TNF-α Contributes to Lymphoid Tissue Disorganization and Germinal Center B Cell Suppression during Intracellular Bacterial Infection Free

Bacterial, viral, and parasitic infections can all lead to major perturbations of immune homeostasis (1). In many cases, these perturbations are accompanied by excessive inflammation, which can result in disorganization of secondary lymphoid tissues. How B cell differentiation is affected by ongoing inflammatory processes is not well characterized, but knowledge of lymphocyte differentiation under such conditions is important for understanding how effective humoral immune responses are generated and maintained.

Host responses to infections and associated alterations in secondary lymphoid tissues are pathogen dependent. For example, Salmonella infection leads to the suppression and delay of germinal center (GC) development via SIP2 T3SS effectors, a process that inhibits bacteria-specific B cell responses (2, 3). During malarial infection, the plasmodium parasite suppresses GC formation by inhibiting the differentiation of T follicular helper (T_FH) cells (4). In contrast, Trypanosoma brucei infection induces splenic remodeling and apoptosis of marginal zone B cells, and this remodeling results in a reduction of Ab-mediated immunity and poor control of infection (5). In yet other studies, repeated administration of CpG oligonucleotides was sufficient to cause lymphoid tissue disorganization (6). We have observed similar effects on immunity following infection with the intracellular bacterium Ehrlichia muris. We have reported that E. muris infection suppresses GC B cell responses and showed that these changes can inhibit humoral immune responses to foreign Ags (7). Moreover, E. muris infection leads to profound IFN-γ–dependent hematological changes that mediate extramedullary hematopoiesis, characterized by increases in megakaryocyte–erythrocyte progenitors, common myeloid progenitors, and granulocyte monocyte progenitors in the spleen (8, 9).

Concurrently, E. muris infection elicits a robust splenic CD11c⁺ CD4 T cell–independent IgM plasmablast response and generates a large population of long-lived CD4 T cell–dependent T-bet⁺ IgM⁺ memory cells (7, 10, 11). These B cells, which are not found in uninfected young C57BL/6 mice, produce pathogen-specific Abs (12, 13, and G.M. Winslow, unpublished data). Unswitched cells dominate this early response to infection, although low frequencies of switched T-bet⁺ memory B cells are also produced (14). Similar, if not identical, T-bet⁺ B cells have been described in a range of chronic conditions, including HIV, hepatitis C virus (15–17), and malaria infections (18–20), and also in autoimmunity (21–23) and age-related immunity (24, 25). Why and how E. muris infection generates such relatively large populations of T-bet⁺ B cells is unclear, but our previous data suggest that this may be a consequence of the disruptions of normal spleen homeostasis and canonical T cell–B cell interactions.

It is not known whether an inflammatory extrafollicular microenvironment is necessary for the generation of T-bet⁺ B cells (26–28), although inflammation is required to induce T-bet expression in T cells (29), and T-bet⁺ T cells can regulate the inflammatory milieu (30). These observations led us to address a role for inflammatory cytokines in the disruption of immune homeostasis during E. muris infection. We focused on TNF-α because this cytokine has been previously implicated in GC development and humoral immunity (31–34). In contrast to studies that have described a role for TNF-α in GC development under homeostatic conditions, we show that TNF-α participates in infection-induced lymphoid disorganization, the disruption of chemokine networks, and the loss of canonical GCs. These findings underscore how bacterial infections can modulate B cell immunity and T-bet⁺ B cell ontogeny by modulating host inflammatory responses, in part via TNF-α.

C57BL/6 and TNF-α–deficient (B6; 129S-Tnf^tm1Gkl/J) mice were obtained from The Jackson Laboratory (Bar Harbor, ME). Mice were used between 6 and 12 wk of age and were age-matched across experiments. All animals were bred and/or maintained at the State University of New York Upstate Medical University (Syracuse, NY) in accordance with institutional guidelines for animal welfare.

Mice were infected via i.p. injection with E. muris at a dose of 5 × 10⁴ copies, as described previously (35). TNF-α blockade was achieved by administration of the anti–TNF-α mAb XT3.11 [10 mg/kg of body weight, every other day (36, 37)] beginning 8 h prior to infection and continuing up to but not including day 16 postinfection. The isotype-matched Rat IgG1 Ab (clone HRPN) was used as a control; the Abs were purchased from Bio X Cell (West Lebanon, NH). CD40L depletion was administered as described elsewhere (11), beginning on day 2 postinfection, every other day, up to day 16 postinfection. CXCL13 was depleted in a similar fashion, using mAb 5378, which was provided by Vaccinex (Rochester, NY). Anti-CXCL13 was administered every other day at 30 mg/kg (38), beginning on day 4 postinfection and up to but not including day 16 postinfection; an isotype-matched mouse IgG2a Ab (clone 2510) was used as a control.

Spleen cells were mechanically disaggregated using a 70-μm cell strainer (BD Biosciences); erythrocytes were removed by hypotonic lysis, using ACK lysing buffer (Quality Biological). Nonspecific binding was blocked by treatment with anti-CD16/32 (2.4G2) prior to incubation with Abs directed against the following Ags: CD73 (clone TY/11.8; BioLegend), IgM (clone R6-60.2; BD Biosciences), GL7 (clone GL-7; eBioscience), CD19 (clone 6D5; BioLegend), CD11c (clone N418; eBioscience), CXCR5 (clone L138D7; BioLegend), B220 (clone RA3-6B2; BD Biosciences), CD138 (clone 281-2; BD Biosciences) CD38 (clone 90/CD38; BD Biosciences), CD95 (clone Jo2; BD Biosciences), PD-1 (clone RMP1-30; BioLegend), CD3 (clone 17A2; BD Biosciences), CD4 (clone GK1.5; BD Biosciences), CD120a (clone 55R-286; BioLegend), and CD120b (clone TR75-89; BioLegend). The cells were stained at 4°C for 20 min, washed, and analyzed for marker expression. Data were acquired on a BD LSRFortessa flow cytometer using Diva software (BD Biosciences) and were analyzed using FlowJo software (Tree Star). CXCL13 was detected in sera from wild-type (WT) and TNF-α–deficient mice using the LEGENDplex Proinflammatory Chemokine Mix and Match Subpanel (740097; BioLegend) in accordance with the manufacturer’s instructions. Data were acquired on a BD LSRFortessa flow cytometer with Diva software (BD Biosciences) and were analyzed using the LEGENDplex Data Analysis Software (BioLegend).

TNF-α concentration in sera was determined using a TNF-α ELISA (88-7324-22; eBioscience) in accordance with the manufacturer’s instructions. Infection-specific IgM and IgG titers were measured by ELISA, using the E. muris Ag outer membrane protein-19 (OMP-19), as previously described (39). The number of Ag-specific IgG and IgM B cells in the spleen was determined using a standard ELISPOT (40). Spots were imaged using a CTL Series 6 Ultra-V Analyzer (Shaker Heights, OH) and enumerated using ImmunoSpot software (Cellular Technology). Spots were normalized to the total population of Ab-secreting cells (ASCs) as determined by flow cytometry.

Spleens were mounted in OCT compound (Tissue-Tek), snap frozen, and sectioned at a thickness of 10–15 μm. Following a 5-min fixation in 4% paraformaldehyde, samples were transferred to a humidified chamber and blocked and permeabilized for 30 min at room temperature using a 3% nonfat milk solution in 1× PBS with 0.1% IGEPAL CA-630. The Ab panel was separately diluted in blocking buffer and added to the specimens for overnight incubation at 4°C. Prior to mounting, the specimens were stained with a 1 μg/ml DAPI solution (D9542; Sigma-Aldrich) in PBS for 5 min. ProLong Gold antifade mountant was used to affix the sections to glass coverslips. Confocal images were obtained at 20× magnification on a 5 × 5 tile scan, using an LSM780 microscope with ZEN imaging software (Zeiss). Images were analyzed using FIJI (41), as necessary, for annotation and cropping. The Abs used were as follows: Alexa Fluor (AF) 488–conjugated GL7 (clone GL7), AF594-conjugated CD4 (clone GK1.5), AF647-conjugated CD35 (clone 7E9), AF647-conjugated F4/80 (clone BM80), all from BioLegend, and AF700-conjugated B220 (clone RA3-6B2; Bio-Rad Laboratories).

Whole spleens were harvested from WT and TNF-α–deficient mice (n = 3 per group per timepoint) at 0 and 16 d postinfection and were fixed in 4% paraformaldehyde for 48 h at 4°C. Samples were briefly stored in 70% ethanol. Histology analysis was performed by HistoWiz (Brooklyn, NY), using in-house standard operating procedures and a fully automated workflow. The samples were processed and embedded in paraffin and sectioned at 5 μm to generate two slides per collected spleen (three to four sections per slide). The sections were stained with hematoxylin, dehydrated, and film-coverslipped using a Tissue-Tek Prisma and Coverslipper apparatus (Sakura). Whole-slide scanning was performed using an Aperio AT2 imaging system (Leica Biosystems). The images were quantified using HALO image analysis software (Indica Labs) with the CytoNuclear module. Samples were diagnosed and evaluated by a HistoWiz pathologist. Data are available upon request. Histology slides are available online at: https://app.histowiz.com/shared_orders/70a17478-3036-493a-ac0e-b7a54223a443/slides/.

Our previous work demonstrated that E. muris infection generates both extrafollicular T-bet⁺ IgM plasmablasts and T-bet⁺ IgM memory B cells (7, 10, 11). These B cells develop in a unique lymphoid environment, which, in the spleen, lacks conventional B cell follicles and GCs (7). Indeed, inflammation associated with E. muris infection likely alters secondary lymphoid homeostasis and contributes to the generation of these B cells, by driving an extrafollicular B cell response. Because TNF-α is required for the proper development of B cell follicles and the generation of GCs in immunized mice, we addressed whether this cytokine was, in part, responsible for modulating the lymphoid environment during E. muris infection.

TNF-α was elicited during E. muris infection, as it was detected in peripheral blood, at its highest levels on day 16 postinfection; by day 30, expression had returned to preinfection levels (Supplemental Fig. 1A). To address a role for TNF-α during ehrlichial infection, we first monitored B cell responses in infected WT and TNF-α–deficient mice. We identified a small population of CD19⁺ GL7⁺ CD38^lo CD95⁺ GC B cells in the spleens of TNF-α–deficient mice as early as day 8 postinfection (Fig. 1A, histogram). These GC B cells were also found in infected WT mice on day 16 postinfection, although the cells were much more abundant in the absence of TNF-α. We also detected CD11c⁺ IgM⁺ CD73⁺ B cells on day 16 postinfection, and this population was similarly expanded in the absence of TNF-α (Fig. 1B). These latter cells likely represent early T-bet+ IgM⁺ memory cells (10), as they express surface markers characteristic of that population (e.g., CD73, CD80) as well as T-bet (Supplemental Fig. 1B). Approximately 50% of the early CD11c⁺ T-bet⁺ memory cells expressed GL7, characteristic of GC B cells (Fig. 1B, right plots). By day 30 postinfection the frequency of CD11c⁺ T-bet⁺ memory cells remained modestly higher in the absence of TNF-α. Similar findings were obtained using Ab-mediated ablation of TNF-α, indicating that the observations were not a consequence of genetic elimination of the cytokine (Fig. 1C).

The detection of GC-phenotype B cells in the absence of TNF-α in infected mice suggested that B cells were able to develop within GCs or within an environment that better promoted GC or GC-like B cells. Although changes in follicular architecture have been described in TNF-α–deficient mice (31), we observed similar ratios of red and white pulp in both uninfected WT and TNF-α–deficient spleens, and these were composed of well-formed follicles (Supplemental Fig. 2A). We also observed that GCs, although rare in WT mice, were not detected in TNF-α–deficient spleens. On day 16 post–ehrlichial infection, WT mice demonstrated notable white pulp disruption, with ∼50% of the white pulp either atrophic or disrupted by red pulp elements (Fig. 2A). In contrast, infected TNF-α–deficient spleens were less disrupted, with only ∼25% of the white pulp elements atrophic or disrupted. Examination of the intact-appearing white pulp demonstrated that, whereas the WT group was composed of small- to medium-sized lymphoid cells similar to those seen in uninfected mice, the white pulp in TNF-α–deficient animals contained medium-sized and large lymphoid cells, reminiscent of centrocytic and centroblastic cells, respectively.

Immunofluorescence analyses revealed that on day 16 postinfection TNF-α–deficient mice exhibited discernable follicular structures (Fig. 2B). Much better distinction between B and T cell zones was observed, and the B cells formed clusters typical of GCs. In contrast, GL7⁺ B cells were not readily detected in WT spleens, and B and T cells tended to disperse randomly throughout the tissue, forming noticeably small follicular clusters at a reduced frequency. These data revealed that TNF-α, in part, contributed to the loss of GCs and overall disruption of typical splenic architecture in infected mice.

We also addressed whether splenic disorganization was a consequence of loss or improper distribution of CD35⁺ follicular dendritic cells (FDCs). Although previous studies of TNF-α–deficient mice reported that FDCs were absent, we detected FDCs in both uninfected and infected TNF-α–deficient mice (Supplemental Fig. 2B). However, FDCs were less abundant in TNF-α–deficient mice, relative to uninfected WT mice. FDCs in infected WT mice clustered in the vicinity of overlapping zones of B and T cells; these zones were more discernable in TNF-α–deficient spleens. On day 16 postinfection, we detected similar frequencies of FDCs in both WT and TNF-α–deficient mice by flow cytometry (Supplemental Fig. 2C). Thus, FDCs did not appear to contribute to the microarchitecture differences we observed in the absence of TNF-α.

In contrast to the FDCs, we observed much higher frequencies of F4/80⁺ macrophages in TNF-α–deficient mice on day 16 postinfection compared with WT mice, and these macrophages clustered in B cell follicles (Fig. 2B). These data also indicated that splenic lymphoid tissue was more organized in the absence of TNF-α during infection.

The more structured lymphoid environment in TNF-α–deficient mice was also associated with a reduction in splenomegaly, suggesting that the loss of TNF-α affected other processes induced by infection such as extramedullary hematopoiesis (Fig. 2C) (42). No apparent changes were detected in bacterial colonization (data not shown), however, which suggests that the effects were likely a consequence of altered inflammatory responses.

We next addressed whether TNF-α altered splenic organization by disrupting cell migration and/or positioning via regulation of chemokine expression. We investigated a role for CXCL13, as this factor is well-known to direct lymphocyte migration and GC development. CXCL13 expression was much higher in serum and spleen cell lysates from WT mice on day 16 postinfection relative to uninfected mice (Fig. 3A). This increase was much reduced in TNF-α–deficient mice, although higher levels of CXCL13 were detected relative to uninfected, gene-targeted control mice (Supplemental Fig. 1C). We also observed a 10-fold increase in CXCR5 expression on CD11c⁺ B cells from infected TNF-α–deficient, but not WT mice (Fig. 3B, Supplemental Fig. 1D). High CXCR5 expression may be a consequence of the much lower CXCL13 expression in TNF-α–deficient mice because of a reduction in ligand-induced endocytosis in the B cells (43, 44).

To address a role for CXCL13 in TNF-α–mediated tissue disorganization, we depleted the chemokine in WT and TNF-α–deficient mice by treating infected mice with an anti-CXCL13 mAb every other day until day 16 postinfection. Loss of CXCL13 markedly reduced the frequency of GC B cells in both WT and TNF-α–deficient mice, although the magnitude of the effect was greater in the absence of TNF-α (Fig. 3C). This latter outcome was not unexpected, as CXCL13 is known to be required for proper GC development (45, 46). We detected only a modest reduction in CD11c⁺ T-bet⁺ memory B cells in TNF-α–deficient mice, although high CXCR5 expression was retained on the B cells. CD11c⁺ B220^lo T cell–independent extrafollicular plasmablasts were increased in frequency in the absence of TNF-α, although they were unaffected by CXCL13 ablation (Fig. 3D). These data suggest that TNF-α mediates the suppression of GC B cells, in part, via the regulation of CXCL13 but that CXCL13 plays. at most, a modest role in the generation of T-bet⁺ IgM memory cells and plasmablasts. These latter data indicate that the normal chemokine cues required for B cell differentiation in GCs are not necessary for the generation of T cell–independent plasmablasts or early T-bet⁺ memory cells during ehrlichial infection.

Differentiation of GC B cells is a T cell–-dependent process, so we next addressed whether TNF-α inhibits GC B cells by suppressing T cell helper functions. In this regard, PD-1⁺ CXCR5⁺ T_FH cells were more abundant in the spleens of infected TNF-α–deficient mice on day 16 postinfection, relative to the same population in WT mice (Fig. 4A). Among the total T_FH population, we detected a higher frequency of GC-phenotype PD-1^Hi CXCR5^Hi T_FH cells under the same conditions. The lower expression of PD-1 and CXCR5 in infected WT mice, relative to uninfected controls, is characteristic of extrafollicular T_FH cells (47, 48). These data are consistent with a partial restoration of GC function in the absence of TNF-α. T cell–mediated helper functions were required to generate GC B cells in TNF-α–deficient mice, as CD40:CD40L blockade reduced the frequency of GC B cells to baseline levels (Fig. 4B). These findings reveal that TNF-α suppresses GC B cell development, directly or indirectly by inhibiting T cell helper functions, even in the presence of a large T_FH cell response.

We also examined whether TNF-α deficiency affected the development of CD11c⁺ plasmablasts that are generated in the spleen following E. muris infection. CD11c⁺ B220^low CD138⁺ plasmablasts were detected at a high frequency on day 16 postinfection, as previously reported (7), and these were increased 2.5-fold in frequency in the absence of TNF-α (Fig. 5A). The Ab-secreting plasmablasts produced ehrlichial OMP-19–specific IgM; however, the number of ASCs was higher in the absence of TNF-α (Fig. 5B). In contrast, many fewer Ag-specific IgG-secreting B cells were detected, and these were lower in number in the absence of TNF-α, presumably as an offset to the expansion of IgM-secreting B cells. We also monitored the number of ASCs 5 d following challenge of mice on day 50 postinfection with purified OMP-19. The number of IgM-producing but not IgG-producing ASCs was greater in TNF-α–deficient mice (Fig. 5C). These data indicate that TNF-α ablation, perhaps by generating higher frequencies of long-term plasma cells or memory cells, can have a long-term effect on the early secondary responses to Ag challenge.

We also addressed whether the increase in IgM-producing cells in the TNF-α–deficient mice was CXCL13 dependent by neutralizing CXCL13, as in the studies shown in Fig. 3. In the absence of CXCL13, IgM-producing cells were unaffected on day 16 postinfection; IgG-producing cells were much reduced, as in Ab-treated WT mice (Fig. 5D). These data indicated that CXCL13 was responsible for promoting the switched GC-like memory cells but not the T cell–independent–unswitched IgM plasmablasts. CXCL13 ablation did not act indirectly via T_FH cells, however, as CXCL13 blockade in the TNF-α–deficient mice did not affect the frequencies of T_FH cells generated during infection (Fig. 5E).

Inflammatory responses have been extensively described relative to their capacity to direct the quality and magnitude of the host response to infection (49–53). This occurs, in part, via the regulation of cytokine and chemokine production, which mediate lymphocyte recruitment and differentiation. These inflammatory responses are pathogen dependent and differ from those elicited by inert Ags, underscoring the importance of studying immunity in natural infections.

Among the many cytokines produced during infections, TNF-α plays a potent immunoregulatory role in the host immune response. TNF-α acts to induce proinflammatory cytokine signaling, which leads to cytotoxicity, cell proliferation, and NF-κB activation (54). Our findings reveal an additional role for TNF-α as an important mediator of lymphoid tissue disorganization during bacterial infection. TNF-α was, in part, responsible for mediating tissue disorganization because blocking TNF-α was sufficient to partially restore GC function and led to the generation of higher frequencies of both T-bet⁺ plasmablasts and IgM memory B cells. These B cells were sustained for at least as long as 30 d postinfection and contributed to better recall kinetics upon secondary challenge.

Our finding that TNF-α is responsible in part for inhibiting GC or GC-like B cells contrasts with early studies using the TNF-α–deficient mice, in which the cytokine was shown to be required for the proper development and maintenance of B cell follicles and GCs (31–34). These disparate findings indicate that TNF-α function is context specific, likely a consequence of yet other initiating factors elicited not only by ehrlichiae, but other bacterial, viral, and parasitic infections that cause lymphoid disorganization (1). These pathogens and factors may function in common by generating a cytokine and chemokine “storm” of mediators that include TNF-α and, together, may cause gross disruptions in lymphoid tissues. How the ehrlichiae trigger innate immunity in this fashion is not known, as these pathogens do not encode classical TLR ligands (55, 56). Nevertheless, the observation that many different infections induce lymphoid tissue disorganization suggests that this process may be of benefit to either the pathogen or host. It would be of obvious benefit to pathogens to limit GC B cell development and the consequent generation of a high-affinity, class-switched Ig response. Under these same conditions, however, E. muris generates highly effective unswitched B cell responses composed of both early IgM plasmablasts and long-term T-bet⁺ IgM memory cells (7, 10). In the absence of evidence to indicate that lymphoid disorganization is caused by the activity of pathogen-derived factors designed to subvert immunity, the most likely explanation is that the lymphoid disorganization we and others have observed results from the loss of positional cues, in part, due to chemokine dysregulation.

The overall improved tissue organization we observed suggests that B cells receive more directed positioning cues in the absence of TNF-α. In this regard, we detected much higher CXCL13 levels in infected WT mice relative to uninfected controls. The interaction of CXCL13 with CXCR5 results in migration that is spontaneously random, driven by the search for Ag (46, 57). Established gradients, however, introduce directionality to movement and have been shown to be responsible for proper B cell migration and GC development (45). Thus, we propose that TNF-α, directly or indirectly, induces excess CXCL13 production in infected mice that alters chemokine gradients and leads to the disruption of normal cell migration and overall follicular and splenic architecture. These dynamics are likely to lead to poor costimulatory help, limited BCR engagement, and subsequent inhibition of the GC reaction. Collectively, these changes may be responsible for driving extrafollicular B cell differentiation during E. muris infection, which may explain why both T-bet⁺ plasmablasts and IgM memory cells are produced in such relative abundance during this infection. Depletion of CXCL13 had a minor effect on the generation of early T-bet⁺ memory cells and plasmablasts in TNF-α–deficient mice. Thus, although TNF-α may inhibit the generation of GC B cells by inducing excess CXCL13, the cytokine regulates T-bet⁺ B cell expansion via a CXCL13-independent pathway.

Although we observed an increase in GC-phenotype T-bet⁺ B cells, this was not accompanied by major changes in either repertoire diversity or mutation frequency (data not shown). We interpret these findings to suggest that the GC-like structures we observed were not fully functional, likely because other components that drive affinity maturation and selection are not intact, even in the absence of TNF-α. Our data also suggest that the IgM plasmablasts and T-bet memory cells do not necessarily represent alternative fates, as both populations were found to increase in frequency in the absence of TNF-α. Thus, the development of the large unswitched CD11c⁺ T-bet⁺ IgM plasmablast population during ehrlichial infection is not solely due to loss of tissue organization and cytokine dysregulation.

Investigation of the roles of TNF-α in antimicrobial immunity is an ongoing area of research. TNF-α receptors are found on many different immune cells, including B cells. Ablation of TNFR1 or TNFR2 only on B cells did not cause major changes in the frequencies of plasmablasts or memory B cells following infection with E. muris (data not shown), indicating that TNF-α acts on other cells, perhaps in addition to B cells. However, TNF-α depletion did not affect the frequency of CD35⁺ FDCs. The lower production of CXCL13 in the absence of TNF-α was also associated with high CXCR5 expression on early T-bet⁺ memory cells. A possible explanation for this observation is that CXCR5 expression is regulated by ligand density and receptor activation (43, 44) such that these dynamics are altered in the absence of TNF-α.

Other possible explanations for improved lymphoid tissue organization following TNF-α ablation are as follows: 1) the apparent large increase in F4/80⁺ macrophages that was observed in B cell follicles and 2) the expansion of T_FH cells. Splenic macrophages have been shown to be necessary for the generation of T cell–dependent B cell responses and GCs (58), so the increased frequency and follicular organization we observed may have contributed to the increased frequency of GL7⁺ T-bet⁺ memory cells that we observed. Moreover, we have demonstrated that CD4 T cells are required for T-bet⁺ B cell development (10), so TNF-α ablation may increase activity by these helper T cells, as has been shown to occur in malarial infection (4). Ehrlichial infection was also associated with increased IFN-γ production, and this was further increased in the absence of TNF-α, but IFN-γ ablation had no apparent effect on either GC B cells or T-bet⁺ B cells (data not shown). Thus, both macrophages and T_FH cells may play important roles in TNF-α–mediated tissue disorganization, although their exact role has not been resolved.

Our findings that TNF-α suppresses GC B cells also contrast with studies of Listeria monocytogenes–infected mice, which concluded that TNF-α was required for GC development (31, 32). Unlike previously published studies of gene-targeted TNF-α deletion, we detected B cell follicles and FDC networks in infected TNF-α–deficient mice. It is possible that TNF-α has a different effect during Listeria infection because this pathogen primarily elicits a CD8 T cell–dependent protective response (59–61). Our findings are, nevertheless, relevant to other infections, including pathogens, such as Salmonella (3, 62) and plasmodium (4, 63, 64), which are likewise associated with a disordered lymphatic microenvironment and a loss of GCs. However, a similar role for TNF-α in these infections has not been addressed in depth.

Our work supports the notion that host responses to infection may benefit from TNF-α inhibition because we have shown that this treatment generates a higher proportion of both Ab-secreting plasmablasts and IgM memory cells. Alternatively, TNF-α inhibition may contribute to autoimmunity by driving autoreactive B cells, as such B cells have been shown to develop extrafollicularly (28, 65). Autoimmunity induced by anti–TNF-α therapy has been documented clinically, although correlations with infections were not reported (66–70). Our findings may warrant a reinterpretation of such studies, especially because T-bet⁺ B cells have been associated with autoimmunity (21–23).

We gratefully acknowledge Dr. Michael Lyon for assistance with cryosectioning and Dr. R. Racine and Maura Jones for helpful preliminary data. We also acknowledge the excellent technical assistance provided by the State University of New York Upstate Medical University's Flow Cytometry and Imaging Core Facilities. We also thank Vaccinex Inc. (Rochester, NY) for generously providing the anti-CXCL13 mAb (clone 5378).

The authors have no financial conflicts of interest.