (A20, 110732, BioLegend), CD45.2-BV650 (104, 109836, BioLegend), CD45-BV785 (30-F11, 103149, BioLegend), CD45-BUV395 (30-F11, 564279, BD Biosciences), CD3-PE/Cy7 (17A2, 100220, BioLegend), TCRvα2-APC (B20.1, 127810, BioLegend), CD279 (PD-1)-BV711 (29F.1A12, 135231, BioLegend), Granzyme B-FITC (GB11, 515403, BioLegend), TIM-3 (CD366)-BUV395 (5D12/TIM-3, 747620, BD Biosciences), Ly108 (SLAMF6)-APC (eBio13G3–19D (13G3–19D), 17–1508-82, eBioscience), FOXP3-FITC (FJK-16s, 11–5773-82, eBioscience, 1:100), TCF1/TCF7 (C63D9, 2203S, Cell Signaling Technology), AF488 Donkey anti-rabbit IgG (Poly4064, 406416, BioLegend), T-bet-APC (eBio4B10 (4B10); 17–5825-82, eBioscience, 1:100), Bcl-xL-PE (54H6, 13835S, Cell Signaling Technology), Phospho-S6 Ribosomal Protein (Ser235/236)-PE (D57.2.2E, 5316S, Cell Signaling Technology, 1:50), Phospho-Akt (Ser473)-PE (D9E, 5315S, Cell Signaling Technology, 1:50), Phospho-4E-BP1 (Thr37/46)-PE (236B4, 7547S, Cell Signaling Technology1:50), Phosph-p44/42 MAPK (Erk1/2) (Thr202/Tyr204)-PE (197G2, 14095S, Cell Signaling Technology, 1:50), Phosph-Stat5 (pY694)-PE (47, 562077, BD Biosciences, 1:50).

Mouse EPO ELISA

Blood serum was collected at different time points, and serum EPO was measured by enzyme-linked immunosorbent assay (ELISA) according to the manufacturer’s instructions (Mouse EPO ELISA Kit, EM28RB, Invitrogen).

Isolation and purification of XCR1⁺CD8α⁺ cDC1s and Epor-tdT⁺ and Epor-tdT⁻ cDC1s

Spleens were minced and digested in 5 ml Iscove’s modified Dulbecco’s media + 10% FCS (cIMDM) with 250 μg/ml collagenase D (Worthington) and 30 U/ml DNase I (Sigma-Aldrich) for 30 min at 37°C with stirring. Cells were passed through a 100-μm strainer before red blood cells were lysed with RBC lysis buffer (420302, BioLegend). 5–10 × 10⁶ cells were used per antibody staining reaction. For further XCR1⁺CD8α⁺ sorting, single spleen cell suspensions were negatively selected with magnetic-activated cell sorting columns with mouse Pan Dendritic Cell Isolation Kit (130–100-875, Miltenyi Biotec). MACS selected DCs were further sorted by fluorescence- activated cell sorting (FACS) (BD, FACSAria II), to obtain B220⁻SiglecH⁻PDCA-1⁻CD11c^high MHCII^highXCR1⁺CD8α⁺ cDC1s purity>99%. Epor-tdT⁺ and Epor-tdT⁻ XCR1⁺CD8α⁺ cDC1s were prepared and sorted from Epor^{tdTomato/tdTomato} mice following TLI/ATS with similar methods.

RNA sequencing analysis

Fresh splenic live cDC1s were purified first from single spleen cell suspensions with negative selection by using mouse Pan Dendritic Cell Isolation Kit (130–100-875, Miltenyi Biotec). MACS -selected DCs were further sorted by FACS (BD, FACSAria II), to obtain live-dead blue⁻Lin⁻ SiglecH⁻PDCA-1⁻CD11c^highMHCII^highCD8α⁺CD11b⁻ cDC1s (purity>98%). FACS-purified cDC1s from untreated (UNT) or TLI/ATS-treated wild-type (WT) C57BL/6J mice (8–10 weeks of age) were used for total RNA isolation with RNeasy Plus Micro Kit (74034, QIAGEN) and submitted for RNA-sequence analysis. The RNA sequencing was performed by the Stanford Functional Genomics Facility. The RNA-seq read count matrix was generated through the following steps: 1) Trimmomatic¹⁰¹ (v0.36) was applied to trim the 76bp paired-end sequencing reads to get rid of low-quality bases and/or adaptor contaminations. 2) HISAT2¹⁰² (v2.1.0, http://daehwankimlab.github.io/hisat2/) was used to map the trimmed FASTQ reads to the Mus Musculus GRCm38 reference genome (the index files of genome_tran were downloaded from https://cloud.biohpc.swmed.edu/index.php/s/grcm38_tran/download). 3) SAMtools (v1.16.1) were used to sort and convert the aligned SAM files to aligned BAM files. 4) Gene-level expression abundance for each sample was quantified from aligned BAM files using featureCounts (v2.0.3). In Fig. 1e,f,k, differential expression analyses between the TLI/ATS-treated and UNT groups were performed using R package DESeq2¹⁰³ (v1.46.0). Genes with adjusted P values < 0.05 (Benjamini-Hochberg correction) and log2 fold changes >1 were considered differentially expressed in comparisons. In, Fig. 1k,l, Epor-tdT⁺ or Epor-tdT⁻Lin⁻CD11c^highMHC-II^high XCR1⁺ CD8α⁺ cDC1s were FACS purified from Epor^{tdTomato/tdTomato} mice following TLI/ATS conditioning. RNA was isolated by using RNeasy Plus Micro Kit (74034, QIAGEN) and subjected to RNA-sequence analysis with Novogene Co. using an Illumina sequencer. In Fig. 4m, pLN migratory cDC1s and CCR7⁺ splenic cDC1s were sorted by flow cytometry directly into lysis buffer and subjected to sequencing by MedGenome.com using an Illumina platform. Heatmaps were generated using R packages ComplexHeatmap¹⁰⁴ (v2.22.0).

GO and GSEA analyses

Gene Ontology (GO) enrichment analysis was performed on the top 500 genes with the highest fold change values and P-value<0.05 (hypergeometric test, corrected with Benjamini-Hochberg method) using enrichGO function provided by R package clusterProfiler¹⁰⁵ (version 4.14.6). The GO Biological Process (GO-BP) terms were used as the reference for functional enrichment analysis. The GO terms were downloaded from the Gene Ontology Consortium (https://geneontology.org/docs/download-ontology/) through clusterProfiler’s internal function and only terms from the “biological_process” parts were used. GSEA software (version 3.0) was run on the Molecular Signatures Database Hallmarks database¹⁰⁶ using the Pre-Ranked Gene List format, and “meandiv” normalization.

qPCR, RNA extraction and cDNA synthesis

Total RNA was extracted from omental tissue using the RNeasy Plus Mini Kit (74134, QIAGEN) and protocol. RNA concentration was determined by optical density and normalized across samples. Equal amounts of cDNA were synthesized using the High-Capacity cDNA Reverse Transcription Kit (4368814, Applied Biosystems) with an RNase Inhibitor (N8080119, Applied Biosystems) according to the manufacturer’s protocol. Each cDNA sample was diluted 1:200 in RNase free water prior to qPCR. qPCR was conducted with TaqMan Gene Expression Assay using probes for the genes Batf3 (Mm01318274_m1), Irf8 (Mm00492567_m1), Epor (Mm00833882_m1), Axl (Mm00437221_m1), Mertk (Mm00434920_m1), Cd5l (Mm00437567_m1), Itgb8 (Mm00623991_m1), Scube3 (Mm01299285_m1), Tgfb1 (Mm01178820_m1), Ccl22 (Mm00436439_m1), Aldh1a2 (Mm00501306_m1), Gapdh (Mm99999915 _g1) and Actb (Mm02619580_g1). Each TaqMan probe was diluted 1:10 in TaqMan Fast Advanced Master Mix (Thermo Fisher, cat. 4444557) to create a TaqMan probe working solution. All qPCR reactions were carried out in a MicroAmp optical 384-well reaction plate. qPCR was performed using the QuantStudio 5 (Applied Biosystems) under the following cycling conditions: 1 cycle at 50°C for 2 min and 95°C for 10 min, followed by 40 cycles at 95°C for 15s and 60°C for 1 min. The average CT value for each gene was calculated and normalized to Gapdh.

TUNEL staining

TUNEL (Terminal deoxynucleotidyl transferase dUTP nick end labeling) staining was performed using the In Situ Cell Death Detection Kit (C10617, Invitrogen) TMR Red according to the manufacturer’s instructions. Briefly, tissue sections were fixed with 4% paraformaldehyde for 20 minutes on ice prior to treatment with 0.1% Triton X-100 in 0.1% sodium citrate for permeabilization. Sections were washed in PBS before incubation for 60 minutes at 37°C with antibodies and TdT enzyme, followed by washing. Images were acquired by tile scanning using a Zeiss LSM 700 confocal laser scanning microscope (Carl Zeiss Microscopy) using the 20X objective and a resolution of 960 × 720 pixels per tile. Scale bar was added in ImageJ (v2.17.0).

Multiplex immunofluorescence imaging by CODEX

Preparation of tissues for CODEX (Co-detection by indexing) imaging was performed as previously described¹⁰⁷, with the following modifications for fresh-frozen mouse tissue. Briefly, spleens were snap-frozen in optimal cutting temperature (OCT) medium (Tissue-Tek, 25680–930, VWR/Sakura), and a 1×1 cm tissue array of spleens was created by trimming and gluing the OCT blocks at −20°C in the cryostat. The array was sectioned to a thickness of 7 μm onto 22×22mm glass coverslips (#1 ½, 12–550-343, Electron Microscopy Sciences) pre-coated with poly-L-lysine (P8920, Millipore Sigma). Sections were stored at −80°C until further use. For staining, sections were equilibrated to room temperature (RT) on drierite desiccant (07–578-3A, Thermo Fisher Scientific) for 2 minutes, followed by incubation in acetone at RT for 10 minutes. Then, sections were dried at RT for 2 minutes, followed by hydration in S1 buffer for 2 minutes after which sections were fixed in 1.6% PFA in S1 buffer at RT for 10 minutes, followed by washing in S1 buffer, and equilibration in S2 buffer. 100μl of antibody cocktail was added, and sections were incubated at RT for 3 hours in a humidity chamber. Then, tissues were washed in S2 buffer, fixed in 1.6% PFA in S4 buffer for 10 minutes, washed in PBS, fixed in ice-cold methanol for 5 minutes, washed in PBS, and fixed in BS3 (21580, Thermo Fisher Scientific) at RT for 20 minutes. Sections were stored in S4 buffer at 4°C until imaging. For CODEX imaging, stained coverslips were mounted onto custom-made acrylic plates (Bayview Plastic Solutions) using mounting gaskets (Qintay, cat. no. TMG-22) and stained with Hoechst 33342 (Thermo Fisher Scientific) at a dilution of 1:1000 in H2 buffer for 1 min, followed by 3 washes in H2 buffer. Automated image acquisition and fluidics exchange were performed using a CODEX PhenoCycler instrument and driver software (Akoya Biosciences) on a BZ-X710 inverted fluorescence microscope (Keyence) equipped with a CFI Plan Apo λ 20x/0.75 objective (Nikon). The following antibodies were used for CODEX: anti-B220 (RA3.3A1/6.1, BE0067, Bio X Cell, 1:100); anti-CD3 (17A2, 555273, BD Biosciences, 1:200); anti-CD169 (MOMA-1, MCA947G, Bio-Rad, 1:50); anti-TER119 (TER-119, 550565, BD Biosciences, 1:400); anti-CD71 (C2F2, 553264, BD Biosciences, 1:400).

Ex vivo analysis of EPO-EpoR downstream signaling in splenic cDC1s

Splenic cDCs were MACS-purified with a pan-DC isolation kit (130–100-875, Miltenyi Biotec) and cultured at 5 × 10⁶ cells/ml full RPMI culture medium supplemented with 10% heat-inactivated fetal bovine serum, 2 mM L-glutamine, 100 units per ml of penicillin, 100 μg/ml of streptomycin sulfate, 1 mM sodium pyruvate, 0.1 mM non-essential amino acids, 10 mM HEPES (all from Gibco), and 50 μM β mercaptoethanol (21985023, Gibco), and then rested overnight. Cells were isolated from untreated (UNT) or TLI/ATS-treated Epor^flox/flox and Epor^ΔXcr1 mice. cDCs from TLI/ATS-treated mice were stimulated ex vivo with recombinant human EPO (rhEPO, PROCRIT^®, epoetin alfa, 10 IU/200 μl) in RPMI full culture medium or PBS (control) overnight. Phosphorylation of downstream signaling molecules was assessed by flow cytometry, gating on Lin⁻SiglecH⁻PDCA-1⁻CD11c^high MHCII^highXCR1⁺SIRPα⁻ splenic cDC1s.

T_reg depletion studies

Foxp3^DTR mice were acquired from Jackson (016958) and bred in our facility at Stanford University. Eight-week-old female Foxp3^DTR mice were treated with TLI/ATS. Mice were injected intraperitoneally (i.p.) with 100μl DT (25ng/g body weight) (D0564, Millipore Sigma) or PBS control every other day (days 1, 3, 5, 7, 9, 11, 13) following allo-BM infusion (day 0), and BM chimerism was measured by blood sampling on day 14, day 28 and day 55. In another group, DT was injected every other day after day 14 (days 15, 17, 19, 21, 23, 25, 27), and BM chimerism was measured by blood sampling on days 14 and 28.

2W1S tetramer enrichment and flow cytometry

Phycoerythrin (PE) MHCII I-Ab 2W1S_55–68 tetramers (NIH Tetramer core facility), and their use with anti-fluorophore-conjugated magnetic beads, anti-PE MicroBeads (130–048-801, Miltenyi Biotec) for enrichment have been described⁵⁶. For analyzing FOXP3⁺ T_regs in 2W1S⁺CD4⁺ T cells, nucleated cells from spleens were collected, enriched using I-Ab 2W1S_55–68 tetramers, and stained for cell-surface MHCI (H-2K^b), TCRβ, CD4, CD44, and intracellular FOXP3, before being analyzed by flow cytometry.

10x Genomics scRNA-seq library preparation

Three different types of scRNA-seq experiments were performed. In the first experiment, spleens were obtained from 7–8-week-old Epor^flox/flox or Epor^ΔXcr1 mice after TLI/ATS treatment or from untreated (UNT) controls. In the second experiment, spleens were obtained from Epor-tdT mice after TLI/ATS treatment. In the third experiment, pLNs were obtained from UNT Epor^flox/flox or Epor^ΔXcr1 mice. For all experiments, single cell suspensions were prepared and subjected to MACS negative enrichment with Pan Dendritic Cell Isolation Kit mouse (130–100-875, Miltenyi Biotec). Samples were then stained with Live-dead aqua, Fc-blocker, and an antibody cocktail used to isolate cDC1s by FACS using a BD FACSAria II instrument. Cells were sorted into phosphate-buffered saline (PBS) supplemented with 0.5% bovine serum albumin and 2.5 mM EDTA. Cell purities of at least 95% were confirmed by post-sort analysis. FACS-sorted splenic cDC1s in Fig. 3a–d,f,g,h, Extended Data Fig. 5a,b,f,g and pLN migratory cDC1s in Fig. 4j,k were then barcoded with unique hashtag antibodies (155841 and 155845, BioLegend), while samples in Extended Data Fig. 5d,e,f,g,h were barcoded with MULTI-seq anchor lipid-modified oligonucleotide (LMO) pre-hybridized to a unique MULTI-seq barcode (2 μM stock, 200 nM labeling concentration). For the third experiment, Epor-tdT⁺ and Epor-tdT⁻ cDC1s were sorted separately from the spleens of TLI/ATS-treated Epor^tdT/tdT mice. Sorted cDC1s were ‘super-loaded’ into 10× Genomics 3′ scRNA-seq V3.1 chips (PN-1000269, 10x Genomics). cDNA, antibody hashing, and MULTI-seq library preparation was performed according to established protocol¹⁰⁸. Library quality control was performed using an Agilent 2100 Bioanalyzer instrument. Pooled cDNA libraries were sequenced using a NovaSeq6000 or NovaSeq X instrument (Illumina). A median sequencing depth of 40,000 and 5,000 reads per cell was targeted for the GEX and HTO/MULTI-seq libraries, respectively.

scRNA-seq data analysis

scRNA-seq library FASTQs were pre-processed using Cell Ranger (v7.0.0) (10x Genomics) and aligned to the mm-10–3.0.0 reference transcriptome. Cell Ranger aggregate was used to perform read-depth normalization. Filtered read depth normalized scRNA-seq count matrices were then read into R and parsed to exclude genes with fewer than 5 counts across all cell barcodes. Parsed scRNA-seq data were then pre-processed using Seurat (v5.0.1)¹⁰⁹ and Speckle (v0.99.7). Cell clusters with low total UMIs and/or high proportion of mitochondrial transcripts were excluded. Cell barcodes passing the first quality-control workflow were then used to pre-process hashtag or MULTI-seq barcode FASTQs and perform sample classification using the ‘deMULTIplex2’ R package (v1.0.1)¹¹⁰. Following MULTI-seq demultiplexing, unclassified cells and clusters enriched with MULTI-seq-defined doublets were removed prior to re-processing. These data were used for unsupervised clustering, differential gene expression testing, and manual annotation of splenic cDC1 subtypes based on the following literature⁷-supported marker genes: immature early (Pdia4, Ncub2, Dnajc3), immature late (Nr4a2, Hfe, Trib1), mature early (Cxcl9, Serpina3g, Slfn5), mature late (Ccr7, Gadd45b, Cd63), and proliferative cDC1s (Stmn1, Mki67, Hells) as well as pre-cDC1s (S100a6, S100a10, Anxa2). Notably, low-quality/doublet cell clusters missed during the initial quality-control workflows were removed during the subtype annotation workflow, after which all datasets were re-processed and used to perform differential gene expression and subtype proportion analyses between all assayed sample groups. The manual annotation of pLN migratory cDC1 clusters was based on unsupervised clustering results of the scRNA-seq data. Four clusters were obtained by using R package Seurat’s FindClusters function with the parameter resolution = 0.3. Each cluster’s identity was determined by analyzing its DEGs (differentially expressed genes) obtained through Seurat’s FindAllMarkers function provided with log2 foldchange >0 adjusted p-value <0.05. The heatmap showing these top DEGs between pLN migratory cDC1 clusters identified in Epor^flox/flox and Epor^ΔXcr1 mice were created using R package Seurat’s DoHeatmap function. The expression density visualization was performed using R package Nebulosa (v1.18.0)¹¹¹. The CD4 T cell signature scores of the cells were calculated using Seurat’s function AddModuleScore, and the gene list shown in Fig. 4k and Supplemental Table 1 (murine) was derived from Lei et al⁷⁶.

Adoptive OT-I and OT-II cell transfer and priming of T cells to cell-associated Ags in vivo

OVA-specific transgenic CD8⁺ (OT-I) or CD4⁺ T (OT-II) cells on CD45.1 background were obtained from LN and spleen cell suspensions of OT-I^{CD45.1/CD45.1} or OT-II^{CD45.1/CD45.1} mice. OT-I cells were isolated by using Naïve CD8α⁺ T Cell Isolation Kit, mouse (130–096-543, Miltenyi Biotec), and enriched CD8⁺ T cells were surface stained and purified by FACS (CD8⁺CD25⁻ CD44^lowCD62L^high). Naïve FOXP3^- OT-II cells (CD4⁺CD25⁻CD44^lowCD62L^high) were isolated by Naïve CD4⁺ T Cell Isolation Kit, mouse (130–104-453, Miltenyi Biotec). 10⁷ cells/ml OT-I or OT-II cells were prelabeled with 5μM CellTrace^™ Violet (C34557, Thermo Fisher Scientific). 1 × 10⁶ naïve OT-I or OT-II cells were adoptively transferred into CD45.2/CD45.2 homozygous Epor^flox/flox or Epor^ΔXcr1 mice by retro-orbital injection under isoflurane gas anesthesia. One day later, 500,000 or 10⁶ apoptotic Act-mOVA thymocytes were injected intravenously to challenge the naïve OT-I or OT-II cells. CTV dilution in adoptively transferred OT-I or OT-II cells was evaluated 4 days later by flow cytometry analysis of splenocytes, following surface staining for CD45.1, CD45.2, TCRvα2 (OVA-specific TCR), CD3, CD8 (OT-I) or CD4 (OT-II).

CellTrace^™ Violet labelling

Naïve OT-I or OT-II cells were resuspended in 1 ml PBS and then incubated with 5 μM Cell Trace^™ Violet (CTV) (C34557, Invitrogen) at 37°C for 20 min. RPMI-1640 medium (5 ml) was added to the cells and incubated for 5 min to remove the free dye in the solution. These cells were then centrifuged and incubated with pre-warmed RPMI-1640 for at least 10 min at room temperature for subsequent analysis.

Preparation and isolation of single cell suspensions from LNs

LNs were suspended in cold full RPMI culture medium. LNs were finely chopped and incubated in Liberase^™ TM (200μg/ml, 5401119001, Roche/Millipore Sigma) and DNaseI (30 μg/ml; D2821, Sigma-Aldrich) in full RPMI culture medium for 25 min at 37°C, 5% CO2. Single cell suspensions were extracted from connective tissue by taking up and resuspending the digests five times.

Digestion and cell isolation from brain, skin, lung and mammary tissue

Brain:

Mice were anesthetized and intracardially perfused with 20 mL Dulbecco’s phosphate-buffered saline (DPBS, pH 7.3–7.4). The brain was then excised. Mechanical dissociation of the brain was performed at 4^oC using a 10 mL Dounce homogenizer and a loose pellet. The homogenate was filtered into a 50 mL conical tube using a 70 μM filter. The filtered homogenate was centrifuged at 300g for 5 min at 4°C. The pellet was resuspended in 10 mL of 30% Percoll (P1644, MilliporeSigma) in complete Hanks’ Balanced Salt Solution (HBSS) (14025092, Gibco) and centrifuged. This Percoll step was repeated a second time. The resulting pellet was then resuspended in complete HBSS for flow cytometry staining.

Whole skin:

Ears were harvested and finely cut with scissors in at least 5ml/4cm² of skin with Liberase^™ TM (200ug/ml, 5401119001, Roche/Millipore Sigma) and deoxyribonuclease I (30 μg/ml; D2821, Sigma-Aldrich) in HBSS (+calcium and magnesium). The suspensions were digested at 37^oC for 1.5– 2 h (under agitation) and then filtered through a 100-μm nylon strainer.

Lung:

Lungs were harvested, cut into small fragments, and digested for 45 min at 37°C with collagenase A (0.6 mg/ml; 10103586001, Sigma-Aldrich) and deoxyribonuclease I (30 μg/ml; D2821, Sigma-Aldrich) in RPMI 1640 medium (Gibco). Digested lungs were mechanically disrupted to obtain single-cell suspensions. Red blood cells were lysed using RBC lysis buffer (420302, BioLegend). Cell suspensions were then filtered through a 100-μm nylon strainer.

Mammary tissue:

The mammary fat pad containing glands was dissected into small fragments and subjected to enzymatic digestion for 20 minutes at 37°C in a CO2-independent medium (Gibco). The remaining tissue pieces were meshed to obtain single-cell suspensions. Red blood cells were lysed using RBC lysis buffer (420302, BioLegend). Cell suspensions were then filtered through a 100-μm nylon strainer.

Efferocytosis assay in vivo and in vitro

For in vivo apoptotic cell engulfment experiments, 50 million thymocytes from CD45.1/CD45.1 C57BL/6 mice were resuspended in 10 ml of RPMI 1640 (21875059, Thermo Fisher Scientific) supplemented with 10% fetal bovine serum (FBS; Bodinco), containing 10 μM dexamethasone (D2915, Sigma-Aldrich), and incubated at 37°C in a humidified atmosphere with 5% CO2 for 4 hours. Apoptotic thymocytes were also generated with 15 Gray radiation. Next, to allow tracking of the apoptotic cells, the cells were labeled with PKH67 (PKH67GL-1KT, MilliporeSigma) for cell membrane labeling according to the manufacturer’s protocol. 2 million apoptotic cells were injected subcutaneously into the 3^rd mammary fat pad or footpad of CD45.2⁺ Epor-tdTomato reporter mice. 12 hrs after injection, the mice were euthanized, and uptake of PKH67-labelled cells in the inguinal or popliteal LN on the injection side and contralateral side was analyzed by flow cytometry. For in vitro efferocytosis induced EpoR expression assay, CD45.2⁺ pLN Epor-tdT⁺ and Epor-tdT⁻Lin⁻CD11c^intMHCII^high migratory XCR1⁺ cDC1s were sorted by FACS and cocultured overnight with CD45.1⁺ apoptotic thymocytes at a 5:1 ratio, and the phenotype of CD45.2⁺cDC1s was analyzed by flow cytometry for the indicated markers. cDC1s were gated as CD45.2⁺CD45.1 ⁻CD11c⁺MHCII⁺.

DEC205-OVA conjugation

2 mg of anti-CD205 (NLDC-145, BE0420, Bio X Cell) was incubated with 0.4 mg EDC (77149, Thermo Fisher Scientific) and 1.1 mg of Sulfo-NHS (24510, Thermo Fisher Scientific) in 1 mL of activation buffer (0.1 M MES, 0.5 M NaCl, pH 6.0) at RT for 15 min. 1.2 μL of 2-mercaptoethanol was added to quench the EDC. 2 mg of Ovalbumin (77120, Thermo Fisher Scientific) was then added for conjugation at room temperature for 2 hr. Hydroxylamine was added to 10 mM final concentration to quench the reaction. The conjugated anti-CD205 was desalted and purified using a Protein G column (45204, Thermo Fisher Scientific).

In vitro OT-II FOXP3⁺ T_reg induction assay

One day following the last dose of TLI/ATS, CD11c^highMHCII^high Epor-tdT⁺ and Epor-tdT⁻ XCR1⁺CD8α⁺ cDC1s were enriched by Pan Dendritic Cell Isolation Kit (130–100-875, Miltenyi Biotec) and further enriched by FACS, achieving >99% purity. cDC1s were cocultured with naïve OT-II T cells, which were isolated from OT-II^{CD45.1/CD45.1} mice using Naïve CD4⁺ T Cell Isolation Kit (130–104-453, Miltenyi Biotec) and FACS as CD45.1⁺CD3⁺TCRvα2⁺CD4⁺CD25⁻CD44^low CD62L⁺. cDC1s were cocultured with naïve OT-II cells in the presence of apoptotic Act-mOVA thymocytes at a ratio of 1:5:2 in 200μl full RPMI culture medium. Where indicated, 20IU/200μl rhEPO (PROCRIT^®, epoetin alfa) was added to the cultures daily for 5 consecutive days. Epor-tdT⁺ and Epor-tdT^- CD11c^intMHCII^high migratory cDC1s were isolated from pLNs with MACS and FACS as described above. pLN migratory cDC1s were cocultured with CTV-labeled naïve OT-II cells in the presence of 2 μg/200μl DEC205-OVA or apoptotic CD45.1⁺ thymocytes at a 1:5:2 ratio. Where indicated, 20IU/200μl rhEPO (PROCRIT^®, epoetin alfa) was added to the cultures daily for 5 consecutive days. FOXP3 expression on OT-II cells prelabeled with CellTrace Violet (CTV) was analyzed by flow cytometry, and OT-II cells were gated as live-dead aqua^-CD45.1⁺CD45.2⁻CD3⁺TCRvα2⁺CD4⁺. CD11c^intMHCII^high migratory cDC1s were isolated from pLNs of Epor^flox/flox or Epor^ΔXcr1 mice with MACS and FACS as described above and cocultured with CTV-labeled naïve OT-II cells in the presence of apoptotic CD45.1⁺ thymocytes at a 1:5:2 ratio (2 × 10⁴ DCs, 1×10⁵ naïve OT-II cells, and 4×10⁴ apoptotic Act-mOVA thymocytes) in RPMI full culture medium. Where indicated, 20IU/200μl rhEPO (PROCRIT^®, epoetin alfa) was added to the coculture daily for 5 consecutive days. FOXP3 expression versus CTV dilution in OT-II cells was analyzed 5 days later by flow cytometry. OT-II cells were gated as live-dead aqua⁻ CD45.1⁺CD45.2⁻CD3⁺TCRvα2⁺ CD4⁺.

Ex vivo Ag-specific FOXP3⁺ T_reg induction by CCR7⁺ cDC1s

12 hrs after i.v. injection of apoptotic Act-mOVA thymocytes (5×10⁶) into Epor^flox/flox or Epor^ΔXcr1 mice, splenic CCR7⁺XCR1⁺SIRPα^- cDC1s (1×10⁴) were sorted by FACS and cocultured with CTV-labeled naïve OT-II cells (5×10⁴) for 5 days. Anti-TGFβ (1D11, 1.25μg/ml. BP0057, Bio X Cell) blocking antibody or PBS as control was added into the coculture with CCR7⁺ cDC1s sorted from the spleens of Epor^flox/flox mice. FOXP3 expression on OT-II cells was analyzed by flow cytometry, and OT-II cells were gated as live-dead aqua⁻CD45.1⁺CD45.2⁻CD3⁺TCRvα2⁺CD4⁺.

In vivo OT-II FOXP3⁺ T_reg induction assay

Naïve CD45.1/CD45.1 background OT-II cells were isolated and sorted as described above and labeled with CTV. 10⁶ CTV-labeled naïve CD45.1⁺ OT-II cells were injected intravenously (i.v.) into Epor^flox/flox or Epor^ΔXcr1 mice. One day later, 10⁶ apoptotic Act-mOVA thymocytes were injected s.c. into the mammary fat pad to challenge the CD45.1⁺ OT-II cells residing in the dLN. Where indicated, 40IU rhEPO (PROCRIT^®, epoetin alfa) was injected i.p. daily for 5 consecutive days or with PBS (w/o) as control. FOXP3 expression versus CTV dilution in adoptively transferred OT-II cells was evaluated 4 days later by flow cytometry analysis of the immune cells in the dLN, and OT-II cells were gated as live-dead aqua^-CD45.1⁺CD45.2⁻CD3⁺TCRvα2⁺CD4⁺.

Tumor models

The MC38 colon carcinoma cell line was a gift from Cornelis J.M. Melief, Leiden University, The Netherlands. EO771 was purchased from ATCC (CRL-3461^™) and B16F10 was purchased from ATCC (CRL-6475^™). MC38-OVA^dim72 and B16F10-OVA⁷² melanoma were from Vijay Kumar Kuchroo, Harvard University. The B16F10-OVA-ZsGreen cell line was created in the lab through lentiviral transduction using LV-EF1a-ZsGreen-IRES-Puro (SL100336, Signagen Laboratories), which were then sorted by FACS to achieve over 98% purity based on ZsGreen expression. All tumor lines were routinely tested for mycoplasma by PCR, and all tests were negative. No additional authentication was performed. Cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) (C11965500BT, Gibco) supplemented with 10% FBS (FBS; Bodinco), 1% penicillin-streptomycin (10378016, Thermo Fisher), 2 mM l-glutamine (A2916801, Gibco), 1 mM sodium pyruvate (11360070, Gibco) and 0.1 mM non-essential amino acids (11140050, Gibco) at 37°C in 5% CO2. Tumor experiments were carried out by subcutaneously implanting tumor cells into gender- and age-matched (8–12 weeks) mice with the following cell numbers: MC38-OVA^dim (0.5×10⁶), B16F10 or B16F10-OVA or B16F10-OVA-ZsGreen (10⁶), MC38 (0.5×10⁶) cells in 100 μl PBS into the flank or EO771 (0.5×10⁶) cells in 100 μl PBS into mammary fat pad. Tumor size was determined by the formula L × W, where L is length and W is width. Anti-mouse PD-1 (RMP1–14, BP0146, Bio X Cell, 100μg/mouse) or rat IgG2a isotype control (2A3, BP0089, Bio X Cell, 100μg/mouse) was injected intraperitoneally (i.p.) on D6, D9 and D12 after tumor cell implantation.

Lymph node and tumor tissue digestion

Tumor draining lymph nodes (tdLNs) were finely minced into small pieces 1–2 mm in size and placed in RPMI-1640 medium containing 1 mg/ml Collagenase IV (Worthington, LS004188), 10 μg/ml DNAse I (Roche, 11284932001), and 3% FBS. The samples were incubated at 37°C for 30 min with stirring. Similarly, the tumor tissues were cut into small pieces 1–2 mm in size and placed in RPMI-1640 medium containing 1 mg/ml Collagenase IV, 20 μg/ml DNAse I, and 3% FBS. The samples were then incubated on a shaker at 37°C for 40 mins. After digestion, the cell suspension was smashed and filtered through a 100 μm filter for subsequent staining.

Graphical illustrations

Graphical elements used in diagrams to illustrate experimental design schemes were created with BioRender.com, licensed by Stanford University.

Statistics and reproducibility

All statistical analyses were performed by Graph Pad Prism (v10) software and R (v4.2.2). P < 0.05 was considered significant. All data are presented as mean ± s.e.m. unless otherwise noted. Data were analyzed using unpaired or paired two-tailed Student’s t-tests for comparisons between two groups; for multiple group comparisons, ordinary one-way ANOVA followed by Tukey’s or Dunnett’s multiple-comparison test, or two-way ANOVA with Tukey’s or Šidák’s multiple-comparison test was used, with P-values corrected for multiple comparisons. The log-rank (Mantel-Cox) test was used to determine P-values for heart survival. Sample sizes were determined based on preliminary data or previous experience with variability in similar experimental settings. Wilcoxon rank sum test was used in scRNA-seq analyses. For differential expression testing between experimental conditions, equal numbers of each cDC1 subtype were subsetted from each condition to control for variations in subtype population structure. Statistically significant shifts in cDC1 subtype proportions were identified using the ‘propeller’ function with bootstrapping in the ‘Speckle’ R package (v0.99.7)¹¹². In bulk RNA-seq analyses, P-values were calculated using hypergeometric tests with Benjamini-Hochberg correction or two-sided generalized linear model (GLM) likelihood ratio tests with Benjamini-Hochberg correction. The following key software packages were utilized in analyses: Seurat (v5.0.1), ggplot2 (v3.5.1), ComplexHeatmap (v2.14.0), reshape2 (v1.4.4), viridis (v0.6.5), viridisLite (v0.4.2), speckle (v0.99.7), RColorBrewer (v1.1–3), deMULTIplex2 (v1.0.1), Nebulosa (v1.18.0), Trimmomatic (v0.36), SAMtools (v1.16.1), HISAT2 (v2.1.0), FeatureCounts (v2.0.3), DESeq2 (v1.46.0), clusterProfiler (v4.14.6), and GSEA (v3.0). Details of the specific tests were noted in the respective figure legends.

Extended Data

Extended Data Fig. 1 |. XCR1⁺CD8α⁺ cDC1s in the spleen following TLI/ATS are bona fide cDC1s.

a, Total splenocyte frequency per spleen (UNT n=5; TLI/ATS n=10). b, Gating of CD11c^highMHCII^high cDCs from the splenocytes. Summary graph of the frequency of cDCs in total splenocytes (UNT, n=6; TLI/ATS, n=7). c, Gating of CD8α⁺CD11b⁻ cDC1s and CD8α⁻CD11b⁺ cDC2s; frequencies in cDCs (UNT, n=6; TLI/ATS, n=6). d, XCR1⁺ and e, Ki67⁺ cDC1 frequencies (UNT, n=5; TLI/ATS, n=5). f, qPCR of Batf3 and Irf8 in XCR1⁺CD8α⁺ cDC1s (n=5/group). g, IRF8 and h, Zbtb46-GFP expression (UNT, n=6; TLI/ATS, n=5). i-k, MFI of MafB-mCherry expression and % of MafB-mCherry⁺ cells in XCR1⁺CD8α⁺ cDC1s (i), cDC2s (j) and red pulp macrophages (RPMΦs) (k). Day 0 (UNT), n = 4; TLI/ATS Day 5, n = 5; TLI/ATS Day 9, n = 5 (i,j,k). Data are shown from one experiment, representative of at least two independent experiments with similar results (a-k). Statistical analysis was performed using unpaired two-tailed Student’s t-test (a,b,c,d,e,h), or one-way ANOVA with Tukey’s multiple-comparison test (f,g,i,j,k). Data are mean ± s.e.m. (a-k).

Extended Data Fig. 2 |. TLI/ATS leads to widespread apoptosis and extramedullary erythropoiesis in the spleen and a marked rise in serum EPO.

a, TUNEL staining of spleen sections, UNT vs. TLI/ATS (scale bar = 200 μm). b, Spleen cell composition after ATS, TLI, or TLI/ATS; pie chart shows mean frequencies of indicated populations (n = 3). T cells (TCRβ⁺ CD19⁻NK1.1⁻), B cells (CD19⁺TCRβ⁻NK1.1⁻), erythroid progenitors (CD11c⁻TER119⁺CD71⁺), cDCs (CD3ε⁻B220⁻SiglecH⁻PDCA-1⁻CD11c^highMHCII^high), other myeloid cells are subdivided into CD11b⁺Ly6C ⁻Ly6G⁺, CD11b⁺Ly6G ⁻Ly6C⁺ and CD11b^intF4/80⁺. c, Serum EPO levels over time after TLI/ATS (ELISA, n = 8). d, CD71⁺TER119⁺ erythroid progenitors in spleen (day 6) (upper) and in spleen/BM (lower) over time after TLI/ATS (n = 3). e, Co-detection by indexing (CODEX) imaging of WT C57BL/6 spleen (UNT vs. 1 day after TLI; scale bar = 500 μm). f, Scheme of EPO treatment in Epor-tdTomato-Cre mice (i.p. × 5 days). g,h, Flow cytometry showing splenic cDC1 frequency among cDCs (g) and Epor-tdT⁺ cDC1 frequency/MFI (h) (+PBS, n = 5; +EPO, n = 5). i-j, Frequencies of Epor-tdT⁺ and Epor-tdT ⁻ TER119⁺ erythroid cells (i) and Epor-tdT⁺CD11b^intF4/80⁺Epor-tdT⁺ RPMΦs (j), (+PBS, n = 5; +EPO, n = 5). k, CCR7 vs. Epor-tdT expression in XCR1⁺CD8α⁺ cDC1s that were gated as live-dead aqua⁻CD3ε⁻CD19⁻B220⁻ SiglecH⁻PDCA-1⁻CD11c^highMHCII^high; histogram overlay for CCR7⁺ cDC1s (UNT, n = 5; TLI/ATS, n = 5). Data are shown from one experiment, representative of at least two independent experiments with similar results (a,b,c,d,g,h,i,j,k) or from one experiment (e). Statistical analysis was performed using unpaired two-tailed Student’s t-test (g,h,i,j,k). Data are mean ± s.e.m. (d,g-k). The diagram in f was created with BioRender by Xiangyue Zhang (2025).

Extended Data Fig. 3 |. EPO-EpoR downstream signaling is activated in cDC1s following TLI/ATS.

a-c, Gene Set Enrichment Analysis (GSEA) of transcriptional profiles using the Hallmark gene set of MSigDB. NES, normalized enrichment score; FDR, false discovery rate. Red: upregulated; Blue: downregulated. TLI/ATS vs. UNT. b, Upregulated gene sets. c, Downregulated gene sets. d-e, Intracellular phospho-flow cytometric analysis of EPO-EpoR downstream signaling in live-dead blue⁻Lin⁻SiglecH⁻PDCA-1⁻CD11c^highMHCII^high. Spleens were harvested on the next day following the last dose of TLI or TLI/ATS. UNT (n=4) vs. TLI (n=4) vs. TLI/ATS (n=4). d, XCR1⁺CD8α⁺ cDC1s and e, XCR1⁻CD8α⁻ cDC2s. f,g, Histograms and MFI of the indicated EPO-EpoR downstream signaling molecules with fluorescence minus one (FMO) as controls by intracellular phospho-flow staining on the next day following the last dose of TLI/ATS treatment. Epor^flox/flox (n=4) vs. Epor^ΔXcr1 (n=5). cDC1s (f) and cDC2s (g). h, Ex vivo analysis of EPO-EpoR downstream signaling in splenic cDC1s. Splenic cDCs were MACS-purified with a pan-DC isolation kit and cultured at 5 × 10⁶ cells/ml, then rested overnight. Cells were isolated from UNT or TLI/ATS-treated Epor^flox/flox (n=4; n=4) and Epor^ΔXcr1 (n=4; n=4) mice. cDCs from TLI/ATS-treated mice were stimulated ex vivo with EPO (10 IU/200 μl) or PBS (control) overnight. Phosphorylation of downstream signaling molecules was assessed by flow cytometry, after gating on XCR1⁺SIRPα⁻ splenic cDC1s. Data are shown from one experiment, representative of at least two independent experiments with similar results (d-h). Statistical analysis was performed using unpaired two-tailed Student’s t-test (f,g), one-way ANOVA Tukey’s multiple-comparison test (d, e and h left) or paired two-tailed Student’s t-test (h right). Data are mean ± s.e.m. (d-h).

Extended Data Fig. 4 |. FOXP3⁺ T_regs play an indispensable role in TLI/ATS-induced cDC1 EpoR-dependent immune tolerance.

a, Representative pseudocolor plots showing FOXP3⁺ T_reg depletion efficiency in recipient mice on day 6 after DT treatment (DT injections on days 0, 2, and 4). b,c, Representative pseudocolor plots of C57BL/6, Batf3^−/−, or Epor^ΔXcr1 recipient conventional CD4⁺ T cell percentages and FOXP3⁺ T_reg percentages in CD4⁺ T cells. d,e, Absolute cell number of indicated cell populations. b,d, Day 0 (UNT, n=5; n=5; n=5 and TLI/ATS, n=6; n=5; n=4) and c,e, Day 14 of UNT or TLI/ATS-treated groups post allo-BM infusion (UNT, n=11; n=6; n=9 and TLI/ATS, n=3; n=11; n=10). f, MHCII expression on cDC1s and cDC2s from MHCII^flox/flox and MHCII^ΔXcr1 spleens. g,h,i, MHCII^flox/flox (n=6) and MHCII^ΔXcr1 (n=6) recipients were given TLI/ATS and i.v. infused with BALB/c donor BM cells. 14 days post BM infusion, the percentages of donor type (H2K^d+) cells among leukocyte populations were determined in the peripheral blood of hosts. g,i, Recipient MHCI (H-2K^b)⁺TCRβ⁺CD4⁺ T cell frequency among total live cells and FOXP3⁺ frequency among CD4⁺ T cells were analyzed on day 14. j, CD45.2⁺Foxp3^WTEpor^flox/flox (+PBS/without DT, n = 8; +DT, n = 8) or Epor^ΔXcr1 (+PBS/ without DT, n = 8; +DT, n = 8) mice were injected with 30 million CD45.1⁺FOXP3^DTR CD4⁺ T cells isolated by MACS. Two consecutive doses of DT or PBS were given on each of the following 2 days. Subsequently, the mice were treated with TLI/ATS, and 2W1S-BALB/c donor BM cells were infused i.v., and 14 days later, 2W1S-tetramer⁺CD44⁺H-2K^b+TCRβ⁺CD4⁺ T cells from the spleens were analyzed for FOXP3 expression by flow cytometry. FOXP3 expression in CD45.1⁺ or CD45.2⁺2W1S-tetramer⁺ CD4⁺ T cells is shown. One experiment (j) or one of two independent experiments with similar results are shown (a-i). Statistical analysis was performed using unpaired two-tailed Student’s t-test (g,h,i), or two-way ANOVA with Tukey’s multiple-comparison test (d,e,j). Data are mean ± s.e.m. (d,e,g,i,j). The diagram in j was created with BioRender by Xiangyue Zhang (2025).

Extended Data Fig. 5 |. Differentially expressed genes (DEGs) in cDC1s in scRNA-seq analysis and ex vivo TGFβ-dependent Ag-specific FOXP3⁺ T_reg induction by CCR7⁺ cDC1s.

a, UMAP of splenic cDC1 gene expression by sample identity. b, Dot plots of top condition-specific DEGs in Epor^flox/flox and Epor^ΔXcr1 mice (TLI/ATS vs. UNT). c, Absolute cDC1 numbers per spleen in UNT vs. TLI/ATS-treated Epor^flox/flox (n=5/condition) and Epor^ΔXcr1 (n=5/condition) mice. d, UMAPs of cDC1 subtypes in Epor-tdT⁺ and Epor-tdT⁻ cells. e-g, Dot plots of top condition-specific DEGs in Epor-tdT⁺ and Epor-tdT⁻ cDC1s (TLI/ATS) and in Epor^flox/flox and Epor^ΔXcr1 mice (TLI/ATS vs. UNT). Dot color = expression, size = % of indicated gene expressed cells (b,e-g). h, Bar charts showing cDC subtype (d) proportions in Epor-tdT⁺ and EpoR-tdT⁻ cDC1s following TLI/ATS. i, Role of TGFβ in FOXP3⁺ T_reg induction by CCR7⁺ cDC1s: 12 h after apoptotic Act-mOVA injection, CCR7⁺ cDC1s (1×10⁴) were cocultured with CD45.1⁺ CTV-labeled naïve OT-II cells ± anti-TGFβ; FOXP3 expression was analyzed by flow cytometry (n=5/group). j,k, Representative flow cytometry analysis and l,m, Absolute cell number of indicated cell populations of Fig. 3j, Itgb8^ΔXcr1 vs. littermate controls. j,l, Day 0 and k,m, Day 14 of UNT (n=5; n=5) or TLI/ATS-treated (n=5; n=5) groups post allo-BM infusion. n,o, Aldh1a2^ΔCD11c: Batf3^−/− (n=6) vs. Aldh1a2^flox/flox: Batf3^−/− (n=7) BM chimeric recipient mice (CD45.1⁺) were given TLI/ATS. 1 day after the last dose of TLI/ATS, 2W1S-BALB/c donor BM cells were infused i.v., and 14 days later, the percentages of donor type (H2K^d+) cells among leukocyte populations in the peripheral blood of hosts were determined (n) and 2W1S-tetramer⁺CD44⁺H-2K^b+TCRβ⁺CD4⁺ T cells from the spleens were analyzed for FOXP3 expression by flow cytometry and FOXP3⁺ T_regs were counted (o). Data are representative of at least three independent experiments with similar results (c,i) or one experiment (j-o). Statistical analysis was performed using unpaired two-tailed Student’s t-test (c,i,n,o), or two-way ANOVA followed by Tukey’s multiple-comparison test with p-value adjusted (l,m), or propeller test, two-sided, no multiple-comparison correction (b), or Wilcoxon rank sum test, two-sided, Bonferroni correction (h). Data are mean ± s.e.m. (c,i-o). The diagram in i was created with BioRender by Xiangyue Zhang (2025).

Extended data Fig. 6 |. Absence of EpoR on cDC1s gives rise to immunogenic cDC1s that promote both CD8⁺ T cell cross-priming and CD4⁺ T cell priming to cell-associated Ags.

a, MFI of indicated molecules on gated cDC1s with fluorescence minus one (FMO) as controls. b, MFI of indicated molecules on gated cDC2s with fluorescence minus one (FMO) as controls. c, Percentages of cDC2s in splenic cDCs. d, Representative flow gating of CCR7⁺XCR1⁺SIRPα⁻ cDC1s in splenic cDC1s (Upper), and percentages and absolute numbers of CCR7⁺ cDC1s (Lower). a-d, Epor^flox/flox (n=5) vs. Epor^ΔXcr1 (n=5) mice. e, MFI of indicated molecules on CCR7⁺ vs. CCR7⁻ cDC1s. CD40 and PD-L1: Epor^flox/flox (n=5); Epor^ΔXcr1 (n=5). CD80 and CD86: Epor^flox/flox (n=6); Epor^ΔXcr1 (n=6). f, Cross-presentation assay: apoptotic Act-mOVA thymocytes injected into Epor^flox/flox (n=5) or Epor^ΔXcr1 (n=5) mice 1 day after transfer of CTV-labeled naïve CD45.1⁺ naïve OT-I cells; spleens analyzed on day 4 for OT-I expansion and proliferation. g, Same setup with OT-II cells; percentages and absolute numbers of OT-II cells and proliferating OT-II cells were assessed. Ag-specific CD4⁺ T cell response: Ag-specific CD4⁺ T cell immune response following i.v. injection of apoptotic Act-mOVA thymocytes 1 day after i.v. injection of CTV-labeled naïve CD45.1⁺ naïve OT-II cells. Spleens were analyzed at day 4 for OT-II expansion and proliferation. Epor^flox/flox (n=5) and Epor^ΔXcr1 (n=5) mice. Data are shown from one experiment, representative of at least three independent experiments with similar results (a-g). Statistical analysis was performed using unpaired two-tailed Student’s t-test (a,b,c,d,f,g) and two-way ANOVA followed by Tukey’s multiple-comparison test (e). Data are mean ± s.e.m. (a-g). The diagrams in f,g were created with BioRender by Xiangyue Zhang (2025).

Extended data Fig. 7 |. Phenotypes of T cells in the spleens of Epor^flox/flox vs. Epor^ΔXcr1 mice and role of EpoR in cDC1-mediated cell-associated Ag-specific CD4⁺ T cell priming and proliferation and FOXP3⁺ T_reg induction.

a-e, Percentages and absolute numbers of CD4⁺ T cells (a), FOXP3⁺CD25⁺ T_regs in CD4⁺ T cells (b), CD44^highCD62L^low effector cells and CD44^low CD62L^high naïve cells in CD4⁺ T cells (c), CD8⁺ T cells (d), and CD44^highCD62L^low effector cells and CD44^lowCD62L^high naïve cells in CD8⁺ T cells (e) in the spleens of Epor^ΔXcr1 and littermate Epor^flox/flox control mice with representative flow cytometric plots. a-e, Epor^flox/flox, n=5; Epor^ΔXcr1, n=5. f,g, Flow cytometry-based measurement of cell-associated Ag-specific CD4⁺ T cell immune response in the spleen following i.v. injection of apoptotic Act-mOVA thymocytes into mice of the indicated genotypes 1 day after i.v. injection of CTV-labeled naïve CD45.1⁺ OT-II cells. f, WT C57BL/6 (n=5) and Batf3^−/− (n=7). g, FOXP3⁺ T_reg induction in Epor^flox/flox and Epor^ΔXcr1 mice. Recombinant EPO or PBS was administered daily, from Day −3 to Day 4. +PBS: Epor^flox/flox (n = 5), Epor^ΔXcr1 (n = 5); +EPO: Epor^flox/flox (n = 5), Epor^ΔXcr1 (n = 5) mice. Data are shown from one experiment, representative of at least three independent experiments with similar results (a-e), or two independent experiments with similar results (f,g). Statistical analysis was performed using unpaired two-tailed Student’s t-test (a-f), or two-way ANOVA followed by Tukey’s multiple-comparison test (g). Data are mean ± s.e.m. (a-g). The diagrams in f,g were created with BioRender by Xiangyue Zhang (2025).

Extended Data Fig. 8 |. Epor-tdT expression on XCR1⁺ cDC1s in selected organs and tolerogenic phenotype of Epor-tdT⁺ migratory cDC1s in pLNs.

a,b,c CCR7- and Batf3-dependent Epor-tdT expression on migratory cDCs in pLNs. Migratory cDCs were gated as CD11c^intMHCII^high from live-dead aqua⁻Lin⁻SiglecH⁻PDCA-1⁻EpCAM⁻ cells, and resident cDCs were gated as CD11c^highMHCII^int from live-dead aqua⁻Lin⁻SiglecH⁻PDCA-1⁻ cells. pLNs including inguinal, axillary, brachial, and superficial cervical LNs were combined for analysis by flow cytometry (a,b). Ccr7^−/−Epor^tdT/+, Batf3^−/−Epor^tdT/+ and WT C57BL/6 mice (a). Histogram overlay of Epor-tdT expression on migratory or resident cDCs from individual mouse strains (b). EpoR-tdT expression on migratory cDCs from individual pLNs of Epor^tdT/+ mice (c). d, Epor-tdT expression on cDC1s obtained from the indicated organs in Zbtb46^GFP/+Epor^tdT/+ mice. cDCs were gated in CD45⁺ cells as CD11c⁺Zbtb46-GFP⁺, in which cDC1s were further gated as XCR1⁺ CD103⁺. e, Flow cytometric analysis of tolerance associated cell-surface molecules on pLN Epor-tdT⁺ migratory cDC1s compared with Epor-tdT^- cDCs and resident cDCs with FMO serving as controls (n=8). Data are representative of at least two independent experiments with similar results (a-e). Statistical analysis was performed using one-way ANOVA Tukey’s multiple-comparison test (e). Data are mean ± s.e.m. (e). The diagrams in a,e were created with BioRender by Xiangyue Zhang (2025).

Extended Data Fig. 9 |. Induction of Ag-specific CD4⁺FOXP3⁺ T_regs by pLN migratory EpoR-tdT⁺ cDC1s.

a, FACS-sorted pLN Epor-tdT⁺ (n=4) or EpoR-tdT⁻ (n=4) XCR1⁺ migratory cDC1s from Epor-tdT mice were cocultured for 5 days with CTV-labeled naïve OT-II cells + DEC205-OVA (ratio 1:5); FOXP3 expression in OT-II cells was analyzed by flow cytometry. b, Same setup as (a) but with apoptotic Act-mOVA thymocytes (ratio 1:5:2) ± EPO (20 IU per well per day for 5 days); FOXP3 expression in OT-II cells was measured. Epor-tdT⁺ (+PBS or w/o, n=4; +EPO, n=4) or EpoR-tdT⁻ (+PBS or w/o, n=4; +EPO, n=3). c, FACS-sorted pLN Epor-tdT⁺ or EpoR-tdT⁻ migratory cDC1s were cocultured for 12 h with apoptotic CD45.1⁺ thymocytes ± EPO; MFIs of surface markers were analyzed. Epor-tdT⁺: n=2. Epor-tdT⁻: n=2. d, Migratory cDC1s from Epor^flox/flox or Epor^ΔXcr1 mice were cocultured with naïve CTV-labeled OT-II cells and Act-mOVA thymocytes (1:5:2) ± EPO (20 IU per well per day for 5 days); FOXP3 induction was assessed. n=6/group. e, Efferocytosis of PKH67-labeled apoptotic thymocytes by migratory cDC1s and cDC2s in dLNs 12 h post-injection. f, Act-mOVA^{CD45.1/CD45.2} mice were reconstituted with either Epor^flox/flox (n=5) or Epor^ΔXcr1 (n=6) BM cells after lethal irradiation. 8 weeks post-reconstitution, naïve CTV-labeled OT-II cells were i.v. infused (day 0), and EPO was administered on days −2 to 2. FOXP3 induction in OT-II cells was assessed in inguinal LNs on day 9. Data are shown from one experiment, representative of two independent experiments with similar results (a-e) or one (f) independent experiment. Statistical analysis was performed using unpaired two-tailed Student’s t-test (a,b,f), or two-way ANOVA with Tukey’s multiple-comparison test (d). Data are mean ± s.e.m. (a,b,c,d,f). The diagrams in a,b,c,e,f were created with BioRender by Xiangyue Zhang (2025).

Extended Data Fig. 10 |. Phenotypes of T cells in the pLNs of Epor^flox/flox vs. Epor^ΔXcr1 mice.

a-e, Percentages and absolute numbers of CD4⁺ T cells (a), FOXP3⁺CD25⁺ T_regs in CD4⁺ T cells (b), CD44^highCD62L^low effector cells and CD44^lowCD62L^high naïve cells in CD4⁺ T cells (c), CD8⁺ T cells (d), and CD44^highCD62L^low effector cells and CD44^lowCD62L^high naïve cells in CD8⁺ T cells (e) in the pLNs of Epor^ΔXcr1 and littermate Epor^flox/flox control mice with representative flow cytometry plots. a-e, Epor^flox/flox (n=9) and Epor^ΔXcr1 (n=9). Data are shown from one experiment, representative of at least three independent experiments with similar results (a-e). Statistical analysis was performed using unpaired two-tailed Student’s t-test (a-e). Data are mean ± s.e.m. (a-e).

Extended Data Fig. 11 |. EpoR expression on tumor-infiltrating leukocytes (TILs), tumor Ag-carrying migratory cDC1s in tdLNs and tumors, and the correlation of tumor growth with systemic EPO levels.

a, Zbtb46^GFP/+ Epor^tdT/+ mice were implanted s.c. with MC38 or B16F10 tumors, or EO771 tumors in the mammary fat pad. On day 12, tumors were harvested for flow cytometric analysis of Epor-tdT on cDCs (live-dead aqua⁻CD45⁺Zbtb46-GFP⁺CD11c⁺); cDC1s were gated as XCR1⁺ and non-cDC1s as XCR1^-. b-d, Mice implanted s.c. with MC38-OVA^dim or B16F10-OVA; on day 10, tumors were analyzed for Epor-tdT expression in TILs. b, Representative gating strategy of individual live-dead blue^- TIL populations. c,d, Histogram overlay showing Epor-tdT expression in individual cell populations. e, Zbtb46^GFP/+Epor^tdT/+ mice with MC38-OVA^dim tumors (day 12) were analyzed for EpoR-tdT on tumor-infiltrating cDCs; CCR7⁺ (population I) and CCR7⁻ (populations II/III by Ly6A) subsets were gated, with XCR1/CD103 staining to define cDC1s and cDC2s. f, Quantification of Epor-tdT expression on individual tumor infiltrating cDC subsets. MC38-OVA^dim (n=4) and B16F10-OVA (n=4) tumors were harvested on day 12 post-s.c. implantation for flow cytometry. Gating strategy as in Fig. 5a and Extended Data Fig. 11e. g, Flow cytometry analysis of Epor-tdT expression on tdLN migratory cDC1s. Overlay of migratory cDC1s with Lin⁻ live cells to show Epor-tdT expression levels. h, Serum EPO levels were measured by ELISA on the indicated days after s.c. implantation of MC38-OVA^dim (n=6) or B16F10-OVA (n=5) tumors in WT mice. i,j,k, B16F10-OVA-ZsGreen cells were s.c. implanted into Epor^tdT/+ mice, and tdLN and tumor were analyzed on day 9 after inoculation. j, Flow cytometry analysis of Epor-tdT expression on tdLN ZsGreen⁺ migratory and resident XCR1⁺ cDC1s. Overlay of migratory ZsGreen⁺ cDC1s or resident ZsGreen⁺ cDC1s with Lin^-live cells to show Epor-tdT expression levels. k, Flow cytometry analysis of Epor-tdT expression on tumor infiltrating ZsGreen⁺ cDC1s. Data are shown from one experiment, representative of at least two independent experiments with similar results (a-e,f,g,h,j,k). Statistical analysis was performed using one-way ANOVA with Dunnett’s multiple-comparison test (f). Data are mean ± s.e.m. (f,h). The diagrams in a,b,e,g,i were created with BioRender by Xiangyue Zhang (2025).

Extended Data Fig. 12 |. Loss of EpoR in cDC1s limits tumor growth and promotes immunogenic function of tumor Ag-carrying cDC1s in both tdLNs and tumor.

a, Growth of MC38-OVA^dim tumor cells implanted s.c. into Epor^flox/flox (n=8) and Epor^ΔXcr1 mice (n=9). b, Growth of B16F10-OVA tumor cells implanted s.c. into Epor^flox/flox (n=6) and Epor^ΔXcr1 mice (n=7). c, Experimental design for phenotyping tumor-Ag carrying ZsGreen⁺ cDC1s in tdLN and tumors in Epor^flox/flox vs. Epor^ΔXcr1 mice. B16F10-OVA-ZsGreen cells were implanted s.c. into Epor^tdT/+ mice, and tdLNs and tumors were analyzed on day 9 after implantation. d, Flow cytometry analysis of Epor-tdT expression on ZsGreen⁺ migratory XCR1⁺SIRPα⁻ cDC1s in tdLNs with summary graph of statistical quantification. Epor^flox/flox (n=7) and Epor^ΔXcr1 (n=8). e, Flow cytometry analysis of CD40, CD80 and CD86 expression on tumor infiltrating ZsGreen⁺ cDC1s with summary graph of statistical quantification. Epor^flox/flox (n=7) and Epor^ΔXcr1 (n=8) mice. (f-l) MC38-OVA^dim tumors were s.c. implanted into Epor^flox/flox (n=8) and Epor^ΔXcr1 (n=7) and 10 days later TILs were analyzed. f. Percentages of CD45⁺ live immune cells and CD8⁺ or CD4⁺ T cells in CD45⁺ TILs. g, Frequency of OVA_257–264-dextramer⁺CD8⁺ T cells among CD8⁺ T cells. h, Representative flow plots and quantification of CD8⁺ T cells expressing TIM-3 and PD-1. i, Representative flow plots and quantification of TCF1⁺TIM-3⁻CD8⁺ T cells. j, Representative histograms and quantification of perforin, granzyme-B, IFNγ and TNFα expression in tumor-infiltrating CD8⁺ T cells. k, Percentage of FOXP3⁺ T_regs in CD4⁺ T cells with representative flow plots (Left). Absolute number of T_regs (Right). l, Representative flow plots and percentages of T-bet⁺CXCR3⁺ T_regs in CD4⁺ FOXP3⁺ T_regs. (f-l). Data are shown from one experiment, representative of at least two independent experiments with similar results (a,b,d,e,f-l). Statistical analysis was performed using two-way ANOVA with Šídák’s multiple comparison test (a,b), or two-tailed unpaired Student’s t-test (d,e,f,g,i,j.k,l), or two-way ANOVA with Tukey’s multiple-comparison test (h). Data are mean ± s.e.m. (a,b,d,e-l). The diagram in c was created with BioRender by Xiangyue Zhang (2025).

Data availability

The data supporting the findings of this study are available in the Article and its Supplementary Information. All transcriptional data generated in the current study were deposited at the NCBI Gene Expression Omnibus (GEO) and are publicly available under the following accession numbers: GSE253056 (bulk RNA-seq) and GSE284080 (scRNA-seq), respectively. Source data are provided with this paper.

Code availability

The scripts for replicating the RNA-seq analyses presented are accessible on GitHub (https://github.com/chansigit/Epor-cDC1-bulkRNAseq). Scripts for reproducing all scRNA-seq analyses presented are accessible on GitHub (https://github.com/chris-mcginnis-ucsf/epor_dc_tolerance) and associated processed data objects are available on Synapse (https://synapse.org/Synapse:syn64330568).