Functional modulation of T follicular cells in vivo enhances antigen-specific humoral immunity

Generation of high-affinity IgG is essential for defense against infections and cancer, is the intended consequence of many vaccines, but can cause autoimmune and inflammatory diseases when inappropriately directed against self (Wang et al., 2018, Ludwig et al., 2017, Chinen et al., 2010). The interplay and balance of T follicular helper cells (TFH) and T follicular regulatory cells (TFR) is critical for production of high-affinity IgG (Wing et al., 2018). Here, we empowered TFH cells and improve antigen-specific IgG responses with two interventions intended to transiently diminish TFR influence. First, adult mice were administered an antibiotic cocktail (ABX) for an extended period to deplete the immunoregulatory intestinal microbiota (Belkaid and Harrison, 2017, Thaiss et al., 2016, Rooks and Garrett, 2016, Honda and Littman, 2016, Perruzza et al., 2017, Teng et al., 2016, Block et al., 2016, Proietti et al., 2014, Slack et al., 2014). This treatment skewed T follicular cell ratios, with increased TFH and reduced TFR numbers. TNP-KLH immunization resulted in higher affinity TNP-specific IgG in ABX mice compared to controls. In a model of IgG-driven inflammatory nephritis, ABX mice had significantly worse nephritis accompanied by higher affinity antigen-specific IgG, and enriched TFH cells compared to controls. Second, we sought to functionally manipulate TFH and TFH cells, which both express the checkpoint inhibitory molecule, PD-1 (Sage et al., 2013), by administration of α-PD-1 during immunization. This intervention enhanced the affinity of antigen-specific IgG and increased in TFH following TNP-KLH immunization and nephritis induction. These results suggest that altering TFH and TFR ratio during immunization is an appealing strategy to qualitatively improve IgG responses.


Introduction
Immunoglobulins (IgG) link the adaptive and innate immune systems with their two domains that combine precise specificity for antigen in the antigen-binding fragment (Fab), and convey effector functions via the constant, crystallizable domain (Fc) interactions with Fc  receptors and complement (Schroeder and Cavacini, 2010, Nimmerjahn and Ravetch, 2008, Ravetch and Bolland, 2001. Production of high-affinity antigen-specific IgG of appropriate subclass is essential for host defense against many infections, and the intended consequence of many vaccines (Zimmermann and Curtis, 2019, Compeer and Uhl, 2020, Robbiani et al., 2020, Yang et al., 2020.
Insights into the cellular and molecular regulation of IgG production have highlighted the role of follicular-resident T cells as crucial gate keepers . T follicular helper cells (TFH) provide B cell help to drive IgG production, whereas T follicular regulatory cells (TFR) set activation thresholds for these responses (Crotty, 2011, Hatzi et al., 2015. TFH cells are identified in by surface expression of CD4, PD1, and CXCR5, and the transcription factor expression, Bcl6 (Hatzi et al., 2015). These markers are shared with TFR, which also express Foxp3 (Sage et al., 2013). Thus, striking a balance between these cells is critical for productive, and not harmful, IgG responses. Indeed, some autoimmune diseases results from IgG generated against host antigens, underscoring the importance of the balance between TFH and TFR cells (Zhang et al., 2017, Sage and Sharpe, 2015.
The intestinal microbiota plays critical roles in development and regulation of the immune system (Rooks and Garrett, 2016). Germ-free mice lacking any bacteria have severely restricted T cell and B cell repertoires (Chen et al., 2018, Wesemann et al., 2013, Li et al., 2020.
Further, the microbiota influences the immune system and its responses through multiple nonmutually exclusive mechanisms. Microbiota-derived metabolites, including short-chain fatty acids (SFCA), have been shown to promote T regulatory cells (Chinen et al., 2010, Atarashi et al., 2011 and also TFR cells (Takahashi et al., 2020). Furthermore, depletion of the microbiota revealed exacerbated or harmful immune responses, generally attributed to T regulatory deficiencies or deficits (Atarashi et al., 2011, Belkaid and Harrison, 2017, Chinen et al., 2010, Honda and Littman, 2016, Maslowski et al., 2009, Rakoff-Nahoum et al., 2004, Rooks and Garrett, 2016. In contrast to this, depletion of the microbiota in humans by a 5-day treatment with broad-spectrum antibiotics led to impaired IgG responses to flu vaccination (Hagan et al., 2019).
Because the effects of dietary SFCA on the immune system required weeks of feeding , Maslowski et al., 2009, Tan et al., 2017, we explored the effect of an extended antibiotic treatment on immunization in addition to characterizing antibody subclass response.
We therefore asked whether treatment with broad-spectrum antibiotics for an extended period would impact TFH and TFR cells, service as an initial platform to influence IgG responses.

Results
We set out to characterize the effect of long-term microbiota depletion on immune response, and orally administered a cocktail of ampicillin, neomycin, metronidazole, and vancomycin (Rakoff-Nahoum et al., 2004), collectively referred to as ABX for several weeks.
Fecal samples from treated animals were cultured under aerobic conditions on Luria broth (LB) agar and anaerobic conditions on brain heart infusion (BHI) agar to reveal a significant reduction in colonies in ABX-mice compared to water-treated wild type (WT) control animals ( Figure 1A).
16S sequencing of fecal pellets confirmed marked perturbations of the microbiota by ABX, as 4 the dominance of Bacteroidetes and Firmicutes in WT was replaced by Firmicutes and Proteobacteria ( Figure 1B, Figures S1A-B). Serum IgG antibody titers in ABX and WT mice were assessed in ELISA, revealing similar total mouse IgG1 (mIgG1, Figure 1C), increased total mIgG2a/c ( Figure 1D), increased total mIgG2b ( Figure 1E), decreased total mIgG3 ( Figure 1F), increased total mIgE ( Figure 1G), similar total mIgM ( Figure 1H), decreased total mIgA ( Figure   1I), as well as total fecal mIgA ( Figure 1J) in ABX-mice compared to WT controls. Intriguingly, the extended ABX treatment shifted homeostatic total mIgG production towards more inflammatory subclasses, including mIgG2a/c and mIgG2b.
We next explored the functional contribution of extended ABX treatment by a model of induced Goodpasture's Disease that culminates in IgG-driven acute nephrotoxic nephritis (NTN). WT or ABX-mice were immunized with sheep IgG and Complete Freund's Adjuvant (CFA; Figure 4A). Four days later, the mice were administered anti-mouse glomeruli basement membrane sera (GBM) raised in sheep (sheep α-GBM), which effectively targets immune complexes to the kidney (Salant andCybulsky, 1988, Kaneko et al., 2006a). On day 7 following sheep α-GBM treatment, serum was collected and assessed, revealing little differences in total IgG across all subclasses in the sera (total mIgG1, Figure 4B; total mIgG2a, Figure 4C; total mIgG2b, Figure 4D; total mIgG3, Figure 4E). However, analysis of antigen-specific α-sheep mIgG titers in sera showed reduced titers of α-sheep mIgG1 and α-sheep mIgG2b in ABXtreated mice compared to WT controls ( Figure 4F-I). Despite this, ABX-mice suffered higher 6 mortality ( Figure 4J) and significantly more kidney damage as measured by blood urea nitrogen (BUN) on day 7 ( Figure 4K) compared to WT controls, indicating microbiome depletion exacerbated IgG-mediated autoimmune nephritis. These results of reduced titers were consistent to a study of responses to flu vaccination during 5-day ABX (Hagan et al., 2019), but indicated the IgG generated during extended ABX-treatment had enhanced effector function/functionality. We next asked whether extended ABX treatment altered effector cells that mediate IgGresponses, including neutrophils and macrophages, thereby compensating for lower IgG titers. Therefore, we passively transferred arthritogenic K/BxN sera (Korganow et al., 1999) to ABX and WT control mice, to induce IgG-mediated joint inflammation. Footpad swelling was monitored over the next several days revealing the transferred sera induced robust arthritic swelling of paws in both treatment groups, although swelling was slightly reduced in ABX animals ( Figure S3A) suggesting WT and ABX animals have similar abilities to elicit IgG functions.
We next examined qualitative differences in mIgG produced by ABX animals during NTN using LPS instead of CFA, because extended ABX treatment enhanced affinity of antigenspecific IgG (Fig. 3N). Analysis of sera mIgG from NTN-induced WT controls and ABX mice revealed similar steady-state binding affinities for sheep IgG ( Figure 5A). Because NTN targets immune complexes to the kidney, we compared mIgG from the kidney of NTN-induced WT controls and ABX-treated mice. Intriguingly, mIgG recovered from the kidney of ABX mice had a marked higher affinity for sheep IgG than kidney-derived mIgG from WT controls ( Figure 5B). This trend held across mIgG subclasses, as kidney α-sheep mIgG1 ( Figure 5C), α-sheep mIgG2a/c ( Figure 5D), and α-sheep mIgG2b ( Figure 5E) from NTN-induced ABX mice had higher affinity for sheep IgG compared to WT controls.

7
To determine whether higher affinity IgG deposited in the kidney of ABX-treated mice following NTN induction was responsible for severity of disease, we passively transferred kidney-deposited IgG ( Figure 5F, Figure S4). NTN was induced in WT controls and ABXtreated mice, and confirmed as day 7 BUN levels enhanced in ABX-treated and reduced α-sheep IgG2b responses compared to controls, and similar α-sheep IgG affinity in sera ( Figure 5G, Figure S4A-F). IgG was purified from the kidneys of WT and ABX mice at day 9 after α-GBM treatment, and transferred to WT control recipients ( Figure 5F, Figure S4E-F). Consistently, WT mice receiving ABX-treated kidney IgG had elevated BUN levels compared to mice administered WT kidney IgG ( Figure 5H), despite transfer of equivalent amounts of IgG. These results confirm IgG deposited in the kidney of ABX-animals after NTN-induction is more pathogenic.

Discussion
Productive immune responses coordinate multiple effector arms, often culminating in high affinity IgG. The humoral arm of the immune system is particularly effective in eliciting defense against a number of infections. It is well established that generating high affinity IgG requires T cell help, and more recently the importance of TFH and TFR as critical cells driving the IgG response is appreciated , Hatzi et al., 2015, Crotty, 2014, Linterman et al., 2011, Crotty, 2011, Zhang et al., 2020, Gowthaman et al., 2019. The balance of these cells ultimately directs antibody production, acting as crucial gatekeepers to IgG responses. Here, we identify two approaches that modulate the balance of TFH and TFR, favoring the generation of high affinity IgG.
The microbiota is well established to play a critical role in maintenance of immunological homeostasis and tolerance. Studies have identified metabolites produced by the microbiota, and induction of regulatory T cells as mechanisms through which the immune system is regulated by the microbiota (Takahashi et al., 2020, Maslowski et al., 2009. Our results uncovered a surprising role in this regard, where the microbiota set the balance of cells intimately involved in IgG responses, and specifically IgG affinity, including TFH, TFR, and GC B cells. Extended ABX leading to transient removal of the microbiota resulted in skewing of these populations in favor of producing high affinity antigen-specific IgG. This held true for two distinct models of immunization, using distinct antigens and adjuvants. The higher affinity IgG resulted in enhanced pathology in the NTN model, despite lower IgG titers and underscores the importance of the quality of the IgG response over quantity. However, it is unlikely that our extended ABX-regiment be implemented as an immunization strategy for multiple reasons. Indeed, short-term antibiotic regiment was previously shown to impair IgG responses to flu vaccination (Hagan et al., 2019). Therefore, we attempted to functionally recapitulate the effect of extended ABX-treatment, namely increasing TFH cells while decreasing TFR cells. Both of these cells express PD-1 which can be readily targeted by checkpoint inhibition immunotherapy. Indeed, blockade of PD-1 enhanced the TFR:TFH ratio in favor of TFH cells. Moreover, this intervention resulted in generation of IgG with higher affinity in two unique immunization models, each with distinct adjuvants. Total and antigen-specific IgG titers were not impaired by α-PD-1 treatment.
Checkpoint inhibition immunotherapy has been a breakthrough for the treatment of many cancers (Gubin et al., 2014, Mamalis et al., 2014. This approach is based on administration of IgG that block inhibitory receptors on immune cells, thereby triggering activating cellular immunity leading to tumor clearance (Wei et al., 2019, Wei et al., 2017, Leach et al., 1996, Iwai et al., 2005, Iwai et al., 2002, Freeman et al., 2000. Although how checkpoint inhibition impacts other aspects of the immune response is in active investigation, reports have indicated intact IgG responses to influenza vaccination in patients undergoing checkpoint therapy, as PD-1 is known to be a negative regulator of B cells (Nishimura et al., 1998)(Ref). We show that PD-1 blockade checkpoint inhibition during immunization with sheep IgG or TNP results in antigen-specific IgG with higher affinity compared to isotype control treatment groups. This intervention did not result in a drop in overall IgG titers, as ABX-treatment did. The availability of PD-1 blockade makes it appealing to speculate it may also improve vaccine responses, enhance monoclonal IgG affinity, or be applied to improve IgG affinity in general. Indeed, potential off-target effects of α-PD-1 are a concern (Nishimura et al., 2001), and a number are associated with cancer immunotherapy. A distinct, and perhaps lower dosing regimen of PD-1 blockade would likely be required. Our experiments showed fewer injections of PD-1 blocking antibodies effectively enhanced the humoral response. Future studies will explore the development of immunological memory responses following PD-1 blockade during vaccination. For ABX, 250uL of an antibiotic cocktail consisting of ampicillin (1g/L), metronidazole (1g/L), vancomycin (0.5g/L) and neomycin (1g/L) was given via oral gavage daily as previously described (Rakoff-Nahoum et al., 2004). After daily treatment mice were housed in new cage with autoclaved fresh bedding, food pellets and water. Treatment was initiated at least 3 prior to analysis, immunization, or initiation of inflammation. Assuming a standard deviation of 9mg/dL for this model, treatment groups of 5/mice would generate p<0.05 at a 95% confidence interval. We therefore randomly assigned five mice to each treatment group for experiments. KRN TCR transgenic mice on a C57BL/6 background (K/B) were gifts from D. Mathis and C. Benoist (Harvard Medical School, Boston, MA) and were bred to NOD mice to generate K/BxN mice (Korganow et al., 1999). K/BxN serum was prepared as described previously (Kaneko et al., 2006b). Inflammatory arthritis was induced by intravenous injection of K/BxN sera (200μL of pooled K/BxN serum per mouse). Arthritis was scored by clinical examination, and the index of all four paws was added (0 = unaffected, 1 = swelling of one joint, 2 = swelling of more than one joint, 3 = severe swelling of the entire paw) as described (Kaneko et al., 2006b). PBS or ABX treated mice via oral gavage 3-4 weeks prior to K/BxN serum injection.

In vivo
For PD-1 blockade, 0.2 mg/kg of α-PD-1 (CD279, Biolegend) was administered intraperitoneally. Treatment was initiated 4 days before TNP-KLH in alum immunization. In the nephrotoxic nephritis model, treatment was initiated 8 days and 4 days before preimmunization with sheep IgG.
Bacterial stool cultures. Mouse stools samples were weight and diluted 1:10 in sterile PBS based on pellet weight. The stool pellet was then vortexed and spun down in microcentrifuge. 20 μL of mix supernatant was then incubated in LB and BHI plates for 2.5-24 hrs. at 37C and the CFU/mL were enumerated. For anaerobe incubation, GasPak EZ Gas Generating Pouch Systems (BD) was used as instructed by manufacturer.
16S Sequencing. Mouse stool samples were collected, and DNA was extracted as per Powersoil 96 kit (Qiagen). 16S rRNA gene libraries targeting the V4 region of the 16S rRNA gene were prepared by first normalizing template concentrations and determining optimal cycle number by qPCR. To ensure minimal over-amplification, samples were normalized to the lowest 13 concentration sample and were amplified with optimal cycle number for the library construction PCR. Four 25 uL reactions were prepared per sample and each sample was given a unique reverse barcode primer from the Golay primer set (Caporaso et al., 2012, Caporaso et al., 2011a, Caporaso et al., 2011b. Replicates were pooled and cleaned via Agencourt AMPure XP-PCR Diversity Index (Shannon, 1948) calculated from relative abundance data at the ASV level.