Interleukin-12 from CD103⁺ Batf3-Dependent Dendritic Cells Required for NK-Cell Suppression of Metastasis

Cancer immunotherapy with immune checkpoint blockade and adoptive T-cell therapies has established the ability of a remobilized immune system to control tumor growth and metastasis and to improve the survival of cancer patients. Much of the success with immune checkpoint blockade and other immunomodulatory strategies has highlighted the critical role that efficient cross-presentation by dendritic cells (DCs) might play with respect to effector T-cell control of tumors (1–4). Yet the role of the DC in the control of metastasis is poorly characterized.

Conventional DCs (cDC) can be divided into two major groups. cDC1 consist of CD8α⁺ tissue-resident or CD103⁺ migratory DCs that are specialized for antigen cross-presentation. cDC2, or CD11b⁺ DCs, directly present antigens via MHC class II. Cross-presenting DCs are of special interest because of their ability to phagocytose cellular proteins, directing them to the MHC class I antigen presentation pathway that will activate CD8⁺ cytotoxic T cells (5, 6). Mice deficient in Batf3, Irf8, Id2, and Nfil3 have defects in the generation of CD8α⁺ and CD103⁺ DCs, suggesting that their development requires these transcription factors. In particular, Batf3-deficient mice show impaired ability to cross-prime cytotoxic T lymphocytes (CTL) against tumor antigens in immunogenic fibrosarcomas and lack virus-specific CTL responses to West Nile infection (7). Batf3-dependent DCs display strong protective Th1 type immune responses against the parasitic infections of Toxoplasma gondii (8) and Leishmania (9, 10) that were dependent on IL12 secretion. Batf3 deficiency also increased resistance to helminth infections by augmenting Th2 type immunity, which was dependent on constitutive expression of IL12 by CD103⁺ Batf3-dependent DCs, but was independent of Toll-like receptors or microbial signals (11). In contrast, the Batf3 deficiency was dispensable for antifungal immune responses (12).

Batf3 DCs are the major producers of IL12 and secretion of IL12 from these DCs is the one of the main mechanisms for controlling infection and cancer (3, 8–10, 13–15). Batf3 deficiency abrogated the antitumor efficacy of the combination of anti-CD137 and anti–PD-1 immunotherapies that was not rescued by IL12 administration (1). Although CD8α⁺ Batf3-dependent DCs were indispensable for TRAIL-based immunotherapy in a mouse model of renal cell carcinoma (16), PD-L1–based immunotherapy in a mouse model of Braf mutant melanoma was dependent on CD103⁺ DCs (2). However, in both cases, the role of IL12 was not assessed. Soluble Flt3 ligand and TLR3 adjuvant poly I:C induced activation and expansion of CD103⁺ DCs in vivo and enhanced antitumor immune responses to immune checkpoint blockade by enhancing CTL responses (1, 2).

The DC populations defined by Batf3 and Irf8 expression, though sporadically present on the tumor margins, are functional and superior stimulators for naïve and activated CD8⁺ T cells (17). Furthermore, the intratumor role of CD103⁺ DCs was supported by The Cancer Genome Atlas (TCGA) analysis showing that a high CD103⁺/CD103⁻ ratio was associated with better survival (17). Though Batf3-dependent DCs play an important role in T cell–dependent immune responses against primary tumors, their role in antimetastatic innate immune responses remains elusive.

In this study, we used two mouse strains of different genetic backgrounds and a series of experimental and spontaneous metastasis models to determine the intrinsic role of cross-presenting DCs in suppressing metastasis. Here, we show that Batf3 deficiency and Irf8 deficiency in DCs enhanced lung metastasis but did not affect the growth of primary tumors. CD103⁺ DCs were the major source of IL12 expression in the lungs of mice and were important for controlling metastasis in a NK cell- and IFNγ-dependent manner. In concert, we found that BATF3 expression was associated with NK cell surface receptors and IFNγ expression and both BATF3 and IRF8 expression were associated with improved survival of triple-negative breast cancer patients.

C57BL/6 and BALB/c wild-type (WT), and CD45.1 congenic C57BL/6 mice were purchased from the Walter and Eliza Hall Institute for Medical Research or bred in house. C57BL/6 Batf3^–/– and BALB/c Batf3^–/– mice were originally provided by Ken Murphy (Washington University School of Medicine, St Louis, MO; ref. 7). C57BL/6 IL12p35^–/– mice were as previously described (18). C57BL/6 Rag2^–/–γc^–/– mice were obtained from The Jackson Laboratory (ID: 014593). C57BL/6 IRF8^fl/flCD11c^Cre mice were provided by Dr. Stephen Nutt (Walter and Eliza Hall Institute of Medical Research, Parkville, Australia), and C57BL/6 IL12-YFP reporter mice were provided by Dr. Geoffrey Hill (QIMR Berghofer, Herston, Australia). All mice were bred and maintained at the QIMR Berghofer Medical Research Institute and used between the ages of 8 to 14 weeks. All experiments were approved by the QIMR Berghofer Medical Research Institute Animal Ethics Committee.

Mouse B16F10 melanoma was obtained from the ATCC in 2007 and LWT1 melanoma was derived from SM1WT1 cells at the QIMR Beghofer in 2013. E0771 and 4T1.2 mammary carcinomas were obtained from Dr. Robin Anderson at the Peter MacCallum Cancer Centre (in 2014 and 2004, respectively). All tumor cell lines were maintained in culture for 0 to 14 days and injected as previously described (19). All cell lines were routinely tested negative for mycoplasma, but cell line authentication was not routinely performed.

For spontaneous metastasis and post-surgery survival experiments, 5 × 10⁴ 4T1.2 tumor cells were inoculated into the fourth mammary fat-pad of BALB/c WT and Batf3^–/– mice. On day 18, mice were anaesthetized to surgically remove primary tumor and the wound was closed with surgical clips. Some mice additionally received anti-CD4 (GK1.5), anti-CD8β (53.5.8) to deplete T-cell subsets and anti-asialoGM1 to deplete NK cells, and neutralizing anti-IFNγ (H22) as indicated. Survival of the mice was monitored. In some experiments, mice were sacrificed at day 30, and lungs were fixed in Bouin's solution for 24 hours and then washed several times in 70% ethanol. Metastatic tumor nodules were counted under a dissection microscope and lung tissues from 6 to 8 mice were randomly pooled from WT or Batf3^–/– mice and embedded in paraffin. Sections were stained with hematoxylin and eosin and slides were scanned through an Aperio slide scanner.

Single-cell suspensions of B16F10 melanoma cells (2 × 10⁵), LWT1 (5 × 10⁵), and 4T1.2 (2 × 10⁴) were injected i.v. into the tail vein of the indicated strains of mice. Some experimental mice additionally received anti-CD4 (GK1.5) and anti-CD8β (53.5.8) to deplete CD4⁺ and CD8⁺ T cells, and anti-asialoGM1 to deplete NK cells, and neutralizing anti-IFNγ (H22) as indicated. Lungs were harvested on day 14 and tumor nodules were counted under a dissection microscope.

Single-cell suspensions were generated from PBS-perfused lungs from WT and Batf3^–/– mice by mincing the lung tissue into small pieces and incubating them with Collagenase type IV (Worthington Chemicals) and DNAse I (Roche) in RPMI for 45 minutes at 37°C. Some samples were incubated with cell-stimulation cocktail (eBioscience) or brefeldin (Biolegend) only for 4 h in complete RPMI medium at 37°C. Samples were washed with PBS and incubated with 2.4G2 (anti-CD16/32) on ice. Cells were stained with antibodies to CD45.2 (104), TCRβ (H57-597), CD8α (53-6.7), CD4 (RM4-5), NK1.1 (PK136), NKp46 (29A1.4), CD49b (Dx5), CD11c (N418), MHC class II (M5/114.15.2), CD103 (2E7), CD11b (M1/70), and CD86 (GL1). The dye Zombie Yellow (Biolegend, BD Biosciences, or eBioscience) was used to distinguish live from dead cells. For intracellular staining, samples were fixed and permeabilized with cytofix/cytoperm kit (BD Biosciences) and stained for anti-IFNγ (XMG1.2) or respective isotype control antibody (Biolegend). To determine absolute number of cells, liquid-counting beads (BD Biosciences) were added directly before samples were run on flow cytometer. All data were collected on an LSR Fortessa 4 (BD Biosciences) flow cytometer and analyzed with FlowJo v10 software (TreeStar, Inc.).

Bone marrow (BM) cells were prepared from femurs and tibias of WT, Batf3^–/– and IL12p35^–/– mice. Flt3 ligand was obtained from Celldex Therapeutics and GM-CSF supernatant was derived from a murine erythroleukemia cell line stably transfected to express mouse GM-CSF. GM-CSF and Flt3 ligand-derived BMDCs were generated as described previously (20). BMDCs were stimulated with 1 μg/mL lipopolysaccharide (LPS) from E. coli (Sigma-Aldrich) and BMDCs supernatant was collected after 24 hours.

IL12p70 concentration was measured from BMDC supernatants samples with the mouse IL12p70 OptEIA ELISA set (BD Bioscience) as per the manufacturer's instructions. Briefly, maxisorp ELISA plated were coated with anti-IL12p70 in the coating buffer (Na₂HPO₄ + NaH₂PO₄) overnight, blocked with 1% BSA for 1 hour and incubated with BMDC supernatants for 2 hours. Samples were subsequently incubated with anti-IL12p70 conjugated to biotin for 2 hours and streptavidin-HRP for 30 minutes before adding TMB substrate. Absorbance was read at 450 nm on a spectrophotometer (BioTek, Millennium Science).

Splenocytes were harvested from WT and Batf3^–/– mice and lysed with ACK lysis buffer for 2 minutes. Splenocytes were stained with Mouse NK Cell Isolation Kit II (Miltenyi Biotec) and enriched by a depleting program using LS columns or on AutoMACS-Pro (Miltenyi Biotec) as per the manufacturer's instructions. A total of 1 × 10⁵ NK cells/well were seeded in a round-bottomed 96-well plates with 1:1 supernatant from BMDCs supernatant in the presence or absence of 10 μg/mL of neutralizing anti-IL12 (C17.8, BioXcell). Supernatant was collected after overnight culture and NK cells were stained for surface markers and intracellular IFNγ as described above. To study NK cell–mediated cytotoxicity, purified NK cells were sorted from WT and Batf3^–/– mice 24 hours post poly I:C (100 μg per mice) i.p. injection. One million target B16F10 or YAC-1 cells were labeled with 100 μCi/1 of ⁵¹Cr and were cocultured for 4 hours with the indicated ratio of primed NK cells.

Mice were injected intraperitoneally with 200 μg LPS from E. coli (Sigma-Aldrich) as described (21). Spleen and lungs were collected 6 hours after the LPS treatment for flow cytometry analysis.

BM from femurs and tibias were harvested and red blood cells were lysed with ACK lysis buffer. Recipient WT, Batf3^–/–, or IL12p35^–/– mice were irradiated with two doses of 5.5 Gy of radiation. After 2 to 4 hours of irradiation, the recipient mice were injected intravenously with 2–4 × 10⁶ BM cells from either a single donor or a 1:1 mixture from two donors as indicated. After 2 months, BM reconstitution was confirmed by flow cytometry analysis on the blood samples using congenic CD45.1/CD45.2 markers of leukocyte populations and chimeric mice were injected i.v. with 2 × 10⁵ B16F10 melanoma cells.

Breast cancer RNAseq data generated by TCGA was obtained from the UCSC Cancer Genomics Hub for analysis. Survival plots for different cancer types were plotted using Kaplan Meier (KM) plotter (22). Reads were then trimmed for adapter sequences using Cutadapt (version 1.11) and aligned using STAR (version 2.5.2a) to the GRCH37 assembly using the gene, transcript, and exon features of Ensembl (release 75) gene model (23). Expression was estimated using RSEM (version 1.2.30). Genes with zero read counts across all samples were removed, leaving 1,226 samples with 50,150 genes. Normalization was performed by dividing each sample's gene count by million reads mapped to generate counts per million (CPM), followed by the trimmed mean of M-values (TMM) method from the edgeR package (24).

Statistical analysis was achieved using GraphPad Prism Software. Data were compared using a Mann–Whitney U test or one-way or two-way ANOVA. Differences in survival were evaluated using a log-rank (Mantel–Cox) test. Data were considered statistically significant where the P value was equal to or less than 0.05.

To understand the role of Batf3-dependent DCs in tumor control, we implanted the spontaneously metastasizing 4T1.2 breast cancer cell line in the mammary fat pad of WT and Batf3^–/– mice. Unlike primary tumor growth, which was similar in the WT and Batf3^–/– mice (Fig. 1A and Supplementary Fig. S1), Batf3^–/– mice had increased lung metastasis (Fig. 1B–E) and poorer survival (Fig. 1F), with median survival of 31 days for Batf3^–/– mice and 38 days for the WT mice, suggesting that Batf3 was required for metastasis control. A similar pattern of increased lung metastasis was observed in B16F10 melanoma (Fig. 2A), LWT1 Braf-mutant melanoma (Fig. 2B), and 4T1.2 breast carcinoma (Fig. 2C) models of experimental metastases, further suggesting the role of Batf3 in innate immune metastatic control.

To further investigate the role of CD103⁺ and/or CD8α⁺ DCs in host control of tumor metastasis, we injected B16F10 melanoma cells in Irf8^fl/flCD11c^cre mice previously reported to lack CD8α⁺CD103⁺ DCs (25). Consistent with the data obtained from Batf3^–/– mice, the lack of Irf8 expression in CD11c⁺ DCs (Irf8^fl/flCD11c^cre+) made no difference to the growth of primary tumors in mice but significantly increased the number of lung metastases compared with their littermate controls (Fig. 2D and Supplementary Fig. S1C), providing further evidence for an important role of CD8α⁺CD103⁺ DC in metastasis control.

Most of the phenotype of Batf3^–/– mice has previously been attributed to the poor priming, activation, and function of CD8⁺ T lymphocytes (7, 26). However, NK cells are the main effectors of antimetastatic activity (27). To investigate the immune effector mechanisms required for Batf3-dependent metastatic control, we depleted CD4⁺ and CD8⁺ T lymphocytes or NK cells in the WT and Batf3-deficient mice and examined the effect on the number of lung metastases. The increased number of lung metastases in Batf3^–/– mice was independent of CD4⁺/CD8⁺ T lymphocytes, as depletion of these cells did not affect the metastatic burden in WT or Batf3^–/– mice (Fig. 3A–C). Depletion of CD8⁺ T cells alone also did not affect survival of 4T1.2 inoculated WT or Batf3^–/– mice post-surgical resection of the primary tumor (Fig. 3D). In contrast, depletion of NK cells enhanced the number of lung metastases to an equivalent level in WT mice and Batf3^–/– mice in the B16F10, LWT1, and 4T1.2 models of experimental metastases (Fig. 3A–C). Furthermore, depletion of NK cells alone, or together with CD8⁺ T cell depletion, reduced the overall survival of WT mice but not Batf3^–/– mice in the 4T1.2 model of spontaneous cancer metastasis post primary tumor resection (Fig. 3D). In addition, we observed a greater number of NK cells, but not CD4⁺ or CD8⁺ T cells, in the lungs of WT mice than Batf3^–/– mice in the B16F10 model of experimental lung metastasis (Supplementary Fig. S2), suggesting that NK cells contributed to the innate immune-dependent control of metastasis mediated by Batf3.

We hypothesized that intrinsic functional defects in NK cells may contribute to the increased lung metastasis in Batf3^–/– mice. However, the frequency of NK cells in the lung, liver, spleen, blood, and BM and their differentiation phenotype with respect to CD11b and CD27 expression were largely similar between naive WT and Batf3^–/– mice (Supplementary Fig. S3).

To assess any functional difference in NK cells between WT and Batf3^–/– mice during inflammation, we examined NK-cell responses after lipopolysaccharide (LPS) injection in vivo. Although LPS activated NK-cell IFNγ production and CD69 expression, the number and activation status of NK cells was similar in the spleens and lungs of WT and Batf3^–/– mice following LPS treatment (Supplementary Fig. S4), suggesting that NK cells from Batf3^–/– mice have no intrinsic defect and that Batf3-dependent DCs may not be critical for NK-cell responses to LPS.

Next, we examined the antimetastatic function of splenic NK cells from WT and Batf3^–/– mice. Adoptive cell transfer (ACT) of splenic NK cells into immunodeficient C57BL/6.Rag2^–/–γc^–/– mice significantly reduces the number of B16F10 lung metastases (28). This activity was independent of Batf3 since the source of NK cells, whether WT or Batf3^–/– mice, equivalently reduced the number of lung metastases following transfer (Supplementary Fig. S5A). The peripheral blood of C57BL/6.Rag2^–/–γc^–/– mice showed a similar NK-cell reconstitution following WT or Batf3^–/– NK-cell transfer (Supplementary Fig. S5B), suggesting that Batf3 expression in NK cells was not required for reconstitution or intrinsic metastatic activity of NK cells. Furthermore, in vitro NK cells from the WT and Batf3^–/– mice displayed similar cytotoxic activity towards YAC-1 lymphoma and B16F10 melanoma target cells after stimulation with Poly I:C (Supplementary Fig. S5C). Overall, these results suggest that there is no fundamental functional defect in NK cells from the Batf3^–/– mice.

As IL12-producing capability of Batf3-dependent DCs is critical to their immune modulatory function (8, 10, 11, 13, 29–31), we hypothesized that IL12 from Batf3⁺ DCs might activate NK-cell antimetastatic activity. As expected, Batf3-deficient mice lacked CD103⁺ lung DCs (Fig. 4A). Using IL12-YFP reporter mice, we observed a higher proportion of DCs and IL12-expressing CD103⁺ DCs in the lungs of mice in the B16F10 model of experimental tumor metastasis (Fig. 4A–C). The IL12 expression was detected in the migratory CD103⁺ DCs; however, resident CD8α⁺ DCs or CD11b⁺ myeloid DCs had little or no expression of IL12 (Fig. 4D). Furthermore, the number of CD8α⁺ DCs in lungs was 10 times lower than CD103⁺ DCs (Supplementary Fig. S6A). Additionally, IL12-expressing CD103⁺ DCs had a higher expression of the CD86 activation marker than IL12-negative CD103⁺ DCs (Fig. 4E). Overall, these results suggest that CD103⁺ migratory DCs increase in numbers, and they express more IL12 than other DC in the mouse model of experimental B16F10 lung metastasis.

IL12 has antimetastatic function (31, 32) and besides CD103⁺ DCs, there are other DCs, macrophages, and neutrophils that can potentially contribute to increased IL12 levels. To test whether IL12 from CD103⁺ Batf3-dependent DCs is the only source of functional protection against metastasis, we generated mixed chimeras using BM from WT, IL12p35^–/–, and Batf3^–/– mice. The mice were bled 2 months after BM transfer to confirm chimerism (Supplementary Fig. S6B). As shown previously, Batf3^–/– mice lacked CD8α⁺/CD103⁺ DCs, but these DCs developed from the BM of IL12p35^–/– mice, despite being unable to produce IL12 (8). In the mixed chimeras reconstituted with IL12p35^–/– and Batf3^–/– BM, CD8⁺/CD103⁺ DCs are the only population that lacks the ability to produce IL12. These mixed chimeric mice displayed very high numbers of B16F10 lung metastases similar to the mice reconstituted with Batf3^–/– or IL12p35^–/– BM (Fig. 4F). In contrast, mixed chimeric mice reconstituted with WT and Batf3^–/– or WT and IL12p35^–/– BM displayed half the number of metastases, in concert with only half of the IL12-producing CD8⁺/CD103⁺ DCs being available to afford protection from metastasis (Fig. 4F). These results suggest that IL12-producing Batf3-dependent DCs are critical for protection against experimental lung metastasis. Conversely, low-dose recombinant IL12 (rIL12) displayed antimetastatic activity in Batf3^–/– mice and WT mice (Fig. 4G), suggesting that even in the absence of the Batf3-dependent DCs, IL12 is capable of providing protection against metastasis.

IFNγ from lung resident NK cells is critical for natural resistance to lung metastasis (27) and is essential for the antimetastatic activity of therapeutic rIL12 (30). To assess the role of IFNγ in Batf3-dependent control of lung metastasis, we neutralized IFNγ in the WT and Batf3^–/– mice. Neutralization of IFNγ increased the number of B16F10 experimental lung metastases (Fig. 5A) in WT mice and decreased the survival of WT mice in 4T1.2 model of spontaneous metastasis (Fig. 5B). However, there was no significant change in either the number of lung metastases or the survival of Batf3^–/– mice after IFNγ neutralization (Fig. 5A and B), suggesting that IFNγ was required for Batf3-dependent control of lung metastasis. Because increased metastasis in Batf3^–/– mice is dependent on both NK cells and IFNγ, we hypothesized that IL12 from CD103⁺ lung DCs primes NK cells to produce IFNγ that is required for protection against metastasis. To test this hypothesis, we generated CD103⁺ BM-derived dendritic cells (BMDC) from WT, Batf3^–/–, and IL12p35^–/– mice as described previously (20). BMDCs generated from Batf3^–/– produced less IL12 than WT BMDCs after LPS stimulation for 24 hours. As expected, BMDCs from IL12p35^–/– mice lacked the ability to produce IL12 (Fig. 5C).

Next, in order to understand whether the IL12 generated from these CD103⁺ BMDCs could prime NK cells, we cultured WT splenic NK cells with or without supernatant from LPS-stimulated CD103⁺ BMDCs derived from WT and Batf3^–/– mice. The number of IFNγ-expressing NK cells increased when cultured with WT BMDCs compared to the Batf3^–/– BMDCs, and the number of IFNγ-expressing NK cells reduced when NK cells were cultured with neutralizing anti-IL12p40 (Fig. 5D). We also observed that NK cells stimulated with WT BMDC supernatant had higher IFNγ expression compared with stimulation with supernatant from Batf3^–/– BMDCs, whereas NK cells stimulated with supernatant from IL12p35^–/– BMDCs had the lowest IFNγ expression (Supplementary Fig. S7). These results suggested that IL12 from CD103⁺ BMDCs activated NK cells to produce IFNγ. To test IFNγ from NK cells in vivo, we collected lungs from WT and Batf3^–/– mice 3 days after B16F10 tumor injection. Control and tumor-bearing lungs in WT mice had more IFNγ⁺ NK cells than did lungs of Batf3^–/– mice (Fig. 5E). Also, after stimulation with cell-stimulation cocktail, IFNγ⁺ NK cells were significantly more numerous in the tumor-bearing lungs of WT mice than in the lungs of Batf3^–/– mice (Fig. 5F). Overall, these results suggest that IL12 from Batf3-dependent DCs may activate NK cells to produce IFNγ.

Immunoscore using a high CD103⁺/CD103⁻ gene expression ratio had a profound effect on increased survival in breast cancer, head and neck squamous cell carcinoma and lung adenocarcinoma (17). Given the fact that Batf3- and Irf8 deficiency promoted mouse lung tumor metastasis, we wished to test whether intratumor expression of Batf3 and Irf8 in humans might predict patient survival (22).

Using TCGA data sets, we observed that a higher expression of BATF3 and IRF8 was associated with a minor, but statistically significant, improvement in relapse-free survival of breast cancer patients (Fig. 6A). Additionally, the high expression of IL12, NK cell–associated genes (CD56, NKP46, and NKG2D) and IFNG (IFNγ) was also associated with improved survival (Fig. 6A). The cDC-specific gene, Zbtb46, is efficiently induced by Irf8 (33), and CCR7 is reportedly expressed by all CD103⁺ DCs (20). We found high expression of CCR7 and ZBTB46 associated with improved patient outcome (Fig. 6A). The correlations between BATF3, CD103 or IRF8 expression and relapse-free survival were even more significant in metastatic triple-negative breast cancer (TNBC) patients amongst all breast cancer patients (Fig. 6B). High expression of BATF3, CD103, and IRF8 correlated (P range = 0.0035–0.046) with relapse-free survival of poorly differentiated high-grade (grade 3) breast cancers, but not for lower grade breast cancers (Supplementary Fig. S8), suggesting that Batf3-dependent DCs are important in controlling highly aggressive metastatic cancers. Additionally, high expression of IRF8 significantly correlated with progression-free survival in gastric cancers (Fig. 6C) and lung cancers (Fig. 6D).

Next, to determine if a Batf3-dependent DCs–NK-cell axis was important for host control of metastasis, we used TCGA data to quantify correlations among relative gene expression for breast cancers. As expected, BATF3 and IRF8 expression was highly correlated with DC markers CD103, CCR7, and CD86 in both normal breast tissues and breast cancers (Supplementary Fig. S9). BATF3 and IRF8 expression was significantly positively correlated with DC effector IL12A only in breast cancers (Supplementary Fig. S10) and was also more positively correlated with NK-cell markers such as KLRK1, NCAM1 (encoding NKG2D), and the effector cytokine IFNG in breast cancer tissues than in normal control tissues (Supplementary Fig. S10). CD103 expression also showed positive correlations with KLRK1 and IFNG in breast cancers compared with the normal tissue. This supports our hypothesis and suggests that IL12-expressing BATF3-dependent DCs and IFNγ from NK cells may form an important functional axis for intrinsic tumor control in breast cancers.

Metastasis is the leading cause of cancer-related deaths worldwide. Better understanding of host mechanisms that can suppress metastasis is required to successfully develop therapies targeting cancer metastasis. NK cells are innate lymphocytes that play a central role in host rejection of tumors and virus-infected cells. We have previously shown that perforin and IFNγ from NK cells is critical for providing natural resistance to tumor metastasis to lung (27, 34). Dendritic cells, which are required for priming and activation of CTL activity, are also responsible for increases in NK cell–cytolytic activity and IFNγ production (35). DC- and NK-cell cross-talk may be dependent on receptor–ligand contacts (35) or IL12 secretion from DCs that primes NK cells for IFNγ production (29–31). In this study, we show that CD103⁺ Batf3-dependent cross-presenting DCs are potent producers of IL12 in lungs and may partially activate host resistance to metastasis in the lung by mechanisms dependent on NK cells and IFNγ.

Batf3 is critical for protection against lung metastasis. Despite similar primary tumor growth, Batf3-deficient mice had enhanced experimental and spontaneous lung metastasis compared with the WT mice. The increased lung metastasis phenotype in mice that specifically lacked Irf8 expression in CD11c⁺ DC subsets suggested that cross-presenting DCs, mainly CD103⁺ DCs, are partially required for NK-cell resistance to metastasis, because Batf3-deficient mice may still develop CD8α⁺ DCs, but not CD103⁺ DCs, under certain conditions (36–39). In support, we found an increased number of IL12-expressing CD103⁺ DCs, but not tissue-resident CD8α⁺ DCs, in the lungs of mice injected with B16F10 melanoma cells. The role of IL12 from Batf3-dependent DCs in suppressing metastasis was confirmed by mixed BM chimeras where the mixed BM from IL12^–/– and Batf3^–/– mice, but not WT and IL12^–/– or WT and Batf3^–/– mice, failed to provide protection against B16F10 lung metastasis. Further, IL12 administration in Batf3-deficient mice partially restored host protection against lung metastasis. Our results in a tumor metastasis model were in concert with a previously published report where mixed BM chimeric mice that specifically lacked IL12 from Batf3⁺ DCs were as susceptible to Toxoplasma gondii infection as Batf3^–/– mice (8).

The majority of Batf3-dependent immune responses against tumors and infection require activation and cytotoxic activity of CD8⁺ T lymphocytes (1, 2, 7, 8, 13, 14, 16, 17, 26). However, both CD4⁺ and CD8⁺ T lymphocytes are dispensable for metastatic control by Batf3-depedent DCs. Instead, Batf3-dependent control of metastasis required NK cells and IFNγ. Dendritic cells can activate NK cells by forming a stimulatory synapse that can polarize IL12 secretion specifically into NK cells (29) or DC-derived exosomes can activate NK cells via a TNF pathway (40). CD103⁺ Batf3-dependent lung DCs were shown to acquire and cross-present apoptotic cell-associated antigen to CD8⁺ T cells (5). It is possible that antigens from metastatic tumor cells in the lungs are taken up by tissue-resident CD103⁺ DCs that then release IL12 to activate lung-resident NK cells for IFNγ production and protection against metastasis. Taken together, our data suggest a pathway comprising Batf3-dependent DCs that produce IL12, which then stimulates NK cells to produce IFNγ and partially control tumor cell metastasis. The partial effect of Batf3 and IL12 was consistent with a demonstrated role of myeloid-derived IFNs α,β, and λ, as well as NK-cell perforin, in experimental metastasis control (34, 41).

CD103⁺ DCs are sparse and proximal to the tumor margins but can interact with intratumor T cells for effective adoptive CTL therapy (17). In humans, the transcript signature ratio of CD103⁺ to CD103^– tumor-associated macrophages provided a strong prognostic signal across multiple tumor types in TCGA data. This ratio was more informative than other commonly used immune-based prognostic gene expression markers (17). In concert, we observed higher expression of BATF3, IRF8, NK-cell markers and IFNγ associated with improved relapse-free survival in breast cancer patients. Furthermore, we found a significant correlation between BATF3 expression and IL12, IFNγ and NK cell-associated markers in tumors from TCGA datasets suggesting that Batf3-dependent DCs in association with NK cells and IFNγ may form an important functional axis for metastasis control.

Immune checkpoint blockade by anti–PD-L1 and the combination of anti-PD-1 and agonistic anti-CD137 requires Batf3-dependent DCs (1, 2), suggesting that impairment of Batf3-dependent DC function may be one mechanism that causes some patients to respond poorly to immune checkpoint blockade. Our studies demonstrate that CD103⁺ Batf3-dependent DCs provide natural resistance to metastasis and encourage the use of reagents, such as TLR3-adjuvant poly I:C and Flt3 ligand, that activate these DCs (42, 43) and may enhance the antimetastatic efficacy of current cancer therapies.

M.J. Smyth reports receiving commercial research grants from Bristol-Myers Squibb, Aduro Biotech, Tizona Therapeutics, and Corvus Pharmaceuticals; is a consultant/advisory board member for Tizona Therapeutics; and has provided expert testimony for Corvus Pharmaceuticals. No conflicts of interest were disclosed by the other authors.

Conception and design: D. Mittal, M.J. Smyth

Development of methodology: D. Mittal, D. Vijayan, A.R. Aguilera

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Mittal, E.M. Putz, A.R. Aguilera, K.A. Markey, S.L. Nutt, M.J. Smyth

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Mittal, A.R. Aguilera, J. Straube, S. Kazakoff, N. Waddell, M.J. Smyth

Writing, review, and/or revision of the manuscript: D. Mittal, K.A. Markey, G.R. Hill, M.J. Smyth

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Takeda

Study supervision: D. Mittal, M.J. Smyth

M.J. Smyth was supported by a National Health and Medical Research Council of Australia (NHMRC) Senior Principal Research Fellowship (1078671) and Program Grant (1013667), and S.L. Nutt was supported by an NHMRC Principal Research Fellowship (1058238) and Program Grant (1054925). E.M. Putz was supported by an Erwin Schroedinger Fellowship of the Austrian Science Fund (J-3635). K.A. Markey is a Queensland Health Junior Clinical Research Fellow. We thank Kate Elder and Liam Town for their technical support with breeding and genotyping of mice. We also thank Fernando Souza-FonsecaGuimaraes and Stuart Olver for their technical help with the cytotoxicity experiments.

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.