It was a great privilege for me to serve on the American Association of Immunologists (AAI) Council and to be President of the AAI during the 2015–16 cycle. I particularly appreciate this opportunity to reminisce a bit about how my early interest in immunology led to the studies in which my laboratory is currently engaged. I titled my AAI presentation “From the Thymus to the Mucosa” because, during the past three decades, my group’s interests progressed from those around development of cells within the thymus to those related to how T cells differentiate in response to gut microbiota.
My interest in science was sparked in my early years. Growing up as a child in Romania at the outset of the space age, I was most excited by the prospect of space exploration. Indeed, after my family emigrated to the United States, I entered Princeton University intent on studying aerospace engineering. My engineering interest was short-lived, however, as I discovered that I wanted a broader-based education, and I switched to majoring in biochemistry after reading some books, including Watson’s Molecular Biology of the Gene. My love of research was sparked by my experience after I joined the laboratory of Marc Kirschner, a new assistant professor who inspired me to spend day and night in a cold room studying the physical properties of polymerization of microtubules. That was my first hands-on exposure to science, and I remember spending all-nighters counting drops for viscosity assays of polymerizing microtubules.
This laboratory experience firmed up my resolve to go into experimental science, but another event led me to immunology. There were no immunologists at Princeton University at that time (and there have been notably few in the following decades) and so Arnie Levine, a virologist at Princeton University, enlisted his friend Norman Klinman, then at the University of Pennsylvania, to organize a course on immunology. They proceeded to invite many of the top immunologists in the world to give lectures. I was fortunate to sit in on the course and became very excited about problems presented by the field of immunology. Big questions back then were based around how diversity is generated and how self versus nonself recognition is achieved. Although we now have a detailed mechanistic understanding of diversity, much still remains to be learned about the regulation of autoreactivity, despite major discoveries of central and peripheral tolerance mechanisms. We still have only a sketchy understanding of why tolerance breaks down in various autoimmune or inflammatory diseases. The question of self/nonself recognition featured prominently in my choice of a laboratory in which to carry out my research project after entering the M.D./Ph.D. program at Washington University in St. Louis. Ben Schwartz and Susan Cullen had recently joined the faculty after completing their postdoctoral fellowships at the National Institutes of Health, and I joined their joint laboratory as their first graduate student. I decided to ask whether recognition of viral Ag and MHC was carried out by one or two receptors on T lymphocytes. To do this, I incorporated MHC and viral envelope glycoproteins into lipid vesicles, to ask whether they could stimulate MHC-restricted virus-specific T cells (1). This approach failed, of course, as we did not consider Ag processing at that early time, but I learned a great amount and concluded that I wanted to next join the laboratory of Richard Axel at Columbia University, where I could take advantage of the new tools becoming available as part of the molecular biology revolution. There, I developed gene transfer approaches and screens for isolation of genes encoding cell surface molecules. Because of my interest in T cell lineages and thymic development, I focused on the cloning of the genes encoding human CD8 and, subsequently, CD4 (2, 3). During that time, it became increasingly clear that the newly discovered HIV specifically attacked CD4+ T cells and that led me to an interest in what accounts for this tropism. Thus, when I started my own laboratory in 1985 in the Department of Microbiology and Immunology at the University of California, San Francisco, my goal was to understand the functions of CD4 and CD8 in T cell development and activation and the potential role of CD4 in HIV entry.
Although our early interest was in how the coreceptors CD4 and CD8 function, particularly how they interact with TCR, MHC class II and class I, and the tyrosine kinase Lck, my laboratory’s long-standing goal has been to understand how immature double-positive thymocytes are specified to become CD4+ MHC class II–restricted Th cells (or regulatory T cells [Tregs]) or CD8+ MHC class I–restricted cytotoxic cells. This is a problem that remains unsolved, but it has led us in interesting directions. We focused on characterizing transcriptional programs that distinguish the lineages, anticipating that this would allow us to work backward toward identifying distinctions in signaling pathways initiated by interaction with different types of MHC molecules. By focusing on Cd4 gene regulation, we identified a lineage-specific silencer that is regulated epigenetically by Runx family transcription factors, particularly Runx3 that solidifies commitment of CD8 lineage cells while conferring heritability of Cd4 silencing function (4). It was in the context of our search for transcription factors that regulate thymocyte lineage specification that we identified an important function for the nuclear receptor RORγt in lymphoid cells shortly after our laboratory relocated to the Skirball Institute at New York University (NYU) (5).
Although the discovery of a RORγt function in the thymus did not provide insight into the lineage problem, it served as a critical pivot toward our subsequent studies of the intestinal mucosa and, in turn, of the role of the gut microbiota in immune responses. Our early characterization of RORγt function showed its importance for survival of double-positive thymocytes, preventing death by neglect and giving the cells an opportunity to undergo positive selection (5). Moreover, it led us to identify its requirement for the development of a class of innate lymphoid cells, the lymphoid tissue inducer cells required for lymphoid organogenesis, including formation of intestinal Peyer’s patches and tertiary lymphoid organs. Shortly thereafter, using a mouse strain in which GFP reports on RORγt expression, Ivo Ivanov, a postdoctoral fellow, found lamina propria T cells that expressed relatively low levels of the transcription factor. Coincidentally, Dan Cua’s group at DNAX found RORγt to be upregulated, along with IL-17A, when activated T cells were treated with IL-23. Th17 cells had just been described and, in collaboration with the DNAX scientists, we showed that IL-17A is expressed in a RORγt-dependent manner in intestinal T cells and in activated CD4+ T cells cultured with a combination of IL-6 and TGF-β, conditions that favor Th17 cell differentiation (6).
Two features stood out in our initial characterization of the RORγt-dependent Th17 cells (6): first, these cells were most abundant among CD4+ T cells in the small intestine, consistent with subsequent findings that they promote tissue repair at mucosal surfaces; and, second, many of the RORγt-dependent Th17 cells required for the autoimmune model experimental autoimmune encephalomyelitis expressed not only IL-17A, but also the Th1 cytokine IFN-γ. As discussed below, we now know that these mixed Th1/Th17 cells are dependent on signaling through IL-23R, are critical in autoimmune disease, and are similarly present in humans. Ivo Ivanov then found that animals treated with antibiotics or rendered germ-free had few, if any, Th17 cells. He soon thereafter made the serendipitous observation that some of the C57BL/6 mice in our colony had few intestinal Th17 cells, whereas others had many, and this was dependent on the vendors from whom the mice were purchased. When the animals were cohoused, they all acquired large proportions of Th17 cells. This led to a collaboration with Kenya Honda that revealed that colonization with a single microbial species, segmented filamentous bacteria (SFB), a Gram-positive spore-forming anaerobe, determined the number of Th17 cells (7). After colonization of germ-free mice with SFB, there was massive induction of IL-17 and IL-22 in CD4+ T cells. This was beneficial to mice, as it protected them from pathogenic bacteria like enteropathic Escherichia coli or Citrobacter rodentium. However, induction of Th17 cells by SFB also predisposed animals to autoimmunity, as shown in the microbiota-dependent K/BxN model of spontaneous arthritis, in collaboration with Diane Mathis and Christophe Benoist (8). This result led us to study fecal microbiota in rheumatoid arthritis patients, and Jose Scher, a clinical fellow at NYU, found that new-onset rheumatoid arthritis patients were much more likely than healthy controls to be colonized with the Gram-negative bacterium Prevotella copri (9). However, it is not yet known whether this bacterium has a causal role in disease, nor is it understood how Th17 cells induced in the gut mucosa can contribute to systemic autoimmunity. This is a central question that is the focus of our current work.
To elucidate the mechanism of SFB-induced autoimmunity, we initially investigated the specificity of the SFB-induced Th17 cells. Benny Yang showed that almost all of these Th17 cells have receptors specific for SFB-encoded Ags (10). In contrast, when we transplanted one of the immunogenic SFB genes into Th1-inducing Listeria monocytogenes, there was a Th1 response specific for the SFB Ag upon colonization with the modified Listeria. This indicated that it is the nature of the bacterium, potentially the niche that they establish or the type of signal that they transmit to APCs in their microenvironment, that determines whether the APCs will direct T cells toward the Th17 or Th1 fate. In studies that nicely complemented our work on Th17 cells, Kenya Honda found that a collection of clostridial species can induce Tregs and, thus, confer some of the protective properties of Tregs in autoimmune diseases or in allergies (11). In our laboratory, Mo Xu showed that Helicobacter hepaticus, which colonizes the large intestine, also induces Tregs that are specific for bacterial Ag (see below). These findings raise the questions of how different CD4+ T cell programs are differentially induced by the diverse noninvasive intestinal bacteria and how induction of effector cells in the draining lymph nodes contributes to systemic inflammatory diseases, such as arthritis.
To investigate how the different types of T cells are induced by luminal microbes, we developed tools to study Ag specificity, including mice with transgenic TCRs specific for Ags encoded by SFB or H. hepaticus and MHC class II tetramers conjugated with bacterial peptides recognized by the T cells. H. hepaticus had been shown by Fiona Powrie’s laboratory to induce an inflammatory response, marked by cells with mixed Th1 and Th17 properties, in mice deficient for IL-10 (12). By studying H. hepaticus Ag-specific responses, Mo Xu demonstrated the induction of Tregs under noninflammatory (IL-10–sufficient) conditions. In mice colonized with both SFB and H. hepaticus, transplanted TCR-transgenic T cells recognizing SFB differentiated into RORγt+ Th17 cells, whereas the T cells specific for H. hepaticus were largely RORγt+Foxp3+ induced Tregs. In contrast, when the T cells were introduced into colonized mice that had been treated with Ab against IL-10R, SFB- and H. hepaticus–specific T cells became RORγt+, but only the latter were highly pathogenic Th17 cells that mediated colitis. These cells produced both IL-17A and IFN-γ and had a very different profile of gene expression than the SFB-specific Th17 cells. Studies from multiple laboratories demonstrated that the “pathogenic Th17” cell profile of gene expression is dependent on IL-23R signaling in T cells (13–15). Moreover, Stockinger’s group performed fate-mapping studies showing that, in autoimmune disease models, there are RORγt+- and IL-23R–dependent IFNγ+ cells derived from IL-17A–expressing T cells (13). Such cells have properties that are similar to those described by Sallusto (16) for CXCR3+CCR6+ Th1* cells in humans. Casanova and colleagues characterized individuals with immunodeficiency who had homozygous mutations in the gene encoding RORγ/γt and phenotypes very similar to those described in mutant mice. Remarkably, these individuals had reduced Th17 and Th1* cells and were particularly susceptible to mycobacteria, because immunization with bacillus Calmette–Guérin resulted in uncontrolled disseminated disease (17). These findings led us to propose that distinct commensal microbes induce nonpathogenic Th17 cells that protect barrier surfaces (as SFB does) and potentially pathogenic Th1* cells (as induced by Candida albicans and by H. hepaticus). Differentiation of the latter is promoted by IL-23 but is restrained by induced Tregs (iTregs) and IL-10, which may be induced by the same bacteria. The ability of bacteria to induce iTregs and pathogenic Th17 cells may be key to establishment of commensalism and, indeed, we have evidence that H. hepaticus–specific RORγt+ Tregs are essential to prevent expansion of pathogenic Th17 cells that can mediate spontaneous colitis (M. Xu, M. Pokrovskii, and D.R. Littman, unpublished observations). These findings highlight a major gap in our understanding as to how commensal microbiota signals direct T cells down different paths of differentiation.
We have begun to investigate how different microbes direct distinct programs of T cell differentiation. Our first efforts were to study how SFB induces Th17 cells (18). We first examined changes in gene expression in intestinal epithelium along the length of the gut following colonization with the bacterium. SFB preferentially associates with epithelium in the terminal ileum, and it was only in this part of the intestine that there were changes in gene expression. The most highly upregulated genes were those encoding serum amyloid A (SAA) 1 and 2, which are secreted molecules known as acute-phase reactants. Expression of these genes occurs within 2–3 d following colonization with SFB and requires a cytokine signaling relay in the small intestinal lamina propria. Thus, IL-23, produced by monocyte-derived cells, activates group 3 innate lymphoid cells and induces their production of IL-22, which binds to receptor on epithelial cells to activate STAT3 phosphorylation–mediated transcription of Saa1/2. In the meantime, APCs drain to the mesenteric lymph nodes, where they activate SFB-specific naive T cells and direct their differentiation toward the RORγt+ Th17 program. We found these RORγt+ T cells distributed throughout the intestine and also in distal lymph nodes and spleen, but they selectively produced IL-17A in the ileum, in response to a signal that is mediated, at least in part, by SAAs. Therefore, we have proposed a two-step process for activation of Th17 cell effector function in response to SFB (18). There is initial specification of the program, marked by RORγt expression, in the draining lymph node, and subsequent induction in target tissues of the effector functions by SAAs and other natural adjuvant-like molecules that have yet to be characterized. This raises the possibility that weak self-antigen agonists may induce inflammatory cytokine production by SFB-specific T cells in tissues in which local SAA levels are high. We have mounting evidence that SAAs may have such functions. For example, when activated CD4+ T cells were exposed to IL-6, there was very little IL-17A and IL-17F production. However, when recombinant SAA1 was included, there was marked upregulation of IL-17 cytokines. This effect of SAAs was also observed under typical in vitro Th17-differentiation conditions, with IL-6 and TGF-β, and with IL-6, IL-1β, and IL-23, which are thought to induce “pathogenic” Th17 cells. We additionally observed a much milder disease course in the experimental autoimmune encephalomyelitis model in mice deficient for SAA1 and SAA2 (J.Y. Lee and D.R. Littman, unpublished observations). We therefore propose that SAAs and other endogenous adjuvants trigger autoimmune inflammation mediated by RORγt+ T cells whose differentiation requires appropriate intestinal microbiota (Fig. 1).
This two-step process for activation of Th17 cell functions may also be important following innate antiviral responses, resulting in protective or pathogenic outcomes. An example of the latter was uncovered in our studies of a mouse model of neuropathology associated with an autism-like syndrome. This model, named the maternal immune activation (MIA) model, was developed over the last several decades in the wake of retrospective studies that reported an increased likelihood of autism in children born to mothers who had various infections during early pregnancy. For this model, either a virus or polyinosinic-polycytidylic acid [poly(I:C)], which activates TLR3, is administered to the mother at day 12.5 of gestation. This results in morphological changes in the CNS of the offspring and in behavioral abnormalities that last through adult life. The late Paul Patterson, who championed this model, found that the phenotype in the offspring required intact IL-6 production in the mother (19). A colleague working at the Simons Foundation for Autism Research brought this article to my attention, and I was fortunate that a new postdoctoral fellow, Jun Huh, expressed interest in the problem and decided to take it on as a side project. Although this was a tangential direction for our laboratory, we succeeded in obtaining a pilot grant from the Simons Foundation to study the pathway involved in the MIA phenotype. We had only recently learned that IL-6 was involved in the induction of RORγt and Th17 cells and, hence, hypothesized that downstream Th17 cells might be involved in the autism spectrum disorder–like phenotype in the offspring. Jun confirmed that there was very rapid upregulation of serum IL-6 after poly(I:C) challenge of the pregnant mice. This was accompanied by a subsequent rise in serum IL-17A, and its level was sustained longer than that of IL-6. We bred conditional Rorc mutant mice with CD4-cre mice to specifically knock out RORγt expression in TCRαβ T cells. As expected, there was no elevation in IL-17A following injection of poly(I:C) in the mutant pregnant dams. Jun (now at the University of Massachusetts Medical Center) then collaborated with his wife, Gloria Choi, a superb neuroscientist who is now at the Massachusetts Institute of Technology, and with our NYU colleague Charles Hoeffer to study behaviors in the offspring. They performed assays for social interaction, communication, and repetitive behaviors. All of these were abnormal after induction of MIA in the mother, but there was complete reversal of the phenotype if the mothers were deficient for RORγt in their T cells or if they were treated with neutralizing anti–IL-17A Ab (20). Jun also performed in situ hybridization for expression of IL-17R in the fetal brain and found it in cells of the developing cortex. Remarkably, its expression was upregulated following MIA induction, consistent with the known positive feedback of IL-17R signaling. Moreover, there was a stereotyped defect in the morphology of the developing fetal cortex observed as early as embryonic day 14.5 and persisting into adult life. This was also eliminated with anti–IL-17A or if the mother was deficient in RORγt. The behavioral and morphological phenotypes were reproduced in the offspring in the absence of MIA if IL-17A was injected through the uterine wall into the fetal brain ventricles at embryonic day 14.5. These phenotypes were not observed if the fetus was deficient in IL-17RA or if IL-6 was injected instead of IL-17A. Based on these results, we propose that inflammation that results in Th17 cell differentiation and IL-17A production in the mother, during a critical stage of gestation, can result in fetal brain developmental abnormalities and in behavioral defects in the offspring (20). But the production of the cytokine is dependent on a priming signal (signal 1), which could be the microbiota, followed by an upsurge in IL-17 only after exposure to signal 2, mediated by infection with a pathogenic microbe. We do not yet understand why IL-17R is expressed in the brain, whether it contributes to normal development of the CNS, or whether it is also present in similar types of cells, which we have not yet fully identified, in humans. The implication of these results is that there may be a subset of mothers who harbor a microbiota prone to elevating tissue Th17 cell numbers, which renders them more susceptible to having high levels of IL-17A.It may thus be possible to reduce the incidence of autism spectrum disorders by prescreening women before or during pregnancy and intervening to reduce their circulating IL-17A.
In closing, I wish to reiterate that the microbiota can have positive and negative influences on our health. There is mounting evidence that it regulates susceptibility to autoimmune disease and that it may even influence development of the CNS. It is also likely that the microbiota can contribute to tumor progression and HIV pathogenesis. But, of course, the microbiota coevolved with us to provide mostly beneficial functions. Our goal is to harness such functions for use in improving barrier defenses, developing better vaccines, and perhaps improving checkpoint immunotherapy, because there clearly is a relationship between microbiota and efficiency of CTLA-4 and PD-1 blockade.
Abbreviations used in this article:
American Association of Immunologists
maternal immune activation
New York University
serum amyloid A
segmented filamentous bacteria
regulatory T cell.
The author has no financial conflicts of interest.