Absence of MHC class II on cDCs results in microbial-dependent intestinal inflammation

Zbtb46 (zDC) is expressed in pre-DCs and their progeny, but not in monocytes, macrophages, and other myeloid cells (Fig. 1 A; Meredith et al., 2012; Satpathy et al., 2012). To express Cre recombinase in cells that actively transcribe zDC, we inserted Cre into the 3′ UTR of zDC (Fig. 1 B; Meredith et al., 2012). Cell type–specific expression of Cre was confirmed by crossing zDC^Cre to Rosa26^lSlYFP mice (zDC^CreRosa^lSlYFP), wherein Cre-mediated excision of a transcriptional Stop element results in YFP expression. The majority of cDCs (Lin⁻CD11c^highMHCII⁺) in the spleen of zDC^CreRosa^lSlYFP mice were YFP⁺. In contrast, B and T lymphocytes, pDCs (CD11c^lowBST2⁺), red pulp macrophages (RPMs; Lin⁻CD11b^lowF4/80⁺), and monocytes (Lin⁻CD11b⁺Ly6C⁺CD115⁺) expressed little or no YFP (Fig. 1, C and D). In contrast, in CD11c^CreRosa^lSlYFP mice, cDCs, as well as the majority of macrophages, and pDCs were labeled in the spleen (Fig. 1 D). Similar results were also obtained after administration of polyinosinic:polycytidylic acid (pIC; Fig. 1 E) and with zDC^Cre+IRF4^wt/fl mice that express GFP upon Cre-mediated IRF4 excision (Klein et al., 2006; Fig. S1).

In the intestine, the phenotype of monocytes and cDCs is altered (Caballero and Pamer, 2015). For example, small intestine lamina propria CD11c⁺MHCII⁺ cells represent a mixture of cDCs and monocytes that can be further subdivided based on expression of CD103, CD11b, and CD64 (Fig. 1 F; Bogunovic et al., 2009; Tamoutounour et al., 2012; Schreiber et al., 2013). Whereas all of these cell types are labeled in CD11c^CreRosa^lSlYFP mice, only pre-DC–derived CD64⁻ cells were YFP⁺ in zDC^CreRosa^lSlYFP indicator mice (Fig. 1, G and H). We did not observe a difference in YFP expression in CD103⁺/CD11b⁻ and CD103⁺/CD11b⁺ CD64⁻ cells in zDC^CreRosa^lSlYFP mice (Fig. 1 F). We conclude that zDC^Cre mice are suitable for cDC-specific conditional gene deletion.

In addition to antigen processing and presentation, cDCs are essential innate sensors of pathogens and other stimuli. To distinguish between innate and adaptive functions of cDCs, we produced a cDC-specific MHCII knockout by crossing zDC^Cre with MHCII^fl/fl mice (zDCΔMHCII; Hashimoto et al., 2002). As expected, zDCΔMHCII mice expressed MHCII on the surface of pDCs, but not on cDCs (Fig. 2 A). Absence of MHCII expression by cDCs did not change the frequency of cDCs among splenocytes (Fig. 2 B), nor did it alter the ratio of CD8α⁺ and CD11b⁺ cDCs (Fig. 2 C). The same was true for the mLNs (Fig. 2, D and E). Moreover, the levels of circulating Flt3L in the serum, which are increased upon cDC depletion (Birnberg et al., 2008; Meredith et al., 2012), were normal in zDCΔMHCII mice (Fig. 2 F).

To determine whether adaptive responses are altered by cDC-specific MHCII deletion, we transferred OVA-specific CD4⁺ T cells (OT-II cells) to control and zDCΔMHCII mice and compared their responses to immunization with OVA. OT-II cells responded by proliferating and up-regulating CD44 expression in control but not in zDCΔMHCII mice (Fig. 2, G and H). In contrast, proliferation and activation of CD8⁺ OT-I T cells in response to immunization with OVA was not altered in zDCΔMHCII mice (Fig. 2, G and H). Thus, deletion of MHCII on the surface of cDCs is sufficient to impair priming of naive CD4⁺ T cells.

To confirm that the innate responses of cDCs are intact in zDCΔMHCII mice, we purified cDCs from the spleen, stimulated them with CpG 1668, LPS, or pIC, and measured expression of co-stimulatory molecules (CD80 and CD86) and production of cytokines (IL-6 and TNF). We found no differences between control and zDCΔMHCII cDCs, indicating that their responses to innate stimuli are intact (Fig. 2 I-M). Thus, zDCΔMHCII cDCs show intact responses to innate stimuli, but are unable to initiate MHCII-restricted adaptive immune responses.

Although zDCΔMHCII mice were born at the expected Mendelian ratio, they showed significantly reduced weight gain (Fig. 3 A) and developed rectal prolapse starting at 6 wk of age (Fig. 3 B). In addition, the mutant mice concomitantly developed splenomegaly and showed increased circulating levels of the acute phase–reactant Csf3 and neutrophilia, which are indicative of a systemic process (Fig. 3, C–E). The levels of Csf1 and Csf2 remained within normal limits (unpublished data).

To quantitate intestinal inflammation, we measured Lipocalin-2, a sensitive biomarker for intestinal inflammation, in the feces of control and zDCΔMHCII mice (Chassaing et al., 2012). zDCΔMHCII mice showed a 100-fold increase in Lipocalin-2 compared with controls (Fig. 3 F). Consistent with this observation, histopathological analysis revealed inflammation that was primarily restricted to the large intestine, with inflammatory infiltrates, crypt hyperplasia, and villus destruction (Fig. 3 G). Neutrophil infiltration in the large intestine was also detected by flow cytometry (Fig. 3 H). In addition, we found elevated levels of the inflammatory cytokine IL-23 in the large intestine of zDCΔMHCII, but not in control mice (Fig. 3 I). To confirm that intestinal innate immune responses were intact, we measured the expression of AMPs in the intestine of zDCΔMHCII and control mice. Consistent with the inflammation found in the intestines of these mice, RegIIIγ and S100A9 were found at higher levels in zDCΔMHCII mice compared with controls (Fig. 3 J). ILC3s, whose IL-22 production is crucial for AMP expression (Sonnenberg et al., 2011; Kinnebrew et al., 2012; Satpathy et al., 2013; Bernink et al., 2015), were not affected and actually occurred at higher frequencies in zDCΔMHCII mice compared with controls (Fig. 3 K). We therefore conclude that intestinal inflammation in zDCΔMHCII mice occurs even in the presence of active cDC-dependent innate immune responses.

Normally, intestinal bacteria are separated from epithelial cells by a mucous layer. Innate immune responses and AMPs are essential for this spatial segregation (Vaishnava et al., 2011; Gallo and Hooper, 2012). To visualize bacteria in the large intestine (colon and cecum) of zDCΔMHCII mice and determine their location and proximity to the epithelial barrier, we performed 16S ribosomal RNA (rRNA) fluorescence in situ hybridization (16S FISH). We found that, in zDCΔMHCII mice, intestinal microbes and gut epithelial cells were in direct contact (Fig. 3 L). However, Muc2 expression in the large intestine was unaltered, and therefore zDCΔMHCII mice do not have a defect in mucus production (Fig. 3 M). Finally, neither orally administrated FITC-Dextran nor LPS were detectable at higher levels in the serum of zDCΔMHCII compared with control mice, indicating no gross defect in the epithelial barrier (unpublished data).

A similar phenotype was also observed in CD11cΔMHCII mice (Fig. 4 A) and in zDCΔMHCII, but not control BM chimeras (Fig. 4, B and C). Finally, BM chimeras reconstituted with a 1:1 mixture of control and zDCΔMHCII BM did not develop disease (Fig. 4 C). Therefore, intestinal inflammation in zDCΔMHCII mice is BM cell–dependent and can be rescued by WT BM cells.

Despite the changes in the anatomical distribution of the microbiome in the intestine, we did not observe dramatic differences in the composition of the microbiome (determined by 16S sequencing of fecal bacteria) between control and zDCΔMHCII mice that were cohoused. UniFrac analysis showed that the microbiome in individual mice clustered by the cage in which they were housed and not by genotype, suggesting that the composition of the microbiome is not dramatically altered in zDCΔMHCII mice compared with controls (Fig. 4 D). These data indicate that in addition to AMPs, MHCII expression by cDCs is required to maintain the spatial segregation between microbes and gut epithelial cells.

DCs and thymic epithelial cells are both involved in the selection of mature T cells in the thymus; therefore, we examined thymic CD4⁺ T cells in zDCΔMHCII mice. Although there was a small increase in mature single-positive CD4⁺ thymocytes in zDCΔMHCII mice compared with control mice, their Vβ chain usage was comparable (Fig. 5, A and B).

To check for the presence of autoreactive CD4⁺ T cells in the periphery, we analyzed T cells in the spleen, mLN, and lamina propria of the large intestine for the presence of activated and cytokine-producing CD4⁺ T cells. zDCΔMHCII mice showed no evidence of increased levels of such cells (Fig. 6). To the contrary, we found reduced frequencies of total CD4⁺ T cells, as well as increased frequencies of naive CD4⁺ T cells and reduced frequencies of activated CD4⁺ T cells, in zDCΔMHCII mice when compared with controls (Fig. 6). In addition, zDCΔMHCII mice showed lower frequencies of peripherally induced regulatory T cells (iT reg cells; defined as Nrp1⁻) in the mLN and lamina propria of the large intestine compared with controls. This is in accordance with a previously described cDC-mediated feedback mechanism controlling T reg cell numbers (Darrasse-Jèze et al., 2009). Despite reduced iT reg cells, the relative frequency of CD4⁺ T cells producing IL-2, IL-17, or IFN-γ was comparable or reduced in zDCΔMHCII mice compared with controls (Fig. 6, D–F). This was also confirmed by quantitative PCR (qPCR) analysis of the large intestine, where we found similar expression levels of IL-17 and IFN-γ (Fig. 3 I). Thus, in the absence of MHCII expression by cDCs, the defect in peripheral T reg cell induction did not result in aberrant levels of CD4⁺ T cell activation.

In contrast to CD4⁺ T cells, we observed a higher frequency of activated CD8⁺ T cells (CD3⁺CD8⁺CD62L⁻CD44⁺) in the spleen and mLN of zDCΔMHCII mice (Fig. 6, D and E). Furthermore, zDCΔMHCII mice showed unaltered or slightly reduced frequencies of γδ T cells in spleen, mLN, and large intestine compared with control mice, ruling out the involvement of pathogenic γδ T cells (Park et al., 2010; not depicted). To investigate whether the phenotype in zDCΔMHCII mice might be a result of insufficient CD4⁺ T cell activation, we depleted CD4⁺ T cells in control (MHCII sufficient) mice. Within 4 wk of depletion, mice developed neutrophilia, and elevated levels of Lipocalin-2 were detectable in their feces, although mice did not develop rectal prolapse within the 4 wk of depletion (Fig. 7 A; and not depicted). To test whether the observed increase in activated CD8⁺ T cells contributed to the phenotype in zDCΔMHCII mice, we depleted CD8⁺ T cells for 4 wk. However, CD8⁺ T cell–depleted zDCΔMHCII mice did not differ from nondepleted zDCΔMHCII mice (Fig. 7 B). Collectively, these data suggest that the phenotype in zDCΔMHCII mice is associated with insufficient adaptive CD4⁺ T cell responses but not pathogenic γδ or CD8⁺ T cells.

To test the possibility that zDCΔMHCII mice develop intestinal inflammation as a result of inefficient effector CD4⁺ T cell responses to intestinal bacteria, we treated the mice with broad-spectrum antibiotics. Lipocalin-2 in the feces, Csf3 in serum, and neutrophil levels in the spleen of zDCΔMHCII mice dropped to baseline levels after 4 wk of antibiotic treatment (Fig. 8, A–C). Consistent with these findings, antibiotic-treated zDCΔMHCII mice were free of intestinal inflammation and inflammatory infiltrate as determined by histology (Fig. 8 E). Moreover, maintaining zDCΔMHCII mice in a germ-free environment completely prevented the development of all signs of disease, including intestinal inflammation, high levels of Lipocalin-2, elevated Csf3 levels, neutrophilia, and increased numbers of activated CD8⁺ T cells (Fig. 8, A–E). These results indicate that the intestinal inflammation in zDCΔMHCII mice is dependent on the microbiota. They also suggest that rather than causing the disease, the activated CD8⁺ T cells in specific pathogen–free (SPF) zDCΔMHCII mice develop in response to microbiota-mediated inflammation.

IgA, the most abundant antibody isotype at mucosal surfaces, is secreted into the intestinal lumen, where it coats bacteria that enter the host and elicit adaptive immune responses (Palm et al., 2014). mLNs, Peyer’s patches (PPs), and isolated lymphoid follicles are the primary sites of adaptive immune responses against intestinal antigens. zDCΔMHCII mice showed normal levels of B cells in the mLNs, PPs, and lamina propria of the large intestine. Whereas GC B cells were reduced in zDCΔMHCII mice, the relative frequency of IgA⁺ B cells was not; the only significant defect in this compartment was in the lamina propria, suggesting that zDCΔMHCII mice do not have an overall defect in class switching. Consistent with the relative paucity of GC B cells, zDCΔMHCII mice showed significantly reduced frequencies of follicular T helper (TFH) cells in both mLNs and PPs (Fig. 9, A–D). These results support the idea that zDCΔMHCII mice have aberrant antimicrobial adaptive immune responses due to insufficient CD4⁺ T cell help (Grajales-Reyes et al., 2015).

To directly measure adaptive immune responses against intestinal bacteria, we colonized germ-free control and zDCΔMHCII mice with a mixture of six different commensal intestinal bacteria (Escherichia coli, Enterococcus faecalis, Bifidobacterium longum, Lactobacillus plantarum, Bacteroides vulgatus, and Ruminococcus gnavus). Colonization was similar for control and zDCΔMHCII mice as determined by PCR 8 wk after colonization (Fig. S2). Although germ-free mice have no detectable GC B or TFH cells in gut-associated lymphoid tissues, newly colonized control mice showed robust GC B and TFH cell responses in mLNs and PPs. In contrast, these responses were barely detectable in colonized zDCΔMHCII mice, even 8 wk after colonization (Fig. 10, A and B).

Consistent with the near absence of GC responses, IgA levels in the feces and serum of colonized zDCΔMHCII mice were significantly lower compared with colonized control mice. Fecal IgA increased progressively in control mice starting after day 7, reaching a mean of 200–400 ng/mg feces. In contrast, fecal IgA reached a plateau of around 10–20 ng/mg feces by day 12 and remained relatively constant thereafter in zDCΔMHCII mice (Fig. 10, C and D). Consistent with the overall lower levels of serum and fecal IgA, bacteria isolated from feces of zDCΔMHCII mice were coated with far lower levels of IgA than bacteria from feces of control mice (Fig. 10 E).

Collectively, these data reveal that zDCΔMHCII mice have defective adaptive immune responses against intestinal commensals, and this defect is associated with a loss of bacterial segregation from the epithelium and severe colitis.

cDCs link innate and adaptive immunity by sensing pathogens and initiating adaptive immune responses. Although the two physiological functions of cDCs are likely to play distinct roles in immune homeostasis, they have not previously been evaluated independently. In the gut, cDCs sense and capture gut microbes in part by extending their processes into the gut lumen (Macpherson and Uhr, 2004; Niess et al., 2005; Chieppa et al., 2006; Vallon-Eberhard et al., 2006). Microbial sensing induces cDCs to produce cytokines such as IL-23, which are required to activate innate lymphoid cells (Kinnebrew et al., 2012; Satpathy et al., 2013). In addition, the ingested microbes are carried to local lymphoid organs, such as the mLNs, processed, and presented to T cells to initiate adaptive immune responses (Macpherson and Uhr, 2004; Niess et al., 2005).

Ablation experiments using CD11c^DTR mice, conditional deletion of genes (e.g., Irf4, Irf8, or Notch2) with CD11c^Cre mice, and Batf3-deficient mice result in loss of cDCs and cDC subsets, and therefore concomitant loss of both innate and adaptive immune functions (Hildner et al., 2008; Lewis et al., 2011; Persson et al., 2013; Schlitzer et al., 2013). For example, CD11c^Cre-mediated DC deletion has been described to activate a feedback loop that produces high levels of circulating Flt3L, resulting in a myeloproliferative syndrome characterized by extramedullary hematopoiesis, splenomegaly, and neutrophilia (Birnberg et al., 2008). Depletion of cDCs in this model likely has additional consequences beyond elevated Flt3L levels. Our data suggests that the phenotype may have been mediated in part by intestinal microbes. In a different study, however, CD11c^Cre-mediated DC depletion was associated with colitis and increased numbers of activated CD4⁺ T cells (Birnberg et al., 2008; Ohnmacht et al., 2009). Although cDCs play an important role for T cell selection in the thymus (Perry et al., 2014), in our model we found no evidence of activated or autoreactive CD4⁺ T cells. In addition to being unable to distinguish between the innate and adaptive functions of cDCs, interpretation of these earlier studies was confounded in part by deletion of CD11c expressing monocytes, macrophages, pDCs, and NK cells. In contrast, deletion of MHCII in cDCs using zDC^Cre selectively ablates the ability of this cell type to initiate adaptive immunity while leaving its innate activity intact. The described phenotype was specific for cDCs because zDC expression is restricted to cDCs and their precursors. zDC is first expressed in committed cDC progenitors (pre-DCs) in the BM and is maintained in all subsequent stages of cDC development, including all major subsets of cDCs (Meredith et al., 2012; Satpathy et al., 2012).

Our experiments show that cDC-induced adaptive immune responses are required to maintain homeostasis between intestinal flora and the host. Other MHCII-expressing antigen-presenting cells, such as activated monocytes, macrophages, or monocyte-derived DCs, are unable to compensate for this requirement, because they are not specialized for CD4⁺ T cell priming. Absence of MHCII on cDCs in zDCΔMHCII mice is associated with intestinal inflammation and elevated serum levels of Csf3, an acute phase reactant. This phenotype is dependent on bacteria, as Csf3 levels dropped to normal levels when the mice were treated with antibiotics or housed under germ-free conditions. These results highlight a previously underappreciated role for cDC-mediated adaptive immune responses to commensal bacteria. Mice lacking MHCII on cDCs may therefore be considered immunocompromised, as they are unable to efficiently prime CD4⁺ T cells. Consistent with our observations, mice lacking CD4⁺ T cells and HIV-infected humans that suffer CD4⁺ T cell loss show increased microbial translocation and systemic immune activation (Brenchley et al., 2006; Sandler and Douek, 2012).

cDC-initiated CD4⁺ T cell responses are likely to help keep the gut microbiome in check by at least several distinct mechanisms. They are required to activate and induce differentiation of T cells, such as protective TH17/TH22 cells, which prevent invasion of enterobacteria (Mangan et al., 2006; Curtis and Way, 2009; Ivanov et al., 2009; Basu et al., 2012). Consistent with this idea, T cell–deficient mice and humans show increased risk for intestinal inflammation (Tamura et al., 2005). In addition, cDCs are also essential in initiating TFH responses that are required to produce germinal centers and high-affinity IgA (Crotty, 2011; Maruya et al., 2013; Kato et al., 2014; Kubinak et al., 2015). Finally, MHCII expression by cDC is required to maintain the physiological number of induced T reg cells (Darrasse-Jèze et al., 2009). In zDCΔMHCII mice, reduced numbers of T reg cells did not result in increased numbers of activated CD4⁺ T cells because their activation requires MHCII expression on cDCs. Nevertheless, decreased numbers of induced T reg cells could contribute to the severity of the phenotype by altering innate immune responses.

Colonization of germ-free mice allowed us to follow the immune response against a limited number of commensal bacteria. This controlled experiment showed that zDCΔMHCII mice have diminished TFH, GC, and IgA responses to intestinal bacteria. Mice that are unable to secrete polymeric antibodies into the intestinal lumen show elevated serum IgG levels specific for commensal bacteria, suggesting more translocation of intestinal bacteria in the absence of luminal IgA (Johansen et al., 1999). In addition, translocation of bacteria was observed in germ-free B cell–deficient mice that were colonized with SPF microbiota (Macpherson and Uhr, 2004), emphasizing the importance of the adaptive response in containing the microbiota. However, the phenotype observed in zDCΔMHCII mice cannot be mediated solely by a defective IgA response against intestinal bacteria. AID^−/− mice, which are unable to switch to IgA, do not develop the phenotype in our SPF facility, possibly because they have unimpaired TFH and germinal center responses and IgM can compensate in part for IgA. In zDC^Cre+MHCII^fl/fl mice, however, CD4⁺ TFH cell priming is defective, resulting in diminished GCs. This idea is supported by the observation that CD4⁺ T cell depletion partially phenocopies zDC^Cre+MHCII^fl/fl mice.

cDCs in zDCΔMHCII mice show intact innate microbial sensing and cytokine production. These features are essential for activation and normal function of innate lymphoid cells (Bernink et al., 2015). The observation that zDCΔMHCII mice develop intestinal inflammation indicates that intact innate mechanisms are insufficient to maintain the normal balance between host and intestinal commensals. Transferable risk for colitis has been described for mice lacking components of the inflammasome, which has been attributed to alterations in intestinal microbiota (Elinav et al., 2011; Palm et al., 2014). In our experiments, control mice did not develop any signs of inflammation, although they were cohoused with zDCΔMHCII mice. The fact that cohoused control mice did not get sick, along with the finding that they share a similar microbiome, indicates that commensal bacteria that can be controlled and are usually harmless in mice with an unimpaired immune system may be responsible for disease in zDCΔMHCII mice that have defective adaptive immune responses. Collectively, the data indicate that in the steady state, cDC-mediated CD4⁺ T cell priming and subsequent adaptive immune responses are essential to maintain homeostasis and prevent intestinal commensals from producing pathology.

zDC^Cre knock-in mice were generated by homologous recombination in C57BL/6 albino embryonic stem cells at The Rockefeller University Gene Targeting Resource Center (New York, NY). The targeting strategy was the same as for the previously described zDC^DTR mice (Meredith et al., 2012). zDC^Cre mice were maintained on a C57BL/6 background. Itgax^Cre (CD11c^Cre; stock number 008068), Rosa26^lSlYFP (Rosa26^lSlYFP; stock number 006148), iab^neo (MHCII^fl/fl; stock number 013181), and Irf4^fl/fl (IRF4^fl/fl; stock number 009380) mice were purchased from The Jackson Laboratory. Mice were housed in The Rockefeller University Comparative Bioscience Center (New York, NY) under SPF and germ-free conditions, respectively. All experiments were performed in accordance with the National Institutes of Health guidelines and approved by The Rockefeller University Animal Care and Use Committee. Germ-free mice were obtained by transfer of newborn pups (Cesarean section) into germ-free isolators. Germ-free status was confirmed by plating feces on a broad range of nutrient agars, as well as by qPCR analysis (16S rRNA). zDCΔMHCII mice were bred with MHCII^fl/fl mice. This way we could compare Cre⁺ (zDCΔMHCII) and Cre⁻ (control) littermates. It is also important to note that zDCΔMHCII males were bred with MHCII^fl/fl females. This way, the fostering female is healthy and does not develop the described phenotype, and development of pups is not affected by the genotype of the fostering female. Littermates were cohoused during all experiments. Mice were generally analyzed at 8–12 wk of age.

BM chimera recipients were irradiated with two doses of 525 rad 3 h apart and injected with 5–10 × 10⁶ BM cells. BM chimeras were analyzed 10 wk after irradiation.

For FACS analysis, cells were stained using antibodies against B220 (RA3-6B2), BST2 (eBio927), CXCR5 (L138D7), F4/80 (BM8), Foxp3 (FJK-16s), Gr-1 (RB6-8C5), IFN-γ (181157), IL-2 (JES6-5H4), IL-17A (TC11-18H10), Ly6C (AL-21), Ly6G (1A8), MHCII (M5/114.15.2), NK1.1 (PK136), Nrp1 (3DS304M), PD-1 (J43), RORg(t) (B2D), Siglec-H (440c), TCRβ (H57-597), CD3 (145-2C11), CD4 (RM4-5), CD8 (53–6.7), CD11b (M1/70), CD11c (N418), CD19 (1D3), CD38 (90), CD44 (1M7), CD45 (30-F110), CD62L (MEL-14), CD64 (X54-5/7.1), CD95 (Jo2), CD103 (2E7), CD115 (AFS98), TCRγδ (GL3), CD80 (16-10A1), and CD86 (GL1). TCR Vβ chain distribution was assessed with a FITC-labeled screening panel (BD). PI (Sigma-Aldrich), DAPI (Invitrogen), or Aqua live/dead (Invitrogen) was added to exclude dead cells from analysis. For intracellular cytokine staining, cells were stimulated for 4 h using Leukocyte Activation Cocktail with GolgiPlug (BD). Surface stained cells were fixed with Cytofix/Cytoperm (BD) and washed with Perm/Wash buffer (BD). Cells were analyzed on a LSR Fortessa flow cytometer (BD). Analysis was performed using FlowJo software (Tree Star).

For isolation of mononuclear phagocytes, spleen and LNs were mechanically disrupted and incubated with HBSS (Gibco) + 5% FBS + 0.5 mg/ml Collagenase D (Roche) at 37°C for 30 min. Cells were filtered through 100 µM cell strainers and washed. Erythrocytes were lysed with ACK Lysing buffer (Gibco). For isolation of T and B cells, spleen, LNs, thymi, and PP were mechanically disrupted, filtered through 100-µM cell strainers, and washed. Erythrocytes were lysed with ACK Lysing buffer (Gibco).

250,000 cDCs (CD11c MACS enrichment) were stimulated with 1 µM CpG 1668, 1 µg/ml LPS, and 1 µg/ml pIC, respectively. Expression of CD80 and CD86 were determined by flow cytometry 16 h later. Cytokines were measured using the Cytometric Bead Array kit (BD).

Small and large intestine were excised and freed of mesentery and fat. Feces were removed and PPs were excised from the small intestine. Both small and large intestine were opened longitudinally and washed in HBSS, followed by HBSS + 1 mM DTT (Sigma-Aldrich). Tissue was cut into 1-cm pieces and incubated with 25 ml of HBSS + 5% FBS + 5 mM EDTA for 15 min at 37°C at 230 rpm. Supernatant containing intraepithelial lymphocytes was decanted and spun down (for RNA isolation from epithelial cells, see next section). Tissue was washed with 10 ml of HBSS + 5% FBS for 15 min at 37°C at 230 rpm. The gut tissues were then finely chopped and digested in HBSS + 5% FBS + 1X Sodium Pyruvate + 25 mM Hepes + 50 µg/ml DNaseI + 0.05 mg/ml Collagenase VIII for 1 h at 37°C at 100 rpm. Cells were separated by discontinuous Percoll gradient (70%/35%) centrifugation. Cells were isolated from the interphase.

Epithelial cells were resuspended in TRIzol (Invitrogen), RNA was isolated according to the manufacturer’s protocol, and cDNA was generated using Superscript II (Invitrogen) and random primers. GAPDH forward, 5′-CTACAGCAACAGGGTGGTGG-3′, GAPDH reverse, 5′-AGATGGGGCATCTTTCTTGG-3′; S100A8 forward, 5′-TGCGATGGTGATAAAAGTGG-3′, S100A8 reverse, 5′-GGCCAGAAGCTCTGCTACTC-3′; S100A9 forward, 5′-CACCCTGAGCAAGAAGGAAT-3′, S100A9 reverse, 5′-TGTCATTTATGAGGGCTTCATTT-3′; Muc2 forward, 5′-ACAAAAACCCCAGCAACAAG-3′, Muc2 reverse, 5′-GAGCAAGGGACTCTGGTCTG-3′. All qPCR reactions were performed with Brilliant SYBR Green (Agilent Technologies) on an Mx3005P system (Agilent Technologies).

Serum levels of Flt3L and Csf3 were determined by ELISA (R&D Systems) according to the manufacturer’s protocol. Lipocalin-2 levels were measured in feces by ELISA according to the manufacturer’s protocol (R&D Systems). IgA concentrations were determined by ELISA according to the manufacturer’s protocol (eBioscience).

Tissue was fixed in 10% Formalin. Tissue was embedded, cut, and stained by the Tri-Institutional Laboratory of Comparative Pathology (New York, NY). A pathologist evaluated the tissue sections in a blinded fashion and took pictures representing the phenotype of the respective group.

CD8⁺ and CD4⁺ T cells were isolated from spleen and LNs of CD45.1⁺ OT-I and OT-II mice, respectively (MACS enrichment), labeled with violet tracem, and transferred into control and zDCΔMHCII mice (CD45.2) by i.v. injection (3–5 × 10⁶ cells/mouse). Mice were immunized (i.p.) with 100 µg OVA + 10 µg pIC. 3 d later, proliferation and expression of activation markers was analyzed by flow cytometry (pooled mesenteric and inguinal LN).

To deplete CD4⁺ T cells, mice were injected i.p. with 100 µg anti-CD4 (GK1.5; Bio X Cell) every 2–3 d for 4 wk. To deplete CD8⁺ T cells mice were injected i.p. with 100 µg anti-CD8 (53–6.72; Bio X Cell) every 2–3 d for 4 wk.

Ampicillin (1 mg/ml), Neomycin (1 mg/ml), Vancomycin (0.5 mg/ml), and Metronidazole (1 mg/ml; Sigma-Aldrich) were administered in the drinking water. Mice were put on Abx drinking water at the age of 8–10 wk and kept on Abx for 4 wk.

Body weight was measured weekly, as well as occurrence of rectal prolapse. Feces and serum samples were collected and stored at –20°C until further use.

16S rRNA FISH was performed with a universal probe (EUB388; Cy3-GCTGCCTCCCGTAGGAGT-Cy3) on sections fixed with Methacarn solution as described previously (Salzman et al., 2002; Swidsinski et al., 2005). In brief, intestinal tissues (∼1 cm), including intestinal contents, were fixed in Methacarn solution (60% methanol, 30% chloroform, and 10% glacial acetic acid) for 6 h at 4°C. After three washes with 70% ethanol, tissues were embedded in paraffin. 5-µm sections were cut and deparaffinized with xylene, and subsequently dehydrated. Probe was diluted (5 ng/µl) in hybridization buffer (0.9 M NaCl, 20 mM Tris-HCl, pH 7.2, and 0.1% SDS), and hybridization was performed at 50°C for 3 h in. After hybridization, tissues were washed with wash buffer (0.9 M NaCl and 20 mM Tris-HCl, pH 7.2) and nuclei were stained with Hoechst stain.

Bacteria were isolated from feces as described previously (Palm et al., 2014). Coating of bacteria with endogenous IgA was determined by staining bacteria with anti–mouse-IgA-PE (clone mA-6E1). Where indicated, MFIs were normalized to the sample with the lowest MFI to be able to compare different time points.

Fecal samples and intestinal contents were frozen immediately after collection in a dry ice/ethanol slurry and stored at −80°C. DNA was extracted using the phenol-chloroform and bead beating method described previously (Ubeda et al., 2010). The V4-V5 region of the 16S rRNA gene was amplified and sequenced with the Illumina Miseq platform, as described previously (Buffie et al., 2015). Sequences were analyzed using version 1.33.3 of the MOTHUR pipeline (Schloss et al., 2009) as described in Buffie et al. (2015). Sequences with distance-based similarity of at least 97% were assigned the same OTU (operational taxonomic unit) and representative OTUs were classified using a modified Greengenes reference database (DeSantis et al., 2006).

Germ-free control and zDCΔMHCII mice were colonized with a mix of six intestinal bacteria (E. coli [LF82], E. faecalis, B. longum, L. plantarum, B. vulgatus, and R. gnavus). Colonization was assessed by PCR using following primers: E. faecalis forward, 5′-CCCTTATTGTTAGTTGCCATCATT-3′ E. faecalis reverse, 5′-ACTCGTTGTACTTCCCATTGT-3′; B. longum forward, 5′-GGGTGGTAATGCCGGATG-3′ B. longum reverse, 5′-TAAGCGATGGACTTTCACACC-3′; L. plantarum forward, 5′-AGCAGTAGGGAATCTTCCA-3′ L. plantarum reverse, 5′-CACCGCTACACATGGAG-3′; B. vulgatus forward, 5′-GGTGTCGGCTTAAGTGCCAT-3′ B. vulgatus reverse, 5′-CGGAYGTAAGGGCCGTGC-3′; R. gnavus forward, 5′-GGAGCAGACCAGATCCAAAT-3′ R. gnavus reverse, 5′-CCAATATACATTCCCGGTCTTT-3′; E. coli forward, 5′-CCATTCATGCAGCAGCTCTTT-3′ E. coli reverse, 5′-ATCGGACAACATTAGCGGTGT-3′. Coating of bacteria with endogenous IgA and amount of total IgA in the serum and feces was determined at indicated time points.

Results are shown as means + SEM. Comparisons between groups was done using Student’s two-tailed t test analysis. Statistical significance was determined as p-values <0.05.

Fig. S1 shows flow cytometric analysis of zDC^Cre+IRF4^wt/fl reporter mice. Fig. S2 show bacteria-specific PCR to confirm colonization of germ-free mice.

We are grateful to members of the Nussenzweig laboratory for helpful discussion, reagents, and critical reading of the manuscript.

This work was supported in part by National Institutes of Health grant number AI 13013. M.C. Nussenzweig is a Howard Hughes Medical Institute investigator.

The authors declare no competing financial interests.