Microbiota regulate the ability of lung dendritic cells to induce IgA class-switch recombination and generate protective gastrointestinal immune responses

We first sought to determine that bona fide LDCs were responsible for IgA CSR. MHCII^hiCD11c⁺Siglec-F⁻ lung cells were collected from enzymatically digested lungs of C57BL/6 mice and FACS sorted. The gating strategy and purity of the populations, consistently >96%, are shown in Fig. S1. Additionally, MHCII^hiCD11c^hi DCs were FACS isolated from the spleen DCs (SpDCs), mesenteric LN DCs (MLN DCs), and skin-draining LN DCs (SkDCs) and co-cultured with CD43^negCD19⁺IgM⁺ B cells at a ratio of 2:1 DC/B cells for 4 d. B cells were stimulated with α-IgM (Wortis et al., 1995) and α-CD40 (Nonoyama et al., 1993; Cerutti, 2008). We observed that the induction of IgA⁺CD19⁺ B cells in LDC–B cell co-cultures was similar to MLN DC–B cell co-cultures and significantly greater than SpDC– and SkDC–B cell co-cultures (Fig. S2). In addition, secreted IgA (SIgA) was also quantified in the culture supernatants using IgA ELISA. In line with the flow cytometry data, the level of SIgA in the LDC cultures was similar to the levels of SIgA in MLN DC cultures and significantly higher than the levels in SpDC or SkDC cultures (Fig. S2). Interestingly, MHCII^loCD11c⁺Siglec-F⁺ cells putative alveolar Mφs (AM) failed to induce IgA CSR (unpublished data).

IgA CSR is mediated by activation induced cytidine deaminase (AID), a well-documented mammalian DNA-editing enzyme (Muramatsu et al., 2000), which is encoded by the Aicda gene. We observed that the level of Aicda transcripts were significantly higher in LDC- and MLN DC-cultured B cells compared with SpDC- and SkDC-cultured B cells, or B cells alone (Fig. S2). Additionally, Cα germ line transcripts (GLT; Stavnezer, 1996) were quantified and found to be significantly higher in B cells cultured with LDC and MLN DCs than those cultured with SpDC, SkDC, or B cells alone, further confirming the capacity of LDCs to induce IgA CSR (Fig. S2).

To define which of the lung APC subsets induce IgA CSR, we examined Siglec-F⁻MHCII^hiCD11c⁺CD103⁺CD11b⁻ cells (LDC CD103⁺), Siglec-F⁻MHCII^hiCD11c⁺CD103⁻CD11b⁺CD24⁺ cells (LDC CD24⁺), and Siglec-F⁻MHCII^hiCD11c⁺ CD103⁻CD11b⁺CD64⁺ cells (LMφ CD64⁺) separately based on recently described functional, ontogenetic, and phenotypic criteria to define lung APCs (Langlet et al., 2012; Schlitzer et al., 2013). MLN DCs, SpDCs, and SkDCs were used as controls. LDC CD103⁺, LDC CD24⁺, and MLN DC were potent inducers of IgA CSR. In contrast, the LMφ CD64⁺ were poor at inducing IgA CSR (Fig. 1, A and B). This was further confirmed by IgA quantification within the supernatants (Fig. 1 C). Additionally, LDC CD103⁺ and LDC CD24⁺ induced higher levels of Aicda (Fig. 1 D) and Cα GLT (Fig. 1 E) in B cells compared with LMφ CD64⁺ and SpDC. We also observed that LDCs induced IgA CSR in the absence of α-CD40 stimulation (which was used here as an in vitro surrogate of TI IgA CSR, albeit with caveats). The levels of IgA produced were expectedly lower in the (–) α-CD40 conditions compared with the (+) α-CD40 conditions (Fig. 1 F). Finally, we also investigated the ability of respective lung APC subsets to induce IgA CSR in the absence of α-CD40 stimulation (Fig. 1 G). Together, these data demonstrate that among the LDC subsets, the LDC CD103⁺ as well as the LDC CD24⁺, a subset of the heterogeneous CD11c⁺MHCII^hiCD11b⁺ population, induce IgA CSR, whereas the LMφ CD64⁺ (also contained within the CD11c⁺MHCII^hiCD11b⁺ cells) do not.

There is strong evidence that lung draining LNs (such as mediastinal LNs [MedLNs]) act as the initial and major sites of B cell response, preceding induction of antibodies in the lungs (Sangster et al., 2003; Sealy et al., 2003). Therefore, we next examined the induction of IgA⁺ B cells in the lung and draining LNs (MedLNs) at steady state and after immunization and observed a progressive increase in IgA⁺ B cells in both MedLNs and lungs after i.n. immunization with cholera toxin (CT). Although these data cannot confirm that MedLNs were the site of initial DC–B cell interaction, the fact that MedLN IgA response preceded the response in the lungs is consistent with previous data wherein antibody forming cells are identified in the cervical and MedLN as early as 2–3 d after infection (Sealy et al., 2003; Choi and Baumgarth, 2008; Fig. 1 H).

Having established that LDC CD103⁺ and LDC CD24⁺ induce IgA CSR, whereas LMφ CD64⁺ do not, we examined the mechanisms involved. Global transcriptome analysis was used to compare gene expression levels in the LDC CD103⁺, LDC CD24⁺ and LMφ CD64⁺ subsets. Using the SAM method (Tusher et al., 2001) with a false discovery rate (FDR) threshold of 5%, genes differentially expressed between all pairs of the three lung APCs were compared. Enriched pathways were identified in these gene lists using the DAVID tool (Huang et al., 2009). Among the enriched pathways, the “Intestinal immune network for IgA production” pathway (LDC CD24⁺ vs. LMφ CD64⁺; P = 0.04; Benjamini q = 0.2) was investigated further. Most notably, Tgf-β1, which was determined to have an unequivocal role in inducing IgA within the GI tract (Stavnezer et al., 2008), was up-regulated in both LDC CD103⁺ (2.3-fold) and LDC CD24⁺ (19.2-fold) compared with LMφ CD64⁺ (Fig. 2 A). Given that TGF-β must be activated to exert its biological effects and that TGF-β mRNA levels do not correspond to active TGF-β protein (Annes et al., 2003), we quantified the expression of TGF-β–activating integrin αvβ8 in sorted LDC/Mφ subsets by qPCR. Both LDC CD103⁺ and LDC CD24⁺ expressed significantly higher levels of integrin β8 compared with LMφ CD64⁺ and SpDCs (Fig. 2 B). These data suggested that LDC CD103⁺ and LDC CD24⁺ were equipped with the machinery to convert latent TGF-β into its active form.

Vitamin A deficiency is associated with a selective loss of IgA-producing cells (Mora et al., 2006), and recent reports show that RA acts as a cofactor in IgA CSR (Seo et al., 2013). We specifically examined the role played by RA in IgA CSR by LDCs and examined RAL DH2 (Aldh1a2) mRNA transcripts in the unstimulated, flow-sorted LDC and LMφ populations. Aldh1a2 expression was higher in the LDC CD103⁺ and LDC CD24⁺ compared with LMφ CD64⁺ (Fig. 2 C). These data are concordant with the Immgen Consortium data (Heng and Painter, 2008). Further, addition of a RAR-β inhibitor (RAR-β), LE540 to the DC/B cell cultures resulted in a significant reduction of IgA levels in both LDC CD103⁺ and LDC CD24⁺ B cell co-cultures (Fig. 2, D–F) across a range of LE540 concentrations (2.5 µM, 250 nM, and 25 nM).

Finally, having observed differential expression of Tgf-β1 gene within LDC and LMφ, we investigated the effect of inhibiting TGF β on IgA CSR by LDCs. An α-TGF-β–blocking antibody was added in a range of concentrations on days 0 and 3. This resulted in reduced IgA CSR by the LDC subsets (Fig. 2, G–I). Interestingly, the effect of TGF-β inhibition appeared to be more pronounced in the LDC CD103⁺ cultures than the LDC CD24⁺ cultures (Fig. 2, G and H). Furthermore, a dose-dependent TGF-β inhibition was seen with LDC CD103⁺ but not with LDC CD24⁺ (Fig. 2 I).

In addition to genes involved in T cell–mediated IgA CSR, we investigated the expression of factors involved in TI IgA CSR, including the TNF family members: APRIL (Hahne et al., 1998) and B lymphocyte stimulator protein (BLys or BAFF; Moore et al., 1999; Litinskiy et al., 2002). Unlike the SpDC, which had negligible levels of APRIL and did not induce IgA CSR, LDC CD103⁺, LDC CD24⁺, and LMφ CD64⁺ all expressed high levels of APRIL protein (Fig. 3 A), with expression being highest in the LDCs CD24⁺ (Fig. 3 B). APRIL expression was further confirmed using immunohistochemistry. Again, all three lung APC subsets expressed APRIL proteins, in contrast to SpDCs, which lacked APRIL expression (Fig. 3 C). These data were further corroborated by comparing mRNA levels of APRIL and BAFF within sorted lung APC subsets (Fig. 3 D). Incidentally, we observed the highest levels of BAFF mRNA in the LMφ CD64⁺ compared with LDC CD103⁺ and LDC CD24⁺ (Fig. 3 D).

We next studied the effect of inhibiting APRIL and BAFF in vitro. A recombinant mouse TACI-Fc decoy receptor (Schneider et al., 1999; He et al., 2010) was added to the DC/B cell cultures on days 0 and 3. Nonsignificant reduction in IgA CSR was noted in both LDC CD103⁺ and LDC CD24⁺ (Fig. 3, E and F). Similar results were observed using TACI^−/− B cells (unpublished data). In addition to BAFF and APRIL, we also investigated IL-10 as a potential mediator of LDC-mediated IgA CSR by adding α-IL-10–blocking antibody to the DC/B cell cultures on days 0 and 3. As shown in Fig. 3 (G and H), this did not significantly affect the induction of IgA mediated by LDC CD103⁺, although a decrease in IgA CSR was observed with the LDC CD24⁺. Thus, whereas LDC CD103⁺ and LDC CD24⁺ induced IgA CSR, and LMφ CD64⁺ did not, all three lung APC subsets expressed APRIL and BAFF.

To definitively understand the difference between the IgA class-switching ability of LDCs and LMφ, we asked whether IgA CSR by the LMφ CD64⁺ could be induced by the addition of RA and TGF-β. To this effect, LMφ CD64⁺ were cultured with IgM⁺ B cells with or without TGF-β or TGF-β and RA. TGF-β alone resulted in increased IgA CSR; however, TGF-β and RA had a synergistic affect and induced significantly higher levels of IgA, similar to LDC CD103⁺ (Fig. 3, I and J). In contrast, addition of APRIL to the LMφ CD64⁺-B cell cultures failed to induce IgA CSR and no additive effect was observed upon addition of APRIL with TGF-β and RA (Fig. 3 K). Combined, these results show that although LDC CD103⁺, LDC CD24⁺, and LMφ CD64⁺ all express APRIL and BAFF, which are powerful TI IgA class switching molecules, the additional ability of LDC CD103⁺ and LDC CD24⁺ to generate TGF-β and RA confers upon them the ability to induce IgA CSR.

Bacterial populations within the gut have been shown to influence IgA CSR via intestinal B cells (Macpherson and Uhr, 2004). To determine the role of microbiota in LDC-mediated IgA CSR, we sorted MHCII^hiCD11c⁺Siglec-F⁻ lung cells from specific pathogen–free (SPF), germ-free (GF), and antibiotic-treated mice and cultured them in the presence of α-IgM– and α-CD40­–stimulated CD43^negCD19⁺IgM⁺ B cells (Fig. 4, A and B). Mice were subjected to 4 wk of orally administered combination of antibiotics comprised of vancomycin, neomycin, metronidazole, and ampicillin. This treatment resulted in significant changes in the composition of culturable commensal bacteria in the gut (unpublished data). Interestingly, IgA CSR was significantly reduced in the GF LDCs and antibiotic-treated LDCs compared with SPF LDCs. This was further confirmed with SIgA quantification within the supernatants (Fig. 4 C). Additionally, the levels of Aicda and Cα GLT were significantly reduced in B cells cultured with GF LDCs compared with SPF LDCs. There was a reduction in antibiotic-treated mice, but the differences did not reach statistical significance (Fig. 4, D and E). These data suggest that commensal microbiota influenced the capacity of LDCs to mediate IgA CSR.

To better understand the mechanisms involved, global transcriptome analysis was used to compare gene expression levels between SPF and GF LDC subsets. Examination of the “Immune network for IgA production” pathway revealed several differentially expressed genes, with the most notably down-regulated genes within GF lung APCs being Tgf-β1 and Aldh1a2. Strikingly, CD24⁺ LDC from GF mice expressed 24-fold lower levels of TGF-β1 compared with CD24⁺ LDC from SPF mice (Fig. 4 F), whereas the expression of Aldh1a2 was decreased by 2.3-fold.

Having observed a significant reduction in the ability of LDC from GF and antibiotic-treated mice to induce IgA CSR, we sought to determine ways to rescue the class switching function of LDCs from antibiotic-treated mice. SPF mice were maintained on the antibiotic cocktail long term (12 wk). At week 12, half of the mice were gavaged with 100 µg of LPS daily for 5 d, whereas the other half were gavaged with 100 µg of PBS daily for 5 d. 12-wk antibiotic treatment further reduced the ability of LDCs to induce IgA CSR (Fig. 4 G) to levels observed in GF mice. Oral administration of LPS completely restored LDC IgA CSR to SPF levels. To corroborate the LPS data, we asked whether reconstitution of GF mice with WT (SFP) microbial flora could restore LDC’s capacity to mediate IgA CSR. Remarkably, reconstitution of GF mice with cecal contents from SPF mice for 2 wk was sufficient to restore the IgA CSR ability (Fig. 4, H and I). Based on this set of experiments, we conclude that microbial signals profoundly affect the ability of LDC to induce IgA class switching.

We next examined the mechanisms by which microbial signals imprint IgA class-switching function on LDCs. Because pattern recognition receptors are essential in cellular response to microbial stimuli, we studied the IgA switching function of LDCs from TRIF^−/−MyD88^−/− mice. Strikingly, in the absence of TLR signaling, the IgA class switching function of LDCs was completely abolished (Fig. 5, A–C). Aicda and Cα GLT levels were also significantly reduced in B cells stimulated with TRIF^−/−MyD88^−/− LDC (Fig. 5, D and E). The adaptor protein MyD88 is recruited by TLR1, TLR 2, TLR 4, TLR5, TLR6, TLR7, and TLR 9, whereas TRIF is recruited by TLR3, TLR4, and TLR5 (O’Neill et al., 2013). To further understand the specific pathways involved, we examined TRIF^−/− and MyD88^−/− single knockout mice. Although inhibition of TRIF signaling partially impaired LDC IgA CSR (Fig. 5, F and G), inhibition of MyD88 signaling alone completely abrogated the IgA class switching function of LDCs. Collectively, these data indicate that TLR signaling, mediated through the MyD88 adaptor protein, is essential for LDCs to induce IgA CSR.

Having observed that the TGF-β1 gene was highly down-regulated in the LDC from GF mice compared with SPF mice, we sought to examine if lack of TLR signaling reduces TGF-β production. The levels of active TGF-β1 protein produced by GF and TRIF^−/−MyD88^−/− LDCs were quantified after a 5-d culture. We found a significant reduction in active TGF-β1 production by both GF and TRIF^−/−MyD88^−/− LDCs compared with SPF LDC in vitro (Fig. 5 H). Reduced levels of active TGF-β1 were also detected in the lung tissues of these animals (unpublished data).

Next, we asked whether the IgA CSR by TRIF^−/−MyD88^−/− LDCs could be restored by the addition of TGF-β1. As demonstrated in Fig. 5 (I and J) supplementation of the TRIF^−/−MyD88^−/− LDC cultures with TGF-β1 restored the level of IgA⁺ B cells to that seen with WT DCs. In contrast, while addition of TGF-β to SpDC increased the IgA⁺ B cells compared with SpDC alone, the levels remained significantly lower than LDC from WT mice. Finally, we asked whether the lack of IgA CSR in TRIF^−/−MyD88^−/− LDC was a result of impaired production of RA. We added RA to TRIF^−/−MyD88^−/− LDC in vitro and observed that the level of CD19⁺IgA⁺ cells increased significantly compared with TRIF^−/−MyD88^−/− DCs alone. Supplementation of TGF-β1 resulted in significantly enhanced production of IgA, approaching the levels seen with WT LDCs (Fig. 5, I and J). Together, these data demonstrate a critical, TGF-β1–dependent role of MyD88-dependent TLR signaling in the IgA class switching function of LDC.

After CSR, B cells need to migrate to the mucosal sites to perform effector functions. Mora et al. (2006) have demonstrated that gut-associated lymphoid tissue (GALT)–resident-DCs metabolize vitamin A to RA and induce gut-homing receptors on B cells. Because we (Ruane et al., 2013) and others (Guilliams et al., 2010) have shown that LDCs can also metabolize vitamin A, we investigated the ability of LDCs to induce gut-homing signals on B cells. We examined total LDCs (identified as MHCII^hiCD11c⁺Siglec-F⁻ cells) and compared with SpDCs, MLN DCs, and SkDCs for their ability to induce gut-homing signals on CD45⁺CD43^negCD19⁺ B cells (DC/B cell ratio of 2:1; cultured for 5–7 d). The expression of α4β7 and CCR9 was significantly up-regulated by LDCs, similar to MLN DCs (Fig. 6, A–C). The addition of an RAR-β antagonist, LE540, decreased the expression of both α4β7 and CCR9 by LDC and MLN DC stimulated B cells (Fig. 6, D–E). Therefore, we conclude that LDCs induce gut-homing molecules on B cells, mediated by RA signaling.

Next, we investigated whether the induction of the gut-homing molecule, α4β7 is specific to IgA class-switched cells. Upon culture of CD43^negCD19⁺IgM⁺ B cells with LDC, SpDC, SkDC, and MLN DC (DC/B cell ratio, 2:1), the expression of α4β7 on IgA⁺ B cells was compared with IgA⁻ B cells. Notably, α4β7 expression was comparable between IgA⁺ and IgA⁻ B cells (Fig. 6 F). Additionally, LDC-induced IgA-positive cells (and total B cells; unpublished data) had higher expression of α4β7 than SpDC and SkDC, but lower than MLN DC–cultured B cells (Fig. 6, F and G). To determine if α4β7 and CCR9 induction occurred in a cell contact–dependent manner, LDC-, SpDC-, SkDC-, and MLN DC–conditioned media were added to α-IgM– and α-CD40–stimulated B cells and the induction of α4β7 and CCR9 was investigated. LDC-conditioned medium significantly up-regulated gut-homing molecules, equivalent to MLN DC–conditioned medium (unpublished data). Therefore, LDC-mediated gut-homing molecules are induced in a contact-independent manner, which could explain why the induction of α4β7 was equivalent between IgA⁺ and IgA⁻ B cells.

In addition to the in vitro studies, we also examined the induction of gut-homing molecules on IgA⁺ B cells in the lung ex vivo at steady state and after i.n. vaccination. As shown in Fig. 6 (H and I), the expression of α4β7 and CCR9 by IgA⁺ B cells significantly increased on days 2, 4, and 8 after i.n. vaccination compared with baseline.

We then asked if there was a difference between the LDC subsets in their ability to induce expression of integrin α4β7 and chemokine receptor 9 (CCR9) on B cells. LDC CD103⁺ and LDC CD24⁺ subsets, as well as LMφ CD64⁺, were cultured with B cells and the expression of α4β7 and CCR9 was quantified. LDC CD103⁺ and LDC CD24⁺ both up-regulated integrin α4β7 and CCR9 on B cells, whereas LMφ CD64⁺ did not (Fig. 6, J and K).

Given the capacity of LDCs to induce gut-homing molecules in vitro and ex vivo, we hypothesized that this mechanism might target B cells to the GI tract in vivo. LDCs and SpDCs, and MLN DCs and SpDCs, were cultured with CD45.2⁺CD43^negCD19⁺IgM⁺ B cells for 5 d, and the respective B cells were labeled with CFSE or cell tracker (CTr) and i.v. co-transferred into CD45.1⁺ recipient mice. Homing index (HI) was calculated as output CFSE:CTr/input CFSE:Ctr for each of the organs examined. As shown in Fig. 6 L, the HI in spleen was 0.9. In contrast, the HI in the small intestinal lamina propria (SILP), colonic lamina propria (CLP), and lung was 3.7, 7.6, and 8.2, respectively. These data demonstrate that, in contrast to SpDC-cultured B cells, LDC-cultured B cells home preferentially to the SILP, CLP, and lung. As previously published (Mora et al., 2006), we found that MLN-stimulated B cells homed more efficiently to the CLP and SILP but not to the spleen or the lung (Fig. 6 L). Thus, LDCs have the capacity to induce gut-homing molecules on IgA⁺ B cells, mediating efficient trafficking of B cells to GI effector compartments.

Having observed that LDCs can induce IgA CSR, confer a gut-homing phenotype, and direct the migration of B cells to the GI tract, we wanted to address whether targeting LDCs can mediate protective immunity in the GI tract. For this we used a CT-induced diarrhea model (Tokuhara et al., 2010) characterized by cAMP-induced secretion of water and chloride ions into the small intestine (Spangler, 1992). CT-specific secretory IgA (sIgA) in the intestinal lumen is responsible for protective immunity against CT-induced diarrhea (Tokuhara et al., 2010). Mice were either left untreated (WT+CT) or immunized with CT delivered intranasally (i.n.), intratracheally (i.t.), s.c., or per orally (p.o.). Notably, the i.t. route was specifically chosen to compare with i.n. immunization to minimize possible GI exposure of the i.n.-delivered vaccine. To determine the optimal volume of vaccine antigen, we conducted targeting experiments where C57BL/6 mice were vaccinated i.n. with 50 µg OVA-AF647 delivered either in a volume of 5 or 40 µl. In a volume of 5 µl OVA-AF647, there was no detectable signal in the LDC subsets. In contrast, 40 µl of OVA-AF647 targeted LDCs effectively (Fig. S3 A). To exclude inadvertent GI exposure of the i.n. delivered vaccine, further targeting experiments were performed where we examined antigen-presenting cells in the spleen, MLN, or GI tissue at various time points after i.n. delivery of 40 µl of a fluorescently labeled antibody. Notably, we did not detect antibody-labeled cells in the spleen, MLN, or GI tissues up to 72 h after administration of the i.n. vaccine; however, labeling in the lung and MedLN remained robust (Fig. S3 B). Guided by these targeting experiments, the i.n. vaccine was delivered in a volume of 40 µl. On day 14, mice were challenged orally by administering CT by gavage. At 2.5 h after CT challenge, mice were sacrificed and the weight of the small intestine and its fluid content was determined. We observed that CT immunization delivered via the s.c. route did not result in protection against CT challenge. In contrast, i.n or i.t. delivery of the CT vaccine generated protective immunity against CT challenge, which was comparable in degree to WT and p.o. vaccinated animals (Fig. 7 A). Additionally, we measured anti-CT antibodies in the serum, spleen, and stool of the vaccinated mice. Although s.c. immunization induced strong IgG responses in the serum, it failed to elicit detectable anti-CT IgA in the stool. In contrast, i.n. and i.t. vaccinated mice developed CT-specific IgA in stool and serum, as well as a serum IgG response (Fig. 7 B).

To exclude the possibility that serum IgA could have been transported in the bile to the GI tract, we examined the frequency of antigen-specific B cells within the lung and GI tissues using CT-specific ELISPOT. Both i.n. and i.t. immunization resulted in the generation of CT-specific B cells in the GI tissue (Fig. 7 C). To further examine if the nasally delivered vaccine was leading to GI tract exposure, we compared i.n. immunized mice with i.t and sublingual (s.l.) immunized mice. Notably s.l. vaccination induced CT-SIgA in the stool and serum, albeit at a significantly lower level than i.n. and i.t. vaccination (Fig. 7 D). The finding that i.n. and i.t. vaccination both induce protective immunity against CT induced-diarrhea prompted us to investigate whether targeting LDCs modifies the outcome of the humoral immune response. To this effect, CT was administered i.n. at a volume of either 5 or 40 µl (with the dose of antigen being constant). We determined that immunization using a volume of 5 µl and targeting only nose-associated lymphoid tissue DCs was insufficient to induce protective immunity in the GI tract (Fig. 7, E and F). Additionally, CT-specific IgA was significantly lower in both stool and serum samples from these animals (Fig. 7 F). Therefore, targeting of the LDC populations is required to induce protective IgA immunity in the GI tract.

To assess the in vivo contribution of LDCs to IgA induction, we immunized Batf3^−/− mice lacking CD103⁺ DCs (Hildner et al., 2008) and WT mice with CT and serially examined the induction of IgA⁺ B cells in the lung and MedLN. A significant reduction in IgA⁺ cells was seen in the lung and MedLN of Batf3^−/− mice compared with WT mice after i.n. immunization (Fig. 7 G). Next, CT-immunized (delivered i.n. or i.t.) WT or Batf3^−/− mice were challenged orally with CT. Batf3^−/− mice showed reduced protection after i.n. and i.t. immunization, which was associated with significantly lower levels of stool and serum IgA compared with WT mice (Fig. 7, H and I). These data demonstrate a critical role for Batf3⁺ DCs in the induction of IgA responses, which cannot be compensated for by CD24⁺ DCs.

To determine if i.n. vaccination could lead to a protective immune response in the gut in the absence of T cell help, TCR^−/− mice were immunized i.n. with CT and challenged orally after 14 d. As shown in Fig. 7 J, whereas unimmunized, CT-challenged mice had significantly greater intestinal weight because of the accumulation of enteric fluid, immunized TCR^−/− mice were protected against CT challenge.

Finally, having determined a profound effect of microbiota on the production of IgA antibodies by LDC, we assessed the physiological contribution of the microbiota during i.n. vaccination. Using the CT-induced diarrhea model, WT (SPF conditions) and GF mice were immunized with CT delivered i.n. and challenged on day 14. Due to the fact that intestinal weight of GF mice is significantly greater than WT mice (because GF mice accumulate fluid in the intestinal lumen; Fig. 7 K), we focused on CT-specific IgA in the stool and observed significantly reduced levels of CT-specific IgA in the stool, lung, and serum of GF mice compared with SPF mice (Fig. 7 L). These data demonstrate that the efficacy of i.n. vaccination is significantly modulated by the microbiota.

Induction of a robust antibody response through mucosal vaccination offers many advantages, including inhibition of microbial adherence, neutralization of toxins, opsonization of bacteria, targeting of antigen to APCs, and antibody-dependent cellular cytotoxicity (Nimmerjahn and Ravetch, 2006; Cerutti and Rescigno, 2008). IgA is the predominant antibody at mucosal surfaces. Therefore, an understanding of the signals that regulate IgA CSR is critical to the design of effective mucosal vaccines. With this as our guiding principle, we undertook this study to determine the factors that influence the production of IgA after i.n. vaccination and have made the following observations.

First, bona fide LDCs, both LDC CD103⁺ and LDC CD24⁺, are proficient in inducing IgA CSR, whereas the closely related LMφ CD64⁺ are not. Second, microbial stimuli, acting through MyD88-dependent TLRs, induce the IgA class-switching function of LDCs via the production of TGF-β. Third, LDCs can up-regulate integrin α4β7 and CCR9 in a RA-dependent fashion, direct antigen-specific B cells to the GI tract, and mediate protection against enteric pathogens after i.n. vaccination. Finally, microbial factors impact the protective effect of i.n. vaccination, demonstrating a bidirectional cross talk between lung and gut immunity.

In contrast to the well-studied process of IgA CSR by gut DCs (Macpherson and Uhr, 2004; He et al., 2007; Tsuji et al., 2008), there are limited studies examining the induction of IgA by LDCs (Wilkes and Weissler, 1994; Naito et al., 2008; Suzuki et al., 2012), with inconsistent results (Wilkes and Weissler, 1994; Naito et al., 2008). This is partly because the description of LDC has been hampered by the overlapping expression of phenotypic markers on bona fide DCs and Mφ in the lung (Ginhoux et al., 2009). Recent studies have clarified this issue such that the expression of CD24 and CD64 on Siglec-F⁻CD11c⁺MHCII^hiCD11b⁺ cells distinguishes LDCs from LMφ contained within the heterogeneous, previously called CD11b⁺ LDC (Schlitzer et al., 2013). Using the same stringent criteria to define LDCs (purity of respective populations in our assays was >96%), we used both molecular indicators of IgA CSR (aicda induction and Cα GLT production) and protein quantification of cellular and soluble IgA to document the IgA CSR function of LDC. Through these complementary but distinct assays, we were able to conclude that whereas LDC CD103⁺ and LDC CD24⁺ induce robust IgA CSR, LMφ CD64⁺ are significantly less proficient. To our knowledge, this is the first study distinguishing bona fide LDC and LMφ in their ability to induce IgA class switching.

Further, we found that LDCs induced IgA CSR in both TD and TI conditions, although TI conditions produced less IgA than TD conditions. Notably, much of the data on respiratory TD versus TI Ig class switching comes from the influenza infection model wherein virus infection elicits a vigorous, CD4⁺ T cell–dependent antibody response in WT mice (Bishop and Hostager, 2001). Although robust, TD IgG and IgM antibodies against influenza are well documented, perhaps the strongest evidence for a TI IgA response in the lung comes from an elegant study by Sangster et al. (2003). This group demonstrated that MHC II–deficient and CD40-deficient B cells produced little influenza virus-specific IgM and IgG, but generated a strong virus-specific IgA response with virus-neutralizing activity. They ascribed the TI IgA response to bystander help from T cells without being able to discern a clear source. Data presented in this study reconcile the observations of Sangster et al. (2003) and demonstrate that LDCs are able to induce IgA CSR, independent of T cell help, in a TGF-β and RA-dependent manner.

IgA plays important and diverse roles in mucosal homeostasis. In particular, opsonization and neutralization of pathogens and toxins results in immune exclusion, reducing antigenic load on the mucosal immune system. This is perhaps best demonstrated in protection models that use bacterial exotoxins (as used in the present study) or for pathogens that reside exclusively in the GI lumen. Several studies demonstrated a protective role for IgA at mucosal surfaces, wherein IgA^−/−mice show impaired clearance of luminal pathogens such as Giardia muris and Giardia lamblia (Langford et al., 2002), Citrobacter rodentium (Simmons et al., 2003), Streptococcus pneumoniae (Sun et al., 2004), and Mycobacterium tuberculosis (Rodríguez et al., 2005). In contrast, other studies have shown that protection against Shigella flexneri was similar in IgA^−/− mice compared with WT mice (Way et al., 1999), implying a less critical role for IgA in mucosal defense. Yet another study showed IgA^−/− mice were equally susceptible to primary infection with reovirus as WT mice, but showed increased susceptibility to reinfection (Silvey et al., 2001), again suggesting less of a role for IgA in defense against acute infection. However, an important and often overlooked caveat in such “negative” studies is that IgA deficiency in IgA^−/− mice is accompanied by a compensatory increase in luminal IgM, perhaps as a result of reduced competition for the polymeric Ig (pIg) receptor (Harriman et al., 1999). Additionally, human data from patients with IgA deficiency clearly shows predisposition to sinopulmonary infections, colitis, and autoimmune diseases (Cunningham-Rundles, 2001; Jacob et al., 2008).

That said, the goal of our study was to examine the class switching function of LDCs in the context of nasal vaccination. To examine the physiological role of LDCs in IgA generation, we immunized Batf3^−/− mice lacking CD103⁺ DCs i.n. and i.t., and observed reduced levels of CT-specific IgA in the stool and serum of Batf3^−/− mice compared with WT mice. Furthermore, the immunized Batf3^−/− mice showed impaired protection against enteric challenge with CT. Combined, these data demonstrate that lung CD103⁺ DCs are critical to the IgA response after CT immunization and are consistent with previous data that demonstrate a protective role of CD103⁺ DCs in influenza infection (Helft et al., 2012).

Next, we looked for mechanistic factors within the lung APC subsets that determined their ability to induce IgA CSR and observed several important differences. First, the TGF-β1 gene was up-regulated 19-fold in LDC CD24⁺ and 2.5-fold in LDC CD103⁺ compared with LMφ CD64⁺. Second, Aldh1a2, encoding the vitamin A–metabolizing enzyme RALDH2, was up-regulated 10-fold in LDC CD24⁺ and sixfold in LDC CD103⁺ compared with LMφ CD64⁺. Third, several B cell relevant genes were up-regulated in the LDC CD24⁺ and LDC CD103⁺ compared with LMφ CD64⁺. Notable examples include NF-κb pathway-related genes that are associated with DC maturation (Nfkb2, Chuk, and Ikbkb), antigen presentation (H2-ab1, H2-dma, H2-dmb1, and H2-dmb2), Icos (Haimila et al., 2009), Map3k14 (Tsuji et al., 2008), and Ltβr (Tsuji et al., 2008). Mutations in all of these genes have been associated with IgA deficiency.

Interestingly, LDC CD24⁺, LDC CD103⁺, and LMφ CD64⁺ all expressed APRIL and BAFF, powerful IgA class-switching molecules (Litinskiy et al., 2002; He et al., 2007; Xu et al., 2007). In fact, whereas LMφ CD64⁺ expressed the highest levels of APRIL and BAFF mRNA, the protein levels were equivalent between all the three APCs, likely reflecting the previously reported, extensive posttranscriptional and posttranslational (Stohl et al., 2004; Kawasaki et al., 2007) modification of these proteins. Thus, even though we did not observe IgA class switching by LMφ CD64⁺ in vitro, given the high expression of APRIL and BAFF, it is possible that LMφ CD64⁺ could cooperate with other cells in vivo and also participate in IgA class switching. Indeed, our data, in which addition of TGF-β and RA to LMφ CD64⁺ B cell cultures resulted in robust IgA class switching, supports this possibility.

Investigation of the factors that modulate LDC ability to induce IgA CSR revealed an unexpected but significant role of the microbiota. Although the commensal microflora have a clear role in modulation of gut (Hill and Artis, 2010) and skin (Naik et al., 2012) immunity, their role in lung immune responses is less clear. Previous studies have linked changes within the microbiome with susceptibility to allergic airway disease (Noverr et al., 2005; Olszak et al., 2012; Russell et al., 2012, 2013) and influenza infection (Ichinohe et al., 2011). Here, we demonstrate that microbiota, acting through TLRs, condition LDCs to produce TGF-β and profoundly influence IgA production in the lung. Thus, the present study offers the first direct evidence linking microbial signals with the IgA class-switching function of LDCs. We are in the process of investigating whether IgA CSR function of LDCs is affected by the enteric microbiome or by oropharyngeal/pulmonary microbial communities.

Our next step was to investigate the migratory properties of LDC-induced B cells. Given that RA and TGF-β, shown here to influence IgA CSR, also induce gut-homing receptors, we hypothesized that B cells will up-regulate α4β7 and CCR9 upon culture with LDC. Contact-independent induction of gut-homing receptors was found to occur with the LDC CD103⁺ and LDC CD24⁺ but not LMφ CD64⁺. In addition, after i.n. immunization, the frequency of IgA⁺α4β7⁺CCR9⁺ B cells in the lung progressively increased over 8 d. In vitro, the expression of α4β7 was seen on both IgA⁺ and IgA⁻ B cells, which is in line with our observation that this was caused by secreted factors. These data are in agreement with previous results from our group, where we reported on the induction of gut-migratory T cells by LDC (Ruane et al., 2013) and support the concept of a common mucosal immune system originally proposed by Bienenstock et al. (1978).

In this study, we observed that LDC-conditioned B cells migrated to both the small intestines and the large intestines. Although we could identify markers known to be associated with small intestinal homing (CCR9; Mora et al., 2003), and panintestinal homing (α4β7; Villablanca et al., 2011), we did not detect the expression of GPR15, the only known large intestine–specific homing molecule (Kim et al., 2013) on LDC-induced B cells.

We undertook several lines of investigation to exclude inadvertent targeting of gut DCs by i.n. vaccination. First, we performed in vivo targeting experiments demonstrating that fluorescent antibodies delivered i.n. did not label gut-resident DCs. Second, we used the i.t. route of vaccine delivery to specifically target LDCs. Here, we found that the induction of intestinal IgA was comparable between i.t. and i.n. vaccination with CT and both routes of vaccine delivery, resulting in subsequent protection against GI challenge with CT. Finally, to exclude that serum IgA was being delivered via bile to the intestines, we found CT-specific IgA⁺ cells in the GI tract using ELISPOT after i.n. vaccination with CT.

Because there is variance in the literature regarding the protective effect of i.n. vaccines, we have explored this further in our study. We hypothesized that i.n. vaccines that target LDCs may have a different effect than vaccines that do not. To test our hypothesis, we delivered the same dose of antigen in two different volumes: 5 and 40 µl. In vivo targeting experiments were done to show that at a volume of 5 µl, nasal DCs were targeted, and at 40 µl, nasal DCs, LDCs, and MedLN DCs were all targeted. Guided by these experiments, we immunized mice with CT (2 µg) delivered in a volume of either 5 or 40 µl. As shown in Fig. 7 (F and G), mice were protected against oral CT challenge if immunized with a volume of 40 µl but not 5 µl. These data, taken together with the data on induction of IgA by LDCs and the ability of LDCs to direct B cells to the GI tract, provide compelling evidence that for a nasally delivered vaccine LDCs should be targeted to induce GI-associated immune responses. This in turn argues for research into more efficacious and safer delivery systems to target LDCs by i.n. vaccines.

Finally, having demonstrated that the microbiota affect the production of IgA by LDC, we hypothesized that microbial signals will play an important role in i.n. vaccine-induced responses. In support of our hypothesis, we observed that GF mice, when immunized i.n., failed to generate specific IgA in the GI tract, unlike WT (SPF conditions) age-matched controls. Thus, efficacy of nasal vaccination, like systemic vaccination (Oh et al., 2014), may be significantly affected by microbial communities. A graphical summary of the major findings of our paper is provided in Fig. S4.

Some important caveats deserve mention here. We observed that CFSE dilution was milder in the LDC-B cell co-cultures compared with other DC–B cell cultures and hypothesized that this is a result of suppressive populations contained within LDCs. Because this was outside of the scope of the present study, we did not examine this question in further detail. We also detected interexperimental variance in the levels of IgA induced by LDCs that appeared to be physiological and related to the batch of mice used. Additionally, this paper does not address the contribution of lung-derived IgA⁺ B cells to the total IgA pool in the gut. As such, significance of IgA originating in the lung to gut IgA responses at steady state remains to be determined.

In summary, the data generated in this study provide fundamental and novel insights on the generation of IgA by LDCs and help pave the way for better preventative and therapeutic strategies against mucosal diseases, both pulmonary and GI.

C57BL/6 mice (B6) mice and were purchased from Taconic Farms or bred at The Rockefeller University or at Icahn School of Medicine at Mount Sinai. Batf3^tm1Kmm/J (Batf3^−/−) mice were purchased from The Jackson Laboratory Laboratories, TCR^−/− mice were provided by the Cerutti laboratory. Mice were maintained in specific pathogen–free conditions. GF adult C57BL/6J mice were maintained in plastic gnotobiotic isolators at the Mount Sinai School of Medicine. TRIF^−/−MyD88^−/− mice were provided by the Nussenzweig laboratory (The Rockefeller University, New York, NY) and the Blander laboratory (Mount Sinai School of Medicine, New York, NY) and were used with the permission of S. Akira (Osaka University, Osaka, Japan). All experiments involving mice were performed in accordance with the guidelines of the animal care and use committee of Mount Sinai School of Medicine and The Rockefeller University.

The following reagents were from BD: anti-CD3 (500A2), anti-CD3 (145–2011), anti-CD103 (M290), anti-CD4 (RM4-5), biotin rat anti–mouse (C10-1), APC Streptavidin (554067), anti-CD64 (X54-5/7.1.1), and anti–Siglec-F (E50-2440). The following were from eBioscience: anti-MHCII (M5/114.15.2), anti-CCR9 (CW-1.2), anti-CD3 (500A2), and anti-CD4 (RM4-5). The following antibodies were from BioLegend: anti-CD45 (30-F11), anti-CD11b (M1/70), anti-CD19 (6D5), anti-CD103 (M290), anti–LPAM-1 (DATK32), anti-CD24 (M1/6a), anti-CCR9 (CW12), anti-CD64 (X54-51711), and anti-CD11c (N418). AQUA (L34957) was from obtained from Invitrogen and CFSE was purchased from Sigma-Aldrich.

Mice were euthanized and blood was obtained from the experimental mice by cardiac puncture. Lung mononuclear cell isolation was performed as previously described (Vermaelen et al., 2001). In brief, after sacrificing the animals, the pulmonary circulation was perfused with saline to remove the intravascular pool of cells. Lungs were carefully separated from thymic and cardiovascular remnants and removed in toto. Organs were thoroughly minced using iridectomy scissors and incubated for 60 min in digestion medium containing collagenase D in a humidified incubator at 37°C and 5% CO₂. In the last 5 min, 10 mM EDTA was added. Tissue fragments were disrupted mechanically by pipetting and passed through a 70-µm cell strainer and mononuclear cells were washed twice in RPMI before using in subsequent experiments as described below.

The spleen and LN were harvested in RPMI media with 5% FCS, injected with collagenase D, and physically disrupted using a 25-gauge syringe. The cell suspension was placed in a humidified incubator at 37°C and 5% CO₂ for 25 min. In the last 5 min, 10 mM EDTA was added. Spleen but not LN samples went through subsequent ACK lysis, were washed twice and counted.

Intestinal lymphocytes were isolated and prepared as previously described (Mucida et al., 2007). In brief, small and large intestines were removed and placed in cold HBSS media containing 5% FCS. The intestines were carefully cleaned from the mesentery and flushed of fecal content. Intestines were opened longitudinally, and then cut into 1-cm pieces. The intestinal tissue was incubated with 1.3 mM EDTA (Cellgro) in HBSS at 37°C for 1 h. The supernatants containing intestinal epithelial cell (IEC) with some superficial villi cells, referred to as the IEC fraction, were not used in the present study. To isolate the lamina propria lymphocytes (LPL), the remaining intestinal tissue was minced and transferred to conical tubes. The minced pieces were resuspended in 20 ml of complete RPMI containing 0.125 mg/ml of collagenase (Sigma-Aldrich) and shaken at 200 rpm for 50 min at 37°C. The tissue suspension was collected and passed through a 70-µm cell strainer, and the cells were pelleted by centrifugation at 1,200 rpm. The cells were then resuspended and layered onto a 40/80% Percoll gradient, centrifuged, and collected, washed, and resuspended in complete RPMI media. These purified cells constituted the SILP or CLP lymphocyte population.

Single-cell suspensions were isolated from the murine lung, spleen, mesenteric, mediastinal, and skin draining LNs as described in the previous section. CD11c⁺ cells were isolated as previously described (Ing et al., 2006) using CD11c magnetic beads (Miltenyi Biotec). In brief, the cells were washed in 1 ml of MACS buffer (Miltenyi Biotec), before 30 min incubation on ice with CD11c⁺ beads. The cellular suspensions were washed twice in MACS buffer and passed through a magnetic column, and then CD11c⁺ cells were isolated by positive selection, washed, resuspended in complete RPMI medium, and counted before co-culture with T cells as described below.

Lung mononuclear cells were isolated as described in the previous subset. Using magnetic beads, CD11c⁺ cells were enriched and sorted on FACSAria (BD) as Siglec-F⁻MHCII⁺CD11c^hiCD103⁺CD11b⁻ (LDC CD103⁺), Siglec-F⁻MHC II⁺CD11c^hiCD103⁻CD11b⁺ CD24⁺CD64⁻ (LDC CD24⁺), and Siglec-F⁻MHC II⁺CD11c^hiCD103⁻CD11b⁺ CD24⁻CD64⁺ (LMφ CD64⁺) subsets. To get sufficient numbers of DCs, LDCs were pooled from 5–10 mice per experiment, as indicated.

Mononuclear cells were isolated as described in Tissue harvesting and cell preparation. Using magnetic beads, CD11c⁺ cells were enriched and sorted on FACSAria (BD) as CD45⁺MHCII⁺CD11c⁺ cells.

The culture medium used for DC/B cell cultures was RPMI (Invitrogen) supplemented with 10% heat-inactivated FCS, 2 mM l-glutamine, 100 U/ml penicillin, 100 µg/ml streptomycin, and 5 mM β-mercaptoethanol (all from Sigma-Aldrich). B cells were negatively isolated using anti-CD43 MACs beads per manufacturer’s instructions. In brief, this method relies on the fact that most leukocytes, except for resting mature B cells, express CD43 antigen. After magnetic bead isolation, cells were sorted and live CD45⁺ CD19⁺ IgM⁺ B cells were obtained. These CD43^negCD19⁺IgM⁺ B cells were added to round-bottom microtest wells at 10⁵/well and mixed with isolated DCs at a DC/B cell ratio of 2:1. B cells were stimulated by adding 10 µg/ml Fab goat anti–mouse IgM (Jackson ImmonoResearch Laboratories) and 25 µg/ml anti-CD40 (BioLegend) After 4 d, the expression of IgA was investigated on B cells; additionally, after 5 d, α4β7 and CCR9 expression on B cells was evaluated by FACS. For the RA inhibition assays, the DC/B cell co-culture were incubated with 3 µM of RAR-β antagonist LE540 (Wako) dissolved in DMSO at a stock concentration of 1 mM and added to cultures at final concentration of 3 µM. Additionally, to determine a dose response, LE540 was added to the DC/B cell cultures in concentrations of 2.5 µM, 250 nM, and 25 nM. For the TGF-β inhibition assay, 70–8.5 µg/ml of TGF-β pan-specific polyclonal antibody (R&D Systems) was added to the cultures. Anti–mouse IL-10 (eBioscience) and recombinant mouse TACI-Fc (BioLegend) at a final concentration of 10 µg/ml and 100 ng/ml, respectively, was added to the cultures on day 0 and 3.

Immunofluorescent microscopy was performed according to standard protocols. In brief, sections were rehydrated in cold 1× DPBS for 5 min, fixed in 4% paraformaldehyde at room temperature for 5 min, permeabilized in 0.2% Triton at room temperature for 10 min. Next, sections were preincubated with blocking buffer (1× DPBS containing 1% BSA supplemented with 1% mouse and 1% rat serum) at room temperature for 1 h. To detect DCs, sections were incubated with CD11c conjugated with FITC at 2 µg/ml (clone N418; BioLegend), and April conjugated with PE at 0.5 µg/ml (clone A3D8; BioLegend) in blocking buffer. Nuclei were visualized by DAPI staining at 1 µg/ml (1 µg/ml DAPI in blocking buffer), and sections were mounted with ProLong Gold Antifade Mountant (Invitrogen). Images were taken using Zeiss Axioplan 2IE controlled by Zeiss AxioVision 4.6 software, and analyzed by ImageJ.

CD45.2⁺CD43^negCD19⁺IgM⁺ B cells were cultured with flow sorted LDC (Siglec-F⁻MHCII^hiCD11c^hi) or MLN DC (CD45⁺MHCII⁺CD11c^hi) or SpDC (CD45⁺MHCII^hiCD11c^hi) in a ratio of 2:1 for 5 d. The respective B cells were labeled with 5 µM CFSE (Invitrogen) for 10 min at 37°C or 10 mM CMTMR (chloromethyl-benzoyl-amino-tetramethylrhodamine; Life Technologies) for 20 min at 37°C. After washing twice with PBS, 5 × 10⁵ CFSE-labeled B cells and 5 × 10⁵ CMTMR-labeled B cells were injected via the retroorbital venous plexus into CD45.1⁺ recipient mice. 18 h after transfer recipient mice were sacrificed and mononuclear cells isolated from the lung, spleen, SILP, and CLP. During FACS analysis, cells were gated on the congenic marker (e.g., CD45.2) and then on the specific B cell marker (e.g., CD19) and analyzed for the ratios between CMTMR- and CFSE-positive cells. The data are expressed as the HI, which is calculated as the ratio CFSE/CMTMR (or CMTMR/CFSE) in each tissue divided by the corresponding input ratio. HI = CFSE_tissue/CMTMR_tissue:CFSE_input/CMTMR_input.

Cells were isolated as described above. Before staining, cells were washed and resuspended in staining buffer containing 1× PBS, 2% BSA, 10 mM EDTA, and 0.01% NaN₃. To block nonspecific staining, the 2.4G2 anti-CD16/32 antibody was added. Antibodies for cell surface markers were added and cells were incubated for 25 min at 4°C. After the staining, the cells were washed twice and analyzed immediately or fixed in PBS containing 1% paraformaldehyde and 0.01% NaN₃ and analyzed later on an LSR II (BD) using multiparameter flow cytometry.

For intracellular cytokine staining, after surface staining, cells were resuspended in Fixation/Permeablization solution (BD Cytofix/Cytoperm kit; BD), and intracellular cytokine staining was performed according to the manufacturer’s protocol. Flow cytometric data were analyzed with FlowJo software (Tree Star).

LDC and LMφ were obtained via flow sorting as described in the previous section. RNA was isolated using the TRIzol (Life Technologies) according to the manufacturer’s instructions. Residual DNA was removed using the RNase-Free DNase Set (QIAGEN) on-column DNase treatment during RNeasy preparation. RNA was amplified by Illumina TotalPrep RNA Amplification kit and hybridized to the Expression BeadChip kit (Illumina). This chip was scanned using an Illumina HiScan. All datasets have been deposited at the National Center for Biotechnology Information Gene Expression Omnibus under accession no. GSE74969.

Stool was collected from the respective mice and placed in 1 ml of protease inhibitor (Roche). Serum was isolated from the blood and stored at −20°C. ELISAs were performed by coating 96-well plates (Nunc) with 2.5 µg/ml CT, followed by blocking with PBS containing 2% BSA. After incubation of sera and stool, IgG and IgA were detected by using biotinylated antibodies, followed by streptavidin peroxidase (Jackson ImmunoResearch Laboratories). For detection of peroxidase activity, o-phenylenedi-amine dihydrochloride (Sigma-Aldrich) was added, and absorbance was measured at 450 nm.

Plates were coated with 10 µg/ml CT (Sigma-Aldrich) overnight. After washing, plates were blocked with B cell media (RPMI containing 10% FBS [Gibco], 1% MEM NEAA [Gibco], 1% anti-anti [Gibco], and 0.1% 2-mercaptoethanol [Gibco]). Cells were plated in a concentration of 1–2 × 10⁶ cells per well (GI) and cultured in B cell media for 24 h at 37°C, washed, and then isotype-specific antibody was detected using HRP-conjugated goat anti–mouse IgA secondary (Southern Biotech). Spots were developed using BD ELISPOT AEC substrate set (551951) according to manufacturer’s protocol. Spots were quantified using a CTL-ImmunoSpot Analyzer and Software.

Protocol as previously described (Ridaura et al., 2013). In brief, cecal contents from WT mice were suspended in PBS (2.5 ml per cecum) and were administered (0.2 ml per mouse) immediately to sterilely packed 7-wk-old C57BL/6 GF mice. Transplanted mice were maintained in sterile, unchanged cages for 4 wk.

Gender-matched wild-type mice were treated for 4 or 12 wk with Vancomycin (500 mg/l), neomycin (1 g/l), metronidazole (1 g/l), and ampicillin (1 g/l) in drinking water. Additionally, Splenda (artificial sweetener) was added (1 g/l) to improve palatability of water. Antibiotic-containing water was changed twice a week.

Gender-matched WT mice, Batf3^−/− mice, and TCR^−/− mice (with appropriate controls) were immunized with 2 µg of CT from Vibrio cholerae (Sigma-Aldrich) in 5–40 µl of PBS i.n., i.t., or s.c. For s.l. vaccination, mice were immunized under the tongue in a volume of 7 µl. In other experiments, mice received 2 µg CT orally by gavage in 200 µl PBS. All animals were given food and water ad libitum. For the CT-induced diarrhea, 10 µg CT in 200 µl PBS was administered by oral gavage. 6 h before challenge, food and water were removed. 2 h after oral gavage, mice were sacrificed, and the weight of the small intestine was determined.

8-12-wk-old WT mice were immunized i.n. with 50 µg of Ova-AF647 (Life Technologies) at two different volumes of either 5 or 40 µl. PBS-treated mice served as controls. After 24 h, LDCs were prepared as described previously and the uptake of OVA by the LDC subsets was examined.

Statistical significance was evaluated using a two-tailed Student’s t test with a 95% confidence interval. Results are expressed as means ± SD. Analysis was performed with a Prism 4 program (GraphPad Software).

For microarray data analysis, images generated by the Illumina Hiscan were uploaded into GenomeStudio Software for primary data extraction and analysis. Spurious probes were filtered by translating the detection p-values into q-values (Storey and Tibshirani, 2003; Robinson et al., 2010) and removing any probe with maximum q-value across all samples ≥10%. Differential expression analyses of the resulting log₂-transformed data were performed with SAM (Tusher et al., 2001) using the two-sample Wilcoxon or Mann-Whitney test statistic (other parameters were held to their default values). Genes assigned a false discovery rate less than 5% were considered differentially expressed. Enriched pathways and other gene sets, such as B cell activation and IgA CSR genes, were identified in the resultant lists of differentially expressed genes using the DAVID tool (Huang et al., 2009).

Table 1 shows the primers used for mouse gene amplification.

Fig. S1 shows the gating strategy to identify and sort LDCs and MACs. Fig. S2 shows that LDCs, SpDCs, SkDCs, or MLN DCs were cultured with CD43^negCD19⁺IgM⁺ B cells in a DC/B cell ratio of 2:1 for 4 d in the presence of α-IgM and α-CD40 to stimulate B cells. Fig. S3 shows targeting of DCs with fluorescent antigens delivered i.n. Fig. S4 is a graphical summary of the paper.

This work was supported by grants from the American Gastroenterology Association Elsevier Award (S. Mehandru), The Rockefeller University Clinical and Translation science award pilot project from National Institutes of Health/NCRR (5UL1RR024143-05; S. Mehandru) and R01 AI57653 (A. Cerutti).

The authors declare no competing financial interests.