Optimization of the Irf8 +32-kb enhancer disrupts dendritic cell lineage segregation

Abstract

Autoactivation of lineage-determining transcription factors mediates bistable expression, generating distinct cell phenotypes essential for complex body plans. Classical type 1 dendritic cell (cDC1) and type 2 dendritic cell (cDC2) subsets provide nonredundant functions for defense against distinct immune challenges. Interferon regulatory factor 8 (IRF8), the cDC1 lineage-determining transcription factor, undergoes autoactivation in cDC1 progenitors to establish cDC1 identity, yet its expression is downregulated during cDC2 differentiation by an unknown mechanism. This study reveals that the Irf8 +32-kb enhancer, responsible for IRF8 autoactivation, is naturally suboptimized with low-affinity IRF8 binding sites. Introducing multiple high-affinity IRF8 sites into the Irf8 +32-kb enhancer causes a gain-of-function effect, leading to erroneous IRF8 autoactivation in specified cDC2 progenitors, redirecting them toward cDC1 and a novel hybrid DC subset with mixed-lineage phenotypes. Further, this also causes a loss-of-function effect, reducing Irf8 expression in cDC1s. These developmental alterations critically impair both cDC1-dependent and cDC2-dependent arms of immunity. Collectively, our findings underscore the significance of enhancer suboptimization in the developmental segregation of cDCs required for normal immune function.

Autoactivation is the process in which an activating transcription factor (TF) directly promotes its own transcription, allowing for bistable expression in either ‘on’ or ‘off’ states¹. When a lineage-determining TF undergoes such a process, two lineages can emerge, with the TF’s ‘on’ state directing gene expression unique to one lineage and its ‘off’ state enabling an alternative lineage^1–3. cDCs exemplify this through their cDC1 and cDC2 subsets^4–6. cDC1s function primarily to direct immune responses against intracellular pathogens, for example, by priming CD8⁺ T cells^7–9. cDC2s are required for effective responses against certain extracellular bacteria^10–12, and T helper type 2 (T_H2) responses against parasites such as the helminth Heligmosomoides polygyrus^13–15.

IRF8 is the lineage-determining TF for the cDC1 subset¹⁶. In concert with basic leucine zipper ATF-like transcription factor 3 (BATF3) and JUNB, IRF8 binds to AP-1-IRF composite elements (AICEs)^17,18 (Fig. 1a). This interaction is critical in cDC1s, where IRF8’s high expression underlies an AICE-dependent transcriptional program essential for maintaining cDC1 identity¹⁹. Conversely, cDC2 cells, characterized by minimal IRF8 levels, lack sufficient IRF8–BATF3 interaction to engage AICEs, precluding the expression of cDC1-specific genes¹⁹. Consequently, the ‘on’ and ‘off’ expression states of IRF8 direct gene expression profiles that define the cDC1 and cDC2 lineages, respectively.

Fig. 1 |. The Irf8 +32-kb enhancer is supported by low-affinity AICEs.

a, A model of the BATF3–JUNB–IRF8 complex binding to an AICE. This model is constructed based on the DNA-binding domain (DBD) and IRF association domain (IAD) of IRF4, and the FOS–JUN heterodimer, derived from Protein Data Bank entries 7O56, 5BV1 and 1FOS, respectively. b, AICE motif identified by MEME analysis of the top 1,000 BATF3 ChIP–seq peaks in cDC1. The analysis yielded an E value of 1.2 × 10⁻¹⁸⁷, with the motif present in 561 of 1,000 peaks. c, Manhattan plot showing the −log₁₀(P value) for the most probable binding sites within each BATF3/IRF8 co-occupancy ChIP–seq peak. The nearest genes to select peaks are noted. P values were computed using the Find Individual Motif Occurrences (FIMO) tool from the AICE PWM in b. No adjustment for multiple comparisons was applied. d, Sequences of EMSA probes derived from highlighted peaks in c. The underlined sequences represent AICE motifs, with associated P values provided for each. P values were computed using FIMO from the AICE PWM in b. No adjustment for multiple comparisons was applied. For the ‘Irf8 s143’ probe, the specific locations and orientations of individual AICEs are annotated beneath the sequence. e, Schematic of the Irf8 +32-kb enhancer indicating EMSA probe locations. f, EMSA performed with ³²P-dCTP-labeled probes on NEs from 293FT cells, either untreated (−) or transfected (+) with Junb and Batf3 (BATF3) or Irf8 (IRF8). Bands corresponding to JUNB–BATF3–IRF8 and JUNB–BATF3 complexes are indicated by the upper and lower arrows, respectively. Data shown are representative of two similar experiments. g, Competitive EMSA performed with the ³²P-dCTP-labeled Ctla4 probe in the presence of a 200-fold excess of various unlabeled competitors. Data shown are representative of two similar experiments. h, Retroviral reporter assay assessing activities of various Irf8 +32-kb enhancer mutants in cDC1s obtained from Kit/Flt3L-cultured BM cells. Shown are enhancer-driven GFP expression in transduced cDC1s (pre-gate: B220⁻CD11c⁺ MHCII⁺CD24⁺Sirpa⁻Thy1.1⁺ cells). Numbers are GFP geometric mean fluorescence intensity (MFI). Data shown are representative of three similar experiments.

cDC1 and cDC2 subsets arise from a shared progenitor, the common dendritic cell progenitor (CDP)^20,21. cDC1 lineage commitment progresses from CDP to pre-cDC1 and subsequently to cDC1, while cDC2 lineage commitment follows a trajectory from CDP to pre-cDC2 and then to cDC2 (ref. 22). Notably, the cDC1 lineage-determining TF IRF8 is highly expressed in CDP, pre-cDC1 and even pre-cDC2 (ref. 22). In the cDC1 developmental route, BATF3 is upregulated early in pre-cDC1, where it facilitates IRF8 autoactivation by binding with IRF8 to AICEs in the Irf8 +32-kb enhancer^19,22,23. In contrast, the cDC2 route is characterized by a more gradual upregulation of BATF3, absent in pre-cDC2 but eventually reaching levels similar to those in cDC1 (refs. 19,22). Despite this increase in BATF3 expression, the Irf8 +32-kb enhancer remains nonresponsive, leading to the loss of IRF8 expression in terminally differentiated cDC2 (refs. 22,23).

The present study was initiated to understand the mechanisms ensuring cDC2 lineage fidelity in the context of concurrent BATF3 and IRF8 expression. In developmental models outside hematopoiesis, the concept of enhancer suboptimization has emerged, highlighting how certain enhancers limit their activities to specific spatial–temporal domains^24–26. By strategically restricting affinities and numbers of TF binding motifs, these enhancers are only active in domains where their binding TFs are highly concentrated^{24,25,27–36}. Our investigation focuses on whether a similar regulatory mechanism operates on the Irf8 +32-kb enhancer to restrict IRF8 autoactivation in the cDC1 developmental pathway while not affecting the cDC2 pathway. We began by assessing the affinities of AICEs within the Irf8 +32-kb enhancer. We then engineered mouse models with high-affinity AICEs inserted into the native Irf8 +32-kb enhancer to examine the impact of motif optimization on the developmental and functional bifurcation of cDC1 and cDC2 subsets. These genetic modifications resulted in two distinct effects: first, a gain of function leading to ectopic Irf8 expression in specified cDC2 progenitors, redirecting them toward the cDC1 lineage and a novel hybrid DC subset with mixed-lineage phenotypes. Secondly, a loss of function leading to diminished Irf8 expression in cDC1s and a partial redirection of cells specified for the cDC1 lineage toward the hybrid DC phenotype. Moreover, mice with an optimized Irf8 +32-kb enhancer failed to reject tumors and exhibited a compromised T_H2 response to helminth infections.

Results

The Irf8 +32-kb enhancer is supported by low-affinity AICEs

Previous research from our laboratory identified four AICEs within the Irf8 +32-kb enhancer, labeled as sites 1 to 4 based on their descending AICE likelihood²² (Extended Data Fig. 1). Sites 1, 4 and 3 are clustered on one end of the enhancer, while site 2 is located on the opposite end²² (Extended Data Fig. 1c). A mouse strain with a deletion across all four AICEs in the Irf8 +32-kb enhancer, as well as a strain with a smaller deletion encompassing sites 1, 4 and 3, exhibits a complete lack of cDC1 development, underscoring the critical role of this enhancer in IRF8 autoactivation²³ (Extended Data Fig. 1c). Intriguingly, none of the four AICEs matches known consensus motifs (TTTCNNNNTGA and GAAATGA)^5,18,37 (Extended Data Fig. 1c). To corroborate this finding, we compared the P values of these four sites against well-characterized AICEs. To this end, we reanalyzed our dataset produced by chromatin immunoprecipitation followed by sequencing (ChIP–seq) to generate an AICE position weight matrix (PWM; Fig. 1b). We then scanned all BATF3/IRF8 co-occupancy peaks in cDC1 for AICE presence using MEME Suite tools (Fig. 1c). This reevaluation led to the rediscovery of previously validated AICEs, such as those near the Ctla4, Snx22, Bcor and Xcr1 loci^17–19,38, with P values below 10⁻⁶ (Fig. 1c,d). Conversely, the four AICEs from the Irf8 +32-kb enhancer exhibited P values exceeding default motif identification threshold of 10⁻⁴, with approximately half of other BATF3/IRF8 co-occupancy peaks containing AICEs with lower P values (Fig. 1c,d). This led us to hypothesize that the Irf8 +32-kb enhancers harbor low-affinity AICEs.

To test this hypothesis, we synthesized electrophoretic mobility shift assay (EMSA) probes encompassing the AICEs from the Irf8 +32-kb enhancer and control probes with high-confidence AICEs near Ctla4, Snx22, Bcor and Xcr1 genes^17–19,38 (Fig. 1c–e). The control probes bound strongly to BATF3–JUNB–IRF8 complexes both as labeled probes and as unlabeled competitors, indicative of high-affinity AICEs (Fig. 1f,g). In contrast, Irf8 +32-kb enhancer probes showed minimal or no detectable binding to BATF3–JUNB–IRF8, even at high concentrations (Fig. 1f,g and Extended Data Fig. 2a). To exclude the possibility that additional TFs in cDC1 enhance BATF3–JUNB–IRF8 complex binding at these Irf8 +32-kb enhancer AICEs, we performed EMSA using nuclear extracts (NEs) from MutuDC, an immortalized cDC1 cell line³⁹. While control probes formed complexes of expected sizes, probes from the Irf8 +32-kb enhancer showed negligible binding (Extended Data Fig. 2b). Together, these results reveal the low-affinity nature of AICEs in the Irf8 +32-kb enhancer.

We then examined the functionality of low-affinity AICEs in the Irf8 +32-kb enhancer by building retroviral reporter constructs lacking individual or combinations of these AICEs (Fig. 1h). When expressed in primary cDC1s, each mutant enhancer showed diminished activity compared to the wild-type (WT) enhancer, with the quadruple AICE mutant displaying virtually no activity (Fig. 1h). These data suggest that each AICE, despite its low affinity, contributes to the overall activity of the Irf8 +32-kb enhancer in cDC1.

Lastly, a comparative genomics approach was used to discern if low-affinity AICE usage is a conserved feature within the enhancer governing IRF8 autoactivation. A manual inspection across several mammalian species found no consensus AICE motif in related sequences of the mouse Irf8 +32-kb enhancer (Extended Data Fig. 2c). Notably, computational predictions identified the mouse enhancer as having the most pronounced AICEs (Extended Data Fig. 2c). We also validated the functionality and low-affinity nature of the human enhancer, located at +48 kb relative to IRF8. Retroviral reporter assays confirmed the cDC1-specific activity of this human enhancer, thereby confirming its functional homology to the mouse Irf8 +32-kb enhancer (Extended Data Fig. 2d). Moreover, EMSA probes derived from the human IRF8 +48-kb enhancer failed to bind BATF3–JUNB–IRF8 complexes (Extended Data Fig. 2e). These observations suggest that the strategic use of low-affinity AICEs in the enhancer driving IRF8 autoactivation is a conserved evolutionary mechanism. Notably, this strategy mirrors a wider biological principle observed in some developmental models, where low-affinity sites in enhancers act as a safeguarding mechanism to prevent ectopic enhancer activity^{24,25,27–36}.

High-affinity AICEs induce a hybrid DC population

The specified pre-cDC2 progenitor is characterized by robust Irf8 expression without concurrent Batf3, contrasting with terminally differentiated cDC2 with diminished Irf8 alongside elevated Batf3 levels²² (Fig. 2a and Extended Data Fig. 3a,b). Thus, during the process of cDC2 commitment, there is a transient period where Batf3 and Irf8 are coexpressed at intermediate levels. We hypothesized that the Irf8 +32-kb enhancer is suboptimized to safeguard against erroneous IRF8 autoactivation throughout cDC2 differentiation by limiting enhancer occupancy.

Fig. 2 |. An optimized Irf8 +32-kb enhancer induces a hybrid DC lineage.

a, Expression levels of Irf8 and Batf3 across different stages of cDC1 and cDC2 development assessed by microarray. b, Diagram illustrating the introduction of high-affinity AICEs at site 3 of the Irf8 +32-kb enhancer, resulting in the Irf8 +32^H allele. The sequence inserted at site 3 is presented. c, Representative flow cytometry plots showing splenic cDC populations from WT, Irf8 +32^H/+ (H/+) and Irf8 +32^H/H (H/H) mice (pre-gate: B220⁻SiglecH⁻CD11c⁺ MHCII⁺ splenocytes). d, Frequencies of splenic cDC1s, cDC2s and hybrid DCs from the indicated mice. Data were pooled from ten independent experiments, presented as mean values ± s.d. Statistical significance was evaluated using one-way analysis of variance (ANOVA) with Dunnett’s multiple-comparisons test. P values are indicated above the graphs. NS, not significant. e, PCA plot derived from bulk RNA-seq data of sorted splenic DC populations for the indicated genotypes. f, Heat map visualization of the bulk RNA-seq data as in e. Shown are the top 464 genes differentially expressed between WT cDC1 and cDC2 (false discovery rate < 0.001, fourfold change). Genes representative of selected clusters are indicated. Irf8 is the direct target gene of the +32-kb enhancer. Genes in bold encode the surface markers used for gating and sorting. sgRNA, single guide RNA; ssODN, single-stranded oligodeoxynucleotide.

We investigated whether integrating high-affinity AICEs into the Irf8 +32-kb enhancer would lower the threshold for IRF8 autoactivation by increasing its sensitivity to BATF3 and IRF8. A mouse model was engineered by inserting several high-affinity AICEs into the endogenous locus of the Irf8 +32-kb enhancer, resulting in the Irf8 +32^H allele (Fig. 2b). Mice heterozygous or homozygous for the Irf8 +32^H allele retained normal frequencies of various DC progenitors and plasmacytoid dendritic cells (pDCs; Extended Data Fig. 3c,d). Unexpectedly, both Irf8 +32^H/+ (H/+) and Irf8 +32^H/H (H/H) mice developed a prominent DC population expressing surface markers of both cDC1 and cDC2, termed hybrid DCs (Fig. 2c,d). Hybrid DCs expressed the TF zinc finger and BTB domain containing 46 (ZBTB46), a cDC lineage-specific TF (Extended Data Fig. 3e). Hybrid DCs could cross-present cell-associated antigen in vitro, resembling cDC1s and unlike cDC2s (Extended Data Fig. 3f,g). To test if differentiated cDC1, cDC2 and hybrid DC subsets can transdifferentiate into each other, we sorted each of these DC subsets from spleens, cultured them in Flt3L or granulocyte–macrophage colony-stimulating factor (GM-CSF) for 1 to 2 days, and then analyzed the cells for XCR1 and Sirpa expression using different colors from sorting (Extended Data Fig. 3h). As expected, sorted cDC populations barely survived for 2 days in culture⁴⁰. During this period, no qualitative change in cell identity was observed, suggesting that no transdifferentiation occurred (Extended Data Fig. 3h).

cDC1 and cDC2 are transcriptionally distinct as they express many genes exclusive to one cell type or the other^19,22. To further characterize the phenotypic identity of hybrid DCs, we compared the transcriptomes of cDC1, cDC2 and hybrid DC subsets by bulk RNA sequencing (RNA-seq). Principal component analysis (PCA) placed hybrid DCs between cDC1s and cDC2s along the first principal component (PC1) for both H/+ and H/H genotypes (Fig. 2e). Notably, hybrid DCs expressed both cDC1-specific and cDC2-specific genes at intermediate levels (Fig. 2f), aligning with their dual expression of XCR1 and Sirpa, the hallmark surface markers of cDC1s and cDC2s, respectively (Fig. 2c). In summary, introducing multiple high-affinity AICEs into the Irf8 +32-kb enhancer gives rise to a novel cDC subset with a composite transcriptional identity intermediate between cDC1 and cDC2.

The optimized enhancer disrupts cDC2 lineage commitment

Further exploration into Irf8 +32^H/H mice revealed major effects of enhancer optimization on cDC2 lineage commitment. We observed a decrease in cDC2 frequencies in both H/+ and H/H mice, suggesting a perturbation in cDC2 differentiation (Fig. 2c,d). DC progenitors, including pre-cDC2s, showed similar IRF8 expression levels in WT and H/H mice (Extended Data Fig. 4a). However, when isolated from bone marrow (BM) and cultured in vitro, WT pre-cDC2s rapidly downregulated IRF8 expression, indicative of their exclusion of cDC1 lineage potential (Extended Data Fig. 4b,c). Conversely, a fraction of H/H pre-cDC2s maintained high IRF8 levels, suggesting an increased susceptibility of the Irf8 +32^H allele to autoactivation (Extended Data Fig. 4b,c). To assess the impact of this lowered autoactivation threshold on cDC lineage determination, we analyzed the in vitro differentiation outcomes of DC progenitors isolated from H/H mice. Compared to WT, H/H CDP produced more cDC1s and fewer cDC2s, aligning with the reduced cDC2 frequencies observed in vivo (Fig. 3a,b). Both H/H and WT pre-cDC1s differentiated exclusively into cDC1s (Fig. 3a,b). Notably, pre-cDC2s from H/H mice produced about 50% cDC1s, in contrast to the near absence of cDC1s from WT pre-cDC2s (Fig. 3a,b). Recently, it has been described that a progenitor population in the spleen, referred to as CD11c^lo transitional DCs (tDC^lo) or pDC-like cells, is another source for cDC2, independent of the CDP–pre-cDC2 pathway^41–43. Similarly to observations with pre-cDC2s, WT tDC^lo exclusively generated cDC2 in culture, whereas H/H tDC^lo generated a substantial cDC1 fraction (Fig. 3c,d). These findings collectively suggest that the Irf8 +32^H enhancer redirects cDC2-committed progenitors toward a cDC1 fate, likely due to enhanced recruitment of IRF8–BATF3–JUNB complexes.

Fig. 3 |. An optimized Irf8 +32-kb enhancer redirects pre-cDC2 to cDC1 and hybrid DC lineages.

a, Representative flow cytometry plot depicting cDCs differentiated from the indicated DC progenitors sorted from WT and H/H BM and cultured in the presence of Flt3L for 4 days (pre-gate: B220⁻SiglecH⁻CD11c⁺M HCII⁺ cells). b, Frequencies of cDC1s and cDC2s differentiated from the indicated DC progenitors. Data were pooled from three independent experiments. Statistical significance was determined by unpaired, two-tailed Student’s t-test. c, Representative flow cytometry plot depicting cDCs differentiated from tDC^lo/pDC-like cells sorted from WT and H/H spleens and cultured in Flt3L + M-CSF for 4 days (pre-gate: B220⁻SiglecH⁻CD11c⁺MHCII⁺ cells). d, Frequencies of cDC1s and cDC2s differentiated from tDC^lo/pDC-like cells. Data were pooled from two independent experiments (n = 4 for WT and H/H mice). Statistical significance was determined by unpaired, two-tailed Student’s t-test. e, Representative flow cytometry plots of BM cells stained for pre-cDC1s, and splenocytes stained for cDCs from WT, H/+, Nfil3^−/− and H/+ Nfil3^−/− mice. Pre-gates: lin⁻SiglecH⁻Flt3⁺ cells for BM, and B220⁻SiglecH⁻CD11c⁺MHCII⁺ cells for spleen. f, Frequencies of pre-cDC1s as percentages of lin⁻SiglecH⁻ BM cells and frequencies of splenic cDC1s across the indicated genotypes. Data were pooled from three (BM) and six (spleen) independent experiments. Statistical significance was determined by one-way ANOVA with Dunnett’s multiple-comparison test. g, Schematics of WT pre-cDC1, WT pre-cDC2 and H/H pre-cDC2 highlighting the relative abundance of the BATF3–JUNB–IRF8 complex and the activation status of the Irf8 +32-kb enhancer. Data in b, d and f are presented as mean values ± s.d. P values are indicated above the graphs.

Interferon regulatory factor 4 (IRF4) is an IRF8 paralog with functional similarities in promoting BATF3-dependent gene expression¹⁹. During the pre-cDC2-to-cDC2 transition, Irf4 is upregulated along with Batf3 (ref. 22; Extended Data Fig. 3a). Therefore, the observed redirection of H/H pre-cDC2 toward a cDC1 fate might be caused by high-affinity AICEs recruiting IRF4 instead of IRF8. To test the role of IRF4 in pre-cDC2 redirection, we generated H/H Irf4^−/− mice and evaluated their cDC development. Hybrid DCs occurred in both H/H and H/H Irf4^−/− mice (Extended Data Fig. 4d,e). Also, the redirection of pre-cDC2 to cDC1 persisted in H/H Irf4^−/− mice. Hence, the aberrant IRF8 autoactivation in H/H pre-cDC2 is driven by low-level coexpression of BATF3 and IRF8 during cDC2 differentiation, independent of IRF4.

Although in vitro cultures of DC progenitors provide valuable insights, they may not fully capture the complexities of in vivo DC development. For example, in our in vitro experiments, a distinct hybrid DC population was not identifiable. Therefore, to test if cDC1s and hybrid DCs could originate from H/H pre-cDC2s in vivo, we turned to two genetic models that are deficient in cDC1 specification^23,44. First, we utilized the knowledge that nuclear factor, interleukin 3 regulated (NFIL3) is required for pre-cDC1 specification and cDC1 development^44,45, by crossing H/H mice onto an Nfil3^−/− background. Similarly to Nfil3^−/− mice with WT Irf8 +32-kb enhancer alleles, H/+ Nfil3^−/− and H/H Nfil3^−/− mice lacked pre-cDC1 in the BM (Fig. 3e, f). Intriguingly, these mice developed cDC1s and hybrid DCs in the spleen, suggesting that the Irf8 +32^H enhancer can be activated in vivo without NFIL3-dependent cDC1 fate specification (Fig. 3e–g).

Furthermore, the Irf8 +41-kb enhancer is integral for pre-cDC1 specification subsequent activation of the +32-kb enhancer^23,46. To test if the Irf8 +32^H enhancer can remain functional in the absence of the +41-kb enhancer, we deleted the Irf8 +41-kb enhancer directly in the Irf8 +32^H allele to derive the Irf8 +32^H +41⁻ allele (Extended Data Fig. 5a,b). Irf8 +32^H/+ +41^−/− and Irf8 +32^H/H +41^−/− mice lacked pre-cDC1 in the BM (Extended Data Fig. 5c,d), yet they developed hybrid DCs in the periphery (Extended Data Fig. 5c,e). In contrast to Irf8 +32^H/H Nfil3^−/− mice, Irf8 +32^H/H +41^−/− mice lacked DC1 and only developed hybrid DCs, suggesting that the Irf8 +41-kb enhancer may contribute to Irf8 expression in terminally differentiated cDC1s (Extended Data Fig. 5c,e). Nonetheless, IRF8 expression was observed in hybrid DCs from Irf8 +32^H/H +41^−/− mice, differing from Irf8 +41^−/− mice, which solely developed cDC2s devoid of IRF8 expression (Extended Data Fig. 5f). This suggests that IRF8 autoactivation can proceed at the Irf8 +32^H enhancer in the absence of the Irf8 +41-kb enhancer. Together, these findings demonstrate that the introduction of high-affinity AICEs within the Irf8 +32-kb enhancer can redirect pre-cDC2 toward cDC1 and hybrid DC phenotypes in vivo, underscoring the importance of enhancer suboptimization for maintaining the fidelity of pre-cDC2 commitment (Fig. 3g).

High-affinity AICEs reduce enhancer activity in cDC1

Having established that the Irf8 +32^H allele initiates ectopic IRF8 autoactivation in the cDC2 developmental pathway, we next examined how high-affinity AICEs affect enhancer activity in the cDC1 lineage. To dissect the roles of AICE affinity and arrangement, we constructed a series of retroviral enhancer reporters with varying numbers and affinities of AICE clusters. These constructs were then introduced to primary cDC1 cultures to assess their activities (Fig. 4a). Increasing the number of low-affinity AICE clusters did not significantly alter enhancer activity (Fig. 4a). Surprisingly, we found that high-affinity AICEs decreased enhancer activity in cDC1 (Fig. 4a). Moreover, increasing the number of high-affinity AICE clusters substantially reduced enhancer activity. This was evident as the ‘high’ construct, designed to replicate the Irf8 +32^H allele, showed diminished reporter activity (Fig. 4a).

Fig. 4 |. High-affinity AICEs paradoxically reduce IRF8 expression in terminally differentiated cDC1s and compromise cDC1 lineage commitment.

a, Retroviral reporter assay assessing activities of various Irf8 +32-kb enhancer mutants in cDC1s obtained from Kit/Flt3L-cultured BM cells. Shown are enhancer-driven GFP expression levels in transduced cDC1s (pre-gate: B220⁻CD11c⁺MHCII⁺CD24⁺Sirpa⁻Thy1.1⁺ cells). The histogram shows data from one representative experiment, and the bar graph shows GFP geometric MFI across four independent experiments. Statistical significance was determined by one-way ANOVA with Dunnett’s multiple-comparison test. b, IRF8 expression in splenic cDC1s, cDC2s and hybrid DCs from WT, H/+ and H/H mice determined by intracellular staining. Numbers are geometric MFI. Data shown are representative of eight similar experiments. c, Representative flow cytometry plots of BM cells stained for pre-cDC2s and splenocytes stained for cDCs from WT, H/+, H/+ Δ1 + 2 + 3, H/H, and H/H Δ1 + 2 + 3 mice. Pre-gates: lin⁻SiglecH⁻Flt3⁺CD11c⁺MHCII⁻ cells for BM, and B220⁻SiglecH⁻CD11c⁺MHCII⁺ cells for spleen. d, Frequencies of pre-cDC2s as percentages of lin⁻SiglecH⁻ BM cells and frequencies of splenic hybrid DCs across the indicated genotypes. Data were pooled from two (BM, n = 6 for WT, 3 for H/+, 3 for H/+ Δ1 + 2 + 3, 6 for H/H, and 3 for H/H Δ1 + 2 + 3 mice) and four (spleen, n = 8 for WT, 6 for H/+, 3 for H/+ D1 + 2 + 3, 4 for H/+ Nfil3^−/−, 6 for H/H, 5 for H/H D1 + 2 + 3 and 5 for H/+ Nfil3^−/− mice) independent experiments. Statistical significance was determined by one-way ANOVA with Dunnett’s multiple-comparison test. e, PCA plot derived from bulk RNA-seq data of sorted splenic DC populations across the indicated genotypes. Data in a and d are presented as mean values ± s.d. P values are indicated above the graphs.

This reduction in enhancer activity was also observed in vivo. Mice heterozygous for the Irf8 +32^H allele demonstrated a reduction in IRF8 expression within cDC1s compared to their WT counterparts (Fig. 4b). This reduction was more pronounced in homozygous mice, where cDC1 IRF8 levels were nearly akin to the intermediate levels expressed in hybrid DCs (Fig. 4b).

We hypothesized that the reduction in IRF8 might prompt some cDC1s in H/H mice to adopt hybrid DC characteristics. To test this possibility, H/H mice were crossed to Zeb2 −165^Δ1+2+3 (Δ1 + 2 + 3) mice, which lack cDC2 development and yield only cDC1s¹⁵. Resembling Δ1 + 2 + 3 mice, H/+ Δ1 + 2 + 3 and H/H Δ1 + 2 + 3 mice lacked pre-cDC2s and cDC2s (Fig. 4c, d). Hybrid DCs were absent in H/+ Δ1 + 2 + 3 mice, suggesting that under heterozygous H/+ conditions, hybrid DCs predominantly originate from pre-cDC2s (Fig. 4c,d). However, hybrid DCs emerged in H/H Δ1 + 2 + 3 mice, suggesting pre-cDC1 as a source for hybrid DCs in homozygous H/H contexts (Fig. 4c,d). Thus, hybrid DCs can develop via both cDC1 and cDC2 pathways.

To determine if hybrid DCs developed from cDC1 and cDC2 routes are transcriptionally similar or distinct, we performed bulk RNA-seq on splenic DC populations across various genotypes, including H/H Nfil3^−/− and H/H Δ1 + 2 + 3, representing our models deficient in pre-cDC1 and pre-cDC2, respectively. PCA revealed that hybrid DCs from all genotypes clustered at an intermediate position between cDC1 and cDC2 along PC1 (Fig. 4e). Heat map analysis corroborated that hybrid DCs from all genotypes expressed both cDC1-specific and cDC2-specific genes at intermediate levels (Extended Data Fig. 6a). Therefore, hybrid DCs originated from either pre-cDC2 or pre-cDC1 pathways display similar transcriptomes, indicative of convergent differentiation (Extended Data Fig. 6b). In addition, we found that cDC1s from H/H Nfil3^−/− mice were transcriptionally similar to cDC1s from NFIL3-proficient genotypes, confirming that the Irf8 +32^H enhancer can bypass the requirement of NFIL3 in authentic cDC1 development (Fig. 3e,f and Extended Data Fig. 6a). Notably, cDC1s from both H/+ and H/H mice exhibited a tendency to downregulate several cDC1-specific genes, including Irf8, while upregulating a subset of cDC2-specific genes (Extended Data Fig. 6a). Collectively, these observations suggest that the enhanced affinities of AICEs within the Irf8 +32-kb enhancer result in decreased enhancer activity within cDC1s, weakening the lineage boundaries between cDC1s and cDC2s.

Analysis of intermediately optimized Irf8 +32-kb enhancers

During the process of generating the Irf8 +32^H allele using CRISPR–Cas9, two unintended alleles, termed Irf8 +32³⁰⁰ and Irf8 +32^I (Int) were also created (Extended Data Fig. 7a). These alleles are characterized by partial disruptions of endogenous low-affinity AICEs and incorporations of two to three high-affinity AICEs, representing intermediate degrees of enhancer optimization (Extended Data Fig. 7a). We also constructed retroviral reporters recapitulating the 300 and Int alleles. When expressed in primary cDC1 cultures, these reporters showed reduced activities compared to the WT reporter, resembling the High reporter but to a lesser extent, which suggests a gradient effect of optimization on enhancer function (Extended Data Fig. 7b). Profiling of splenic DC populations in 300/+ and 300/300 mice revealed the presence of hybrid DCs and slight reductions in cDC2 numbers (Extended Data Fig. 7c,d). Analyses of I/I mice revealed that they develop small but statistically significant numbers of hybrid DCs coexpressing XCR1, Sirpa and CD11b (Extended Data Fig. 7e,f). I/I hybrid DCs express higher levels of IRF8 than cDC2s (Extended Data Fig. 7g). In line with the results from retroviral reporters, cDC1s from I/I mice exhibit reduced IRF8 expression compared to WT cDC1s (Extended Data Fig. 7h). In summary, these findings further support the conclusions that optimizing the Irf8 +32-kb enhancer results in suboptimal levels of IRF8 in terminally differentiated cDC1s, the emergence of a hybrid DC population and, in some cases, compromised cDC2 development.

Single-cell RNA-seq of WT and Irf8 +32^H/H tissues

To evaluate the impact of the Irf8 +32^H enhancer on the transcriptomes of DC progenitors at single-cell resolution, we conducted single-cell RNA sequencing (scRNA-seq) on sorted BM lin⁻Flt3L⁺Kit^int–lo cells from WT and H/H mice. After in silico removal of contaminating lin⁺ cells, the remaining cells clustered into 18 populations (Extended Data Fig. 8a–d). These populations were classified by comparing their transcriptional profiles with published scRNA-seq datasets on DC progenitors (Extended Data Fig. 8a,c)^43,44,47. For example, clusters 11 and 12 correspond to pre-cDC1s expressing Ifi205, Batf3, Rab7b and Id2. Clusters 5, 13 and 16 likely represent pre-cDC2s expressing Klf4, Upb1 and Cd7. Clusters 6, 8 and 10 likely represent CDPs expressing Mpo and Igfbp4. Overall, WT and H/H samples exhibit similar clustering patterns (Extended Data Fig. 8b–d). Notably, H/H samples showed slight increases in Irf8 levels in pre-cDC2s and decreases in and pre-cDC1s (Extended Data Fig. 8e,f). These observations are in line with the finding that the Irf8 +32^H enhancer exhibits ectopic activity in pre-cDC2s and reduced activity in cDC1s.

We also conducted scRNA-seq on sorted splenic CD11c⁺ cells from WT and H/H mice. After in silico removal of lin⁺ cells and cells that do not express Itgax (encoding CD11c), the remaining cells clustered into 11 populations (Fig. 5a). These populations were classified by comparing their transcriptional profiles with bulk RNA-seq datasets generated in this study, as well as published scRNA-seq datasets on DCs (Fig. 5a,b and Extended Data Fig. 8g,h)^42,47. A hybrid DC population (cluster 1) expressing both cDC1 and cDC2 marker genes was present exclusively in H/H samples (Fig. 5b–d and Extended Data Fig. 8g), corroborating our flow cytometry and bulk RNA-seq results. Besides validating the hybrid DC population, we confirmed the reduction of Irf8 expression and the gain of several cDC2 marker genes in H/H cDC1s at single-cell level (Fig. 5d–g). Notably, both our initial clustering analysis and reclustering of cDC1 populations showed that H/H cDC1s occupied distinct uniform manifold approximation and projection (UMAP) spaces compared to WT cDC1s. In summary, these results confirm that the Irf8 +32^H enhancer induces a hybrid DC population and compromises cDC1 identity.

Fig. 5 |. scRNA-seq profiling of WT and Irf8 +32^H/H spleens.

a, Clustering of splenic CD11c⁺ cells from WT and H/H mice projected onto a UMAP space (7,535 total cells, with 3,779 from WT and 3,756 from H/H mice). b, Normalized and scaled expressions of selected cDC1 marker genes (upper) and cDC2 marker genes (lower). c, Distributions of WT and H/H cells across the UMAP in a. d, Violin plot of normalized and scaled Irf8 expression across clusters. WT and H/H cells are displayed side by side for each cluster. e, Normalized and scaled Irf8 expression in WT and H/H cells. f, Projection of cDC1 clusters from a onto a new UMAP space. g, Violin plot of normalized and scaled expressions of selected genes in cDC1s from f. h, Projection of the DC3 cluster from a onto a new UMAP space. i, Violin plot of normalized and scaled expressions of selected genes in DC3s from h. j, Gating strategy for identifying DC3s. k, IRF8 expression in DC3s from WT and H/H mice determined by intracellular staining. Data shown are from two independent experiments (n = 4 for WT and H/H mice). Statistical significance was determined by unpaired, two-tailed Student’s t-test. P values are indicated above the graphs.

Recent studies on cDC2 heterogeneity have characterized a DC3 subset that, besides expressing numerous cDC2 markers, also expresses CD16/CD32 and originates from monocyte and DC progenitor (MDPs) independently from the CDP–pre-cDC2 pathway^43,47. Our scRNA-seq analysis revealed that H/H DC3s occupied distinct UMAP spaces compared to WT DC3s (Fig. 5c,h and Extended Data Fig. 8g), driven by gains of Irf8 and several other cDC1 marker genes (Fig. 5i). The acquisition of IRF8 expression in H/H DC3s was confirmed at the protein level by intracellular staining (Fig. 5j,k). These results further illustrate that optimizing the Irf8 +32-kb enhancer expands the domain of IRF8 expression. This optimization not only sustains IRF8 expression in some cDC2 progenitors, leading to their redirection to cDC1s and hybrid DCs, but also induces IRF8 expression in DC3.

The optimized enhancer compromises T_H2 responses

cDC2s are required for T_H2 responses against helminth infections^13–15. Given that insertion of high-affinity AICEs into the Irf8 +32-kb enhancer diverted a portion of pre-cDC2s away from the cDC2 lineage (Fig. 3), we asked whether this impaired the effectiveness of T_H2 responses against the intestinal helminth H. polygyrus. In the uninfected state, we noted a pronounced reduction in migratory cDC2 numbers within the mesenteric lymph nodes (MLNs) of H/H mice compared to their WT counterparts (Fig. 6a,b). Following H. polygyrus infection, H/H mice produced fewer small intestinal granulomas and had an increased fecal egg burden compared to WT mice (Fig. 6c). In line with these findings, the frequency of T_H2 cells in MLNs of H/H mice was reduced by approximately 50% compared to WT (Fig. 6d,e). These results demonstrate that the Irf8 +32^H enhancer leads to a decreased abundance of cDC2s, compromising immune responses against infections by intestinal helminths.

Fig. 6 |. An optimized Irf8 +32-kb enhancer reduces cDC2 development and impairs T_H2 responses.

a, Representative flow cytometry plots showing MLN migratory cDC populations from WT and H/H mice (pre-gate: B220⁻F4/80⁻CD11 c⁺MHCII^hi cells). b, Frequencies of MLN migratory cDC1s, cDC2s and hybrid DCs from WT and H/H mice. Data were pooled from two independent experiments (n = 4 for WT and H/H mice). Statistical significance was determined by unpaired, two-tailed Student’s t-test. c, Analysis of WT, H/H and Stat6^−/− mice following H. polygyrus infection. The left graph presents the count of granulomas in the small intestine (SI), and the right graph shows the fecal egg burden after 2 weeks. Data were pooled from two independent experiments (n = 8 for WT, 8 for H/H, and 5 for Stat6^−/− mice). Stat6^−/− mice served as negative controls for immune responses to H. polygyrus infection. Statistical significance was determined by one-way ANOVA with Dunnett’s multiple-comparison test. d, Representative flow cytometry plots of MLN cells from H. polygyrus-infected mice stained for T_H2 cells (pre-gate: CD45⁺CD3e⁺CD4⁺ cells). e, Frequencies of T_H2 cells from MLNs of H. polygyrus-infected mice. Data were pooled from two independent experiments (n = 8 for WT, 8 for H/H and 5 for Stat6^−/− mice). Statistical significance was determined by one-way ANOVA with Dunnett’s multiple-comparison test. Data in b, c and e are presented as mean values ± s.d. P values are indicated above the graphs.

The optimized enhancer impairs CD8⁺ T cell responses to tumors

Finally, we tested the impact of optimizing the Irf8 +32-kb enhancer on cDC1-mediated immune responses using the subcutaneous fibrosarcoma 1956-mOVA model⁴⁸. At steady state, skin-draining lymph nodes (SLNs) from H/H mice exhibited the presence of hybrid DCs in both resident and migratory compartments (Fig. 7a,b). In line with observations in the spleen, SLN-resident cDC1s from H/H mice expressed lower levels of IRF8 compared to their WT counterparts (Fig. 7c). Migratory cDC1s showed diminished IRF8 levels, consistent with previous publications^49,50 (Fig. 7c). 1956-mOVA cells are typically rejected by WT mice through cDC1-mediated antitumor CD8⁺ T cell responses but grow progressively in cDC1-deficient Irf8 +32^−/− mice^8,23,48. We inoculated Irf8 +32^H/H mice, along with WT and Irf8 +32^−/− controls, with 1956-mOVA, and monitored subsequent tumor growth. Despite the presence of cDC1s, Irf8 +32^H/H mice showed uncontrolled 1956-mOVA growth comparable to that in Irf8 +32^−/− mice (Fig. 7d). Furthermore, the expansion of tumor-specific CD8⁺ T cells, as identified by SIINFEKL-K^b tetramer staining, was significantly attenuated in Irf8 +32^H/H mice to levels similar to those in naive WT mice (Fig. 7e,f). This suggests that the normal antitumor functionality of cDC1 is compromised in Irf8 +32^H/H mice, potentially due to a hybridization of cDC1 and cDC2 phenotypes.

Fig. 7 |. An optimized Irf8 +32-kb enhancer reduces IRF8 expression in cDC1 and impair antitumor immunity.

a, Representative flow cytometry plots showing SLN-resident and migratory cDC populations from WT and H/H mice. Pre-gates: B220⁻EpCAM^int–negCD11c⁺MHCII^int cells for resident cDCs, and B220⁻EpCAM^int–negCD11c⁺MHCII^hi cells for migratory cDCs. b, Frequencies of SLN-resident and migratory cDC1s, cDC2s and hybrid DCs from WT and H/H mice. Data were pooled from two independent experiments (n = 6 for WT and H/H mice). Statistical significance was determined by unpaired, two-tailed Student’s t-test. c, IRF8 expression in SLN-resident and migratory cDC1s and hybrid DCs from WT and H/H mice determined by intracellular staining. Numbers are geometric MFI. Data shown are representative of three similar experiments. d, Tumor growth curves for WT, Irf8 +32^H/H and Irf8 +32^−/− mice inoculated with 10⁶ 1956-mOVA fibrosarcoma cells. Irf8 +32^−/− mice served as negative controls for antitumor responses. Data were pooled from two independent experiments. e, Representative flow cytometry plots of SLN cells from WT and H/H mice inoculated with 1956-mOVA 6 days earlier, or naive WT mice, stained for CD8⁺ T cells reacting with SIINFEKL-K^b tetramer (pre-gate: B220⁻TCRb⁺CD8a⁺ cells). f, Frequencies of SIINFEKL-K^b tetramer⁺ CD8⁺ T cells in SLNs from the indicated mice. Data were pooled from two independent experiments (n = 5 for WT and 5 for H/H tumor-bearing mice, n = 4 for WT tumor-naive mice). Statistical significance was determined by one-way ANOVA with Dunnett’s multiple-comparison test. Data in b and f are presented as mean values ± s.d. P values are indicated above the graphs.

Discussion

Our findings elucidate that optimal divergence of cDC lineages relies on suboptimization of the Irf8 +32-kb enhancer. This suboptimization achieves two critical yet opposing functions: it prevents unwanted IRF8 autoactivation in developing cDC2 while maximizing IRF8 expression in developed cDC1. These processes enable the clear distinction between the cDC1 and cDC2 lineages, which is crucial for protective responses against distinct immune challenges.

Our study suggests that low-affinity TF binding sites may be an integral feature in autoactivation of lineage-determining TFs. Besides IRF8, many other lineage-determining TFs display ‘on’ or ‘off’ expression patterns controlled by autoactivation, including deformed, myoblast determination protein 1 (MyoD1) and GATA binding protein 3 (GATA3)^1–3,22,51. Beyond regulating their own expression, these lineage-determining TFs often regulate hundreds of target genes to define cell identities¹⁹. These target genes can use a spectrum of binding affinities within their enhancers, facilitating diverse expression levels or spatial–temporal patterns^52,53. Our findings with the Irf8 +32-kb enhancer led us to hypothesize that low-affinity tuning may be a widespread strategy among autoregulatory enhancers of lineage-determining TFs. By avoiding high-affinity binding sites, autoregulatory enhancers can have a good dynamic range in response to differing levels of binding TFs, allowing for clear ‘on’ and ‘off’ states instead of being constantly ‘on’. While previous studies have suggested that low-affinity motifs can act to restrict the spatial and temporal domains of gene expression^{24,25,27–36,54}, our study is unique in showing an in vivo phenotype directly caused by alterations at the endogenous locus of a lineage-determining TF. Future studies could focus on experimentally modifying affinities of autoregulatory enhancers for other lineage-determining TFs and examine the developmental outcomes.

We found that low-affinity AICEs paradoxically maximized IRF8 expression in the intended cDC1 lineage. We initially anticipated that introducing high-affinity AICEs to the Irf8 +32-kb enhancer would increase IRF8 expression in cDC1s, based on previous research reporting that increasing motif affinity typically increases enhancer activity in their native domains^{24,25,27,28,32,33}. However, contrary to our expectation, high-affinity AICEs decreased the activity of the Irf8 +32-kb enhancer. This reduction of enhancer activity not only diminished IRF8 levels in cDC1s but also contributed to the emergence of the hybrid DC lineage, characterized by intermediate expression levels of both cDC1 and cDC2 subset-specific genes. This inverse relationship between motif affinity and enhancer activity was previously observed in a few studies but not mechanistically explained^34,55,56. Thus, our results highlight the necessity for future molecular investigations to decipher the cis-regulatory code for enhancer activity.

Enhancer suboptimization typically involves the use of low-affinity TF binding sites, as well as suboptimal numbers and arrangements of TF binding motifs^{24–26,28,57}. Our results suggest that suboptimization of the Irf8 +32-kb enhancer is crucial for the normal cDC1/cDC2 segregation, as introducing higher-affinity and increased numbers of AICEs in this enhancer resulted in several abnormalities. However, our experiments did not definitively establish whether low affinity is required to prevent ectopic activity of the Irf8 +32-kb enhancer. To directly address this, future studies could generate mouse strains with specific increases in the affinity of AICEs in the Irf8 +32-kb enhancer, without altering the number and arrangement of these motifs. Nevertheless, our current study demonstrates that low affinity is a conserved characteristic of this enhancer, and our reporter assays suggest that low affinity is required for the enhancer’s maximal activity in differentiated cDC1s.

It has been demonstrated in other systems that enhancers regulate the frequency of transcriptional bursts without physically contacting target promoters^58–60. In light of this, a plausible mechanism for the observed reduction of Irf8 levels in Irf8 +32^H/H cDC1s could be that the optimized enhancer sequesters IRF8 and BATF3, along with coactivators, at the high-affinity AICEs. This sequestration may limit their availability to induce transcriptional bursts at the Irf8 promoter, thereby reducing Irf8 expression. While a thorough characterization of the chromatin state using CUT&RUN for acetylated histone H3 Lys27, IRF8 and BATF3 in WT versus Irf8 +32^H/H cDC1s could provide some insights, such experiments are beyond the scope of the current study. Future work will aim to explore these aspects to better understand the complex interplay at the Irf8 locus.

In conclusion, our study uncovers two critical aspects of the Irf8 +32-kb autoregulatory enhancer: its role in establishing a threshold ensuring the fidelity of cDC2 commitment and its capacity to optimize IRF8 expression in cDC1s. These dual functions contribute to the distinct and nonredundant immune roles of the two cDC lineages. This discovery raises an intriguing question about whether similar mechanisms of enhancer suboptimization regulate the autoactivation thresholds of other lineage-determining TFs.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41590-024-01976-w.

Methods

ChIP–seq analysis

ChIP–seq reads in .sra format were downloaded and converted into .fastq format for reanalysis. Reads were aligned to the mouse genome (mm10 assembly) using Bowtie 2 (version 2.2.5), followed by conversion to .bam format and sorted using SAMtools (version 1.3.1) with the command: ‘bowtie2 −5 1 -k 1 --no-unal -p 26 -x < mm10_indexes > -U <unaligned_fastq > | samtools view -bhS - | samtools sort -m 3000000000 -o <aligned_bam > –’. Peak calling was conducted via MACS2 (version 2.2.9.1) with the following command: ‘macs2 callpeak -t <aligned_bam > -c <aligned_input_control_bam > -f BAM --call-summits -q 0.01 -n <peak_files>‘. For motif analysis to generate an AICE PWM, the top 1,000 BATF3 peaks in cDC1 with the lowest P values were selected, extracted 200 bp centered on the summit for each peak and analyzed using MEME (version 5.5.5 MEME Suite). BATF3/IRF8 co-occupancy peaks were identified using BEDtools (version 2.30.0) with the following command: ‘bedtools intersect -a < BATF3_peaks > -b < IRF8_peaks > <co-occupancy_peaks>. The occurrence of AICEs in BATF3/IRF8 co-occupancy peaks was found with FIMO⁶¹ (version 5.5.5 MEME Suite) at a P-value threshold of 1 × 10⁻³. FIMO calculates a log-likelihood score for a given motif with respect to each sequence position and converts the scores to P values using dynamic programming. Peak annotation and Manhattan plot generation were performed in R, utilizing the ChIPseeker and ggplot2 packages.

EMSA

Oligonucleotide sequences used for EMSA are provided in Supplementary Table 1. Complementary oligonucleotide pairs were annealed to generate probes that were subsequently labeled with ³²P-dCTP using Klenow (exo-) polymerase. 293FT cells were either left untreated or co-transfected with retroviral vectors for Batf3 and Junb, or single-transfected with retroviral vectors for Irf8 using TransIT-LT1, 2 days before harvesting for NE preparation. 293FT cells and MutuDCs were lysed with buffer A (10 mM HEPES-KOH, pH 7.9, 1.5 mM MgCl₂ and 10 mM KCl) containing 0.2% NP40 and protease inhibitors. The nuclei were then pelleted, resuspended in buffer C (20 mM HEPES-KOH, pH 7.9, 420 mM NaCl, 1.5 mM MgCl₂, 0.2 mM EDTA and 25% glycerol) containing protease inhibitors, and centrifuged to obtain NEs. Protein concentration of NEs was determined using the Bradford assay.

EMSA was performed essentially as previously described³⁸. Specifically, 0.0125 pmol of labeled probes were mixed with approximately 6 μg of NEs in EMSA buffer (20 mM HEPES-KOH, pH 7.9, 50 mM KCl, 1 mM EDTA, 2 mM dithiothreitol, 2 mM MgCl₂, 0.25 mg ml⁻¹ BSA, 5% glycerol and 50 μg ml⁻¹ poly(dI-dC)). For competition with unlabeled oligonucleotides, 200-fold excess (2.5 pmol) of unlabeled oligonucleotides was added to the reaction unless otherwise noted. The results were analyzed by autoradiography.

Microarray analysis

Microarray expression raw files (.cel format) were downloaded and subsequently reanalyzed using the Transcriptome Analysis Console (TAC) Software (version 4.0.3, Applied Biosystems). Probe annotation, probe summarization, quality-control analysis and statistical tests for differential expression were all conducted according to the official guidelines provided by Applied Biosystems.

Mice

C57BL/6 (referred to as WT) and C57BL/6-Tg(TcraTcrb)1100Mb/J (OT-I) mice were purchased from The Jackson Laboratory. Irf8 +32^−/− and Zeb2 −165^Δ1+2+3 mice were generated in-house and reported previously^15,23. Nfil3^−/− mice were originally provided by A. Look and T. Mak⁶². Irf4^−/− mice were generated by crossing Irf4^f/f mice first to CMV-Cre mice and then to CMV-Flp1 mice, as reported previously^19,63. All mice were maintained on the C57BL/6J background in specific pathogen-free facilities according to institutional guideline protocols approved by the Animal Studies Committee of Washington University in St. Louis. All mice were maintained on 12-h light cycles and housed at 21 °C and 50% humidity. Mouse chow was from LabDiet (53WU Irradiated PicoLab Rodent Diet 20). Mice aged 6–14 weeks olds that were age and sex matched were used for all experiments.

Antibodies and flow cytometry

Gating strategies used in the current study are summarized in Supplementary Fig. 1, and antibodies are listed in Supplementary Table 3. Flow cytometry was performed using a FACSAria Fusion instrument (BD) and an Aurora (Cytek) instrument. Flow cytometry data were collected using BD FacsDiva software version 8.0 and SpectroFlo software version 2.2.0, with subsequent data analysis using FlowJo version 10. The FACSAria Fusion instrument was also utilized for cell sorting. Surface staining was performed at 4 °C in the presence of Fc block (2.4G2) in magnetic-activated cell-sorting buffer (PBS + 0.5% BSA + 2 mM EDTA). For panels involving CD16/CD32, FITC-conjugated anti-CD16/CD32 (2.4G2, BD) was used instead of Fc block at a dilution of 1:200. Intracellular staining for IRF8, Foxp3 and GATA3 was performed using the Foxp3 staining kit (00-5523-00; eBioscience). For SIINFEKL-K^b tetramer staining, 2 × 10⁶ Fc-blocked cells were incubated with fluorochrome-conjugated tetramers at a dilution of 1:100 each at 37 °C for 30 min. The tetramer-stained cells were then proceeded to surface staining at 4 °C without washing.

Retroviral reporter assay

Sequences of enhancer variants used in retroviral reporter assays are cataloged in Supplementary Table 2, with all variant sequences validated through Sanger sequencing, provided by GENEWIZ. The retroviral reporter vector Thy1.1 pA GFP CMVp RV (referred to as ‘empty’ vector) was described previously²³. Irf8 +32-kb enhancer reporters (5′ and full-length, both containing all four AICEs) were described previously²³. The 5′ version was used in experiments shown in Fig. 1, while the full-length version was used in subsequent figures. New enhancer variants were generated by PCR or ordered from Twist Bioscience. Some inserts were further processed by digesting with HindIII and BglII or HindIII and BamHI to produce compatible overhangs for ligation into HindIII-digested and BamHI-digested empty vector with T4 DNA ligase. Alternatively, other variants were cloned into HindIII-digested and BamHI-digested empty vector by HiFi assembly.

KitL/Flt3L culture and retroviral transduction of BM cells were performed as previously described⁶⁴. Retroviruses were produced by transfecting retroviral vectors into Plat-E cells using TransIT-LT1 and collecting viral supernatants 2 days later. BM cells, isolated from WT mice and erythrocyte-depleted via ammonium chloride-potassium bicarbonate (ACK) lysis, were filtered through a 70-μm nylon mesh. The cells were then cultured in Iscove’s Modified Dulbecco’s Medium (IMDM) supplemented with 10% FBS, 1% penicillin–streptomycin solution, 1% sodium pyruvate, 1% MEM non-essential amino acid, 1% l-glutamine solution and 55 μM β-mercaptoethanol (I10F) with 5% KitL-conditioned medium. Following overnight incubation, BM cells were then transduced with viral supernatant by harvesting from transfected Plat-E cells. The spin infection was executed at 1,800g for 1 h in the presence of 2 μg ml⁻¹ polybrene. Culture conditions were switched to Flt3L-supplemented medium on day 3 of culture and maintained for an additional 6 days before the cells were harvested and analyzed by flow cytometry.

Generation of Irf8 +32^H/H, Irf8 +32^I/I, Irf8 +32^300/300 and Irf8+32^H/H +41^−/− mice

Sequences of the guide RNAs (gRNAs)²³, the ssODN containing high-affinity AICEs and PCR primers for mouse genotyping are provided in Supplementary Table 4. gRNAs (ordered in the form of sgRNA), ssODN and Cas9 nuclease were ordered from IDT.

For generating Irf8 +32^H/H mice, Irf8 +32 mid sgRNA and S-BXC-SXC ssODN were conjugated with Cas9 to form ribonucleoprotein (RNP) complexes. Day 0.5 single-cell zygotes from WT mice were isolated and RNP complexes were introduced via electroporation by the Department of Pathology & Immunology Transgenic Mouse Core at Washington University in St. Louis. Around 60 single-cell zygotes were electroporated with 8 μM RNP using a 1-mm gap cuvette (Bio-Rad). Electroporated zygotes were then transferred into the oviducts of day 0.5 pseudo-pregnant recipient mice. The resulting pups were screened by PCR with ‘founder_f’ and ‘founder_r’ primers and validated by Sanger sequencing through GENEWIZ to confirm integration of the high-affinity AICE motifs. Mice with the desired integrations were outcrossed to WT mice, and the resulting heterozygous Irf8 +32^H/+ mice were intercrossed to yield homozygous Irf8 +32^H/H mice. In instances of unintended genetic modifications, specifically those referred to as ‘Int’, the mice were similarly outcrossed to WT mice. These mice were then bred to homozygosity, resulting in Irf8 +32^I/I mice. The process for generating Irf8 +32^300/300 mice involved Irf8 +32 mid sgRNA and C-B ssODN, similarly to the above methodology.

Irf8 +32^H/H +41^−/− mice were generated using a similar strategy. sgRNAs Irf8 +41 5′ and Irf8 +41 3′ were complexed with Cas9 nuclease. The two RNP complexes were then combined to mitigate biased sgRNA incorporation. The RNP complexes were then electroporated into Irf8 +32^H/H zygotes. The resulting pups were screened by PCR using ‘41_f’ and ‘41_r’ primers followed by Sanger sequencing to identify those that had the desired deletion. Mice with the correct deletion were then crossed to Irf8 +41^−/− mice. Irf8 +32^H/+ +41^−/− offspring were interbred to generate homozygous Irf8 +32^H/H +41^−/− mice.

DC preparation

Spleen, MLNs and SLNs were minced and digested in 5 ml of I10F with 250 μg ml⁻¹ of collagenase B (Roche) and 30 U ml⁻¹ of DNase I (Sigma) for 30 min at 37 °C with stirring. After digestion, red blood cells were lysed with ACK lysis buffer and then the remaining cells were passed through 70-mm nylon filters to create single-cell suspensions.

Cross-presentation assay

In vitro cross-presentations were performed as described previously⁶⁵. Briefly, 10,000 sorted DCs and 25,000 CellTrace Violet-labeled OT-I T cells were co-cultured with various concentrations of HKLM-OVA (a gift from H. Shen, University of Pennsylvania) for 3 days. OT-I cell division was assessed by analyzing CellTrace Violet dilution via flow cytometry.

Bulk RNA-seq experiment and analysis

Splenic cDC1s, cDC2s and hybrid DCs were sort purified after CD11c enrichment (CD11c MicroBeads UltraPure, mouse; Miltenyi Biotec). Total RNA was extracted using NucleoSpin RNA XS (Macherey-Nagel). Library preparation, Illumina sequencing, read alignment and transcript compilation were conducted by the Genome Technology Access Center at the McDonnell Genome Institute (GTAC@MGI) at Washington University School of Medicine as follows. Total RNA integrity was determined using Agilent Bioanalyzer or 4200 Tapestation. Library preparation was performed with 10 ng of total RNA with an RNA integrity number greater than 9.0. Double-stranded complementary DNA (cDNA) was prepared using the SMARTer Ultra Low RNA kit for Illumina Sequencing (Takara-Clontech) per the manufacturer’s protocol. cDNA was fragmented using a Covaris E220 sonicator using a peak incident power of 18, a duty factor of 20% and 50 cycles per burst for 120 s. cDNA was blunt ended, had an A base added to the 3′ ends and then had Illumina sequencing adaptors ligated to the ends. Ligated fragments were then amplified for 12–15 cycles using primers incorporating unique dual-index tags. Fragments were sequenced on an Illumina NovaSeq 6000 using paired-end reads extending 150 bases. RNA-seq reads were then aligned and quantified to the Ensembl release 101 primary assembly with an Illumina DRAGEN Bio-IT on-premise server running version 3.9.3–8 software.

Raw gene counts were imported to R using the EdgeR package for downstream analyses. After filtering out lowly expressed genes, counts were normalized using the trimmed mean of M-values method for differential expression analysis, identifying differentially expressed genes with greater than fourfold changes and false discovery rate below 0.001. For PCA and heat map construction, counts were normalized by the counts per million method and transformed to log₂. PCA plots were generated using the ggplot2 package, and heat maps were generated using the pheatmap package. Data from two sequencing experiments involving DCs from different genotypes for Fig. 4e and Extended Data Fig. 6a were integrated using Combat-seq⁶⁶.

Isolation and culture of DC progenitors

BM progenitors were isolated as previously described^15,23. BM cells were isolated and depleted of lin (CD3e, CD19, B220, CD105, CD127, Ly-6G and TER-119)-expressing cells by staining with the corresponding biotinylated antibodies, followed by depletion with Mojosort Streptavidin Nanobeads (BioLegend). Subsequently, the lin⁻ BM cells were then stained with fluorochrome-conjugated antibodies for sorting. As illustrated in Supplementary Fig. 1, CDPs were characterized as lin ⁻Siglec-H⁻CD117^intCD135⁺CD115⁺CD11c⁻MHCII⁻, pre-cDC1s as lin⁻Siglec-H⁻CD135⁺CD226⁺CD117^int and pre-cDC2s as lin⁻Siglec-H⁻CD135⁺CD11c⁺ MHCII⁻CD115⁺CD117⁻. After sorting, progenitors were cultured in I10F supplemented with 5% Flt3L conditioned medium. Unless specified otherwise, progenitors were maintained in culture for 4 days before being analyzed for lineage outputs by flow cytometry. For experiments with tDC^lo/pDC-like cells, splenocytes were isolated and depleted of lin (CD3e, CD11b, CD19, NK1.1, Ly-6G)-expressing cells. The remaining lin⁻ splenocytes were stained with fluorochrome-conjugated antibodies for sorting. tDC^lo/pDC-like cells were characterized as lin⁻Flt3⁺XCR 1⁻CX3CR1⁺CD11c⁺Ly-6C⁺ cells. After sorting, the progenitor cells were cultured in I10F supplemented with 5% Flt3L conditioned medium and 20 ng ml⁻¹ M-CSF (PeproTech) for 4 days.

scRNA-seq experiment and analysis

BM and spleens were harvested from two WT and two Irf8 +32^H/H mice. The BM samples were depleted of lin⁺ cells expressing CD3e, CD8b, CD11b, CD19, B220, ly-6G, TER-119 and NK1.1. Each sample was stained with fluorochrome-conjugated antibodies and labeled with a unique Antibody Capture TotalSeq B antibody (BioLegend) before sorting for lin⁻Flt3⁺Kit^int–lo cells. Two samples from each genotype were pooled and resuspended in PBS with 0.04% BSA at a final concentration of approximately 1,000 cells per μl. The spleen samples were enriched for CD11c-expressing cells, similarly stained with fluorochrome-conjugated antibodies and labeled with a unique Antibody Capture TotalSeq B antibody (BioLegend), before further purification by sorting CD11c⁺ cells. All four samples were pooled and resuspended in PBS with 0.04% BSA at approximately 1,000 cells per μl. These were loaded on a 10× Genomics Chromium Single Cell Controller at GTAC@MGI at Washington University School of Medicine. Library preparation was performed using the 10× Genomics Next GEM Single Cell 3′ Reagents Kit v3. Libraries were sequenced using an Illumina NovaSeq 6000 sequencer, producing 150-bp paired-end reads. Sequencing targeted depths of 1 billion reads for the gene expression (GEX) library and 100 million reads for the hashtag oligonucleotide library. Cell Ranger v8.0.0 was used for aligning reads to the mm10 reference genome, single-cell counting, barcode processing and sample demultiplexing.

scRNA-seq downstream analysis was conducted using Seurat v5.0.3 in R. After removing unwanted cells, unique molecular identifier counts in each cell were normalized using log transformation with a scale factor of 10,000. For each dataset, 2,000 highly variable genes were identified and used for PCA. For the BM dataset, cell cycle phase scores were calculated, and the difference between G2M and S phase scores was regressed out. Dimensional reduction and clustering were performed using the FindNeighbors, RunUMAP and FindClusters functions, utilizing the top 30 PCs. The clustering resolution was set at 0.6 for the BM dataset and 0.5 for the spleen dataset.

Helminth preparation and infection

H. polygyrus third-stage larvae (L3) were propagated as previously described⁶⁷. H. polygyrus L3 viability was checked by microscope for motility, and their numbers were quantified before use. Mice were gavaged with 100 H. polygyrus L3 using 20-gauge × 38-mm plastic feeding tubes (Instech). Fourteen days after inoculation, fecal samples were collected from individual mice to enumerate egg burden with the saturated sodium chloride salt flotation method using McMaster counting chamber.

Tumor growth experiments

The immunogenic fibrosarcoma expressing membrane ovalbumin (1956-mOVA) was described previously⁴⁸ and used for all tumor experiments described herein. 1956-mOVA fibrosarcoma cells derived from frozen stocks were propagated for 2 days in RPMI medium supplemented with 10% FBS, washed three times with PBS and resuspended at 1 × 10⁷ cells per ml in PBS. Tumor cells (10⁶; 100 μl) were injected subcutaneously into the flanks of each mouse. Tumor growth was monitored for 16 days before the mice were euthanized. Tumor growth was measured with a caliper, and tumor area was calculated by the multiplication of two perpendicular diameters. All experimental procedures adhered to Institutional Animal Care and Use Committee guidelines, ensuring that the maximum tumor burden did not exceed 20 mm in any single dimension, as stipulated by the approved protocol.

Statistical analysis

Statistical analysis was performed using GraphPad Prism software (version 10). For determining significant differences between two groups, unpaired, two-tailed Student’s t-test was used. For determining significant differences across multiple groups, ordinary one-way ANOVA with Dunnett’s multiple-comparisons test was used. Differences with P values ≥ 0.05 were considered not significant. All data are presented as means ± s.d.

Extended Data

Extended Data Fig. 1 |. The Irf8 + 32 kb comprises four AICE motifs.

(a) ATAC-seq and ChIP-seq tracks for the indicated cell types are shown for the Irf8 locus. The locations of the Irf8 + 32 kb, +41 kb and +56 kb enhancers are highlighted. The tracks are displayed using the UCSC genome browser with vertical auto scale. (b) A zoomed-in view of a 1,168 bp region surrounding the Irf8 + 32 kb enhancer highlighting the locations of AICEs. The previously deleted 149 bp (blue) and 421 bp (red) regions from Irf8 + 32 5′^−/− mice and Irf8 + 32^−/− mice, respectively²³, are indicated relative to the AICEs. (c) Sequence of the 421 bp region deleted previously in Irf8 + 32^−/− mice²³, with AICEs underlined and highlighted in blue. The AICEs are displayed from left to right in the order of site 1, site 4, site 3, and site 2.

Extended Data Fig. 2 |. AICEs in mouse and human IRF8 enhancers show low binding affinities to the BATF3-JUNB-IRF8 complexes.

(a) Competitive EMSA showing binding of ³²P-dCTP-labeled Ctla4 probe with the indicated unlabeled competitors at increasing amounts (0, 25, 100, 400, and 1600⁻fold excess). The NEs used were combined from 293FT cells transfected with Junb and Batf3 and 293FT cells transfected with Irf8. Bands corresponding JUNB-BATF3-IRF8 complexes and free probes are indicated by the upper and lower arrows, respectively. Data shown are representative of two similar experiments. (b) EMSA performed with ³²P-dCTP-labeled probes on various NEs: NEs from 293FT cells transfected with Junb and Batf3 (B), combined NEs from 293FT cells transfected with Junb and Batf3 and 293FT cells transfected with IRF8 (B + I), and NEs from MutuDCs. Bands corresponding JUNB-BATF3-IRF8 and JUNB-BATF3 complexes are indicated by the upper and lower arrows, respectively. Data shown are representative of three similar experiments. (c) Alignment of a 161 bp region encompassing the four AICEs in the mouse Irf8 + 32 kb enhancer (chr8:120,768,539–120,768,699, mm10) across several mammalian species, colored by conservation (Jalview). The six most probable AICEs (defined using FIMO and the AICE PWM from Fig. 1b) are indicated. (d) Retroviral reporter analysis of the mouse Irf8 + 32 kb enhancer and the human IRF8 + 48 kb enhancer in cDC1s and cDC2s obtained from KitL/Flt3L cultured mouse BM cells. Shown are enhancer-driven GFP expression in transduced cDC1s (pre-gate: B220⁻ CD11c⁺ MHCII⁺ CD24⁺ Sirpa⁻ Thy1.1⁺ cells) and transduced cDC2s (pre-gate: B220⁻ CD11c⁺ MHCII⁺ Sirpa⁺ Thy1.1⁺ cells). Numbers are GFP geometric MFI. Data shown are representative of two similar experiments. (e) Competitive EMSA performed with ³²P-dCTP-labeled Ctla4 probe in the presence of 200-fold excess of various unlabeled competitors. Competitors derived from the human +48 kb enhancer are denoted as ‘Hu’. Bands corresponding JUNB-BATF3-IRF8 complexes and free probes are indicated by the upper and lower arrows, respectively. Data shown are representative of two similar experiments.

Extended Data Fig. 3 |. Characterization of DC progenitors and splenic DC populations.

(a) Expression levels of selected TF-encoding genes in different stages of DC development assessed by microarray. (b) IRF8 Expression across different stages of DC development in WT mice as determined by intracellular staining. Numbers are geometric MFI. Data shown are representative of two similar experiments. MDP, monocyte-dendritic cell progenitors. (c) Frequencies of the indicated DC progenitors from WT, H/+, and H/H mice as percentages of lin⁻ SiglecH⁻ BM cells. Data are pooled from three independent experiments (n = 5 for WT, 3 for H/+, and 4 for H/H mice). Statistical significance was determined by one-way ANOVA. (d) Frequencies of splenic pDCs (gated as B220⁻ SiglecH⁺ splenocytes) from WT, H/+, and H/H mice. Data are pooled from three independent experiments (n = 9 for WT, 5 for H/+, and 7 for H/H mice). Statistical significance was determined by one-way ANOVA. (e) ZBTB46 expression in the indicated cell types from H/H spleens measured by intracellular staining. B cells serve as a negative control for ZBTB46 expression. (f) In vitro cross-presentation assay with different DC populations. Shown are representative flow cytometry plots depicting CellTrace Violet-labeled OT-I CD8 T cells co-cultured with the indicated DC subsets and HKLM-OVA antigen for three days (pre-gate: CD45.2⁻ CD45.1⁺ CD8a⁺ CD4⁻ Va2⁺ cells). Divided OT-I cells indicate successful cross-presentation and T cell activation. (g) Frequencies of divided OT-I cells from in vitro cross-presentation assays. Data are pooled from two independent experiments. (h) Overlaid flow cytometry plots showing splenic cDC1s (cyan), cDC2s (red), and hybrid DCs (purple) before (top panel) and after (middle and bottom panels) culturing in the specified conditions. The data presented are representative of three similar experiments. Data in c and d are presented as mean values +/− SD.

Extended Data Fig. 4 |. An optimized Irf8 + 32 kb enhancer maintains IRF8 expression in pre-cDC2, impeding their lineage commitment to cDC2.

(a) IRF8 Expression in the indicated DC progenitors from WT and Irf8 + 32^H/H (H/H) mice as determined by intracellular staining. Numbers are geometric MFI. Data shown are representative of two similar experiments. (b) Representative flow cytometry plots depicting intracellular IRF8 in WT and H/H pre-cDC2s after 18 hours of culturing in the presence of Flt3L. (c) Frequencies of IRF8^hi pre-cDC2s after 18 h of culturing. Data are pooled from three independent experiments (n = 5 for WT and H/H mice). Statistical significance was determined by unpaired, two-tailed Student’s t test. (d) Representative flow cytometry plots showing splenic cDC populations from WT, Irf4^−/−, H/H, and H/H Irf4^−/− mice (pre-gate: B220⁻ SiglecH⁻ CD11c⁺ MHCII⁺ splenocytes). (e) Frequencies of splenic cDC2s and hybrid DCs from the indicated mice. Data are pooled from five independent experiments. Statistical significance was evaluated using one-way ANOVA with Dunnett’s multiple comparisons test. (f) Representative flow cytometry plot depicting cDCs differentiated from the indicated DC progenitors sorted from WT and H/H Irf4^−/− BM and cultured in the presence of Flt3L for four days (pre-gate: B220⁻ SiglecH⁻ CD11c⁺ MHCII⁺ cells). (g) Frequencies of cDC1s differentiated from WT and H/H Irf4^−/− pre-cDC2s. Data are pooled from three independent experiments. Statistical significance was determined by unpaired, two-tailed Student’s t test. Data in c, e, and g are presented as mean values +/− SD. P values are indicated above the graphs.

Extended Data Fig. 5 |. The Irf8 + 32^H enhancer drives hybrid DC development independently of the Irf8 + 41 kb enhancer.

(a) Diagram illustrating the deletion of the Irf8 + 41 kb enhancer on the Irf8 + 32^H/H background using CRISPR/Cas9. (b) Sequence of a 658 bp region surrounding the Irf8 + 41 kb enhancer. Nucleotides deleted in both the Irf8 + 41^−/− line²³ and the Irf8 + 32^H/H + 41^−/− line are highlighted in red. Additional nucleotides deleted only in the Irf8 + 32^H/H + 41^−/− line highlighted in blue. sgRNAs target sites are underlined. (c) Representative flow cytometry plots of BM cells stained for pre-cDC1 (upper panel), and splenocytes stained for cDCs (lower panel) from WT, Irf8 + 41^−/− (+41^−/−), Irf8 + 32^H/+ + 41^−/− (+32^H/+ + 41^−/−), and Irf8 + 32^H/H + 41^−/− (+32^H/H + 41^−/−) mice. Chromatin phasing diagrams are displayed above the plots for context. Pre-gates: lin⁻ SiglecH⁻ Flt3⁺ cells for BM, and B220⁻ SiglecH⁻ CD11c⁺ MHCII⁺ cells for spleen. (d) Frequencies of pre-cDC1s as a percentage of lin⁻ SiglecH⁻ BM cells across the indicated genotypes. Data are pooled from three independent experiments (n = 4 for WT, 5 for +41^−/−, 5 for +32^H/+ + 41^−/−, and 4 for +32^H/H + 41^−/− mice). Statistical significance was determined by one-way ANOVA with Dunnett’s multiple comparison test. (e) Frequencies of splenic cDC1s, cDC2s and hybrid DCs across the indicated genotypes. Data are pooled from six independent experiments. Statistical significance was determined by one-way ANOVA with Dunnett’s multiple comparison test. (f) IRF8 expression in cDC2s and hybrid DCs from the indicated genotypes as determined by intracellular staining. Data shown are representative of two similar experiments. Data in d and e are presented as mean values +/− SD. P values are indicated above the graphs.

Extended Data Fig. 6 |. An optimized Irf8 + 32 kb enhancer drives convergent differentiation of pre-cDC1 and pre-cDC2 toward a hybrid DC phenotype.

(a) Left panel: heatmap visualization of the bulk RNA-seq data as in Fig. 4e. Shown are the top 464 genes differentially expressed between WT cDC1 and cDC2 identified in Fig. 2f. Right panel: zoomed-in views of selected clusters. (b) Schematics illustrating the differentiation of CDP into cDC1, cDC2, and hybrid DC in WT, H/+, and H/H mice. Cell types are color-coded based on IRF8 and BATF3 expression.

Extended Data Fig. 7 |. Analyses of intermediately optimized Irf8 + 32 kb enhancers.

(a) Diagrams and sequences of the different Irf8 + 32 kb enhancer alleles generated in this study. (b) Retroviral reporter activities of various Irf8 + 32 kb enhancer constructs in cDC1s obtained from Kit/Flt3L-cultured BM. Shown are enhancer-driven GFP expression in transduced cDC1s (pre-gate: B220⁻ CD11c⁺ MHCII⁺ CD24⁺ Sirpa⁻ Thy1.1⁺ cells). Numbers are GFP geometric MFI. Data shown are pooled from two independent experiments (n = 3 for WT mice). Statistical significance was determined by one-way ANOVA with Dunnett’s multiple comparison test. (c) Representative flow cytometry plots of splenocytes stained for cDCs (pre-gate: B220⁻ SiglecH⁻ CD11c⁺ MHCII⁺ splenocytes). (d) Frequencies of the indicated DC subsets as percentages of B220⁻ SiglecH⁻ CD11c⁺ MHCII⁺ cDCs. Data are pooled from six independent experiments. Statistical significance was evaluated using one-way ANOVA with Dunnett’s multiple comparisons test. (e) Representative flow cytometry plots of CD11c-enriched splenocytes stained for cDCs (pre-gate: B220⁻ CD11c⁺ MHCII⁺ splenocytes). (f) Frequencies of the indicated DC subsets as percentages of B220⁻ CD11c⁺ MHCII⁺ cDCs. Data are pooled from two independent experiments (n = 4 for WT and I/I mice). Statistical significance was determined by unpaired, two-tailed Student’s t test. (g) Representative histograms showing intracellular IRF8 levels in DC subsets gated in (e). Numbers are geometric MFI. (h) IRF8 geometric MFI in cDC1s gated in (e). Statistical significance was determined by unpaired, two-tailed Student’s t test. Data shown are pooled from two independent experiments (n = 4 for WT and I/I mice). Data in b, d, f, and h are presented as mean values +/− SD. P values are indicated above the graphs.

Extended Data Fig. 8 |. scRNA-seq analysis of WT and Irf8 + 32^H/H BM and spleen.

(a) Clustering of BM lin⁻ Flt3⁺ Kit^int-lo cells from WT and H/H mice projected onto a UMAP space (22,839 total cells, with 11,441 from WT and 11,398 from H/H mice). (b) Distributions of WT (blue) and H/H (red) cells across the UMAP in (a). (c) Dot plot of top 10 differentially expressed genes for each cluster in (a). (d) Clustering of BM lin⁻ Flt3⁺ Kit^int-lo cells in (a) split by individual biological replicates. (e) Normalized and scaled Irf8 expression in WT (upper panel) and H/H (lower panel) cells. (f) Violin plot of normalized and scaled Irf8 expression across clusters. WT (blue) and H/H (red) cells are displayed side by side for each cluster. Statistical significance was determined by a two-tailed Wilcoxon rank sum test, applying a log₂ fold-change threshold of 0.1. P values are indicated above the graphs for statistically significant comparisons. (g) Clustering of splenic CD11c⁺ cells from Fig. 5a split by individual biological replicates. (h) Normalized and scaled expressions of selected cluster marker genes in splenic CD11c⁺ cells.

Data availability

The bulk RNA-seq data generated in this study on splenic DC populations are available in the Gene Expression Omnibus under accession number GSE241341. The scRNA-seq data generated in this study are available under accession number GSE270060. The following datasets were previously published and reanalyzed in this study: microarrays on DC progenitors and splenic DCs (GSE66565)²², and ChIP–seq for BATF3 and IRF8 in cDC1 (GSE66899)²². The following datasets were reanalyzed previously¹⁹ and used in the current study: ChIP–seq for p300, H3K4me1 and H3K27ac in DC progenitors and DCs (GSE66899)²², the assay for transposase-accessible chromatin with sequencing (ATAC-seq) of DC progenitors (GSE132240)²³, and Immunological Genome Project (ImmGen) ATAC-seq on splenic DCs (GSE100738)⁶⁸. Mouse genome (mm10 assembly) Bowtie 2 indexes were downloaded from the Bowtie 2 website (https://genome-idx.s3.amazonaws.com/bt/mm10.zip). Source data are provided with this paper. All other data are available in the paper or Supplementary Information.