Batf3-Dependent Genes Control Tumor Rejection Induced by Dendritic Cells Independently of Cross-Presentation

Conventional dendritic cells (cDC) develop as two major lineages that are dependent on distinct transcriptional programs for their development (1). The cDC1 lineage is dependent on the transcription factors IRF8 and BATF3 for development and expresses certain unique markers such as CD8α, CD103, and XCR1 in various tissues (1). BATF3-dependent cDC1s are specialized for antigen cross-presentation and are required for antiviral and antitumor CD8⁺ T-cell responses (2–4). In particular, BATF3-dependent cDC1s are required for rejection of immunogenic syngeneic fibrosarcomas (2), and type I interferon signaling supports this capacity (5, 6). BATF3-depdendent cDC1s are also required for T-cell priming in response to DNA vaccines (7) and their abundance in humans tumors correlated with improved tumor regression (8, 9). The ability of checkpoint blockade to mediate antitumor responses against progressively growing sarcomas was shown to also require BATF3-dependent cDC1s (10–12). However, despite the clearly important role of cDC1 cells in antitumor immunity, much remains unclear about how they survey tissues and initiate immune responses.

IRF8 and BATF3 play distinct roles in cDC1 development, which proceeds through distinct specification and commitment stages (13). IRF8 is required for development of a cDC1-specified bone marrow (BM) progenitor, which is missing in Irf8^−/− mice. In contrast, this specified progenitor develops in Batf3^−/− mice, but fails to commit to mature cDC1s and instead diverts to the cDC2 lineage due to the inability to sustain the normal high Irf8 expression (13). In this cDC1-specified progenitor, BATF3 and IRF8 cooperate in binding to an enhancer containing several AP-1/IRF consensus elements (AICE; ref. 14) that functions to sustain Irf8 autoactivation initiated earlier in development (13). Thus, at least one function of BATF3 is exerted in the commitment stage of cDC1 development.

Because there is currently no system for conditional deletion of Batf3, it has been difficult to determine whether BATF3 also functions in cDC1 cells during immune responses. The normal requirement for Batf3 for cDC1 development can be bypassed under some conditions, although the mechanism is not completely understood. Splenic CD8α⁺ cDC1 were restored in Batf3^−/− mice during infection by Mycobacteria tuberculosis (15) or by administration of IL12, and this restoration was blocked by in vivo neutralization by IFNγ (15). Molecularly, Batf can compensate for Batf3 in cDC1 development (15), and cDC1 development induced by IL12 in Batf3^−/− mice was reduced in Batf^−/− Batf3^−/− mice, suggesting some role for Batf in the mechanism of restoration (15). cDC1 development is transiently restored after transfer of Batf3^−/− BM into irradiated recipients (16), although the basis for this effect is unclear. In addition, Batf3^−/− mice on the C57BL/6 genetic background frequently retain a population of cells resembling cDC1 in skin draining lymph nodes, but not in the spleen or other tissues (15). The basis for this strain- and tissue-specific phenomenon has not been established, but may be specific to microbiota, as it is not observed in all colonies of C57BL/6 Batf3^−/− mice.

Another system for BATF3 compensation is based on the particular molecular mechanism of cDC1 development (13). Crossing a transgenic Irf8^VENUS reporter strain (17) with Batf3^−/− mice also restores cDC1 development (13). In this case, cDC1 development appears to result from the increased IRF8 expression arising due to the three intact copies of IRF8 present in the BAC reporter transgenes, leading to autonomous Irf8 autoactivation (13). In the present study, we used this Irf8^VENUS reporter system to allow an examination of cDC1 cells that develop and maintain IRF8 expression in the absence of BATF3 for their capacity to function in cross-presentation and tumor rejection, and to identify transcriptional gene targets requiring BATF3.

Wild-type and Batf3^−/− mice on C57BL/6 background were crossed to IRF8^VENUS (17) as described (13). Experiments using Batf3^−/− Batf^−/− mice used a mixed 129S6/SvEV/C57Bl/6 background (18). C57BL/6-Tg(TcrαTcrβ)1100Mjb/J (OT-1) mice were purchased from The Jackson Laboratories. MHCI KO mice (K^b−/−D^b−/−β2m^−/−; TKO) were a gift from Herbert W. Virgin and Ted Hansen, Washington University, St. Louis (19). Mice harboring a conditional allele of Itga8 (Itga8^F/F) were obtained from The Jackson Laboratories as B6.129S6-Itga8tm1.1Rdav/J and crossed to CD11c-Cre (20). All mice were maintained in a specific pathogen-free animal facility following institutional guidelines and with protocols approved by the Animal Studies Committee at Washington University in St. Louis, an Institutional Animal Care and Use Committee. Experiments were performed with sex-matched mice 6–20 weeks of age without randomization or blinding.

Cells were stained at 4°C in the presence of Fc Block (2.4G2; Bio X Cell) in flow cytometry buffer (0.5% BSA/PBS). Antibodies to the following proteins were used: from Becton Dickinson (BD): MHCII- IA/IE (M5/114 15.2, CD4 (RM4-5), and CD11b (M1/70); from BioLegend: XCR1 (ZET), CD24 (M1/65), CD326 (G8.8), TCR-Vα2 (B20.1), and B220 (RA3-6B2); from Tonbo Biosciences: CD11c (N418) and CD45.1 (A20); from eBioscience: CD172a, CD45 (30F-11), and CD44 (IM7). Cells were analyzed on a FACSCanto II or FACSAria Fusion flow cytometer (BD) and data were analyzed with FlowJo software (TreeStar). Unstained cells or genetic controls that were missing populations of interest were used as negative controls for antibody staining.

Minced spleens, skin draining lymph nodes (pooled inguinal, cervical, and brachial), and tumors, harvested 9–10 days after transplantation, were digested in collagenase B (0.25 mg/mL) and DNAse1 (30 u/mL) in complete IMDM (Iscove's modified Dulbecco's medium with10% FCS, 2ME, penicillin/streptomycin, NEAA, and glutamine) for 40 minutes at 37°C with stirring and subjected to ACK lysis. Prior to sorting, spleen cells were enriched for CD11c⁺ cells (Miltenyi). Ears were split into dorsal and ventral halves prior to mincing and digestion in Liberase (0.26 u/mL) and DNAse1 (30 u/mL) in complete IMDM for 40 minutes at 37°C with stirring. After addition of 5 mmol/L EDTA, suspensions were filtered through 70-μm nylon mesh, pelleted, subjected to ACK lysis, washed, and used for flow cytometry analysis.

cDC1 (XCR1⁺ CD24⁺ CD172a⁻ MHCII⁺ CD11c⁺ B220⁻) or (CD24⁺ CD172a⁻ MHCII⁺ CD11c⁺ B220⁻) were sorted from spleens, and migratory (MHCII^hi CD11c^int) and resident (MHCII^int CD11c^hi) cDC1s (CD326^lowCD24⁺ CD172a⁻ B220⁻) were sorted from skin draining lymph nodes. Total RNA from spleen DCs was extracted using RNAqueous-Micro Kit (Ambion), amplified with the Ovation Pico WTA System (NuGEN) and hybridized to GeneChip Mouse Gene 2.0 ST microarrays (Affymetrix). Total RNA from lymph node DCs was extracted using Nucleo-spin RNA XS kit (Macherey-Nagel), amplified with GeneChip WT Pico Kit (Applied Biosystems) and hybridized to GeneChip Mouse Gene 1.0 ST microarrays (Affymetrix). Data were normalized by robust multiarray average summarization and quartile normalization with ArrayStar software (DNASTAR).

Gene-expression microarray data have been deposited in the Gene-Expression Omnibus (accession no. GSE111034).

The MCA-induced fibrosarcoma 1969 was a gift from Robert Schreiber, Washington University School of Medicine, in 2014. It was generated in a female C57BL/6 Rag2^−/− mouse, was tested for mycoplasma, and was banked at low passage as previously described (5, 21). Tumor cells derived from frozen stocks were propagated for 1 week with one intervening passage in vitro in RPMI media supplemented with 10% FCS (HyClone), were washed three times with PBS, resuspended at a density of 6.67 × 10⁶ cells/mL in endotoxin-free PBS, and then 150 μL was injected subcutaneously into the flanks of recipient mice. Cells were not reauthenticated in the past year. Tumor growth was measured with a caliper and expressed as the average of two perpendicular diameters. An immunogenic fibrosarcoma expressing membrane ovalbumin was generated from the MCA-induced progressor fibrosarcoma 1956, also a gift from Robert Schreiber in 2014. An mOVA fragment pCI-neo-mOVA (Addgene #25099) was ligated into MSCV-IRES-Thy1.1 vector (22) to generate MSCV-mOVA-IRES-Thy1.1. 1956 tumor cells retrovirally transduced with this vector were sorted for expression of Thy1.1, and surface OVA expression was validated using flow cytometry (Millipore AB1225).

Cross-presentation assays were as described (23, 24). Briefly, MHCI-deficient splenocytes were ACK-lysed, loaded with ovalbumin (Worthington) in hypertonic medium (0.5 M sucrose, 10% w/v polyethylene glycol 1,000, 10-mm Hepes, RPMI 1640 pH 7.2) and irradiated (13.5Gy). OT-1 T cells (CD4⁻ B220⁻ CD11c⁻ CD8a⁺ Vα2⁺) were sorted from ACK-lysed, B220-macs depleted OT-1 splenocytes, and labeled with CFSE. For in vivo cross-presentation, mice were injected intravenously (i.v.) with 500,000 CFSE-labeled OT-1 T cells and on the next day with PBS or varying numbers of ovalbumin-loaded MHCI-deficient splenocytes. Three days later, spleens and inguinal lymph nodes were harvested for analysis of CFSE dilution by flow cytometry. For in vitro cross-presentation, cDC1 (CD24⁺ CD172a⁻ CD11c⁺ MHCII⁺ B220⁻) were sorted from splenocytes that had been positively enriched for CD11c⁺ cells (Miltenyi). Sorted cDC1 were placed in 96-well round bottom plates with ovalbumin-loaded MHCI-deficient splenocytes or heat killed Listeria monocytogenes expressing ovalbumin HKLM-ova (24) and CFSE-labeled OT-1 T cells in Iscove's MEM at 37°C in a CO₂ incubator. Alternatively, migratory and resident DCs were sorted from mice 4 or 7 days after subcutaneous implantation of 1956-ova tumor cells and were placed in 96-well round bottom plates with CFSE-labeled OT-1 T cells. CFSE dilution was analyzed on day 3.

All statistical analyses were performed using Prism (GraphPad Software). One-way analysis of variance (ANOVA) was used to compare means of population percentages between mice using Sidak or Holm–Sidak multiple comparisons test with a cutoff of 0.05.

Transgenic Irf8^VENUS reporter mice possess three cointegrated copies of a phage artificial chromosome containing a 130-kb Irf8 genomic region with an internal ribosome entry site and sequence encoding the yellow fluorescent protein VENUS into the Irf8 3′ untranslated region (17). Thus, mice with one transgenic reporter allele have a total of five functional Irf8 loci, resulting in increased IRF8 expression that was approximately 2-fold higher in Irf8^VENUS+ cDC1 cells compared with Irf8^VENUS− cDC1 cells (13). Despite this slight IRF8 overexpression, expression of the Venus reporter is similar to endogenous Irf8 expression in several ways, such as being highly expressed in cDC1 and pDC lineages, and being reduced in cDC2 cells to the low expression typical of Irf8^VENUS− cDC2 (13).

We have previously shown that Batf3^−/− Irf8^VENUS+ mice had normal cDC1 development in contrast to Batf3^−/− Irf8^VENUS− mice, which have impaired cDC1 development (13). Conceivably, this effect of compensation could be dependent on endogenous expression of Batf in DCs, in the manner of IL12-dependent compensation observed previously (15). To address this, we asked whether restoration of cDC1in Batf3^−/− Irf8^VENUS+ mice was due to compensation by endogenous Batf (Fig. 1). In Batf^−/− Batf3^−/− Irf8^VENUS− mice that lack both Batf and Batf3, CD24⁺ CD172a⁻ cDC1s were reduced by 94% in spleen compared with wild-type (WT) Irf8^VENUS− mice (Fig. 1A). This cDC1 population was restored by introduction of the Irf8^VENUS transgene in Batf^−/− Batf3^−/− Irf8^VENUS+ mice to 60% of that in WT mice (Fig. 1A and B). Similarly, using XCR1 instead of CD24 to identify the cDC1 population (25), cDC1s were restored in spleens of both Batf3^−/− Irf8^VENUS+ mice and doubly deficient Batf^−/− Batf3^−/− Irf8^VENUS+, and percentages of these were also reduced by 40% compared with WT (Fig. 1C and D). These results indicate that cDC1 restoration in Batf3^−/− Irf8^VENUS+ mice bypasses dependence on either Batf3 or Batf, unlike restoration induced by IL12 observed previously (15). Further, these results suggest that in spleen cDC1s, the expression of XCR1 is not completely dependent on Batf3 or Batf.

We analyzed the ability of Batf3^−/− cDC1s restored by the Irf8^VENUS transgene to cross-present cell-associated antigen in vitro (Fig. 2). cDC1 purified from WT spleens, that were either Irf8^VENUS− or Irf8^VENUS+, were able to cross-present to OT-1 T cells (Fig. 2A and B), as expected. cDC1 purified from spleens of Batf3^−/− Irf8^VENUS+ (Fig. 2A and B) and Batf^−/− Batf3^−/− Irf8^VENUS+ mice (Fig. 2C and D) cross-presented cell-associated ovalbumin as efficiently as WT cDC1 in vitro. As a control, cDC2 cells purified from either WT or Batf^−/− Batf3^−/− Irf8^VENUS+ mice were unable to cross-present cell-associated antigen (Fig. 2C and D), as expected. We next compared the efficacy of cross-presentation using a dose titration of HKLM-ova (Supplementary Fig. S1). cDC1s purified from spleens of Batf3^−/− Irf8^VENUS+cDC1s were able to cross-present HKLM-ova to OT-1 T cells as well as those purified from WT spleens that were either Irf8^VENUS− or Irf8^VENUS+ (Supplementary Fig. S1A). As a control, cDC2s did not cross-present HKLM-ova, as expected (Supplementary Fig. S1B). Thus, cDC1 cells that express IRF8 but lack Batf3 and Batf proteins are capable of in vitro cross-presentation.

We have previously shown that cDC1s restored by IL12 treatment of Batf3^−/− mice were capable of cross-presentation in vivo (15). We next tested whether Batf3^−/− Irf8^VENUS+ mice could cross-present cell-associated antigen in vivo. Transferred OT-1 CD8 T cells proliferated in spleens and inguinal lymph nodes of WT mice, either with or without Irf8^VENUS (Fig. 2E and F). No OT-1 proliferation was observed in spleens or lymph nodes of Batf3^−/− mice without Irf8^VENUS, as expected. However, OT-1 proliferation was observed in spleens and lymph nodes of Batf3^−/− Irf8^VENUS+ and was equivalent to that in WT mice. These results show that cDC1s do not require Batf3 for in vivo cross-presentation function. We reported that cross-presentation of cell-associated antigens by monocyte-derived DCs (mo-DC) is also independent of Batf3 and Batf (26). Similarly, cross-presentation in cDC1s is not dependent on Batf3 itself, although the development of the cDC1 lineage is.

Previously, we showed that transplanted immunogenic fibrosarcomas cannot be rejected by Batf3^−/− mice, but that IL12-treated Batf3^−/− mice reacquire a population of cDC1 cells and can also mount antitumor responses (2, 15). However, using IL12-induced cDC1 restoration, it is possible that IL12 might independently augment tumor rejection by actions on targets cells other than cDC1s in Batf3^−/− mice. Thus, there was a need to evaluate the intrinsic capacity of the Batf3^−/− cDC1s to mediate tumor rejection in another system. Therefore, we next asked whether Batf3^−/− Irf8^VENUS+ mice could also reject tumors, using a C57BL/6 regressor fibrosarcoma, 1969, that is rejected by WT mice (refs. 5, 21; Fig. 3).

WT mice were able to reject the 1969 fibrosarcoma independently of being Irf8^VENUS− or Irf8^VENUS+ (Fig. 3A). By contrast, tumors grew progressively in Batf3^−/− Irf8^VENUS− mice, as expected (Fig. 3A). However, tumors also grew progressively in Batf3^−/− Irf8^VENUS+ mice (Fig. 3A), despite the presence of cDC1 cells capable of cross-presentation (Fig. 1). Tumors were of similar sizes in WT and Batf3^−/− mice, for both Irf8^VENUS− and Irf8^VENUS+ mice, 7 to 8 days after implantation. However, on day 18, tumors in WT mice were rejected completely, whereas large tumors persisted in all Batf3^−/− mice, both Batf3^−/− Irf8^VENUS− and Batf3^−/− Irf8^VENUS+ genotypes (Fig. 3B).

We also examined the infiltration of 1969 tumors by DCs and T cells in these mice. There was an infiltration of MHCII⁺ CD11c⁺ cells into tumors of all mice, and tumors in WT Irf8^VENUS+ mice contained XCR1⁺ cDC1s and CD8α⁺ T cells (Fig 3C–E), as expected. By contrast, progressively growing tumors in both Batf3^−/− Irf8^VENUS− and Batf3^−/− Irf8^VENUS+ mice lacked XCR1⁺ cDCs and CD8α⁺ T cells. The lack of CD8α⁺ T cells in these tumors is consistent with the lack of tumor rejection in these mice. Thus, despite restoration of cDC1 development and cross-presentation in Batf3^−/− Irf8^VENUS+ mice, these Batf3^−/− Irf8^VENUS+ cDC1s are insufficient for tumor rejection.

Conceivably, cDC1 might mediate tumor rejection by priming T cells, though cross-presentation of tumor antigens delivered to lymph nodes through lymphatics. Alternately, rejection might require cDC1 to acquire antigens directly from tumors and traffic to lymph nodes to prime T cells. We therefore analyzed the cDC populations in dermis of non–tumor-bearing mice (Fig. 4A). First, dermis of both Irf8^VENUS− Batf3^+/+ (WT) and Irf8^VENUS+ Batf3^+/+ mice contained the two major DC populations, the CD11b⁺ cDC2 and XCR1⁺ CD24⁺ CD11b⁻ cDC1s (Fig. 4B). In addition, dermal DCs from Batf3^−/− Irf8^VENUS− mice included CD11b⁺ DC2, but not XCR1⁺ CD24⁺ CD11b⁻ cDC1s. However, dermal DCs from Batf3^−/− Irf8^VENUS+ mice included the CD11b⁺ cDC2 and CD24⁺ CD11b⁻ cDC1s, but unexpectedly these CD24⁺ CD11b⁻ DCs did not express XCR1 (Fig. 4C).

Next, we analyzed resident and migratory cDC1 populations in skin draining lymph nodes (SDLN; Fig. 5). The resident gate, identified as CD11c^hi MHCII^int, contained CD24⁺ CD172a⁻ cDC1s in WT mice, both Irf8^VENUS−and Irf8^VENUS+. These cDC1s also highly expressed XCR1 (Fig. 5A and B). The migratory DC gate, identified as CD11c^int MHCII^hi, also contained CD24⁺ CD172a⁻ cDC1s in WT mice, both Irf8^VENUS−and Irf8^VENUS+. However, XCR1 expression on these migratory cDC1s was lower than XCR1 expression on resident cDC1 cells, consistent with increased maturation in migratory compared with resident DCs (ref. 27; Fig. 5A and C).

We have previously observed that SDLN from Batf3^−/− mice on the C57BL/6 background can retain a CD24⁺ CD172a⁻ population resembling cDC1s (28). Their development has been attributed to compensation by Batf (15). As expected, Batf3^−/− Irf8^VENUS− mice retain this residual cDC1 population in the resident DC gate of the SDLN, and have lower XCR1 expression and increased CD172a expression compared with WT CD24⁺ CD172a⁻ resident cDCs (Fig. 5B). In contrast, in Batf3^−/− Irf8^VENUS+ mice, the CD24⁺ CD172a⁻ cDC1 population has similar CD24 and CD172a expression compared with WT mice (Fig. 5B), but lower expression of XCR1, being expressed on 50% of the CD24⁺ CD172a⁻ cDC1s.

The residual CD24⁺ CD172a⁻ cDC1s are much less abundant in the migratory gate of SDLNs from Batf3^−/− Irf8^VENUS− mice and here they do not express XCR1 at all (Fig. 5C). Migratory CD24⁺ CD172a⁻ cDC1s were restored in Batf3^−/− Irf8^VENUS+ mice, but these also did not express XCR1. Thus, in contrast to spleen, the Irf8-Venus transgene does not fully restore XCR1 expression by cDC1 in the dermis and SDLN in the absence of Batf3. These results may suggest that there may be parallel pathways for inducing XCR1 expression, one of which is Batf3-dependent, and another that is dependent on a signal present in the spleen and LN, but lacking in the dermis. This conditional regulation of XCR1 could be similar to our previous demonstration of the expression of CD103, which can be both Batf3-dependent and induced by GM-CSF in the absence of Batf3 (28).

We wanted to test whether cDC1s from skin and from the migratory gate of SDLN of Batf3^−/− Irf8^VENUS+ mice are able to cross-present cell-associated antigen. As expected, migratory DCs from naïve mice were already matured and had lost the capacity to cross-present newly acquired antigen to OT-1 T cells (Supplementary Fig. S2A; refs. 27, 29). Not surprisingly, our isolation procedure for skin dendritic cells resulted in their maturation, so we could not directly assay their cross-presentation capacity (30). Therefore, to answer this question, we implanted a fibrosarcoma expressing membrane-anchored ovalbumin, 1956-ova, and asked whether migratory cDCs purified from SDLN could stimulate proliferation of OT-1 T cells in vitro (Supplementary Fig. S2B). Because we have shown previously that cDC1 and not cDC2 are uniquely capable of cross-presentation (Supplementary Fig. S1) (24), we did not separate cDC1 from cDC2 in this experiment. Migratory, but not resident, cDCs from 1956 ova tumor-bearing WT Irf8^VENUS− and WT Irf8^VENUS+ mice stimulated robust OT-1 proliferation. Migratory cDCs from Batf3^−/− Irf8^VENUS+ mice also stimulated OT-1 proliferation at approximately 40% of WT, but significantly more than their counterparts from the resident gate (Supplementary Fig. S2B). We conclude that migratory cDC1s from SDLN in Batf3^−/− Irf8^VENUS+ can cross-present antigen. However, we cannot be sure that these migratory cDC1s acquired antigen while they resided in the skin, and it remains a possibility that migration of skin cDC1s or their antigen processing may be altered.

To determine Batf3-dependent genes that may contribute to tumor rejection, we used microarray analysis to compare cDC1s purified from spleens of WT Irf8^VENUS−, WT Irf8^VENUS+, and Batf3^−/− Irf8^VENUS+ mice (Fig. 6). First, increased IRF8 provided by the Irf8^VENUS transgene in WT cDC1s did not cause an overall increase in gene expression for cDC1-associated genes (Fig. 6A). Irf8 expression was increased 1.6-fold in Irf8^VENUS+ compared with Irf8^VENUS− cDC1s, presumably a result of the Irf8 transgenes, but other cDC1-associated genes, such as Btla, Itgae, CD8a, Clec9a, and Xcr1, were not affected by Irf8^VENUS. Batf was not induced by Irf8^VENUS, which indicates that compensation by Batf is unlikely to be the mechanism by which Irf8^VENUS restores cDC1 in Batf3^−/− mice.

To determine which cDC1-associated genes were dependent on Batf3 for their expression, we compared cDC1s from WT Irf8^VENUS+ mice and Batf3^−/− Irf8^VENUS+ mice (Fig. 6; Supplementary Fig. S3). Approximately 10 genes had decreased expression in Batf3^−/− Irf8^VENUS+ cDC1 from spleen compared with WT Irf8^VENUS+ cDC1 (Fig. 6C). Itga8 had the highest fold change (10.4×) between WT Irf8^VENUS+ and Batf3^−/− Irf8^VENUS+ cDC1s. Other Batf3-dependent genes with 3-fold or greater changes between Batf3^+/+ Irf8^VENUS+ and Batf3^−/− Irf8^VENUS+ cells included the signaling adaptor molecule Clnk, the phosphatase Ppef2 and Gcsam, among others. We found no genes that were increased by the Irf8^VENUS transgene in WT cells, with the exception of Irf8 itself. Among a list of genes important for development or function in DCs, including Id2, Ciita, Nfil3, Zeb2, and Irf4, none were Batf3-dependent (Supplementary Fig. S3A). Compared with WT Irf8^VENUS+ cDC1s, Batf3^−/− Irf8^VENUS+ cDC1 had higher expression of Ccr7 and Ccr5, which are known to be involved in DC migration to lymph nodes (31, 32), but no difference in expression of other chemokine receptors, including Xcr1 (Supplementary Fig. S3B). Itga8 was the only integrin for which expression was Batf3-dependent (Supplementary Fig. S3C).

We had observed that XCR1 expression was reduced in Batf3^−/− Irf8^VENUS+ migratory cDC1 from SDLN compared with WT mice (Fig. 5C), so we investigated whether other genes were also reduced in the lymph node cDC1s. We analyzed cDC1s from SDLNs of WT Irf8^VENUS+ mice and Batf3^−/− Irf8^VENUS+ mice by microarray (Fig. 6C; Supplementary Fig. S3D). All but one of the spleen Batf3-dependent genes, Ctla2b, were also Batf3-dependent in resident cDC1s from SDLN (Fig. 6C). These genes were also reduced in migratory, i.e., mature, cDC1s compared with resident cDC1s from WT Irf8^VENUS+ mice and were not further reduced in Batf3^−/− Irf8^VENUS+ migratory cDC1s. A total of 34 genes, including 8 listed for spleen, were reduced in Batf3^−/− Irf8^VENUS+ resident cDC1s compared with WT Irf8^VENUS+ resident cDC1s (Supplementary Fig. S3D). Twenty-eight of these genes are among the approximately 900 genes that are 3-fold or greater reduced in migratory cDC1s compared with resident cDC1s from WT Irf8^VENUS+ mice. Fifteen genes that maintained expression in WT Irf8^VENUS+ migratory cDC1s were reduced in Batf3^−/− Irf8^VENUS+ compared with WT Irf8^VENUS+ migratory cDC1s. Xcr1 is 9.1-fold reduced in migratory, compared with resident, cDC1s from WT Irf8^VENUS+mice, and stands out as being the only gene that is Batf3-dependent in both resident and migratory cDC1s. In summary, the most Batf3-dependent genes are also genes that are reduced upon the maturation of cDC1 and so are more expressed in resident cDC1 compared with migratory cDC1.

We analyzed IRF8 and BATF3 ChIP-seq binding in genomic loci of Batf3-dependent genes (Fig. 7). Shown are some examples, Clnk, Gcsam, Itga8, and Xcr1, whose expression is specific to cDC1 compared with cDC2, is unaffected by the presence of the Irf8^VENUS transgene in WT mice, and is Batf3-dependent (Fig. 7A). For each of these genes, we could find strong BATF3 and IRF8 binding located in open chromatin as assessed by peaks of H3K27Ac, either in the gene body (Gcsam) or nearby (Clnk, Itga8, and Xcr1; Fig. 7B), suggesting these genes may be direct targets of Batf3 and Irf8. To begin to address which of the Batf3-dependent genes might be required for tumor rejection, we implanted 1969 fibrosarcoma cells into mice with conditional deletion of Itga8 (Supplementary Fig. S4). As controls, WT mice could reject 1969 tumor cells (Supplementary Fig. S4A) but Batf3^−/− mice could not (Supplementary Fig. S4B). However, the conditional loss of Itga8 in CD11c⁺ cells was not sufficient to abrogate tumor rejection (Supplementary Fig. S4C). Additional analysis will be required to determine which components of the Batf3-dependent pathway are required for tumor rejection.

Our aim was to determine which cDC1-specific genes rely on Batf3 for their expression and whether any of these are required for tumor rejection mediated by cDC1. The requirement for Batf3 in maintaining Irf8 expression during cDC1 development has prevented a direct determination of Batf3 target genes in cDC1, because cDC1 normally fail to develop in Batf3^−/− mice. In this study, we developed a system to allow cDC1 development in the absence of Batf3 and identified a relatively small number of cDC1 genes rely on Batf3 for their expression. This set of Batf3-target genes includes genes required for cDC1 to mediate tumor rejection.

One Batf3 transcriptional target gene is Irf8. The high IRF8 expression that is required for cDC1 development is achieved by Irf8 autoactivation that depends on the cooperative interaction between IRF8 and BATF3. This interaction is likely mediated through an Irf8 +32 kb enhancer that contains multiple AICEs where BATF3 and IRF8 cooperatively bind (13). We took advantage of the phenomenon that the Irf8-Venus reporter can maintain cDC1 development in the absence of BATF3 and BATF to discover whether the function of cDC1 is entirely a property of high IRF8 expression or whether other genes in mature cDC1 that may require BATF3 for their expression might contribute to cDC1 function.

First, we found that the cDC1 that develop in Batf3^−/− Irf8^VENUS+ mice were capable of cross-presentation of cell-associated antigen, a process that is specific to the cDC1 lineage (24). Thus, the transcriptional program for cross-presentation of cell-associated antigens may rely solely on high Irf8 expression, and not directly on Batf3-dependent gene expression. However, we found that Batf3^−/− Irf8^VENUS+ mice, like Batf3^−/− mice, cannot reject immunogenic fibrosarcomas that are normally rejected by WT mice, despite functional cross-presentation in vivo. This result is in contrast to the case in which cDC1 development was restored in Batf3-deficient mice by IL12 treatment, in which cDC1 restoration led to restored resistance to Toxoplasma gondii, restoration of cross-presenting cDC1 and to capacity for tumor rejection (15). Thus, IL12 treatment may provide additional or different signals that may not be present in the Batf3^−/− Irf8^VENUS+ mice that compensates for the absence of Batf3 in cDC1 for these functions.

All of the identified candidate genes for which expression requires Batf3 are expressed more highly in cDC1 compared with cDC2 cells. Such genes may be required for tumor rejection but would not be predicted to be required for cDC1 development or cross-presentation. For example, Itga8 as a heterodimer with Itgb1, α8β1, has several reported ligands including osteopontin, fibronectin, vitronectin, nephoronectin, tenascin C, and MFGE8 (33–36). Although Itga8 expression is not limited to the immune system, its expression on cDC1s could facilitate interactions with the tumor microenvironment (25, 37). However, a single gene defect in Itga8 expression did not block tumor rejection, but this may be due to redundancy with other partners of Itgb1, which will require future studies.

In addition to cDC1s, Gcsam and its human counterpart HGAL are also highly expressed in germinal center B cells. Gcsam is a cytoplasmic protein that contains a putative immune tyrosine activation motif. Gcsam's biological role is unknown, but it has been shown that Gcsam is dispensable for the germinal center reaction (38). Another Batf3-dependent candidate is Clnk, a member of the Blnk/SLP-76 adapter family. In addition to cDC1s, Clnk is expressed in activated T cells, IL2-activated NK cells, and mast cells (39). No immune defects were observed in Clnk-deficient mice, but this was attributed to potential compensation by SLP-76. cDC1-specific functions for these and other BATF3-target genes have not been evaluated in the context of tumor rejection.

XCR1 has been recognized as a robust marker of cDC1 (40). In our study, Xcr1 was not initially seen to be dependent on Batf3 in splenic cDC1, but was more Batf3-dependent in cDC1s from SDLNs. Xcr1 expression was also diminished upon cDC1 maturation in SDLNs, being reduced by 9-fold in migratory cDC1s relative to resident cDC1s. These observations could imply that the Xcr1 gene is regulated differentially in different tissues. XCL1 produced by CD8 T cells can act as a chemoattractant for XCR1⁺ cDC1s (41). It is possible that lack of XCR1 on migratory cDC1s could help explain lack of tumor rejection in the Batf3^−/− Irf8^VENUS+ mice. In a vaccinia infection model, Xcr1 was required for the clustering of CD8⁺ T cells with cDC1s (42). Another study showed that NK cell release of XCL1 in tumors may recruit cDC1s into tumors, but stated that loss of Xcr1 was not sufficient to block intratumor cDC1 accumulation (43). However, as far as we are aware, neither these studies nor others (44, 45) reported whether loss of XCR1 abrogated tumor rejection, so future studies will be required to test this.

No potential conflicts of interest were disclosed.

Conception and design: D.J. Theisen, S.T. Ferris, N. Kretzer, T.L. Murphy

Development of methodology: D.J. Theisen, N. Kretzer, T.L. Murphy

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.J. Theisen, S.T. Ferris, C.G. Briseño, A. Iwata, T.L. Murphy

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.J. Theisen, T.L. Murphy

Writing, review, and/or revision of the manuscript: S.T. Ferris, K.M. Murphy, T.L. Murphy

Study supervision: T.L. Murphy

This work was supported by the Howard Hughes Medical Institute (K.M. Murphy). We thank the Alvin J. Siteman Cancer Center at Washington University School of Medicine for use of the Center for Biomedical Informatics and Multiplex Gene Analysis Genechip Core Facility. We thank Robert Schreiber for the 1969 and 1956 fibrosarcoma lines.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.