Research ArticleMUCOSAL IMMUNOLOGY

Early-life programming of mesenteric lymph node stromal cell identity by the lymphotoxin pathway regulates adult mucosal immunity

See allHide authors and affiliations

Science Immunology  20 Dec 2019:
Vol. 4, Issue 42, eaax1027
DOI: 10.1126/sciimmunol.aax1027

Early-Life Licensing of Antiviral IgA Immunity

Mucosal immune responses after exposure to pathogens are influenced by multiple factors, including key developmental milestones achieved before birth or during the neonatal period. Li et al. investigated when during mouse immune system development signals initiated by lymphotoxin (LT) expression are needed to enable an effective mucosal IgA antibody response after rotavirus infection of adult mice. Transient interruption of signaling by the LT receptor during late embryonic life was sufficient to induce persistent defects in stromal cells resident in the mesenteric lymph node that impaired the IgA antirotavirus response of adult mice without interfering with IgA responses to commensal bacteria. These findings highlight the pivotal role early-life exposures play in configuring the mammalian immune system.

Abstract

Redundant mechanisms support immunoglobulin A (IgA) responses to intestinal antigens. These include multiple priming sites [mesenteric lymph nodes (MLNs), Peyer’s patches, and isolated lymphoid follicles] and various cytokines that promote class switch to IgA, even in the absence of T cells. Despite these backup mechanisms, vaccination against enteric pathogens such as rotavirus has limited success in some populations. Genetic and environmental signals experienced during early life are known to influence mucosal immunity, yet the mechanisms for how these exposures operate remain unclear. Here, we used rotavirus infection to follow antigen-specific IgA responses through time and in different gut compartments. Using genetic and pharmacological approaches, we tested the role of the lymphotoxin (LT) pathway—known to support IgA responses—at different developmental stages. We found that LT-β receptor (LTβR) signaling in early life programs intestinal IgA responses in adulthood by affecting antibody class switch recombination to IgA and subsequent generation of IgA antibody-secreting cells within an intact MLN. In addition, early-life LTβR signaling dictates the phenotype and function of MLN stromal cells to support IgA responses in the adult. Collectively, our studies uncover new mechanistic insights into how early-life LTβR signaling affects mucosal immune responses during adulthood.

INTRODUCTION

Lymphotoxin (LT)–β receptor (LTβR) signaling is required for the development of secondary lymphoid organs in utero (14) and for the maintenance of lymphoid tissue architecture during adulthood (5, 6). The membrane-bound form of the LTβR ligand LTα1β2 is expressed by activated T and B lymphocytes as well as retinoic acid receptor–related orphan receptor γt (RORγt+) innate lymphoid cells, whereas the corresponding LTβR is expressed by dendritic cells (DCs), macrophages, and radioresistant stromal cells (5, 7, 8). LTα−/− and LTβR−/− mice lack Peyer’s patches (PPs) and all lymph nodes (LNs) (1, 9, 10), whereas LTβ−/− mice lack PPs and most LNs but retain some mucosal-associated LNs including mesenteric lymph nodes (MLNs) (2, 11). Whereas serum immunoglobulin M (IgM) and IgG levels are relatively normal in LT-deficient mice, both serum and intestinal IgA levels are severely reduced (1012). The selective reduction in IgA is not due to a lack of secondary lymphoid organs in LT-deficient mice (12, 13), raising the key question of defining the mechanism of LTβR-dependent control of IgA responses.

IgA is the most abundant antibody isotype in the body, and more than 80% of human plasma cells (PCs) produce IgA (1416). IgA production at mucosal sites is a major means for containing the intestinal microbiota and preventing the colonization of infectious agents, such as pathogenic viruses (1517). Intestinal IgA can be generated via both T cell–dependent and T cell–independent mechanisms (14, 18, 19). T cell–dependent IgA responses are initiated within gut-associated lymphoid tissue (MLN and PP) via the formation of germinal centers (GCs), which support interactions between B cells and follicular helper T (Tfh) cells (14). Post-GC IgA+ antibody-secreting cells (ASCs) egress through efferent lymphatics and subsequently enter the blood via the thoracic duct. Because of their expression of homing chemokine receptors and adhesion molecules such as CCR9, CCR10, and α4β7 integrin (20, 21), ASCs primed in the MLNs or PPs ultimately circulate back to the intestinal lamina propria effector site where they secrete IgA that is transported into the gut lumen.

Recent studies have reported that DC-intrinsic LTβR signaling contributes to local IgA production within PP in response to the microbiota (22), whereas other studies point to a role for LTβR signaling in the radioresistant compartment (12, 13). Whether intestinal IgA responses are initiated by T cell–dependent versus T cell–independent mechanisms and, in what location these responses are initiated (MLN, PP, or LP), are key variables that likely influence how LT and related pathways affect mucosal IgA responses (2325). Because of these variables, model systems that measure de novo IgA responses to foreign enteric pathogens are a useful strategy for answering questions about how IgA responses are generated in the host. When and how the LT pathway shapes the IgA response to a foreign mucosal pathogen has not been fully addressed.

One very important and emerging parameter that influences mucosal immune responses is exposure to perinatal signals, which comprise inputs from both environmental and genetic factors (26, 27). How such signals play a role in influencing mucosal IgA responses is unclear. Given that LTβR plays a critical role in supporting lymphoid tissue development during fetal life (28), we hypothesized that early-life LTβR signals are likewise important for programming the identity of key support cells that influence immune responses within intact lymphoid tissues in the adult animal. To test this, we took advantage of genetic and pharmacological approaches that inhibit LTβR signaling yet spare the development of MLN. We combined this approach with a rotavirus (RV) infection model that we show is dependent on B cell priming within the MLN. Using this strategy, we found that LTβR signaling during early life, but not during adulthood, is critically required to generate a de novo antigen-specific IgA response in the gut by programming the MLN stromal cell environment. Collectively, our studies provide new insights into how early-life signaling events affect the identity of gut-resident stromal cells and responses to foreign pathogens during adulthood.

RESULTS

Mice that lack LTβ in early-life exhibit impaired RV-specific IgA responses during adulthood

IgA levels in LT-deficient mice, which lack PP and LN, are profoundly suppressed (1012), and introduction of LT-sufficient bone marrow (BM) into LT-deficient adult mice has been shown to rescue polyclonal IgA responses via signaling through LTβR in the gut (12). Although these results imply that at least some IgA responses can be rescued by restoring LTαβ/LTβR signaling in the adult animal, it is unclear whether this is also the case for IgA responses to mucosal pathogens. We selected RV as a model pathogen for several reasons: In mice, RV is highly tropic for small intestinal epithelial cells and induces a predominant IgA response that peaks in the first 9 to 13 days after infection (29, 30). Although RV infection in neonates can cause mild small intestinal pathology and diarrhea, RV infection in adults is asymptomatic (31). Furthermore, there are redundant mechanisms for clearance of RV in adult mice, with B cell–deficient mice not exhibiting impaired fecal RV antigen clearance unless CD8+ T cells are codepleted (29, 32). This is an advantage that allows us to explore the mechanisms of B cell activation in an acute, rather than chronic, setting. Thus, RV infection is an appealing model for studying a de novo IgA response in the adult gut without accompanying pathology.

To test the role of LTβR signaling on polyclonal and RV-specific IgA responses, we used a number of approaches to limit LTβR signaling to specific stages of life. We started with inhibition of the pathway during development up to sexual maturity (6 weeks of age) and then subsequently narrowed down this window to more discrete stages of early life. First, we created BM chimeric mice whereby wild-type (WT) or Ltb−/− BM was introduced into Ltb+/+ or Ltb−/− recipient mice, followed by cohousing (Fig. 1A and fig. S1A) (33). We then examined levels of polyclonal IgA in the feces versus IgA responses to RV. As expected, Ltb−/− mice that received Ltb−/− BM had very low levels of fecal IgA. Moreover, consistent with the results of Kang et al. (12), we observed comparable levels of total fecal IgA in WT➔Ltb−/− versus WT➔Ltb+/+ chimeric mice (fig. S1B), confirming that restoration of LTβR signaling in adult mice (by introduction of Ltb+/+ BM) has the capacity to rescue IgA levels in the gut. We next measured anti-RV IgA levels in mice that lack LTβR signaling in early life. Adult WT mice housed in our vivarium exhibit fecal anti-RV IgA levels at about day 7 (d7) after infection concomitant with RV antigen clearance (fig. S1, C and D). Because there is no anti-RV antibody standard available for this enzyme-linked immunosorbent assay (ELISA), all the fecal samples were tested in one ELISA plate, and results were confirmed using enzyme-linked immune absorbent spot (ELISpot) (see below). In contrast to the observed normalization of polyclonal fecal IgA levels upon transplantation of WT BM into Ltb−/− mice (fig. S1B), WT➔Ltb−/− chimeras exhibited a significant reduction in their ability to mount an anti-RV IgA response that persisted from d8 to d49, when compared with WT➔Ltb+/+ controls, and the low level of anti-RV IgA in WT➔Ltb−/− chimeras was comparable with Ltb−/−Ltb−/− chimeras (Fig. 1B). In addition, we used an RV-specific ELISpot to assay the frequency of IgA+ RV-ASC in the small intestinal lamina propria (SILP) at d8 and d49 after RV infection. In comparison with WT➔Ltb+/+ controls, the accumulation of IgA+ RV-ASC in the SILP was also significantly reduced in both WT➔Ltb−/− and Ltb−/−Ltb−/− chimeras at d8 and d49 after infection, respectively (Fig. 1C).

Fig. 1 LTβR signaling during fetal life in mice with intact MLN is critically required for the generation of an RV-specific IgA response.

(A) Diagram depicting experimental setup. (B) ELISA analysis of fecal anti-RV IgA levels in co-caged BM chimeric mice. Two independent experiments were performed with n = 4 mice per group per experiment, and one representative experiment is shown here. Data were analyzed by two-way ANOVA. **P < 0.01, ***P < 0.001. n.s., not significant; OD450, optical density at 450 nm; un, uninfected. All the samples were tested in one ELISA plate. (C) Enumeration of IgA+ RV-ASC in the SILP of BM chimeric mice at d8 (left) and d49 (right) after infection. At least two independent experiments were performed with n = 4 to 6 mice per group per experiment, and one representative experiment is shown here. (D) Representative picture of MLN from WT➔Ltb+/+ versus WT➔Ltb−/− chimeric mice at d8 after infection. (E) Enumeration of IgA+ RV-ASC in the MLN of chimeric mice at d8 after infection. Three independent experiments were performed with n = 4 to 6 mice per group per experiment, and one representative experiment is shown here. (F) Depiction of experimental setup for inhibiting LTβR signaling during both fetal and neonatal periods (E11 to d19). (G) Representative picture of MLN from E11 to d19 LTβR-Ig or control Ig (Ctrl Ig)–treated WT mice at d8 after infection. (H) Analysis of IgA+ RV-ASC in the MLN of E11 to d19 LTβR-Ig or control Ig–treated WT mice (n = 8 to 12 mice per group; data were pooled from three independent experiments). We included all the male and female mice that successfully develop MLN in this graph. (I) Depiction of experimental setup for in utero (E14/E17) inhibition of LTβR signaling in WT mice. (J) Representative picture of MLN from E14/E17 LTβR-Ig– or control Ig–treated WT mice at d8 after infection. (K) Enumeration of IgA+ RV-ASC in the MLN at d8 after infection of WT mice that received LTβR-Ig or control Ig at E14/E17. (L) Enumeration of IgA+ RV-ASC in the SILP at d8 (left) and d49 (right) after infection of WT mice that received LTβR-Ig or control Ig at E14/E17. Three independent experiments were performed with n = 4 to 6 mice per group per experiment, and one representative experiment is shown in (K) and (L). Data in (C), (E), (H), (K), and (L) were presented as means ± SEM and were analyzed using two-tailed unpaired Student’s t test.

In addition to cohousing WT➔Ltb+/+ and WT➔Ltb−/− chimeras to normalize the microbiota, we also assessed whether separately reared and maintained lines would exhibit the same results in terms of polyclonal and anti-RV IgA responses (fig. S2A). Unexpected, we observed that both total fecal IgA and anti-RV IgA responses were defective in WT➔Ltb−/− chimeric mice when compared with WT➔Ltb+/+ chimeric mice when recipient mice were reared and maintained separately (fig. S2, B to E). This suggested that Ltb−/− BM recipients may harbor an altered microbiota that impairs the generation or maintenance of homeostatic mucosal IgA responses that can be normalized upon co-caging Ltb−/− and Ltb+/+ BM recipients. 16S ribosomal RNA (rRNA) microbiota sequencing of fecal material revealed that Ltb−/− mice exhibit a markedly distinct intestinal microbiome profile compared with separately reared Ltb+/+ mice (fig. S3A). In particular, we observed that the relative abundance of Akkermansiaceae, which has been shown to inversely correlate with anticommensal IgA levels in Il33−/− mice (34), was greatly increased in the gut of Ltb−/− mice when compared with Ltb+/+ mice (fig. S3B). By contrast, Ltb−/− mice exhibited lower abundance of pro-IgA commensal gut microbes including Proteobacteria taxa (fig. S3C) (35). The differences in intestinal microbiota observed in Ltb+/+ versus Ltb−/− mice were largely normalized when Ltb+/+ and Ltb−/− mice were cohoused after WT BM transplantation (fig. S3, A, B, and D).

Together, these results demonstrate that LTαβ/LTβR signaling in the first 6 weeks of life is essential for anti-RV IgA responses during adulthood, and restoration of LTβR signaling in the adult animal cannot rescue this response. In addition, unlike polyclonal IgA responses, anti-RV responses rely on early-life LTβR signaling regardless of whether Ltb+/+ versus Ltb−/− BM recipient mice were kept separate or cohoused. Anticommensal polyclonal B cell responses are thought to be chronic and can occur independent of T cell help (24, 25). In contrast, anti-RV IgA production results in antigen clearance (fig. S1D) and is largely T cell–dependent (fig. S4, A and B), indicating that housing conditions have different impacts on the role of early-life LTβR signaling on polyclonal versus anti-RV IgA responses. Nevertheless, we used the co-caging modality for the remainder of our studies on anti-RV IgA responses.

LTβR signaling is critically required for the generation of an anti-RV IgA response in the adult MLN

We next asked whether the defect in production of RV-specific IgA observed in the SILP and the fecal pellet of WT➔Ltb−/− mice was due to a problem at the effector site (SILP) or due to a defect in the priming site. To ascertain which possibility applied, we first determined the location of B cell priming against RV in our mice. Because activation-induced cytidine deaminase (AID) is triggered by B cell receptor (BCR) activation (36), we used an AID–green fluorescent protein (GFP) reporter mouse to identify the location of B cell priming against RV (37, 38). Because fecal anti-RV IgA was induced at about d7 after infection in WT mice (fig. S1C), we assayed the expression of AID-GFP in B cells within this time frame. AID-GFP transgenic mice that were infected with RV showed evidence of AID-GFP expression mainly in the MLN rather than the PP or the SILP (fig. S4, C to E). In addition, using our RV-specific ELISpot readout, we found that the MLN contained about 1000-fold more IgA+ RV-ASC compared with the PP at d8 after infection and 4-fold greater compared with the SILP at this time point (fig. S4F). Because there was limited AID-GFP induction in the SILP, presumably the IgA+ RV-ASC in the SILP represents recent immigrants from the MLN inductive site. Together, these data demonstrate that the MLN is the main inductive site for the initiation of an IgA class–switched B cell response against RV.

We next determined whether the defect in anti-RV IgA responses in WT➔Ltb−/− mice was due to impaired generation of RV-specific IgA ASC in the MLN. We were able to assess this because WT➔Ltb−/− mice retain a small MLN (Fig. 1D) (11). Accordingly, we assayed the number of IgA+ RV-ASC in the MLN of WT➔Ltb−/− and Ltb−/−Ltb−/− BM chimeras, comparing with WT➔Ltb+/+ controls. Compared with WT➔Ltb+/+ controls, the number of RV-IgA+ ASC in the MLN was significantly reduced in WT➔Ltb−/− chimeric mice and, to a similar extent, in Ltb−/−Ltb−/− chimeric mice (Fig. 1E). This suggests that the MLN environment in WT➔Ltb−/− chimeric mice is incapable of supporting an RV-specific IgA response.

Thus far, our results indicate that LTβR signals derived from an unknown cell type in the first 6 weeks of life are required for the induction of an anti-RV IgA response in an adult that has had LTβR signaling restored via WT BM transplantation. To more accurately narrow down the relevant period of LTβR signaling that is required for generating an anti-RV IgA response in the MLN, we treated pregnant WT dams and their neonatal offspring with LTβR-Ig, starting from embryonic d11 to postnatal d19 (E11 to d19) (Fig. 1F) (39). We observed that, although most males develop MLN (six of seven; compared with seven of seven for control Ig group), most female littermates failed to develop MLN when treated with LTβR-Ig from E11 to d19 (two of six females have MLNs; compared with six of six for control Ig group; P < 0.05). Of those mice receiving E11 to d19 LTβR-Ig treatment that successfully developed MLN, the size of their MLN was comparable with control Ig–treated mice (Fig. 1G). Nevertheless, despite a grossly normal-appearing MLN, the generation of IgA+ RV-ASC in MLN was significantly impaired at d8 after RV infection in E11 to d19 LTβR-Ig–treated mice compared with the control Ig treatment group (Fig. 1H).

To further narrow down the relevant LTβR-dependent developmental period that is required for an optimal IgA response to RV during adulthood, we treated pregnant WT mice with LTβR-Ig at E14 and E17 (in utero treatment) (Fig. 1I). For more than 50 mice examined, all E14/E17 LTβR-Ig–treated mice successfully developed MLNs, and MLN sizes were comparable with those derived from the control Ig treatment group (Fig. 1J and fig. S4G). Consistent with previous findings (12), we observed comparable anticommensal polyclonal IgA responses in E14/E17 LTβR-Ig versus control Ig treatment groups (fig. S4H). However, we found that the frequency of antigen-specific IgA+ RV-ASC at d8 after RV infection was significantly reduced in the MLN of mice that received LTβR-Ig in utero compared with those that received control Ig (Fig. 1K). Consistent with this observation, the accumulation of IgA+ RV-ASC in the SILP was also significantly diminished at d8 and d49 after RV infection in E14/E17 LTβR-Ig–treated WT mice (Fig. 1L). In contrast, inhibition of LTαβ/LTβR signaling in adulthood, either by introduction of Ltb−/− BM into adult WT mice or by treatment of adult mice with LTβR-Ig, had no impact on the anti-RV IgA response (fig. S5). In summary, our results show that in utero LTβR signaling from E14 to birth is required to support an optimal anti-RV IgA response during adulthood. Moreover, LTβR signaling from E11 (but not E14) is partially required for the development of MLN particularly in female mice. The underlying mechanism(s) for the observed sex difference in MLN development in the setting of LTβR signaling inhibition warrants further investigation.

Early-life LTβR signaling is not required for the initiation of RV-induced GC responses in the adult MLN

There are many checkpoints that lead to the production of IgA-switched ASC during a mucosal immune response—including the initiation of a GC response, up-regulation of AID, class switch of B cells, and their differentiation into PC that can home appropriately to the SILP. We therefore wished to understand how early-life LTβR signaling affects the different steps that ultimately generate gut-homing anti-RV IgA-producing PCs. Because the LN conduit system participates in the timely delivery of virions to the LN environment (40), we first examined the impact of LTβR signaling inhibition at E14/E17 on the gross appearance of the antigen-delivery conduit system. Applying immunofluorescence staining for collagen I, a component of the conduit system (41), we found that the abundance of collagen I+ conduits in the MLN was comparable between mice treated in utero (E14/E17) with LTβR-Ig and those that received control Ig both during the steady state and at d8 after RV infection (fig. S6, A, C, and D).

We next examined the frequency of discrete GC B cell subsets in the MLN as a percentage of CD19+ B cells (fig. S7, A and B) (22). To do this, we introduced BM from B lymphocyte–induced maturation protein-1 yellow fluorescent protein (Blimp-1YFP) reporter mice (42) into Ltb+/+ or Ltb−/− recipient animals to identify and gate out plasmablasts and PCs from our analysis (this strategy was later used to specifically analyze these cells). Compared with uninfected controls, the frequency of GC B cells was significantly increased in the MLN of WT➔Ltb−/− chimeras at d8 after RV infection (fig. S7C), and WT➔Ltb−/− chimeras did not exhibit a significant defect in the percentage of GC or pre-GC/memory B cells in the MLN when compared with WT➔Ltb+/+ control chimeras (Fig. 2, A and B). A comparable frequency of GC or pre-GC/memory B cells was also observed in mice treated in utero (E14/E17) with LTβR-Ig versus control Ig (Fig. 2C). In addition, we performed immunofluorescence staining on the MLN from E14/E17 LTβR-Ig– or control Ig–treated WT mice and found that the number and average size of GC were comparable between these two groups (fig. S6, A and B). Consistent with the notion that the defect in the MLN of WT➔LTβ−/− mice is not due to an impaired GC response, RV infection induced a significant accumulation of Tfh cells in the MLN at d8 after infection, and no differences in the accumulation of Tfh cells in the MLN were observed in WT➔Ltb−/− versus control WT➔Ltb+/+ mice (fig. S7, D and E). These data demonstrate that early-life LTβR signaling is dispensable for the initiation of a GC response to RV infection in the MLN.

Fig. 2 Early-life LTβR signaling is not required for the initiation of RV-induced GC response in the adult MLN.

(A) Representative FACS plots of GC B cells in the MLN of WT➔WT versus WT➔LTβ−/− chimeras at d8 after infection. (B) Frequency of GC B cells and pre-GC/memory B cells in MLN CD19+ cells were compared between Blimp-1YFPLtb+/+ and Blimp-1YFPLtb−/− chimeras. Two independent experiments were performed with n = 5 to 6 mice per group per experiment, and one representative experiment is shown here. (C) Frequency of GC B cells and pre-GC/memory B cells in MLN CD19+ cells were compared between in utero (E14/E17) LTβR-Ig and control Ig–treated WT mice. Two independent experiments were performed with n = 4 mice per group per experiment, and one representative experiment is shown here. (D) Frequency of AID-GFP+ B cells during the first 9 days after infection in the MLN of the indicated BM chimeras (n = 3 mice for uninfected group; n = 4 to 6 mice for the RV-infected group per time point). %, frequency. Data in (B) and (C) were presented as means ± SEM and were analyzed using two-tailed unpaired Student’s t test, whereas data in (D) were analyzed by two-way ANOVA.

Transcription of AID in activated B cells is essential for antibody class switch recombination (CSR) from IgM to IgA in response to RV infection (36). To test whether the MLN of mice that lack LTβR signaling can support B cell–intrinsic expression of AID, we assayed the kinetics of AID-GFP expression in MLN B cells derived from AID-GFP➔Ltb+/+ versus AID-GFP➔Ltb−/− chimeric mice. Despite the lack of LTαβ/LTβR signaling during early life, we observed that the frequency of AID-expressing B cells in response to RV infection in the MLN of AID-GFP➔Ltb−/− mice was comparable with that of AID-GFP➔Ltb+/+ mice (Fig. 2D). Thus, early-life LTβR signaling is unlikely to be required for the expression of AID in MLN-intrinsic B cells in response to RV infection.

In summary, although there is a profound reduction in IgA+ RV-specific ASC in the MLN of mice that lack LTβR signaling in early life, this cannot be explained by defective GC initiation. Moreover, despite the observed reduction in the size of MLN in WT➔Ltb−/− chimeric mice (Fig. 1D), the Ltb−/− MLN (as well as MLN from mice treated with LTβR-Ig in utero) is capable of supporting at least the initial stages of a GC response, including the accumulation of Tfh, pre-GC, and GC B cells, and the B cell–intrinsic induction of AID expression in response to RV infection.

Early-life LTβR signaling is required for IgA class switch and the accumulation of gut-homing PCs in response to RV infection in the adult mouse

Although AID expression is required for a class-switched response to IgA, it is not sufficient. Thus, despite the normal GC response and AID induction in the Ltb−/− MLN, it is possible that CSR to IgA in mice that lack early-life LTβR signaling may still be impaired. We therefore assessed class switch to IgA in WT➔Ltb+/+ versus WT➔Ltb−/− chimeras as well as in mice treated in utero (E14/E17) with LTβR-Ig. Compared with WT➔Ltb+/+ controls, we observed a marked increase in the frequency of IgM+ RV-ASC and a corresponding reduction in the frequency of IgA+ RV-ASC in the MLN of WT➔Ltb−/− chimeric mice that translated into an overall reduction in the ratio of IgA+ versus IgM+ RV ASC, indicating a severe IgA CSR defect in MLN B cells from WT➔Ltb−/− chimeric mice (Fig. 3, A and B). In addition, compared with control Ig treatment, the IgA CSR defect was also observed when mice were treated with LTβR-Ig from E11 to d19 as well as during the narrower E14/E17 treatment window (Fig. 3, C and D). Because AID expression was normal in AID-GFP➔Ltb−/− chimeric mice, we reasoned that the expression of the μ and α switch regions located upstream of the μ and α constant regions must be affected by the absence of early-life LTβR signaling to account for the observed defect in IgA CSR. To test this, we sorted out AID-GFP+B220+ cells from the MLN of AID-GFP➔Ltb+/+ versus AID-GFP➔Ltb−/− chimeric mice and evaluated the levels of Iμ-Cμ and Iα-Cα germline transcripts at d8 after RV infection. Consistent with the observation of defective class switch in mice that lack early-life LTβR signaling, we found that the level of Iα-Cα germline transcripts in MLN B cells from AID-GFP➔Ltb−/− chimeras was significantly reduced and accompanied by a trend toward reduced levels of Iμ-Cμ germline transcripts compared with AID-GFP➔Ltb+/+ controls (Fig. 3E). Collectively, these data demonstrate that LTβR signaling from E14 to birth is likely to be required for the induction of Iα-Cα and Iμ-Cμ germline transcripts in AID+ B cells and subsequent IgA CSR in the adult MLN in response to RV infection.

Fig. 3 Early-life LTβR signaling is required for IgA class switch in the adult MLN.

(A) Enumeration of IgM+ RV-ASC (left) and IgA+ RV-ASC (right) in the MLN of BM chimeric mice at d8 after infection. (B) The ratio of IgA+ versus IgM+ RV-ASC in the MLN of chimeric mice at d8 after infection. Two independent experiments were performed with n = 3 to 4 mice per group per experiment, and one representative experiment is shown in (A) and (B). (C) The ratio of IgA+ versus IgM+ RV-ASC in the MLN of WT mice that received E11 to d19 LTβR-Ig versus control Ig at d8 after infection. (D) The ratio of IgA+ versus IgM+ RV-ASC in the MLN of WT mice that received E14/E17 LTβR-Ig versus control Ig at d8 after infection. Data in (C) and (D) were pooled from three independent experiments (n = 8 to 13 mice per group). (E) Levels of Iμ-Cμ and Iα-Cα germline transcripts (GLTs) in AID-GFP+/B220+ cells isolated from the MLN of AID-GFP➔Ltb+/+ or AID-GFP➔Ltb−/− chimeras at d8 after infection (n = 5 mice per group). HPRT, hypoxanthine guanine phosphoribosyltransferase. Data in (A) to (E) were presented as means ± SEM and were analyzed using two-tailed unpaired Student’s t test.

To further characterize the phenotype of ASC generated in the MLN of Ltb−/− chimeras, we next assessed the capacity for B cells to up-regulate Blimp-1, a transcription factor that is essential for PC differentiation (42, 43). To test this, we transferred Blimp-1YFP BM into WT and Ltb−/− hosts and tracked the expression of Blimp-1 in MLN-resident B cells at d8 after RV infection (fig. S8, A and B). We observed that the induction of YFP in MLN B cells was not impaired and increased in Blimp-1YFPLtb−/− chimeras when compared with Blimp-1YFPLtb+/+ controls (Fig. 4, A and B), which is consistent with our earlier work, whereby the GC response is slightly enhanced/accelerated at early time points in the absence of LTβR signaling in peripheral LN (38). However, because the reduction in frequency of IgA+ RV-specific ASC was also quite pronounced in the SILP of mice that lack early-life LTβR signaling (Fig. 1, C and L), we hypothesized that subsequent maturation of Blimp-1+ RV-induced PC may be impaired in WT➔Ltb−/− chimeric mice. Because the differentiation of IgA+ PC involves the progressive down-regulation of B220, CD19, and major histocompatibility complex class II (MHC II) molecules and up-regulation of CCR10 (20, 42, 44, 45), we used low B220 levels as a distinguishing mark of late-stage PC. RNA sequencing (RNA-seq) analysis of B220highYFPhigh and B220lowYFPhigh B cell subsets from RV-immunized Blimp-1YFPLtb+/+ chimeras validated that B220lowYFPhigh cells have a phenotype consistent with that of late-stage PC. Specifically, elevated expression of Ccr10 concomitant with reduced levels of Cd19 and MHC II molecules H2-Aa, H2-Ab1, H2-DMb2, and H2-Ob was observed in B220lowYFPhigh B cells compared with B220highYFPhigh B cells, consistent with a late-stage PC phenotype (Fig. 4C) (20, 42, 44, 45). Using this strategy, we analyzed the ratio of B220lowYFPhigh versus B220highYFPhigh B cells and found that PC differentiation in response to RV immunization was impaired in Blimp-1YFPLtb−/− chimeras when compared with control Blimp-1YFPLtb+/+ chimeras (fig. S8D). In addition, we assayed expression of genes such as Ccr9, Ccr10, Itga4, and Itgb7, all of which are involved in IgA+ ASC migration to the SILP. We observed a significant decrease in Ccr9, Ccr10, and Itgb7 transcripts and a trend toward diminished Itga4 transcripts in the B220lowYFPhigh B cells derived from the MLN of Blimp-1YFPLtb−/− chimeras at d8 after immunization when compared with control Blimp-1YFPLtb+/+ chimeras (Fig. 4D). Furthermore, we examined the surface expression of gut-homing receptors on late-stage PC and found that, at d8 after RV infection, the levels of α4β7 and CCR9 were significantly reduced in B220lowYFPhigh B cells from the MLN of Blimp-1YFPLtb−/− chimeras when compared with control Blimp-1YFPLtb+/+ chimeras (Fig. 4, E and F, and fig. S8, A, C, and E). Collectively, these data indicate that defects in the overall PC differentiation program occur in mice where LTβ expression is absent in early life. In summary, although early-life expression of LTβ is not required for the initiation of an RV-driven GC response and AID induction in adult MLN B cells, it is required for CSR to IgA and the up-regulation of gut-homing markers on Blimp-1–expressing B220low B cells.

Fig. 4 Early-life LTβR signaling is required for the accumulation of gut-homing PCs in adults.

(A) Representative FACS plots of B220lowYFPhigh and B220highYFPhigh B cells derived from the MLN of uninfected versus d8 after infection of Blimp-1YFPLtb+/+ chimeric mice. (B) Frequency of YFPhigh B cells in the MLN of Blimp-1YFPLtb+/+ versus Blimp-1YFPLtb−/− chimeric mice. Three independent experiments were performed with n = 3 to 5 mice per group per experiment, and one representative experiment is shown here. (C) Whole-genome RNA-seq of B220lowYFPhigh and B220highYFPhigh B cells sorted from MLN of Blimp-1YFPLtb+/+ chimeric mice at d8 after infection (n = 3 samples per group, and each sample is pooled from two to three mice). (D) Digital droplet PCR analysis of indicated genes in B220lowYFPhigh B cells sorted from MLN of Blimp-1YFPLtb+/+ versus Blimp-1YFPLtb−/− chimeric mice at d8 after infection (n = 4 samples per group, and each sample is pooled from two to three mice). (E) Representative FACS plots of surface expression of α4β7 and CCR9 in B220lowYFPhigh B cells from the MLN of Blimp-1YFPLtb+/+ versus Blimp-1YFPLtb−/− chimeras at d8 after infection. (F) Frequency of α4β7+CCR9+-expressing B220lowYFPhigh B cells in the MLN of Blimp-1YFPLtb+/+ versus Blimp-1YFPLtb−/− chimeras at d8 after RV infection. Two independent experiments were performed with n = 4 to 6 mice per group per experiment, and one representative experiment is shown here. Data in (B), (C), (D), and (F) were presented as means ± SEM and were analyzed using two-tailed unpaired Student’s t test.

Early-life LTβR signaling dictates the phenotype of MLN-resident stromal cells during adulthood

To test whether the anti-RV IgA CSR defect in WT➔Ltb−/− chimeras is B cell intrinsic or extrinsic, we sorted MLN B cells from uninfected WT➔Ltb+/+ and WT➔Ltb−/− chimeras and induced IgA and IgG1 CSR ex vivo in response to cytokines [transforming growth factor–β (TGF-β), interleukin-5 (IL-5), IL-4, and anti-IgD dextran for IgA CSR and IL-4 for IgG1 CSR] as well as Toll-like receptor 4 stimulation, conditions that have been shown to induce class switch (46, 47). We found that MLN B cells derived from Ltb−/− chimeras exhibited no CSR defect ex vivo (Fig. 5A). Therefore, consistent with the notion that LTβR is not expressed by B cells but rather by DCs, macrophages, and lymphoid stromal cells (5, 7, 8), the defect in CSR we observe in the MLN of mice that lack early-life LTβR signaling is B cell extrinsic.

Fig. 5 Early-life LTβR signaling dictates the phenotype of lymphoid stromal cells in the adult MLN.

(A) Frequency of IgG1 and IgA ex vivo class-switched B cells derived from the MLN of naïve WT➔Ltb+/+ or WT➔Ltb−/− chimeric mice. Two independent experiments were performed with n = 4 to 5 mice per group per experiment, and one representative experiment is shown here. (B) Representative flow cytometry plots of MLN stromal cells. LECs, lymphatic endothelial cells; BECs, blood endothelial cells. (C) Frequency of FRCs among Lin stromal cells in the MLN comparing WT➔Ltb+/+ and WT➔Ltb−/− chimeric mice at d8 after infection. Two independent experiments were performed with n = 5 to 6 mice per group per experiment, and one representative experiment is shown here. (D to F) Frequency of PDPN+CD31 FRCs among Lin cells in the MLN of d8-RV–infected WT mice that received LTβR-Ig or control Ig at E11 to d19 (D), at E14/E17 (E), or during adulthood (F). (G) Frequency of PDPN+CD31 FRCs among Lin cells in the MLN of uninfected WT➔WT and WT➔LTβ−/− chimeric mice. (H) Frequency of PDPN+CD31 FRCs among Lin cells in the MLN of uninfected WT mice that received LTβR-Ig or control Ig at E14/E17. At least two independent experiments were performed with n = 3 to 6 mice per group per experiment, and one representative experiment is shown in (D) and (H). Data in (G) were acquired with a different FACS machine than data in (C) to (F) and (H). Data in (A) and (C) to (H) were presented as means ± SEM and were analyzed using two-tailed unpaired Student’s t test.

Because it has been demonstrated that lymphoid stromal cells play an important role in modulating B cell and PC responses in peripheral lymphoid tissues (42, 4850), we reasoned that stromal cell–intrinsic LTβR signaling during early life may be important for generating IgA+ RV-specific ASC within the adult MLN. To test this, we first examined the expression of podoplanin (PDPN) and CD31 on lineage-negative (Lin) MLN stromal cells in WT➔Ltb+/+ versus WT➔Ltb−/− chimeric mice infected with RV to identify PDPN+/CD31 fibroblastic reticular stromal cells (FRCs), PDPN+/CD31+ lymphoid endothelial cells, PDPN/CD31+ blood endothelial cells, and PDPN/CD31 double-negative cells (DNs) (see fig. S9A) (51). Compared with WT➔Ltb+/+ controls at d8 after RV infection, the frequency of FRCs among Lin cells was significantly reduced in the MLN of WT➔Ltb−/− mice, whereas the representation of DNs was proportionally increased (Fig. 5, B and C). We did not calculate absolute numbers of FRCs in MLNs of WT➔Ltb−/− mice, because the decreased cellularity in WT➔Ltb−/− MLN would likely skew the analysis (as opposed to mice receiving LTβR-Ig in utero; see below). This observation of disproportionate loss in FRCs in the inflamed MLN after RV infection in chimeric mice was recapitulated in mice treated with LTβR-Ig from E11 to d19 (Fig. 5D and fig. S9B) and from E14/E17 (Fig. 5E and fig. S9C). In contrast, the frequency and number of FRCs increased in the MLN when LTβR-Ig was administered in adulthood (Fig. 5F and fig. S9D). This may due to compensatory effects in the FRC compartment because inhibition with LTβR-Ig in adult mice results in a marked reduction of follicular dendritic cells (52).

In terms of the phenotype of FRCs, LTβR-Ig treatment at E14/E17 reduced the expression of MLN stromal cell activation markers (53), including vascular cell adhesion molecule–1 and MHC I (H-2Kb) as measured by flow cytometry (fig. S10, A to C). We also used RNA-seq on purified CD31 stromal cells (FRCs plus DNs) derived from MLNs of mice that were treated at E14/E17 with LTβR-Ig versus control Ig. Among the known positive regulators of gut IgA responses (34, 5457), we found a significant reduction in Mmp2, Mmp9, Cxcl12, Il33, Il1b, Tnfsf12, Rdh10, Aldh1a3, and Aldh1a1 transcripts and a trend toward reduced Tnfsf13b (Baff) and Il6 transcripts in mice that received LTβR-Ig in utero (E14/E17) (fig. S10, D and E), and similar results, with the exception of Aldh1a1, were observed when MLN stromal cells were subjected to quantitative polymerase chain reaction (qPCR) (Fig. 6A).

Fig. 6 Early-life LTβR signaling dictates the gene expression of lymphoid stromal cells in the adult MLN.

(A) Quantitative PCR analysis of Mmp2, Mmp9, Cxcl12, Il33, Tnfsf13b, Il6, Tnfsf12, Rdh10, Aldh1a3, and Aldh1a1 transcripts of MLN CD45/Ter-119/EpCAM/CD19/CD31 stromal cells from d8-RV–infected WT mice that were treated at E14/E17 with LTβR-Ig or control Ig. (B) qPCR analysis of Mmp2, Mmp9, Cxcl12, Il33, Tnfsf13b, Il6, Tnfsf12, Rdh10, Aldh1a3, and Aldh1a1 transcripts of MLN CD45/Ter-119/EpCAM/CD19/CD31 stromal cells from uninfected WT mice that were treated at E14/E17 with LTβR-Ig or control Ig. Data in (A) and (B) were combined from two independent experiments (n = 7 to 8 samples per group, and each sample is pooled from two mice), were presented as means ± SEM, and were analyzed using two-tailed unpaired Student’s t test.

In addition to the inflamed MLN at d8 after RV infection, we also examined the effect of early-life LTβR signaling on MLN FRCs at steady state in WT➔Ltb−/− chimeras and E14/E17 LTβR-Ig–treated mice compared with controls. We found that the marked reduction in FRCs that we observed in the MLN of RV-infected mice was not observed during the steady state (Fig. 5, G and H). In contrast, when we perform the same gene expression analysis for IgA switch factors as we did for d8-RV–infected mice (Fig. 6A), the expression of these genes (with the exception of Aldh1a1) was still reduced in the uninfected MLN stromal cells of E14/E17 LTβR-Ig–treated mice, when compared with the control Ig group (Fig. 6B). In addition, we observed that the numbers of FRCs in the MLN of E14/E17 control Ig–treated mice were reduced at d8 after RV infection when compared with uninfected conditions (fig. S9, C and E).

Because stromal cell fibers that stain positive with the ER-TR7 monoclonal antibody are known to be produced by FRCs (58), we also performed ER-TR7 immunofluorescence staining on the uninfected and d8-RV–infected MLN from E14/E17 LTβR-Ig– or control Ig–treated mice as an independent measure of FRC abundance. We found that, similar to our flow cytometry results enumerating PDPN+/CD31 cells, the density of ER-TR7+ cells was reduced in the reactive MLN, but not in the uninfected MLN, in mice treated with LTβR-Ig at E14/E17 when compared with control Ig–treated mice (fig. S6, C and E). In summary, early-life LTβR signaling is required to maintain FRC numbers and ER-TR7 levels in the MLN during RV infection and is likewise critical for sustained gene expression of pro-IgA switch factors irrespective of RV infection.

Conditional deletion of LTβR in stromal cells impairs mucosal anti-RV IgA responses

Because a reduction in FRCs was observed in the adult MLN of mice treated in utero with LTβR-Ig during RV infection and because these stromal cells exhibited defects in many pro-IgA switch factors, we hypothesized that specific deletion of LTβR in FRCs would result in an impaired IgA response to RV in the adult mouse. Our previous finding that hematopoietic expression of LTβR (in myeloid cells) has no impact on the RV-IgA response (59) further supports this hypothesis.

To test this, we examined the RV-specific IgA response in Ccl19-Cre Ltbrfl/fl mice that lack LTβR signaling specifically in FRCs (58). Consistent with previous studies examining the inguinal LN in the steady state (58), the frequency of FRCs in inguinal LNs is markedly reduced in uninfected Ccl19-Cre Ltbrfl/fl mice when compared with littermate controls (fig. S11A). Moreover, consistent with E14/E17 LTβR-Ig–treated mice and WT➔Ltb−/− chimeras, compared with littermate controls, Ccl19-Cre Ltbrfl/fl mice exhibited normal numbers of FRCs in the MLN in the steady state (fig. S11B). These data suggest that the FRCs in the MLN have differential reliance on LTβR signaling when compared with peripheral (inguinal) LN under homeostatic conditions.

Next, we examined the impact of Ltbr deletion in FRCs during RV infection. Unlike during the steady state, we found that the frequency of FRCs in the reactive MLN of Ccl19-Cre Ltbrfl/fl mice among lin stromal cells was significantly reduced when compared with littermate controls (Fig. 7, A and B), similar to what we observed in E14/E17 LTβR-Ig–treated mice and WT➔Ltb−/− chimeras. However, unlike mice treated with LTβR-Ig at E14/E17 or WT➔Ltb−/− chimeric mice, Ccl19-Cre Ltbrfl/fl retain most of their PPs (Fig. 7C), thus allowing us to definitely disentangle a putative role for PPs in the anti-RV IgA immune response. When we examined the RV response in Ccl19-Cre Ltbrfl/fl mice, we observed a significant reduction in fecal anti-RV IgA titers when compared with littermate controls (fig. S11C). Moreover, compared with littermate controls, we observed a defect in IgA CSR in the MLN at d8 after RV infection and a corresponding reduction in the accumulation of IgA+ RV-ASC in the SILP in Ccl19-Cre Ltbrfl/fl mice at d8 and d45 after infection (Fig. 7, D to F, and fig. S11D), further demonstrating that LTβR signaling in stromal cells is required for the induction and/or maintenance of an intestinal anti-RV IgA response. In summary, these data provide direct evidence that LTβR expression in FRCs is required for the anti-RV IgA response.

Fig. 7 Deletion of LTβR in FRCs impairs the anti-RV IgA response.

(A) Representative FACS plots of MLN stromal cells from Ccl19-Cre Ltbrfl/fl versus Ltbrflox littermate controls (Cre-negative Ltbrfl/fl or Ltbrfl/+) at d8 after infection. (B) Frequency of FRCs and DNs among Lin cells was compared between d8-RV–infected Ccl19-Cre Ltbrfl/fl and Ltbrflox littermate controls. Three independent experiments were performed with n = 3 to 5 mice per group per experiment, and one representative experiment is shown here. (C) Number of PPs per mouse in d8-RV–infected Ccl19-Cre Ltbrfl/fl and Ltbrflox littermate controls (n = 3 to 5 mice per group). (D) Ratio of IgA+ versus IgM+ RV-ASC in MLN of Ccl19-Cre Ltbrfl/fl versus Ltbrflox littermate controls at d8 after infection. Three independent experiments were performed with n = 4 to 5 mice per group per experiment, and the pooled data from all three experiments are shown here. (E) Enumeration of IgA+ RV-ASC in the SILP of Ccl19-Cre Ltbrfl/fl versus Ltbrflox littermate controls at d8 after RV infection. Two independent experiments were performed with n = 4 to 5 mice per group per experiment, and one representative experiment is shown here. (F) Enumeration of IgA+ RV-ASC in the SILP of Ccl19-Cre Ltbrfl/fl versus Ltbrflox littermate controls at d45 after RV infection (n = 5 to 6 mice per group). Data in (B) to (F) were presented as means ± SEM and were analyzed using two-tailed unpaired Student’s t test.

DISCUSSION

In this study, we found a previously unidentified role for early-life LTαβ/LTβR signaling in shaping the humoral immune response to RV infection during adulthood. We further show that early-life LTβR signaling influences MLN stromal cell gene expression and phenotype. Consequently, although the initiation of a GC response and up-regulation of AID within B cells in the adult MLN is normal, subsequent class switch to IgA and the acquisition of gut-homing molecules on post-GC Blimp-1+ PC is strongly impaired. Last, we found that deletion of Ltbr in FRCs recapitulates this phenotype. We postulate that the combination of a reduction of FRCs, as well as diminished expression of key pro-IgA factors in remaining FRCs in the MLN of mice that lack early-life LTβR signaling, accounts for the pronounced IgA CSR defect in response to RV infection in these mice (fig. S12).

PPs have been thought as the major site for homeostatic commensal-specific IgA induction and for initiating the intestinal immune response to pathogens (60). Antigen and immune cells from PPs can reach the MLNs via lymphatics, with MLNs serving as a second line of defense (61, 62). Mice that lack LTβ in utero, such as those we studied herein, lack PPs (11). This could potentially complicate the interpretations of our results. Nevertheless, we found that during RV infection, the induction of AID-GFP expression and the generation of IgA+ RV-specific ASC were all focused within the MLN rather than the PP. This is consistent with work from Greenberg and colleagues (21) who observed that IgA+ RV-ASC peaked in the MLN at about d7 after infection compared with the PP at d270 after RV infection, suggesting that during the early phases of the RV IgA response, generation of RV-specific IgA ASC occurs largely in the MLN. In addition, although PPs are present in Ccl19-Cre Ltbrfl/fl mice, the RV-specific IgA response is markedly impaired. Thus, PPs are neither required nor sufficient for induction of an anti-RV IgA response, at least in the first weeks after RV infection. However, we cannot completely exclude the possibility that there is some relevant event occurring in the PP under normal physiological conditions that optimizes anti-RV IgA responses. For example, the PP may play a role in shuttling RV antigen into the MLN where the immune response is then initiated, and RV has been shown to localize to the dome of the PP (30). Thus, it would be interesting to examine whether immune cells in the PP migrate to the MLN during RV infection using a photoconversion approach, for example, in Kaede mice (63).

Having shown that PPs are not necessary for generating anti-RV IgA responses, it was curious to note that although the MLN in WT➔Ltb−/− chimeras is smaller (Fig. 1), it is nevertheless capable of supporting Tfh cell expansion, generation of GC B cells, and up-regulation of Blimp-1 in B cells. This is consistent with studies performed by Koni and Flavell (64) several years ago where they noted the presence of GC in the MLN of Ltb−/− mice, demonstrating that the small MLN in Ltb−/− mice is nevertheless “GC competent.” Thus, it seems that mice with multiple disadvantages (no PP and small MLN) are hard-wired to initiate a GC response to RV. Although a small Ltb−/− MLN is sufficient to induce an anti-RV IgA response, we nevertheless do not know whether it is necessary. Surgical removal of the MLN would allow one to determine whether other organized LN tissues beyond the MLN can support the RV IgA response.

Despite the ability to initiate the anti-RV IgA response in the MLN of mice that lack LTβR signaling in early life, subsequent steps of this response—namely, the expression of Iα-Cα germline transcripts, class switch to IgA, accumulation of Blimp-1+B220low B cells, and the expression of Ccr9, Ccr10, and Itgb7 in Blimp-1+B220low B cells— are all impaired. These defects are accompanied by a marked reduction in RV-specific IgA+ ASC in the SILP and RV-specific IgA in the feces. Therefore, although the GC response occurs in the Ltb−/− MLN, the capacity of the GC to support class switch, as well as the quality of the emerging Blimp-1+ cells, including their potential to reach the SILP effector site, is significantly deranged.

Using Ccl19-Cre mice to eliminate LTβR specifically in stromal cells, we found that the anti-RV IgA phenotype in Ccl19-Cre Ltbrfl/fl mice recapitulates that observed in the WT➔Ltb−/− chimeras and E14/E17 LTβR-Ig–treated mice. These results imply that stromal cell–intrinsic LTβR signaling is required to support the anti-RV IgA response. In mice that lack early-life LTβR signaling, gene expression of multiple IgA switch factors was reduced in MLN stromal cells in both the steady state and during RV infection. For mRNA measurements, we assessed a bulk population of CD31 stromal cells, including DNs that can up-regulate PDPN and differentiate to FRCs in an LTβR-dependent manner (58). Analyzing pooled FRCs and DNs allowed us to evaluate how early-life LTβR signaling affects the gene expression of all FRC stages/subsets, including their precursors. However, the bulk RNA-seq approach cannot distinguish which subsets of FRCs were affected by early-life LTβR inhibition. Considering the high degree of heterogeneity among FRCs (65, 66), it would be very interesting to perform single-cell RNA-seq of the MLN to test whether early-life inhibition of LTβR signaling affects the development and transcriptional signature of selected MLN stromal cell subsets.

In addition to changes in gene expression, we noted that the abundance of FRCs was also reduced in mice that lack early-life LTβR signaling. However, this was only observed in the MLN during RV infection and not in the steady state. In control Ig–treated mice, we noticed that the numbers of FRCs were reduced in the MLN of d8-RV–infected mice when compared with uninfected animals, implying that MLN FRCs are somehow sensitive to RV infection. In other settings where FRCs are reduced during inflammation [e.g., lymphocytic choriomeningitis virus (LCMV) infection], LTβR signaling is required for FRC recovery (67). Although the mechanisms of FRC loss are likely different in the two systems (LCMV versus RV), it is possible that the LT pathway plays a key role in FRC recovery after inflammation in a variety of settings.

Our previous studies found that DC- and macrophage-intrinsic LTβR signaling are dispensable for anti-RV IgA responses (59), further supporting a stromal cell–intrinsic role for LTβR signaling in supporting anti-RV IgA responses. This contrasts with recent findings showing that DC-intrinsic LTβR signaling is required for polyclonal IgA responses generated in the PP (22). However, when the microbiota is normalized in co-caged mice, we found that the polyclonal IgA response in mice that lack early-life LTβR signaling is normal, implying that the stromal cell–dependent mechanism that is required for the anti-RV IgA response described in this study is not operational in maintaining polyclonal IgA responses and presumably a different mechanism is at play (such as DC-intrinsic LTβR signaling). PPs harbor distinct subsets of DCs that play unique roles in inducing anticommensal IgA responses in this location (68, 69), and these anticommensal promoting DCs are presumably not relevant for the anti-RV IgA response.

The reasons why anti-RV IgA versus polyclonal IgA responses have different requirements vis-à-vis the LT pathway are not clear, but it is likely due to the fact that these immune responses are fundamentally different in multiple ways: First, as shown by Reboldi et al. (22), polyclonal IgA responses that depend on DC-intrinsic LTβR signaling are initiated in the PP, whereas we show here that B cell priming against RV is initiated in the MLN. Second, reports have demonstrated that polyclonal IgA responses largely do not require T cell help (24, 25), or even organized GC reactions such as one would observe in the MLN (13, 70). This is in sharp contrast to anti-RV IgA responses, which are several fold lower in T cell–deficient (αβ/γδ TCR−/−) mice (71), agreeing with our own data whereby CD4+ T cell depletion results in a significant reduction in anti-RV IgA. Moreover, we observed an induction in Tfh cells in the MLN and a robust and coordinated GC response upon RV infection. Third, it has been reported that some anticommensal IgA can be produced by innate B1 cells (72). Again, this is in sharp contrast to the RV response where B2 cells, not B1 cells, are the precursors for anti-RV IgA-producing cells (73). Last, alterations in the gut microbiota have been shown to affect the levels of fecal IgA (33), and we observed differences in polyclonal IgA levels in separately caged WT➔Ltb+/+ versus WT➔Ltb−/− chimeric mice that were normalized upon co-caging. Because co-caging of Ltb−/− and Ltb+/+ mice also equilibrates the microbiome of Ltb−/− mice, this further confirms that IgA levels are sensitive to animal husbandry practices, and by extension the composition of the microbiota. In sharp contrast, the defect we observed in the anti-RV IgA response in WT➔Ltb−/− chimeric mice persisted in both the separately caged and co-caged setups. In summary, given the multiple distinctions in the location, initiation, and regulation of polyclonal IgA versus anti-RV IgA responses, it is perhaps expected that the LT pathway would differentially regulate such responses.

The direct mechanism for how stromal cells affect the anti-RV IgA response is likely related to the variety of IgA switch factors that are expressed by MLN stromal cells in an (early-life) LTβR–dependent manner. Retinoic acid, a vitamin A metabolite produced by retinal dehydrogenases, is important for promoting IgA CSR and for the up-regulation of gut-homing molecules including α4β7 on IgA-ASC (19, 74, 75). Our data revealed that the expression of some components of the retinoic acid synthesis pathway (i.e., Aldh1a1, Aldh1a3, and Rdh10) in MLN stromal cells was significantly reduced by E14/E17 LTβR-Ig treatment when examined with RNA-seq. The reductions in expression of both Aldh1a3 and Rdh10 were confirmed by qPCR but not Aldh1a1. This discrepancy may be due to the multiple mRNA splice variants for Aldh1a1, leading to different results with the two techniques. Nevertheless, although the oxidation of retinal to retinoic acid is performed by multiple overlapping enzymes including Aldh1a1 and Aldh1a3, Rdh10 activity is essential for the upstream conversion of retinol to retinal (76). Thus, it is likely that optimal generation of retinoic acid in the MLN requires LTβR signaling.

TGF-β is a very key pro-IgA factor (77); thus, the impaired induction of Iμ and Iα germline transcripts in the absence of early-life LTβR signaling might be due to limited availability of the active cleaved form of TGF-β1 within the MLN microenvironment. Mmp2 and Mmp9, which encode the matrix metalloproteinases (MMPs) that cleave TGF-β1 from its latent complex (54), were both markedly reduced in MLN stroma after in utero (E14/E17) inhibition of LTβR signaling. In addition to these MMPs, the expression of well-known pro-IgA switch factors Tnfsf13b and Il6 (55), as well as Il33, Il1b, and Tnfsf12, was reduced in MLN stromal cells derived from mice that received LTβR-Ig in utero (E14/E17). IL-33 has been shown to synergize with TGF-β to promote IgA generation, and mice deficient in IL-33 have much lower levels of intestinal IgA (34). IL-1β–deficient mice also manifest decreased levels of IgA-producing cells in the SILP and reduced levels of intestinal IgA, concomitant with decreased mRNA expression of inducible nitric oxide synthase in the small intestine (56). It has been reported that patients with autosomal dominant deficiency in TNFSF12 exhibit reduced levels of IgA, although the molecular mechanisms by which TNFSF12 regulates IgA production are unclear (57). Collectively, we reason that a putative reduction in active TGF-β1 along with a reduction in other pro-IgA switch factors expressed by MLN stromal cells conspires to create an environment in the adult MLN that fails to promote IgA class switch and the proper up-regulation of mRNA encoding LP homing molecules (Ccr9, Ccr10, Itga4, and Itgb7) (20, 21). This “compound” FRC defect results in a profound reduction in RV-specific IgA production in the gut, and to our knowledge, this is the first example of a molecular signal during fetal life that imprints the phenotype of MLN-resident FRCs in the adult. In contrast, although some stromal cells such as follicular dendritic cells are highly sensitive to LTβR-Ig treatment in the adult mouse (52), treatment of adult mice with LTβR-Ig did not affect the frequency of FRCs proportional to other lin MLN cells or the IgA response to RV. It remains to be investigated whether the LTβR signaling–dependent stromal cells we have implicated in the IgA response to RV are involved in the immune response against other intestinal pathogens.

Early-life exposures including the maternal microbiota and microbiota-associated metabolites, maternal nutrition, and maternal antibodies have all been shown to have important effects on the development of the mucosal and peripheral immune system (78). These exposures may ultimately influence immune responses to vaccines. Vaccination against RV, the leading cause of childhood diarrhea worldwide (79), works well in North American children, yet is less efficacious in resource-limited countries (80, 81). Identifying molecular pathways that operate in the perinatal period to poise the organism for an optimal immune response to pathogens later in life (such as the LT pathway) will help inform vaccine design. Moreover, in addition to IgA production, several recent studies including our own have shown that IgA-producing PCs have other functions, notably the production of cytokines and the expression of checkpoint inhibitory molecules that play immunosuppressive roles in the immune system (63, 8284). Thus, an understanding of how the MLN environment fosters the generation and function of IgA-producing B cells is important for gaining insights into mechanisms of immune system homeostasis.

MATERIALS AND METHODS

Study design

The goal of this study was to investigate which stages of LTβR signaling affect gut IgA responses to mucosal pathogens (i.e., RV) in adults and to explore the underlying mechanisms. We used several approaches (e.g., BM chimeras and LTβR-Ig treatment) to restrict LTβR signaling to specific stages of life, starting with inhibition of the pathway during development up to 6 weeks of age and then narrowing down the interval of inhibition to a short stage in early life. The experiments were performed using age- and sex-matched mice. The sample size per group and experimental repeats were indicated in each figure legend, and samples were not double-blinded or randomized during experiments or analysis.

Mice

C57BL/6 WT mice (Charles River Laboratories, St. Constant, QC, Canada) were bred in the Division of Comparative Medicine Animal Vivarium facility at the University of Toronto. Ltb−/− mice were originally from B&K Universal and bred in our animal facility (85). AID-GFP mice were obtained from R. Casellas, and Blimp-1YFP mice were purchased from the Jackson Laboratory. Ccl19-Cre Ltbrfl/fl mice were generated as previously described (58). BM chimeras were generated as we previously described (38, 86); irradiated mice were rested for 8 to 12 weeks to allow for BM reconstitution and were kept on neomycin sulfate water (2 g/liter) for the first 2 weeks. The mice were maintained under specific pathogen–free conditions. The experimental procedures were approved by the Animal Care Committee of University of Toronto and by the Cantonal Veterinary Office (St. Gallen, Switzerland).

Tissue harvest and cell isolation

At the indicated time points, all MLNs or PPs were harvested and ground between glass slides, followed by filtration with a 70-μm cell strainer. BM cells were flushed out from femurs and tibia of mice, followed by red blood cell lysis. Single-cell suspensions from PP were prepared by grinding PP between glass slides, followed by filtration with a 70-μm cell strainer. For AID-GFP fluorescence-activated cell sorting (FACS) and RV-ELISpot with SILP cells, PPs were first removed from the small intestine. The segmented small intestine was washed with CMF buffer (Hanks’ balanced salt solution, 2% fetal bovine serum, and 15 mM Hepes) and then vigorously shaken in CMF/EDTA buffer (5 mM EDTA), followed by digestion with freshly made enzyme mix composed of RPMI 1640 containing Collagenase IV (0.25 mg/ml; Sigma-Aldrich) and deoxyribonuclease I (0.025 mg/ml; Roche) in a 37°C water bath.

Fecal polyclonal IgA ELISA

Fecal pellets were collected from the indicated groups at about 2 months after BM transplantation. Fecal supernatant was prepared (10%, w/v) with phosphate-buffered saline (PBS). The Ig isotype–specific ELISAs were performed as we previously described (87). Briefly, 96-well ELISA plates (Nunc) were coated overnight at 4°C with goat anti-mouse Ig (SouthernBiotech). After blocking with 2% bovine serum albumin, one-fifth diluted fecal samples were added into the wells, followed by serial threefold dilutions. The bound antibodies from fecal supernatants were detected with goat anti-mouse IgA that was conjugated with alkaline phosphatase (SouthernBiotech), followed by development with tetramethylbenzidine (TMB) substrate (BioShop).

Assessment of the fecal microbiome

For fecal microbiome analysis, fecal pellets were collected from sex-matched 2-month-old Ltb+/+ or Ltb−/− mice, as well as WT➔Ltb+/+ and WT➔Ltb−/− BM chimeric mice, followed by DNA extraction using the NucleoSpin Soil kit (MACHEREY-NAGEL). Bacterial 16S rRNA genes were sequenced and analyzed as we previously described (88). Briefly, 16S ribosomal DNA sequencing reads were processed in R version 3.4.4 using the DADA2 version 1.6.0 bioinformatics pipeline (89). Amplicon sequence variants were inferred, paired-end sequences were merged, and chimeras were removed. Phylogeny was assigned using a custom R script that used VSEARCH to search the EzTaxon database (January 2017 update) and assigned taxonomy down to the species level, using the highest identity match greater than 97%. The phyloseq package was used to perform principal coordinates analysis using Bray-Curtis dissimilarity. Differential abundance analysis of bacterial taxa between study groups was performed using DESeq2.

RV-specific ELISA and ELISpot

WT mice, AID-GFP mice, or BM chimeric mice were infected with RV as we previously described (47, 59). Fecal pellets were collected from each mouse 1 day before RV challenge and on the indicated days after infection. Fecal supernatant was prepared (10%, w/v) with PBS, and one-half dilution of fecal supernatant was used for anti-RV IgA and RV antigen ELISAs, as we previously described (47, 59). For anti-RV IgA ELISA, serial dilutions of RV-infected samples were used to validate the assay, and all the samples shown in the individual graph were tested in one ELISA plate. At the indicated days after RV infection, an RV-specific ELISpot assay was performed, measuring the numbers of RV-specific IgA+ or IgM+ ASC from indicated tissues. Lymphocytes from the SILP were prepared by Percoll gradient as previously described (59). Briefly, MultiScreen-HTS-HA filter plates were coated overnight with inactive antigen (Microbix) and then blocked with complete RPMI 1640 media. The plates were incubated with twofold serial dilutions of MLN, BM, or SILP lymphocytes overnight, followed by detection with horseradish peroxidase (HRP)–conjugated goat anti-mouse IgA (SouthernBiotech). 3-amino-9-ethylcarbazole (AEC) (Vector Laboratories) was then used as the colorimetric precipitating substrate for HRP to develop the plate. Positive spots on the membrane within each well were counted in a blinded manner with a Nikon stereomicroscope. For the ASC data presented in each graph, all the experimental and control mice were infected with RV at the same time and harvested together at indicated time points.

Pharmacological inhibition of LTβR signaling

To pharmacologically inhibit LTβR signaling in adult mice, we intraperitoneally injected WT mice with 100 μg of LTβR-Ig or isotype control antibodies on the indicated days. To inhibit LTβR signaling in utero, we injected pregnant WT dams with 200 μg of LTβR-Ig or control antibodies at d14.5 and d17.5 of pregnancy via tail vein injections, as previously described (39). The E14/E17 LTβR-Ig–treated pups were co-caged with control Ig–treated pups since weaning (i.e., 3 weeks old). To inhibit LTβR signaling during both fetal life and neonatal periods, we injected pregnant WT dams with 200 μg of LTβR-Ig or control antibodies at d11.5, d14.5, and d17.5 of pregnancy via tail vein injections, and the delivered offspring received injections of LTβR-Ig or control antibodies (100 μg of antibody per 20 g of mice) at neonatal d5.5, d12.5, and d19.5. The E11 to d19 LTβR-Ig–treated pups were co-caged with control Ig–treated pups since weaning.

Ex vivo CSR

B cells were sorted from the MLN of co-caged WT➔Ltb+/+ and WT➔Ltb−/− chimeras, and IgG1 or IgA isotype switching was induced ex vivo, as we previously described (47). Briefly, single-cell suspensions of MLN were stained with anti-B220 (BV605) antibody in the presence of Fc block, and B cell sorting was performed with a FACSAria IIu sorter (BD Biosciences). LN B cells were cultured in complete RPMI 1640 with lipopolysaccharide for 4 days. IL-4 was added to induce IgG1 switching, whereas IL-4, TGF-β, IL-5, and anti-IgD dextran were added for IgA switching.

Statistical analysis

Data are presented as means ± SEM. All data were analyzed using two-tailed unpaired Student’s t test except for the anti-RV IgA ELISA kinetics that were analyzed by two-way analysis of variance (ANOVA) and the frequency of MLN formation in LTβR-Ig–treated mice that was analyzed by chi-square test. P < 0.05 was considered significant.

SUPPLEMENTARY MATERIALS

immunology.sciencemag.org/cgi/content/full/4/42/eaax1027/DC1

Methods

Fig. S1. Intestinal polyclonal IgA responses in co-caged BM chimeric mice and kinetics of RV infection in WT mice.

Fig. S2. Intestinal anticommensal and anti-RV IgA responses in the separately caged BM chimeric mice.

Fig. S3. Analysis of intestinal microbiome in separate versus co-caged Ltb+/+ and Ltb−/− mice.

Fig. S4. Anti-RV IgA response is mainly T cell dependent, and the MLN is the major initiation site of AID induction in response to RV infection.

Fig. S5. LTβR signaling during adulthood is dispensable for the induction and maintenance of an IgA response against RV.

Fig. S6. Immunofluorescence stains of the MLN from mice that received LTβR-Ig or control Ig at E14/E17.

Fig. S7. Analysis of GC B cells and Tfh cells in the MLN.

Fig. S8. PC analysis in the MLN at d8 after RV infection.

Fig. S9. Analysis of lymphoid stromal cells in the MLN.

Fig. S10. Activation status and gene expression of MLN stromal cells and the effects of CD4+ T cell depletion on the anti-RV IgA response.

Fig. S11. Analysis of lymphoid stromal cells in the inguinal LN and MLN, and anti-RV IgA responses in Ccl19-Cre Ltbrfl/fl mice.

Fig. S12. Schematic of proposed mechanisms for early-life LTβR signaling in regulating mucosal anti-RV IgA responses during adulthood.

Table S1. Raw data file (Excel spreadsheet).

References (90, 91)

REFERENCES AND NOTES

Acknowledgments: We thank Biogen Idec for providing LTβR-Ig and isotype control antibodies and H. Greenberg for providing the murine RV (EC strain). We thank the Gommerman lab members for advice and discussion during this study and D. White for the help in flow cytometry. Funding: This study was supported by a foundation grant from the Canadian Institutes of Health Research to J.L.G. (#15992), project grants from the Canadian Institutes of Health Research (PJT-153307 to A.M., #144628 to W.N., and PJT-156035 to S.L.), a Princess Margaret Cancer Foundation grant to H.H.H., and a Swiss National Science Foundation grant 166500 to B.L. C.L. is a recipient of the Canadian Institutes of Health Research Postdoctoral Fellowship. Author contributions: C.L. designed the experiments, performed the research, analyzed data, and wrote the manuscript. E.L., C.P.-S., L.A.W., J.Z., D.L., A.N., K.V.K., and T.S. performed the research. M.A. and H.H.H. analyzed the RNA-seq data. E.B. and W.N. analyzed the microbiome sequencing data. O.R. edited the manuscript. S.L., A.M., and B.L. analyzed data and edited the manuscript. J.L.G. is the principal investigator who designed the experiments, analyzed data, and wrote the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: The raw RNA-seq data files for this study have been deposited in the Gene Expression Omnibus repository under the accession number GSE138457. All other data needed to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.

Stay Connected to Science Immunology

Navigate This Article