Salmonella infection induces the reorganization of follicular dendritic cell networks concomitant with the failure to generate germinal centers

Summary Germinal centers (GCs) are sites where plasma and memory B cells form to generate high-affinity, Ig class-switched antibodies. Specialized stromal cells called follicular dendritic cells (FDCs) are essential for GC formation. During systemic Salmonella Typhimurium (STm) infection GCs are absent, whereas extensive extrafollicular and switched antibody responses are maintained. The mechanisms that underpin the absence of GC formation are incompletely understood. Here, we demonstrate that STm induces a reversible disruption of niches within the splenic microenvironment, including the T and B cell compartments and the marginal zone. Alongside these effects after infection, mature FDC networks are strikingly absent, whereas immature FDC precursors, including marginal sinus pre-FDCs (MadCAM-1+) and perivascular pre-FDCs (PDGFRβ+) are enriched. As normal FDC networks re-establish, extensive GCs become detectable throughout the spleen. Therefore, the reorganization of FDC networks and the loss of GC responses are key, parallel features of systemic STm infections.


INTRODUCTION
A hallmark of the mammalian immune response is the induction of adaptive immune responses to pathogens. An important contribution to the protection provided by the adaptive immune response is the generation of antibodies, which is a typical consequence of infection and a key aim of vaccination. 1 Antibodies can be generated through two predominant, interconnected pathways. In primary responses, extrafollicular (EF) responses, which develop in the red pulp of the spleen or the medulla in lymph nodes, provide the first wave of IgM and IgG, and these antibodies are typically of modest affinity because there is limited affinity maturation of B cells that enter this pathway. 2 Moreover, antibodies from plasma cells generated through the EF pathway is typically supplanted after a week 3 or so by antibodies secreted by plasma cells derived from the germinal center (GC) response. 4 GCs form in the B cell follicles of secondary lymphoid organs (SLOs) such as the spleen or lymph nodes. 5,6 A key difference between the EF and GC responses is that antibodies generated from the GC tend to be of higher affinity and most memory B cells and the longest lived plasma cells derive from this response. 7 The generation of these productive outputs from the GC requires the interplay of multiple cell types at different sites within SLO, with all processes dependent on the microarchitecture of SLO, including interactions between T and B cells in the T zones and the follicles. 8,9 Although the organization of cells within T cell zones (TCZ) is dependent on CCL21/CCL19 secreted by fibroblastic reticular cells (FRC), 10,11 the generation of B cell follicles relies on CXCL13 secreted by follicular dendritic cells (FDCs) and marginal reticular cells (MRCs). [12][13][14] Moreover, FDCs play a critical function in the GC by holding antigens in their native conformation on their surface as immune complexes driving the selection of B cell clones with the highest affinity. [15][16][17][18] Therefore, the organization of lymphoid tissues is essential for the efficient generation of productive immune responses and FDCs are crucial for normal follicle architecture and the GC reaction. 8,19 Antibody responses induced during natural infection can help moderate the spread of the pathogen and secondary superinfections. 20 Nevertheless, many pathogens can modulate the capacity to mount antibody responses during infection, potentially affecting the capacity of the host to deal with current or later infectious threats. 21 Bacterial and parasitic infections such as those caused by Ehrlichia muris, Salmonella

Salmonella Typhimurium infection perturbs the organization of the white pulp microarchitecture
Systemic infection of susceptible C57BL/6 mice with attenuated STm SL3261 results in a self-resolving infection characterized by rapid colonization of the spleen and bacterial numbers peaking from the first week before gradually decreasing after the third week ( Figure 1A). In parallel, STm infection induces a marked splenomegaly which peaked at 21 days, when spleens were around 10 times the mass of non-infected control mice ( Figure 1B). As reported previously, 24 GCs are not a feature of early STm infections and are only consistently detected at day 42 after infection when the infection and associated splenomegaly has largely resolved ( Figures 1C and 1D). We hypothesized that GC responses are delayed during STm infection because infection induces a perturbed white pulp (WP) topography thus inhibiting the generation of productive responses. Assessment of the WP containing the combined T and B cell compartments in the spleen showed that the proportion of the spleen that is WP decreased between 7 and 30 days after infection before recovering to similar proportions as non-infected mice by day 42 after infection (Figures 2A and  2B). Moreover, at day 21, when the effects of infection on splenic architecture are most pronounced, the absolute area of individual WPs was significantly smaller than in control mice and contained poorly defined T and B cell areas, with a relative paucity of T and B cells within their respective compartments (Figures 2A-2E). Features that characterized the WP after infection included the altered distribution of dendritic cells (DCs; DEC205+ CD11c+ cells). These cells are mostly restricted to the TCZ in non-infected mice, but were found throughout the WP area after infection, including in and around B cell follicles (Figures 2F and S1A).
The marginal zone (MZ) borders the WP and discrete populations of macrophages and B cells reside in this site, including CD169+ metallophilic macrophages (MMMFs) and SIGN-R1+ MZ macrophages (MZMFs) as well as MZ B cells. Moreover, the cell populations in the MZ can regulate GC B cell responses. 38,39 By day 7 after infection, immunofluorescence (IF) microscopy showed a reduced detection of CD169+ MMMFs, which was more apparent from day 21 and afterward, and a near absence of signal for SIGN-R1+ MZMFs from day 7 (Figures 3A-3C and S1B-S1D). In addition, STm infection induced a reduction in B cells in the MZ as assessed by both imaging and flow cytometry ( Figures 3A and 3D-3F). Although MZ B cells recovered by day 42, this was not the case for CD169+ MMMFs and SIGN-R1+ MZMFs (Figures S1C and S1D). Therefore, STm infection results in a significant remodeling of the WP and MZ splenic microarchitecture.

STm-induced loss of organized FDC networks correlates with the lack of GCs
The perturbed microarchitecture observed after infection suggested that stromal cells which orchestrate the migration of cells in SLO might be affected by STm. FDCs play a critical role in the antigen-mediated  Figure S3D). Next, the lack of organized FDC networks and the development of GCs were evaluated in parallel. Spleen sections from non-infected mice and STm-infected mice stained with peanut agglutinin (PNA), anti-IgD, and anti-Mfge-8. GCs (PNA+ IgD-) were only detected when classical FDC networks were recovered; for instance, at 28 days few GCs were detected only in follicles displaying a more compact FDC network and by 42 days, nearly all follicles contained a GC and an organized FDC network ( Figure 5A). The maintenance of the FDC network in the adult spleen is dependent on lymphotoxin (LT)/LTb receptor (LTbR) and tumor necrosis factor (TNF)/TNF receptor (TNFR) signaling. [44][45][46] Gene expression of Ltb, Lta, Ltbr, Tnf, and Tnfr1 was investigated by RT-PCR in microdissected WP isolated from the spleens of non-infected mice and mice infected for 21 days. Ltb expression was significantly downregulated ( Figure 5B), whereas Ltbr and Tnfr gene expression was significantly upregulated in WP of infected mice compared to non-infected mice ( Figures 5C and 5D). Tnf and Lta expression was not different between the groups ( Figures 5E and 5F). Overall, the transient absence of GC is associated with the lack of FDC networks and perturbations in gene expression in the LT and TNF pathways.

Perivascular and marginal sinus FDC precursor-like expand during infection
We assessed the effects of infection on other stromal cells that may contribute to follicle organization. One such cell-type are MRCs, which express the adhesion molecule MadCAM-1, secrete CXCL13 and have been described as marginal sinus pre-FDCs as they also share the expression of Mfge-8 in the MZ in steadystate. 14,47 Spleen sections from non-infected mice or mice infected with STm were stained for MadCAM-1, Mfge-8, CD21/35 and laminin and assessed by IF ( Figure 6A; presented as a merge of all staining on the top row or individual markers in grayscale on the bottom rows). STm infection induced a noticeable expansion of MadCAM-1+ cells from 7 days up to 28 days after the infection, and these cells were detected not only in the MZ but also in the follicles, and some were positive for Mfge-8 ( Figures 6A and S4A and S4B). Analysis by flow cytometry confirmed that there was an expansion in the frequency and number of MRCs at day 21 post-infection, which was not observed at day 42 ( Figures 6B and S3A). In addition, FDC-M1+/Mfge-8+ cells were observed around CD31 + blood vessels ( Figure 6C). These cells also expressed plateletderived growth factor receptor beta (PDGFRb; Figure 6D) which has been described as a marker of perivascular pre-FDCs in the spleen but not in the lymph node. [47][48][49] Therefore, MadCAM-1+ cells and perivascular PDGFRb+ cells expand during STm infection.

MadCAM-1-expressing cells become major producers of CXCL13 during infection
Given the reorganization of the WP and the lack of FDC networks induced after STm infection, we hypothesized that STm infection perturbs the expression of chemokine profiles within the WP. Key amongst these chemokines are CCL21 and CXCL13, which are required for normal T zone and follicle segregation and 24 h after STm infection changes in expression of these chemokines have been reported. 50 The distribution of these chemokines was examined by IF microscopy and gene expression from microdissected WP in spleens from non-infected mice and after 21 days of STm-infection. CCL21 gene expression was reduced in the WP and this was also reflected at the protein level in T zones of infected mice compared to naive controls ( Figures 7A and 7B and S5A). In contrast to early time points, 50 and despite perturbed FDC networks being

DISCUSSION
Here, we show how systemic Salmonella infection induces the disorganization of the splenic WP and loss of GC formation, and the later detection of GCs correlates with the induction and resolution of these changes. The capacity of STm infection to impact GC induction and maintenance is not restricted to STm-specific responses but also impacts GC induced to a diverse spectrum of other pathogens and antigens, including model antigens, bacterial flagellin, the microbiota, influenza virus and the helminth Nippostrongylus brasiliensis. [32][33][34]37 The diversity of these antigens and the observation that STm infection impacts pre-existing and ongoing GC responses indicates the antigen independence of these effects. Because B cells from STm-infected mice maintain the capacity to develop into GC B cells, 37 it indicates that a significant contributory reason for the absence of GC in the first weeks of infection is that STm disrupts the niches in which GCs develop.
Previous studies have shown how non-B cell-intrinsic mechanisms including the recruitment of discrete Sca-1+ monocyte populations, TLR4 expression and IL-12-mediated suppression of Tfh cells can impair GC formation after STm infection. 34,37 The impact of IL-12 induced to Salmonella is likely to act early, possibly in the first 24 h after infection because this cytokine is produced by conventional and monocyte-derived dendritic cells from 2 h post-infection to promote Th1 differentiation. 33 In contrast to these early events, the identification of perturbed FDC and MZ organization occurs later and indicates that the structures needed to support GC development and function are not maintained during the peak of the infection. Moreover, MZ B cells, MMMFs and MZMFs, which support GC responses, 38,39 were detected less readily after day 7 after infection when GCs were absent. Unexpectedly, only MZ B cells were found at levels comparable to non-infected mice when GCs developed. The two macrophage populations had not returned to normal by then, therefore, more work is needed to understand the potential role of MZ populations in GC responses after STm infection. Despite the multiple and quite marked effects of Salmonella on the host adaptive immune response, it is important to contextualize the lack of GCs as being a selective effect and not representative of a general impairment in the capacity of the host to induce antibody responses to the pathogen. Many B cell responses remain active in the host during the period when GC are not detectable. For instance, extensive EF B cell responses are detectable from the first days after infection. 24 IgM responses are induced in a T-independent manner whereas B cell switching to IgG dependent upon BCL6 + PD1 lo CXCR5 lo T cells and CD40L. 24,51,52 Moreover, antibodies derived from the EF response can moderate bacteremia and subsequent re-infection. 24 Therefore, the tissue reorganization observed in primary infection is not a barrier to productive EF responses, and so the impairment of the GC response to Salmonella is unlikely to have evolved simply as a strategy to impair antibody responses per se. Therefore, whether there is a selective advantage to the host or to the pathogen remains unclear. In contrast to the lack of GCs observed after live infection, GC responses are detected within the first four days after immunization with outer membrane vesicles (OMVs) from STm. 53 These OMVs contain many of the immunodominant cell surface antigens from STm, including LPS and porins. 54 This rapid induction of Tfh cells and GCs 53 suggests that exposure of the host to multiple STm antigens in their natural conformation within the bacterial iScience Article membrane, concomitant with significant TLR4 ligating activity, is not sufficient to explain why GCs are not induced to this pathogen.
A hallmark of the GC is the organization of multiple cell types within follicles. This includes FDCs which are essential not only for providing chemokines to recruit B cells but also for antigen-driven selection. 13,19,55 The STm-induced changes in FDCs and the downregulation of key phenotypic markers, alongside the timing of these changes, is fully consistent with changes in FDCs being a key reason why GC do not   59 and this factor has also been shown to be important for the localization of pneumococcal bacteria to FDCs. 60 The role of complement in the context of GCs and Salmonella infections is not well explored, and in the presence of antibodies (which are induced rapidly in this infection 24 ), C3 from mice is not deposited as efficiently as human C3 on the surface of Salmonella in vitro. 61 This may also partially explain why STm were only rarely associated with FDCs, or there may be other reasons such as the rate of bacteria capture by phagocytic cells. Therefore, how complement may contribute, or otherwise, to GC formation during STm infection requires further investigation.
FDCs are plastic stromal cells that originate from perivascular cells and require B cell-derived TNFR and LTbR signaling for their maintenance. [45][46][47][62][63][64][65] The de-clustering of FDC induced by STm infection is unlikely to reflect the death of FDCs as although the density of FDCs in the spleen was significantly reduced after infection, the total numbers were not and these cells are known to be long-lived and resistant to stresses such as radiation exposure. 13,16,40 Another potential outcome is that FDCs lose maturity and partially de-differentiate as FDC-like cells were readily detectable in the follicles and expansion of MadCAM-1+ MRCs in the MZ and throughout the follicle was observed, possibly because of exposure to TLR4 ligands. 66 Likewise, the capacity to form normal FDC networks could also occur because of the reduced density of B cells in the follicles or/and an intrinsic decrease of LTab expression, which may limit LTbR signaling. The other LTbR ligand, LIGHT (TNFSF14) may contribute to this process. 67 Here, we have shown that the expression of genes within the TNFR and LTbR signaling pathways in the WP were modulated by STm infection. TNF is produced by multiple innate immune cell types and T cells after infection and plays pleiotropic roles in controlling infection in tissues such as the spleen and liver. [68][69][70] Less is known about the contribution of LT and TNF to the organization of WP after STm infection. We did attempt to modulate FDCs by targeting the LTbR and modulation of TNF using agonistic and neutralizing antibodies, respectively. Targeting LTbR did not provide consistent results, with some mice showing accelerated induction of FDCs and GCs, but not all. The reasons for this are unclear but we speculate that they may be consequences of trying to maintain agonistic LTbR signaling for a sufficiently long time to have lasting effects. In contrast, neutralizing TNF during E. muris infection enhances GC responses 25 but we observed an enhanced loss of FDCs and no GC response during STm-infection. The loss of FDCs, GCs and impaired humoral immunity after interference with the TNF signaling using either TNF-neutralizing antibodies or gene-targeted mice has been described extensively in steady-state conditions 46,64,65,71 and under inflammatory environments such as sepsis. 72 These results suggest that during active infection, the targeting of specific cell types may be more difficult than targeting soluble factors such as TNF and highlights the potential difficulty in using single agent interventions to target different bacterial infections as each infection may have a distinct immune signature.

Limitations of the study
A limitation of this study is the lack of formal demonstration that changes in FDC organization are responsible for the lack of early GC induction and other mechanistic insights. Probably, there are multiple active

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

DECLARATION OF INTERESTS
The authors declare no competing interests.

INCLUSION AND DIVERSITY
One or more of the authors of this paper self-identifies as an underrepresented ethnic minority in their field of research or within their geographical location. One or more of the authors of this paper received support from a program designed to increase minority representation in their field of research. While citing references scientifically relevant for this work, we also actively worked to promote gender balance in our reference list. We support inclusive, diverse, and equitable conduct of research.