CD103+CD11b+ mucosal classical dendritic cells initiate long-term switched antibody responses to flagellin

Antibody responses induced at mucosal and nonmucosal sites demonstrate a significant level of autonomy. Here, we demonstrate a key role for mucosal interferon regulatory factor-4 (IRF4)-dependent CD103+CD11b+ (DP), classical dendritic cells (cDCs) in the induction of T-dependent immunoglobulin G (IgG) and immunoglobulin A (IgA) responses in the mesenteric lymph node (MLN) following systemic immunization with soluble flagellin (sFliC). In contrast, IRF8-dependent CD103+CD11b− (SP) are not required for these responses. The lack of this response correlated with a complete absence of sFliC-specific plasma cells in the MLN, small intestinal lamina propria, and surprisingly also the bone marrow (BM). Many sFliC-specific plasma cells accumulating in the BM of immunized wild-type mice expressed α4β7+, suggesting a mucosal origin. Collectively, these results suggest that mucosal DP cDC contribute to the generation of the sFliC-specific plasma cell pool in the BM and thus serve as a bridge linking the mucosal and systemic immune system.


INTRODUCTION
Flagellin is the filament protein component of bacterial flagella. Extracellular flagellin is recognized primarily through Toll-like receptor 5 (TLR5) and this can induce profound responses in innate and adaptive immune cells. 1 Immunization with purified, soluble flagellin (sFliC) protein from Salmonella Typhimurium is sufficient to drive T-and B-cell responses against itself and coimmunized antigens in the absence of additional adjuvant. [2][3][4][5] This autoadjuvant activity of flagellin has led to its use as a carrier protein in a number of vaccine strategies, [6][7][8] including an influenza fusion vaccine tested in humans. 9,10 Additionally, immunization of sFliC in mice has been shown to enhance protection against viral infections 11 and radiation exposure, 12 promote antigen presentation through major histocompatibility complex class-II (MHC-II), 13 and reduce T helper cell type 1 (Th1) differentiation after coimmunization with Salmonella Typhimurium. 14 Although such findings indicate that flagellin is an important modulator of the adaptive immune system, the cellular mechanism(s) underlying its mode of action remain unclear.
Previously, it has been shown that systemic immunization with sFliC, given subcutaneously in the footpad or intraperitoneally, induces immunoglobulin G (IgG) responses in the spleen and concurrent IgG and IgA responses in the intestinal draining mesenteric lymph nodes (MLNs). 15 This unexpected induction of intestinal responses after systemic immunization was TLR5 dependent and associated with the rapid and extensive recruitment of antigen-loaded CD103 þ classical dendritic cells (cDCs) into the MLN. This coincided with a decrease in the frequency of these cells in the small intestine lamina propria (SI-LP) and suggests that the autoadjuvant activity of sFliC may, in part, be mediated through the activation of mucosal CD103 þ cDCs.
The intestinal mucosa contains three major subsets of cDCs: CD103 þ CD11b þ , CD103 þ CD11b À , and CD103 À cDCs, 16,17 that require different transcription factors for their development and survival. Deletion of the transcription factors interferon regulatory factor-8 (IRF8), BATF3, or ID2 results in a loss of intestinal and MLN CD103 þ CD11b À cDCs, [18][19][20] whereas deletion of IRF4 and NOTCH-2 results in a loss of intestinal-derived CD103 þ CD11b þ cDCs in the MLN. 21,22 We, and others, have recently demonstrated that these subsets play key non-redundant roles in regulating intestinal immune homeostasis. For example, IRF4-dependent cDCs play an important role in intestinal Th17 (refs. 21,23 ) and Th2 responses 24 and for driving postoperative ileitis. 25 In contrast, IRF8-dependent CD103 þ CD11b À cDCs are required for the maintenance of T cells within the small intestinal epithelium and for the generation and maintenance of intestinal interferon-g À producing Th1 cells. 26,27 In this study, we assessed the role of mucosal cDCs in the generation of sFliC-specific IgG and IgA responses in the MLN following systemic immunization, and the impact of this response on the accumulation of plasma cells in the bone marrow (BM). We demonstrate that mucosal CD103 þ CD11b þ but not CD103 þ CD11b À cDCs are essential for the generation of sFliC-specific responses in the MLN, and that the absence of this response affects long-term systemic antibody (Ab) response in the BM. Collectively, these results suggest that mucosal CD103 þ CD11b þ cDCs act as a bridge to link adaptive immune responses of the intestinal mucosa to serological memory and systemic protection.

RESULTS
DP cDCs recruited to the MLN after direct stimulation by sFliC are functional Intraperitoneal (IP) or subcutaneous immunization with sFliC drives a TLR5-dependent accumulation of CD103 þ cDCs in intestinal draining MLN. 15 To determine which CD103 þ cDC subset accumulates in the MLN in response to sFliC, wild-type (WT) mice were immunized IP with sFliC and numbers of CD103 þ CD11b þ (DP), CD103 þ CD11b À (SP), and CD103 À cDCs in the MLN and SI-LP were assessed 24 h later by flow cytometry (for gating strategy see Supplementary Figure S1a online). sFliC immunized mice had an increased frequency of CD11c þ MHC-II hi cDCs in MLN (Figure 1a). In the steady state this population has been suggested to contain cDCs that have migrated from the SI-LP 28 and potentially some resident CD8a þ and CD11b þ cDCs that have upregulated MHC-II upon activation. Within the CD11c þ MHC-II hi population, DP but not SP or CD103 À cDC numbers increased in the MLN in response to sFliC (Figure 1a) that paralleled a selective loss of DP cDCs in the SI-LP (Figure 1b). Despite the selective increase in DP cDC numbers in MLN, sFliC immunization induced upregulation of CD40 and CD86 in all MLN cDC subsets ( Figure 1c).
sFliC immunization failed to induce DP cDC accumulation in the MLN of mice lacking MyD88 in CD11c þ cells (Cd11ccre.MyD88 fl/fl mice 28 ) (Figure 1d), indicating that sFliC may directly drive DP cDC recruitment to MLN. To assess this possibility, mixed BM chimeras were generated with BM from Cd11c-cre.Irf4 fl/fl mice, which lack DP cDC in MLN, 21 and Cd11c-cre.MyD88 fl/fl mice (Figure 1e). In these chimeras, DP cDCs in the MLN are MyD88 deficient, whereas other CD11c þ cells are a mixture of MyD88-sufficient and -deficient cells. sFliC also failed to induce DP cDC accumulation in the MLN of these mice (Figure 1e). Deletion of IRF4 in Cd11c-cre.MyD88 fl/ fl leads to the expression of green fluorescent protein (GFP) 21 in CD11c þ cells and thus can be used to discriminate between CD11c þ cells derived from different donor mice. Assessment of GFP expression in cDCs from the SI-LP of the chimeras showed that both MyD88 fl/fl (GFP À ) and Cd11c-cre.MyD88 fl/fl cDCs (GFP À ) were found in similar proportions in comparison with cDCs derived from Cd11c-cre.Irf4 fl/fl (GFP þ ) donor BM (Supplementary Figure S1b). Collectively, these results demonstrate that sFliC signaling in DP cDCs is required for their accumulation in the MLN.
It has previously been shown that CD103 þ cDCs are responsible for T-cell priming in the MLN following IP immunization with sFliC. 15 To determine which of the two MLN CD103 þ cDC subsets underlie this response, SP and DP cDCs were fluorescence-activated cell sorted (FACS) from the MLN 24 h after IP immunization with sFliC and co-cultured with SM1 transgenic T cells that are specific for an epitope in Salmonella Typhimurium FliC (amino acids 427-441). 29,30 DP cDCs were far more efficient than SP cDCs at inducing SM1 T-cell division and activation as assessed by carboxyfluorescein succinimidyl ester dilution, downregulation of CD62L, and total cell counts ( Figure 1f). Importantly, ex vivo addition of sFliC to the SP and DP cDC-T cell co-cultures resulted in similar SM1 T-cell division (Figure 1f), demonstrating that the diminished capacity of SP cDCs to present sFliC in vivo was not due to an inability of these cells to present antigen. Thus, DP cDCs represent the major sFliC peptide-presenting cells in the MLN.
CD103 þ CD11b þ DP cDCs are required for the generation of mucosal anti-sFliC IgA and IgG responses To assess the role of MLN CD103 þ cDC subsets in sFliCspecific Ab responses, we used a prime-boost system as previously described. 15 First, we determined whether priming with sFliC could interfere with the accumulation of DP cDCs (for example, through the induction of antibodies) after secondary immunization. Secondary immunization induced a similar and selective accumulation of DP cDCs in the MLN as observed after primary immunization (Figure 2a), despite the presence of sFliC-specific IgG in the serum (Figure 2b).
To address the role of DP cDCs in the sFliC-specific Ab response, Cd11c-cre.Irf4 fl/fl mice were immunized twice with sFliC and the response was examined 4 days after boost. In marked contrast to control Irf4 fl/fl mice, the number of plasma cells did not increase in the MLN of Cd11c-cre.Irf4 fl/fl mice following boosting with sFliC ( Figure 3a; detailed gating strategy is shown in Supplementary Figure S1c), suggesting a reduced sFliC-specific Ab response at this site. Consistent with this, sFliC-specific IgG and IgA Ab-secreting cells (ASCs) were readily detected in the MLN of Irf4 fl/fl but not Cd11c-cre.Irf4 fl/fl mice as assessed by ELISPOT (Enzyme-Linked ImmunoSpot) ( Figure 3b). Plasma cells derived from the MLN can migrate to the SI-LP contributing to the specific response in this site. To examine whether this element of the response was also affected, cells were isolated from the SI-LP and sFliC-specific ASCs assessed as above. As with the MLN, sFliC-specific IgG þ and IgA þ ASCs were detected in the SI-LP of Irf4 fl/fl but not Cd11ccre.Irf4 fl/fl mice ( Figure 3b, lower panels). To exclude the possibility that the reduced numbers of sFliC-specific ASCs observed in Cd11c-cre.Irf4 fl/fl mice was a result of an intrinsic defect in B-cell function, 31 mixed BM chimeras were generated using a 50:50 mix of BM cells from Irf4 fl/fl or Cd11c-cre.Irf4 fl/fl mice with BM cells from Rag-1 À / À mice. In these chimeras Irf4-dependent cDCs derive from Rag-1 À / À BM, whereas the B-cell compartment derives from Cd11c-cre.Irf4 fl/fl BM. RAG-1 À / À BMs fully rescued the defect in plasma cell numbers observed in single Cd11c-cre.Irf4 fl/fl chimeras ( Figure 3c lower graph) as well as sFliC-specific IgG-and IgA-producing ASCs (Figure 3c), confirming that the absence of Ab response in the Cd11c-cre.Irf4 fl/fl mice was not due to an intrinsic B-cell defect.
In marked contrast to Cd11c-cre.Irf4 fl/fl mice, Cd11c-cre.Irf8 fl/fl mice, which lack migratory SP cDCs and LN-resident CD8a þ cDCs in the MLN, 26 induced equivalent numbers of plasma cells and anti-sFliC IgG and IgA ASCs in the MLN as control Irf8 fl/fl mice, following sFliC prime-boost (Figure 3d,e). Collectively, these results demonstrate that the generation of sFliC-specific Ab responses in the MLN requires DP cDCs. First, we evaluated the cDC response in the spleen after immunization with sFliC. In contrast to the MLN, immunization with sFliC did not affect the frequency or total number of CD4 þ or CD8a þ cDCs (Figure 5a), although it led to increased expression of CD86 and CD40 by both subsets (Figure 5b). As expected, 21 splenic CD4 þ cDC numbers were reduced in Cd11c-cre.Irf4 fl/fl compared with Irf4 fl/fl mice in steady state (Figure 5a), and this difference was maintained after sFliC immunization.
When assessing the Ab response 4 days after secondary immunization, Cd11c-cre.Irf4 fl/fl mice displayed a reduction in sFliC-specific IgG þ ASCs and virtually no IgA þ ASCs when compared with Irf4 fl/fl mice (Figure 5c). sFliC-specific GCs Thus, in contrast to the MLN, IRF4-dependent cDCs are not required for Tfh cell generation and GC induction in the spleen in response to sFliC immunization.
We hypothesized that Tfh cells were still detectable in the spleens of Cd11c-cre.Irf4 fl/fl mice because CD8a þ CD11b À splenic   cDCs could potentially contribute to antigen presentation. To address this possibility, WT mice were immunized with sFliC for 24 h and CD4 þ CD11b þ and CD8a þ CD11b À splenic cDCs were FACS sorted and co-cultured with SM1 T cells. Both CD4 þ CD11b þ and CD8a þ CD11b À cDCs induced SM1 T-cell proliferation, although CD8a þ CD11b À cDC did so less efficiently (Figure 5g). The difference in T-cell proliferation after co-culture probably reflects differences in antigen capture in vivo rather than an intrinsic difference in their capacity to present antigen as T-cell proliferation was similar when sFliC was added to the cultures ex vivo (Figure 5g). Collectively, these results suggest that IRF4-dependent and -independent cDCs contribute to the splenic sFliC-specific response.

Mucosal DP cDCs contribute to sFliC-specific Ab responses in the BM
To study the persistence of the anti-sFliC Ab response, we next examined the BM, representing an important site for maintenance of long-lived plasma cells. Strikingly, although sFliC-specific IgG þ and IgA þ ASCs were readily detected in the BM of Irf4 fl/fl mice, they were completely absent in the BM of Cd11c-cre.Irf4 fl/fl mice (Figure 6a). In contrast, the BM of Cd11c-cre.Irf8 fl/fl mice had a similar number of sFliC-specific ASCs as control Irf4 fl/fl mice (Figure 6b). The complete lack of sFliC-specific ASCs in the BM of Cd11c-cre.Irf4 fl/fl mice suggested that mucosal DP cDCs may contribute to the long-term Ab response in the BM. To assess this possibility, expression of the intestinal-associated signature integrin a 4 b 7 (refs. 32,33 ) was examined on BM sFliC-specific plasma cells from WT mice primed-boosted with sFliC. A sFliC-specific population of plasma cells was detected in the BM of sFliC immunized but not unimmunized mice, as assessed by flow cytometry (Figure 6c). Furthermore, a proportion of these sFliC-specific but not non-sFliC-specific BM plasma cells expressed a 4 b 7 (Figure 6c). Consistent with these findings, sFliC-specific cells expressing a 4 b 7 were detected on cytospins preparations of enriched CD138 þ BM cells from the same mice ( Figure 6d). Finally, the data suggest that loss of DP cDCs should have an effect on the total serum antibody response to sFliC. IgG titers were 490% reduced in Cd11c-cre.Irf4 fl/fl mice compared with control mice and IgA was undetectable (Figure 6e). Collectively, these results suggest that mucosal DP cDC priming of antibody responses in the MLN can enhance sFliC-specific ASC numbers in the BM following IP immunization with sFliC.

DISCUSSION
Previously, it has been shown that systemic immunization with sFliC induces parallel Ab responses in systemic and mucosal secondary lymphoid tissues, with the latter responses associated with a rapid TLR5-dependent accumulation of antigencarrying CD103 þ cDCs in the intestinal-draining MLN. 15 Here we demonstrate that mucosal DP cDCs are required for the anti-sFliC response in the MLN and provide evidence that plasma cells derived from this response can ultimately take up residence in the BM and contribute to the systemic antibody pool. The ability of sFliC to efficiently engage the mucosal immune system after a systemic immunization may confer a significant advantage to sFLiC-containing vaccines, such as the influenza-flagellin fusion vaccine that has shown safety and potential not only in healthy adults but also in the elderly 9,10 who often induce poor Ab responses. 34 Although multiple MLN cDC subsets can induce T-cell responses to sFliC after antigen loading in vitro, 35 we demonstrate here that only DP cDCs drive T-and B-cell responses in the MLN in vivo. The reasons why IRF4dependent DP cDCs are so critical in driving sFliC-specific T-cell responses in the MLN remain to be fully elucidated but are likely multifactorial. Previous studies have demonstrated that IRF4-dependent cDCs have an enhanced intrinsic capacity to prime CD4 þ T cells compared with IRF8-dependent cDCs. 23,36 Interestingly, sFliC induced MyD88-dependent accumulation was similar after secondary immunization, even in the presence of high-affinity sFliC Ab, 2,3 indicating a high efficiency of sFliC capture by TLR5. These results are consistent with prior observations that small amounts of sFliC are required for its immune modulatory effects 37 and further suggest that inclusion of sFliC in boost as well as in prime vaccines could further promote mucosal immune system engagement. Notably, recognition of intestinal microbiotaderived flagellin is also TLR5 dependent and boosts vaccine responses through the action of clodronate-sensitive (presumably monocyte derived) cells. 38 Given our current findings, the cross-talk between these cells and DP cDCs in driving the microbiota-derived flagellin response warrants further study. sFliC-specific responses in the spleen were reduced in Cd11ccre.Irf4 fl/fl mice. Splenic CD4 þ cDCs from immunized WT mice efficiently primed sFliC-specific T cells ex vivo and these cells were selective reduced in Cd11c-cre.Irf4 fl/fl mice, collectively indicating that IRF4-dependent CD4 þ cDCs are required for optimal sFliC-specific responses in the spleen. Despite these findings, sFliC-induced Tfh cell accumulation was not significantly altered, and sFliC-specific GCs were readily detectable in the spleen in the absence of IRF4dependent cDCs, suggesting that additional antigen-presenting cells contribute to the sFliC-specific response at this site. Although the identity of these IRF4-independent antigenpresenting cells remains to be identified, splenic CD8 þ cDCs isolated from immunized mice induced limited sFliC-specific T-cell generation ex vivo, indicating that IRF8-dependent cDCs may contribute to this response. Collectively, these findings indicate that the cellular and molecular mechanisms driving sFliC-specific immune responses in the spleen and MLN are, in part, distinct.
These findings are consistent with prior studies demonstrating that immune responses induced by sFliC are both complex and site specific. Thus, although sFliC drives both an IgG and IgA response in the MLN, it drives primarily an IgG response in the spleen. 15 sFliC-specific IgA responses are TLR5 dependent, 15 whereas sFliC-specific IgG responses, in particular IgG1 responses, can occur through TLR5-and inflammasomeindependent pathways. 13,39 Furthermore, sFliC-specific T-cell responses in the spleen are also less dependent on TLR5 than in the MLN 15 and splenic cDCs have been reported to express lower levels of TLR5 in comparison with cDCs from the intestinal mucosa. 40 Thus, the features of sFliC that enable it to have these effects are likely to relate to it being the ligand for TLR5 and its modest molecular size. The accumulation of CD103 þ cDCs in the MLN after sFliC immunization is TLR5 dependent 41 and SI-LP cDCs with a similar phenotype as DP cDC (CD103 þ CD11c hi CD11b hi ) have been shown to express high levels of TLR5. 42 In contrast, splenic cDCs do not appear to express as high levels of TLR5 as SI-LP cDCs. 40 Indeed, consistent with the need for TLR5 expression, we show that MyD88 expression in DP cDCs is needed for their accumulation in the MLN after immunization. Furthermore, TLR5 itself can enhance presentation of antigen through MyD88-independent mechanisms, indicating its function in promoting antigen presentation may be multifactorial. 13 Although these points may help explain why DP cDCs have an enhanced capacity to capture and present sFliC, it does not explain how sFliC gets to this site. This could relate to the use of highly purified, monomeric flagellin in these studies. sFliC has a relatively modest size of B50 kDa, meaning it can disseminate readily through the host after systemic immunization. Other studies have found molecules of a similar or greater size are also able to disperse rapidly throughout the host and prime responses in multiple sites. 43 While not addressed in the present study, it remains possible that DP cDC promote sFliC responses in part through direct interactions with B cells. A more detailed understanding of the molecular pathways by which antigen-presenting cell subsets orchestrate these diverse responses to sFliC should provide important information regarding sFliC usage in future vaccines.
Interestingly, IgA þ and IgG þ sFliC-specific ASCs were completely absent in the BM of sFliC immunized Cd11ccre.Irf4 fl/fl mice and we found that a proportion of sFliC-specific ASCs accumulating in the BM of sFliC immunized WT mice expressed the intestinal homing receptor a 4 b 7 . This suggests that some of these ASCs were initially generated in intestinal lymph nodes. 44 Nevertheless, this does not necessarily mean all ASCs in the BM derive from the mucosa, nor does it exclude the possibility that long-lived ASCs can be generated in other sites, as we detected sFliC-specific ASCs in the spleen. For instance, it has been shown that long-lived plasma cells can be found in the spleen, as well as the BM, indicating multiple reservoirs can exist. 45 Collectively, these results suggest that the priming of responses in the MLN by DP cDCs may contribute to the accumulation of sFliC-specific plasma cells in the BM and hence to sFliC-specific serological memory. In support of this, sFliC-specific IgA þ plasma cells are detected in the MLN and not the spleen during a primary response 15 and long-lived BM plasma cell responses can be induced in splenectomized mice after immunizing mice orally with high doses of ovalbumin and cholera toxin. 46 The ability of sFliC to engage DP cDCs, and in so doing promote both mucosal and serological antibody responses, could mean that optimal serological memory requires engagement of the mucosa. Alternatively, this effect could be restricted to the response to sFliC, or just last for the period assessed in this study. Indeed, the ablated ASC response in the MLN and SI-LP observed when mucosal DP cDCs were reduced was not observed in the spleen, indicating that other pathways for the generation of ASCs remain. Thus, this reduction in ASC in the BM may not be permanent or absolute. Either way, it suggests it may be possible to use flagellin to direct mucosally induced plasma cells to the BM. Understanding this may help identify how to enhance the longevity of responses to vaccines that can be dramatically different depending upon their nature and how they are administered. 47 In summary, our results show how sFliC, by targeting DP cDCs, can overcome the difficulties in inducing a mucosal response after systemic immunization. This highlights the interplay between the mucosal and systemic immune systems and offers an alternative and highly desirable approach to drive long-lived systemic immunity by engaging mucosal DP cDCs.

METHODS
Mice. Cd11c-cre.MyD88 fl/fl , 28 Cd11c-cre.Irf4 fl/f , 21 and Cd11c-creIrf8 fl/fl mice 26 were maintained at the Biomedical Center at Lund University (Lund, Sweden). SM1 transgenic 29 mice were maintained at the University of Birmingham Biomedical Service Unit (Birmingham, UK). Specific pathogen-free 6-8-week-old C57BL/6 mice were purchased from Harlan Sprague-Dawley (Huntington, UK). Littermates or age-matched mice were used for all experiments where appropriate. All animal procedures were carried out in strict accordance with the Lund/Malmö Animal Ethics Committee, the University of Birmingham Ethics Committee, and the UK Home Office approval (project license 30/2850).
Antigen preparation and immunizations. sFliC was generated as previously described. 3 Briefly, the FliC gene from Salmonella Typhimurium was cloned into pETT22b þ via ndeI and xhoI sites to incorporate a poly-histidine tag. After induction, His-tagged recombinant protein was enriched by nickel affinity chromatography resulting in a purity of 95%. Following dialysis, the protein was further purified by immunoprecipitation with a sFliC-specific monoclonal. Each batch was tested by immunization of WT and TLR5-deficient mice. Batches were only accepted if cDCs in the spleen and MLN matured in WT but not TLR5-deficient mice. Mice were immunized IP with 20 mg recombinant sFliC, IP boosted 21 days later with the same dose, and responses were assessed 4 days after boost.
Cell isolation and flow cytometry. Single-cell suspensions from spleen and MLN and BM were generated by mechanical disruption. When evaluating cDCs, MLN and spleen were enzymatic digested with collagenase IV digestion (400 U ml À 1 ; 25 min; 37 1C). Cell suspensions from SI-LP were generated as previously described 48 using Liberase (0.2-0.3 WunchU ml À 1 , Roche, Basel, Switzerland). Cells were processed for flow cytometry according to standard procedures. 15 Data acquisition was performed on a LSRII (BD Bioscience, San Jose, CA) or a CyAn ADP (Beckman Coulter, Brea, CA) and analyzed using FlowJo software 9.8.2. (Tree Star, Aghland, OR). All antibodies used are listed in Supplementary Table S1.
Mixed BM chimeras. BM chimeras were generated by transferring BM cells (1.5 Â 10 6 cells) intravenously from the indicated donor mice into lethally irradiated (900 rad) recipient mice. Mixed BM chimeras were generated by transferring a 50:50 mix of BM from the indicated strains.
Mice were used for immunization experiments 8 weeks after BM transfer.
In vitro co-culture to evaluate cDC antigen loading in vivo. Spleen and MLN were collected 24 h post IP immunization with sFliC (20 mg), and single-cell suspensions prepared as described above. The cDCs were pre-enriched using MACS beads (anti-CD19, CD5 and DX5). Cell suspensions were subsequently stained with anti-CD11c, MHC-II, CD103, and CD11b (MLN) or CD11c, MHCII, CD4, and CD8 (spleen) and the indicated cDC subsets FACS sorted on a BD FACSAria Fusion. The following subsets were isolated: SP (CD103 þ CD11b À ) and DP (CD103 þ CD11b þ ) cDCs from the MLN from the CD11c hi MHC-II þ population. From the spleen using again a pregate CD11c hi MHC-II þ CD8a þ CD11b À and CD4 þ CD11b þ cDCs were sorted to a purity of at least 97%. SM1 T cells were MACS enriched (CD5 þ selection) and carboxyfluorescein succinimidyl ester labeled. The cDCs and T cells were co-cultured for 4 days (1:30 ratio cDC/T) before flow cytometry analysis.
ELISPOT analysis. ELISPOT was performed as described previously. 15 In brief, 5 Â 10 5 cells were added per well in triplicates in a sFliC precoated plate and cultured for 6 h at 37 1C. After incubation, plates were incubated overnight at 4 1C with alkaline phosphateconjugated anti-IgG and IgA (Southern Biotech, Birmingham, AL). Reaction was developed with SIGMA Fast BCIP/NBT (Sigma Aldrich, St. Louise, MO). Spots were counted using the AID ELISPOT Reader System and AID software version 3.5 (Autoimmune Diagnostika, Strassberg, Germany). Counts were expressed as spot-forming units per 5 Â 10 5 cells.
Immunohistochemistry and confocal microscopy. Immunohistology was performed as described previously. 3 Cryosections were incubated with primary unconjugated Abs for 45 min at room temperature before addition of either horseradish peroxidase-conjugated or biotin-conjugated secondary antibodies. sFliC-binding cells were identified as previously described 3,15 using biotinylated sFliC. When using cytospins, 1 Â 10 5 cells were used per cytospin and cells were fixed in cold acetone for 10 min. Signal was detected using diaminobenzidine for horseradish peroxidase activity and naphthol AS-MX phosphate with Fast Blue salt and levamisole for alkaline phosphatase activity. Images were acquired using a Leica (Milton Keynes, UK) microscope DM6000 using 10 Â and 20 Â objectives or the Zeiss (Jena, Germany) Axio ScanZ1 Slide Scanner using 10 Â objective. Quantification of GC area was performed using Zen 2012 (blue edition, Jena, Germany) software.
FliC-specific enzyme-linked immunosorbent assay. Enzyme-linked immunosorbent assay plates were coated with 5 mg ml À 1 of sFliC (2 h at 4 1C) and blocked with 1% bovine serum albumin overnight at 4 1C. Serum, diluted 1:100 in phosphate-buffered saline-0.05% Tween, was added and diluted stepwise. To measure total Ab titers serum was diluted in 1:1,000 and added and further diluted stepwise on uncoated plates. Following incubation for 1 h at 37 1C, plate-bound antibodies were detected using alkaline phosphatase-conjugated goat anti-mouse IgG, IgG1, and IgA (Southern Biotech). Reaction was developed with Sigma-Fast p-nitrophenylphosphate (Sigma Aldrich). Relative reciprocal titers were calculated by measuring the dilution at which the serum reached a defined OD 405 .
Statistics. Statistics were calculated using the nonparametric Mann-Whitney sum of ranks test or two-way analysis of variance as appropriate using the GraphPad Prism software (GraphPad, La Jolla, CA).