The innate immune system is responsible for initiating the host immune response to infection. The study of microbial virulence has uncovered numerous mechanisms for microbes to evade innate immune detection. In contrast, relatively little is understood about the strategies employed by the host to prevent microbes from evading innate immunity. The NAIP innate immune receptors provide an intriguing case study to investigate these strategies. In mice, Naip has undergone gene duplication and drift, recombination, and pseudogenization1, all of which can be signatures of a co-evolutionary arms race with targeted pathogens2. This duplication and specialization has allowed NAIP paralogs to recognize several distinct bacterial proteins: NAIP5 binds bacterial flagellin (FlaA), and NAIP2 detects the inner rod protein (PrgJ) of the pathogen-associated type III secretion system (T3SS)3,4.

I sought to address how gene duplication and drift enabled functional specialization by first defining which NAIP domains bind to bacterial ligands. I analyzed a panel of chimeric proteins, in which homologous domains of NAIP5 and NAIP2 were swapped, to determine which domains conferred the ability to recognize FlaA or PrgJ. A long-standing expectation in the field was that the auto-inhibitory C-terminal leucine-rich repeat (LRR) domain mediates ligand binding. Surprisingly, I found that the LRR was dispensable for ligand specificity. Instead, ligand recognition was mediated by several alpha-helical domains in the center of the protein. Strikingly, these domains are specifically evolving under positive selection, in which non-synonymous mutations are repeatedly selected to provide altered ligand binding surfaces. Separation of sensing and auto-inhibition functions into different domains may allow NAIPs to sample ligand recognition-altering mutations without disrupting steady-state auto-inhibition.

These data suggested that NAIPs are engaged in a co-evolutionary arms race with bacteria over innate immune detection. However, bacterial ligands can evolve much more rapidly than mammalian NAIPs. To determine how NAIPs can successfully compete in such an arms race, I conducted alanine scanning screens of FlaA and PrgJ to comprehensively identify the ligand motifs recognized by NAIPs. Both NAIP5 and NAIP2 recognized multiple conserved surfaces, near the N- and C-termini, of their respective ligands. This multi-surface recognition strategy conferred NAIPs with robust detection of their bacterial ligands, as single point mutations in any recognition motif did not affect NAIP recognition. Rather, bacterial immune evasion required simultaneous mutation of multiple recognition motifs. However, highly mutated ligands that escaped immune detection also lost their native function, suggesting that multi-surface recognition serves to constrain bacterial immune evasion.

To verify these biochemical results, we have determined the cryo-EM structure of NAIP5 bound to FlaA, an event which triggers oligomerization with the adapter protein NLRC4 into a large (>1 MDa) signaling complex. The structure reveals direct contacts between the NAIP5 ligand recognition domains and both recognition surfaces of FlaA. The extensive and largely hydrophobic contacts between NAIP5 and FlaA are consistent with a lack of “hot spot” binding sites and likely contribute to the robust recognition of NAIP5 for FlaA single point mutants. Additionally, our structure reveals how binding to FlaA triggers a conformational change in NAIP5 to expose its oligomerization surface, allowing the recruitment of NLRC4. The polymerization of NLRC45 and subsequent recruitment of the signaling effector, CASPASE-1, illustrates the switch-like mechanism by which the detection of a single ligand monomer is amplified into oligomerization-induced signaling.

Collectively, this dissertation elucidates the biochemical mechanism of NAIP innate immune detection of bacterial ligands. Furthermore, it has provided surprising insights into strategies employed by innate immune receptors to compete with bacteria in an evolutionary arms race over host defense.

The innate immune system detects pathogens and initiates adaptive immune responses. Inflammasomes are central components of the innate immune system, but whether inflammasomes provide sufficient signals to activate adaptive immunity is unclear. In this dissertation, I present the work I have done using a genetic mouse model system that allowed me to simultaneously express the model antigen ovalbumin (Ova) and activate the NAIP–NLRC4 inflammasome in specific cells throughout the mouse. Chapter One begins with an introduction to innate and adaptive immunity. I then provide an overview of inflammasome activation, followed by a discussion of what is currently known about how inflammasomes influence adaptive immunity. This section discusses the roles inflammasome-driven lytic cell death (termed pyroptosis) might play in antigen release, evidence for inflammasome activation driving CD4+ and CD8+ T cell responses, and instances where inflammasome activation appears to inhibit adaptive immunity. This chapter closes with evidence for inflammasomes influencing adaptive immunity in vaccines, anti-tumor immunity, and autoimmunity. Overall, Chapter One provides a foundation for appreciating why we need better understanding of the role inflammasome activation plays in driving adaptive immune responses. In Chapter Two, I present my early doctoral work, where I began to explore what role(s) NAIP–NLRC4 inflammasome activation might have on adaptive T cell immunity. Here I introduce the OvaFla mouse model previously, which was described by the Vance lab. These mice use the Cre-Lox system to inducibly express a fusion protein containing Ova antigen and the 166 amino acid C-terminal of flagellin, which will activate NAIP–NLRC4 but not an alternative flagellin sensor called TLR5. For experiments in this chapter, I crossed these “OvaFla” mice with mice containing a tamoxifen-inducible Cre driver, Cre-ERT2, that results in systemic OvaFla expression following tamoxifen administration. I found that systemic OvaFla can drive cross priming of CD8+ T cells in both WT and NLRC4-deficient mice. However, because Cre-ERT2 is expressed throughout the mouse, we remain unsure where and how this cross priming is occurring. I did determine, however, that signaling through the IL-18R on cross presenting cells is not required for CD8+ T cell activation. One potential benefit to the OvaFla Cre-ERT2 system is that localized tamoxifen application can be used to drive a more focused Cre expression. Additionally, bone marrow-derived cells from these mice retain the ability to activate OvaFla, which may be useful for future in vitro studies. In all, the work presented in this chapter provides some initial insights into the OvaFla Cre-ERT2 system, with suggestions on how they may be a useful tool for others. Chapter Three describes the bulk of my doctoral work, which specifically focused on activation of the NAIP–NLRC4 inflammasome in intestinal epithelial cells (IECs). NAIP–NLRC4 activation in these cells results in IEC pyroptosis, followed by an expulsion of the IEC into the intestinal lumen. One of my original hypotheses was that pyroptosis, which is mediated by the pore-forming protein Gasdermin D, provides an opportunity for cytosolic antigen to escape into the underlying lamina propria. In the lamina propria, the antigen is theoretically available to be cross-presented on dendritic cells (DCs), which can then drive antigen-specific CD8+ T cell activation. To test this hypothesis, I crossed the OvaFla mice with the Villin-Cre-ERT2 mice, thereby creating animals where tamoxifen administration results in robust OvaFla expression in IECs. These OvaFla Villin-Cre-ERT2 mice were crossed onto Gasdermin D-, ASC-, and NLRC4-deficient backgrounds. In support of my hypothesis, my work showed that IEC-derived antigens can be cross presented to CD8+ T cells in vivo, but we were surprised to find that this cross presentation occurred in both WT and NLRC4-deficient OvaFla mice. Additionally, Gasdermin D-mediated pyroptosis played only a partial role in CD8+ T cell cross-priming. My project then shifted to understanding whether there were any mechanistic differences in antigen cross-presentation between inflammatory conditions (NLRC4-dependent) and steady state (NLRC4-independent). By using two separate genetic knockout mouse lines, I found that cross presentation of IEC antigens during non-inflammatory conditions (in NLRC4-deficient mice) relies on a subset of classical DCs (cDCs) that require the Batf3 transcription factor (cDC1s)—these findings align with previously published data. However, in the presence of inflammasome activation, a Batf3-independent cDC population (likely cDC2s) can cross present IEC-derived antigen. Altogether, these data provide a better understanding of the complex interactions between IECs, DCs, and CD8+ T cells in the gut. In Chapter Four, I close my dissertation with a discussion on some of the remaining questions generated from my work. These questions center around the mechanism(s) of antigen acquisition functional maturation of the cross presenting cDCs. In all, my dissertation work has provided a stripped-down approach to understanding how inflammasome activation influences adaptive immunity. Unlike previous studies that rely on infection models, the OvaFla system allowed me to selectively activate the NAIP–NLRC4 inflammasome and uncover a Batf3+ cDC1-independent pathway of IEC antigen cross presentation.

The innate immune system can detect the presence of microbial infection by employing germline-encoded receptors specific for conserved microbial ligands. For example, the extracellular presence of the bacterial protein flagellin is sensed by Toll-like receptor 5, leading to activation of the transcription factor NF-κB, cytokine expression, and immune responses. In addition, certain flagellated bacterial pathogens, also appear to be sensed by an inflammasome. The inflammasome is a cytosolic molecular complex that contains a host Nod-like protein, Ipaf, and triggers caspase-1 and macrophage death. We used the intracellular pathogen Legionella pneumophila as a model intracellular bacterium to study the activation of the Ipaf inflammasome. Macrophage resistance to Legionella growth is entirely dependent upon the host expression of Ipaf and the bacterial expression of flagellin making it a useful tool to probe this pathway.

Using a retroviral expression vector, we found that the cytosolic expression of bacterial flagellin, in the absence of other bacterial contaminants and virulence factors was sufficient to trigger macrophage cell death. Through the use of genetic knock-out macrophages we found that this cell death is dependent upon caspase-1 and Ipaf. Taken together these results indicate that flagellin expressed in the cytosol is sufficient to activate the Ipaf inflammasome. We then mapped the region of flagellin required to activate Ipaf to the highly conserved C-terminus of flagellin. This finding indicates that the region of flagellin recognized by Ipaf is distinct from the region recognized by Toll-like receptor 5. Indeed, when the residues recognized by Toll-like receptor 5 are mutated the flagellin is still capable of activating Ipaf. The minimal region of flagellin that is required for sensing by Ipaf is the C-terminal 35 amino acids including three conserved C-terminal leucine residues that are critcal for sensing. Interestingly, activation of the inflammasome in response to this minimal peptide also requires the host protein Naip5, a unique member of the Nod-like and inhibitor of apoptosis (IAP) gene families.

Consistent with this result, the restriction of L. pneumophila growth within macrophages is dependent upon both Naip5 and Ipaf. However, Naip5 was dispensable for the response to other Ipaf-dependent stimuli including retroviral transduction of full-length flagellin. Activation of caspase-1 in response to S. typhimurium infection is also only partially dependent on Naip5. By expressing Salmonella flagellin in L. pneumophila we were able to show that it is not the type of flagellin, but rather the mode of delivery that results in a Naip5-dependent or -independent response. Indeed, Salmonella expresses an Ipaf activator, prgJ, that is Naip5 independent, explaining why some of the inflammasome activation in response to Salmonella is Naip5 independent. We have also determined that the N-terminus of flagellin plays a critical role in determining the differential roles of Naip5 and Ipaf in flagellin-recognition. Our results provide a molecular framework for understanding the cytosolic recognition of flagellin by host cells.

The innate immune system responds to infectious threats by detecting specific molecular structures conserved among microbes, such as bacterial lipopolysaccharide or flagellin. However, these conserved molecules are found on harmless and pathogenic microbes alike. In order to discriminate between harmful and harmless microbes, it has been proposed that the innate immune system may also sense `patterns of pathogenesis', the disruptions to host physiology orchestrated specifically by pathogens to infect, replicate within, and spread among their hosts. Immune recognition in plants is known to be based in part upon recognition of specific pathogen-associated activities, but few analogous examples have been described in mammals. The intracellular bacterial pathogen Legionella pneumophila can infect macrophages in the mammalian lung, causing a severe inflammatory pneumonia called Legionnaires' Disease. For virulence, L. pneumophila requires a Dot/Icm Type IV secretion system that translocates bacterial effectors to the host cytosol. In this dissertation, L. pneumophila was used as a tool to reveal two novel immunosurveillance mechanisms that can discriminate between virulent and avirulent bacteria. The two distinct pathways integrate detection of both microbial molecules and pathogen-associated activities to generate specific responses to Dot/Icm+ L. pneumophila. The first of these novel mechanisms leads to a potent transcriptional response, termed the `Effector-Triggered Response' (ETR), in macrophages infected with virulent L. pneumophila, but not an avirulent Dot/Icm- mutant. I demonstrate that this unique transcriptional response is due to secretion of five bacterial effector molecules that inhibit host protein synthesis. Upon infection of macrophages with Dot/Icm+ L. pneumophila, these five effectors caused a global decrease in host translation, thereby preventing synthesis of IkB, an inhibitor of the NF-kB transcription factor. Thus, macrophages infected with wildtype L. pneumophila exhibited

prolonged activation of NF-kB, which was associated with transcription of ETR target genes such as Il23a and Csf2. L. pneumophila mutants lacking the five effectors still activated TLRs and NF-kB, but because the mutants permitted normal IkB synthesis, NF-kB activation was more transient and was not sufficient to fully induce the ETR. Translation inhibition also activated other host pathways, including MAP kinase signaling. L. pneumophila mutants expressing enzymatically inactive effectors were also unable to fully induce the ETR, whereas multiple compounds or bacterial toxins that inhibit host protein synthesis via distinct mechanisms recapitulated the ETR when administered with TLR ligands. Thus, a a pathogen-encoded activity, namely translation inhibition, can elicit a specific immune response, both in cultured macrophages and in vivo.

The second novel mechanism consists of two different TNF-inducible inflammasomes that initiate an inflammatory host cell death, called pyroptosis, in macrophages infected with virulent L. pneumophila but not with an avirulent Dot/Icm- mutant. One of these inflammasomes begins with the previously reported detection of bacterial flagellin by the host proteins Naip5 and Nlrc4, but then leads to the activation of a novel downstream `death effector'. The other inflammasome involves activation of the protease Caspase-11 by cIAP1, a host protein that has not previously been implicated in inflammatory death. This latter form of cell death is antagonized by a bacterial effector, SdhA, which is required for growth of L. pneumophila within macrophages. The data presented here are consistent with a model in which increased host cell death upon infection with L. pneumophila aids in restriction of bacterial growth within these macrophages. The activity of these novel inflammasomes may explain the long-standing observation that TNF is crucial for complete restriction of L. pneumophila growth in macrophages.

Previous studies have demonstrated that the host response to bacterial infection is induced primarily by specific microbial molecules that activate TLRs or cytosolic pattern recognition receptors. Our results add to this model by providing several striking illustration of how the host immune response to a virulent pathogen can also be shaped by pathogen-encoded activities, such as (1) inhibition of host protein synthesis and (2) delivery of ligands to the

cytosol via specialized secretion systems. Elucidation of these immunosurveillance pathways increases our understanding of how the innate immune system may integrate multiple signals to sense a microbe, determine whether that microbe is a pathogen, and finally generate an appropriately scaled response.

Inflammasomes are cytosolic protein complexes that serve as platforms for the recruitment and activation of the pro-inflammatory CASPASE-1 protease (CASP1). CASP1 activation leads to processing and maturation of the cytokines interleukin-1b and -18, and a lytic form of cell death termed pyroptosis. Inflammasome assembly is initiated by cytosolic proteins in response to microbial infections, and many of these sensor proteins belong to the Nucleotide-binding domain, Leucine-rich Repeat containing protein (NLR) family. NLRP1 (NLR family, Pyrin domain containing 1) was the first NLR described to form an inflammasome, but until recently, its mechanism of activation and physiological functions in host defense have remained unclear.

In Chapter 1 we extensively review the literature on the proposed mechanisms of NLRP1 activation in humans and rodents. We discuss the activation of NLRP1 by various stimuli, including Bacillus anthracis Lethal Toxin, Toxoplasma gondii, muramyl dipeptide (MDP), and host intracellular ATP depletion. The role NLRP1 plays in pathogen recognition and resistance during infection is also discussed, as is the regulation of NLRP1 by host and viral proteins. We also discuss the unexpected differences in the mechanism of NLRP1 inflammasome activation as compared to the activation of other inflammasomes, such as the NAIP/NLRC4 inflammasomes.

In Chapter 2 we cover our discovery regarding the mechanism of mouse NLRP1B activation by Lethal Toxin (LeTx), which is composed of Lethal Factor (LF) and Protective Antigen (PA). We made the critical observation that LF cleaves NLRP1B directly near the N- terminus. We then demonstrated that LF cleavage of NLRP1B is required for NLRP1B activation and inflammasome formation. Most importantly, we were able to show that proteolysis is sufficient to activate NLRP1B when we replaced the activity of LF with a Tobacco Etch Virus (TEV) protease. We conclude by proposing that NLRP1B has evolved to respond to other proteases derived from other pathogens.

In Chapter 3 we test the hypothesis that other variants of NLRP1 expressed in mice and human are also activated by proteolysis. We first demonstrate that another allele of NLRP1B, which has no known agonist, is able to respond to cleavage and induce inflammasome formation. We then demonstrate that this activity can also be seen with the NLRP1A paralog. Lastly, we extend this work by analyzing the human ortholog of NLRP1. Surprisingly, proteolysis is also a conserved mechanism of activating human NLRP1. Collectively, these results suggest that NLRP1 might be broadly conserved as a protease sensor in mammals, and provide an important host defense mechanism.

In Chapter 4 we report on the development of a new method of analyzing CASP1 activation. We generated a CASP1 dimerization reporter that generates the fluorescent Venus protein when an inflammasome is activated and does not induce cell death. We generated a cell line that stably expresses the reporter, NLRP1B, and ASC. This line responds to LeTx, and can be analysed by flow cytometry and live cell microscopy. This reporter also works in macrophage-like cells that endogenously express inflammasome components. As a proof of principle we demonstrate that cells with an activated inflammasome can be enriched by FACS. We believe this reporter is an ideal tool for the discovery of novel positive and negative regulators of inflammasomes. The largest advantage of the reporter is that it does not induce cell death. This feature allows for the recovery of cells from a complex mixture that would be amenable for a high thorough put screen.

Shigella species, the causative agents of bacillary dysentery or shigellosis, infect more than 250 million people each year and are a major driver of diarrheal associated morbidity and mortality world-wide. Shigella induces severe gastrointestinal disease by colonizing and disseminating within the epithelial lining of the human colon and rectum, events that drive significant inflammation in the gut. Despite this heavy disease burden, little is known about the molecular determinants of pathogenesis during infection and the adaptive immune correlates of protection up re-infection, due largely to the lack of an inexpensive, tractable, and physiologically relevant mammalian model of shigellosis. Indeed, a detailed understanding of the immune response to this pathogen is critical for the development of vaccines to Shigella, none of which has been effective enough at eliciting protection to be licensed for use in humans. Mice are an ideal system to model human specific disease because of their short generation time, the relatively low cost and effort to maintain a mouse colony for research, and the extensive immunological and genetic tools that have been established in this organism over the past half-century. Wild-type laboratory mice, however, are intrinsically resistant to high doses of oral Shigella challenge and the genetic and mechanistic underpinnings of this resistance have remained largely unknown. My dissertation examines the host innate immune factors in mice that dictate resistance to oral Shigella infection. In the first chapter I review what is currently known about Shigella pathogenesis, discuss the animal models that have been used to establish this understanding, and provide background on innate immune pathways in humans and mice that sense Shigella. In the second chapter I describe the genetic basis of resistance to Shigella infection in mice and show that NAIP–NLRC4 inflammasome-deficient mice are susceptible to Shigella infection and serve as a physiologically relevant model of dysentery. In the third chapter, I use this model to expand our understanding of innate immune resistance to Shigella infection and show that a redundant, layered hierarchy of cell death in intestinal epithelial cells is key to this resistance. Finally, in the fourth and final chapter, I describe how our findings advance the field and comment on how our new model of Shigella infection can be leveraged to learn more about the immune response to this deadly intestinal pathogen.

Initial detection of invading microorganisms is one of the primary tasks of the innate immune system. However, the molecular mechanisms by which pathogens are recognized remain incompletely understood. I used the intracellular gram-negative Legionella pneumophila to study mechanisms by which the innate immune system distinguishes virulent bacteria from avirulent bacteria. I have made the surprising observation that a cytosolic RNA immunosurveillance pathway (called the RIG-I/MDA5 pathway), thought primarily to detect viruses, is also involved in the innate immune response to the intracellular vacuolar bacterial pathogen, Legionella pneumophila. In the response to viruses, the RIG-I/MDA5 immunosurveillance pathway has been shown to respond to viral RNA or DNA. We found that the RIG-I pathway was required for the response to L. pneumophila RNA, but was not required for the response to L. pneumophila DNA. Thus one explanation of my results is that L. pneumophila RNA accesses the host cell cytosol via its type IV secretion system, where it triggers the RIG-I/MDA5 pathway. This is unexpected since bacteria have not previously been thought to translocate RNA into host cells. I was able to isolate IFN-stimulatory activity by immunoprecipitating RIG-I from cells infected with T4SS-competent Legionella. In the future, I will utilize deep sequencing technology to pinpoint the origin and identity of RIG-I bound ligands during T4SS+ L. pneumophila infection.

I also found that L. pneumophila suppresses the RIG-I/MDA5 pathway by a translocated effector protein, SdhA. Several viral repressors of the RIG-I/MDA5 pathway have been described, but bacterial repressors of RIG-I/MDA5 are not known. Thus, this study provides novel insights into the molecular mechanisms by which the immune system detects bacterial infection, and conversely, by which bacteria suppress innate immune responses.

While all bacteria are capable of inducing type I interferon, many species do so independently of the cytosolic RNA sensing pathway that responds to Legionella. Therefore, I, along with many Vance lab members, investigated the mechanism by which cyclic dinucleotides (c-di-GMP and c-di-AMP), bacterial second-messenger molecules, activate a robust and specific host response in macrophages. c-di-GMP has been shown to activate TBK-1, IRF3, NF-κB, and MAP kinases to induce type I interferon, in manner independently of known TLR and cytosolic nucleic acid sensing pathways. In parallel studies in the lab, ENU mutagenesis of mice identified a mouse mutant that completely abrogates the host response to cyclic dinucleotides. Sequencing identified a missense mutation in the open reading frame of Sting, which converts an isoleucine to aspargine in the C-terminal globular domain, rendering STING protein undetectable in mutant macrophages. Previous reports of Sting demonstrated a role for this multiple transmembrane domain containing protein that localizes to the endoplasmic reticulum and/or mitochondrial associated-membrane (MAM) in cytosolic DNA and RNA sensing pathways. I found that overexpression of the ENU-induced mutant Sting allele failed to induce type I interferon, despite robust expression of the protein. Surprisingly, studies in the Vance lab have shown that wild type STING is capable of binding c-di-GMP, in contrast to the ENU mutant allele, which does not bind. I found that a soluble C-terminal truncation of STING (amino acids 138-379), which removes most predicted transmembrane domains, is sufficient to bind c-di-GMP. Taken together, genetic studies have demonstrated an essential role for Sting in the innate immune response to cyclic dinucleotides and biochemical data shows that the C-terminal region of the protein functions as the direct sensor of cyclic dinucleotides.

The experimental line of investigation presented in my thesis dissects the pathways by which the innate immune system recognizes infection of virulent bacteria. Both stories discussed herein demonstrate the importance of innate immune detection of nucleic acids; molecules microbes cannot live without. Interestingly, it is the compartment in which nucleic acids are present that ultimately triggers innate immune receptors. The demonstration that cytosolic RNA sensors detect secretion system competent Legionella illustrates the consequence of breaching the phagosomal barrier. While bacterial replication may be restricted to membrane bound compartments, accessing the host cell cytosol via the T4SS alerts host sensors to the pathogen's presence. Previously thought only to recognize viral infection, I have demonstrated the breadth of cytosolic RNA sensors in recognizing not only viruses, but also bacteria. STING's ability to detect cyclic dinucleotides delivered to the host cell cytosol exemplifies how the innate immune system has honed in on a uniquely bacterial nucleic acid molecule. Cyclic dinucleotides are not only structurally distinct, but their role in regulating virulence factor expression makes them an excellent target for innate immune detection of pathogens.