During antigen ‘cross-presentation’ (1), antigens generated in one cell are presented by MHC class I molecules of a second cell. It remains unclear whether all antigen presenting cells (APCs) use cross-presentation and whether this pathway plays a role in immune responses in vivo (2). Dendritic cells (DCs) are a heterogeneous group of APCs with two major subsets, plasmacytoid dendritic cells (pDCs) and conventional CD11c⁺ dendritic cells (cDCs) (3). Subsets of cDCs include CD8α⁺, CD4⁺, and CD8α^-CD4^- populations that may exert distinct functions in immune responses. Evidence has suggested that CD8α⁺ cDCs are important for cross-presentation during infections, but is based on ex vivo analysis (4-6) or in vitro antigen loading (7). Evidence both for and against a role for cross-presentation in responses against tumors has been reported (8-10).

Attempts have been made to study the in vivo role of dendritic cells by selective depletion. Diphtheria toxin treatment can deplete all CD11c^hi cells in one transgenic mouse model (11), but affects splenic macrophages and activated CD8⁺ T cells (12). Gene targeting of transcription factors (e.g.,Irf2,Irf4,Irf8,Stat3 andId2) has caused broad defects in several DC subsets, T cells and macrophages (13). To identify genes regulating DC development, we performed global gene expression analysis across many tissues and immune cells (fig S1A).Batf3 (p21SNFT) (14) was highly expressed in cDCs, with low to absent expression in other immune cells and non-immune tissues. Thus, we generatedBatf3^-/- mice lacking expression of the Batf3 protein (fig. S1B-D).

In spleens ofBatf3^-/- mice we found a selective loss of CD8α⁺ cDCs, without abnormalities in other hematopoietic cell types or architecture (Fig. 1,fig. S2-S11). CD8α⁺ cDC coexpress DEC205, CD24, and low levels of CD11b (3,15).Batf3^-/- mice lacked splenic CD11c^hi CD8α⁺ DEC205⁺ cells (Fig. 1A), showed a loss of CD11c^hi CD11b^dull cells and CD11c^hi CD8α⁺ CD24⁺ cells (Fig. 1B), but had normal populations of CD4⁺ and CD8α^-CD4^- cDC subsets (Fig. 1B). Lymph nodes and thymi ofBatf3^-/- mice lacked CD8α⁺ DCs but had normal distributions of CD8α^- CD11c⁺ cells (Fig. 1C). DEC205^int and DEC205^hi DCs were present in lymph nodes draining the skin ofBatf3^-/- mice (Fig. 1C), and showed normal migration from skin to lymph node after topical application of fluorescein-5-isothiocyanate (fig. S3A).Batf3^-/- mice had normal development of pDCs (CD11c^int CD11b^- B220⁺) (fig. S3B), interstitial DCs of pancreatic islets (CD11c⁺ CD8α^-) (fig. S3C, D), monocytes, neutrophils (fig. S3E) and SIGN-R1⁺ and MOMA-1⁺ marginal zone macrophages (Fig. 2A). CD8α⁺ cDCs developed normally in heterozygousBatf3^+/- mice (fig. S4A), and were absent inRag2^-/-Batf3^-/- mice (fig. S4B).

This loss of CD8α⁺ cDCs could result from a cell-autonomous hematopoietic defect or a cell-extrinsic requirement forBatf3. To distinguish these possibilities, we generated chimeras in which CD45.2⁺Batf3^+/+ or CD45.2⁺Batf3^-/- bone marrow (BM) was transplanted into lethally irradiated CD45.1⁺CD45.2⁺ recipients (Fig. 2B). Upon reconstitution (fig. S5A), we found CD8α⁺ cDCs developed only fromBatf3^+/+ donor BM cell (Fig. 2B), indicating a cell-intrinsic hematopoietic defect inBatf3^-/- mice.

Treatment of mice with fms-like tyrosine kinase 3 (flt3) ligand-Fc (FL-Fc) expanded CD8α⁺ cDCs, CD8α^- cDCs and pDCs inBatf3^+/+ mice, but failed to expand CD8α⁺ cDC inBatf3^-/- mice (Fig. 2C). In vitro culture of BM with FL generates cell populations corresponding to pDCs (CD11c⁺CD45RA⁺) and cDCs (CD11c⁺CD45RA^-) (3,16) (Fig. 2D). These in vitro-derived cDCs do not express CD8α or CD4, but contain a CD24⁺Sirp-α^lo-int population corresponding to CD8α⁺ cDC (16).Batf3^+/+ orBatf3^-/- BM cells treated with FL produced similar ratios of pDCs and cDCs (Fig. 2D andfig. S5B). However,Batf3^-/- BM generated far fewer CD24⁺ Sirp-α^- cells compared toBatf3^+/+ BM (Fig. 2D), corresponding to loss of CD8α⁺ cDCs. Finally, DCs generated fromBatf3^-/- BM were selectively deficient in TLR3-induced IL-12 production (fig. S5C), a specific feature of CD8α⁺ cDCs (16). Similarly, CD11c⁺ cDCs from the spleens ofBatf3^-/- mice were selectively deficient in TLR3-induced IL-12 production but had normal responses to TLR4 and TLR9 ligands (fig. S6A).

We next tested whether APCs fromBatf3^-/- mice could prime CD4⁺ and CD8⁺ T cell responses. Similar proliferative responses of OT-II transgenic CD4⁺ T cells (17) occurred with soluble ovalbumin presented byBatf3^+/+ andBatf3^-/- cDCs (fig. S6B). However,Batf3^-/- cDCs were defective in an assay for cross-presentation of cellular antigen to CD8⁺ T cells (2,18) (Fig. 3A). OT-I T cells proliferated in response toBatf3^+/+ cDCs cocultured with ovalbumin-loaded cells, but failed to proliferate in response toBatf3^-/- cDCs in this assay.

We examined responses ofBatf3^-/- mice to West Nile virus (WNV) (19,20).Batf3^-/- mice showed normal WNV-specific antibody responses (Fig. 3B), memory B cells (fig. S6C) and CD4⁺ T cell responses (fig. S6D) but had a dramatic reduction in WNV-specific CD8⁺ T cell responses (Fig. 3C) and in vivo CTL killing of WNV peptide-loaded target cells (fig. S7A, B).Batf3^-/- mice lacked WNV-specific memory CD8⁺ T cells and had impaired formation of CD8⁺ CD44^hi CD62L^low cells (fig. S7). Adoptive transfer ofBatf3^-/- CD8⁺ T cells intoRag2^-/- mice generated normal WNV-specific CD8⁺ T cell response (Fig. 3D), but adoptive transfer ofBatf3^+/+ CD8⁺ T cells intoBatf3^-/-Rag2^-/- mice generated an impaired WNV-specific CD8⁺ T cell response (fig. S7C). This shows that impaired WNV-specific CTL responses inBatf3^-/- mice results from a defect of DCs rather CD8⁺ T cells.

We challengedBatf3^+/+ andBatf3^-/- mice with syngeneic fibrosarcomas that normally are rapidly rejected in a CD4⁺ and CD8⁺ T cell-dependent manner (21,22) (fig. S8A). Two independent fibrosarcomas were rapidly rejected byBatf3^+/+ mice, but grew progressively inRag2^-/- mice andBatf3^-/- mice (Fig. 4A,fig. S8B and C). Moreover,Batf3^-/- mice failed to develop tumor-specific CTLs (Fig. 4B). Tumor-infiltrating CD8⁺ T cells, but not CD4⁺ T cells, were significantly reduced inBatf3^-/- mice (Fig. 4C). The failure ofBatf3^-/- mice to reject these tumors was not due to defective NK cell development or function (fig. S2B and S9A-C). We considered whetherBatf3^-/- T cells have an intrinsic dysfunction, since overexpression studies had suggestedBatf3 might affect IL-2 transcription (14). WhileBatf3 overexpression reduces IL-2 reporter activity in Jurkat T cells (fig. S10B),Batf3^-/- CD4⁺ T cells showed normal IL-2 production (fig. S10D) and normal T_H1, T_H2 and T_H17 differentiation (fig. S10C-E, S11B). Finally,Batf3^-/- CD8⁺ T cells showed normal allospecific effector responses (fig. S11A) and cytokine production (fig. S11B).

Other DC subsets may cross-present, although less efficiently than CD8α⁺ DCs (23-26), suggesting there may be residual cross-presentation capacity inBatf3^-/- mice. We therefore challenged mice using reduced tumor-cell numbers which might allow effective responses in the setting of reduced cross-presentation (fig. S8). While 10⁴ and 10⁵ tumor cells grew in allRag2^-/- mice, someBatf3^-/- mice controlled this lower tumor burden (fig. S8D, 8E) and developed tumor-specific CTL response (fig. S8F). Whereas adoptive transfer of wild type DCs led to partial control of tumor growth inBatf3^-/- mice, transfer ofBatf3^-/- DCs did not (fig. S12).

Subsets of cDCs have recently been described with functional similarities to CD8α⁺ cDCs. Migratory Langerin⁺ dermal and lung DC subsets express DEC205⁺ and CD103⁺, and like CD8α⁺ cDCs, are CD11b^lo/- (27,28). CD8α⁺ cDC and migratory CD103⁺ DC populations share the distinctive properties of TLR3 responsiveness (27) and capacity for cross-presentation (26), further supporting the idea that these CD103⁺ subsets may be related. In spleen, CD103 is co-expressed with CD8α on cDCs (fig. S13A) (29) and is selectively expressed by the ‘CD8α equivalent’ CD24⁺Sirp-α^lo-int cDC subset derived from FL-treatedBaft3^+/+ BM (fig. S13C), but is not expressed byBatf3^-/- splenic cDCs (fig. S13B) or FL-treatedBatf3^-/- BM cells. This suggests that CD103 expression may requireBatf3. In agreement,Batf3^-/- mice showed reduced CD103 expression on DEC205⁺CD8α^-CD11b^lo/- dermal DCs in skin draining lymph nodes (fig. S14).

This study describes a transcription factor that controls development of CD8α⁺ cDCs.Batf3^-/- mice exhibit impaired antigen cross-presentation, impaired CTL responses against viral infection, and impaired responses to tumor challenge. These results suggest an important role for in vivo cross-presentation in CTL responses and provide support for therapeutic approaches that utilize CD8α⁺ cDCs for the induction of effective immune responses.