Washington University Medical Center
Washington University St. Louis: Open ScholarshipNot a member yet
26344 research outputs found
Sort by
Towards Graph Foundation Models: Few-shot and Zero-shot Learning on Graphs
No full textGraphs naturally model complex relationships and interactions across various domains, including social networks, biological systems, and recommender platforms. Graph Neural Networks (GNNs) have emerged as powerful tools for learning effective graph representations through iterative message passing, significantly improving performance in tasks such as node classification, link prediction, and graph classification. However, the success of GNNs largely depends on abundant labeled data, posing challenges in practical scenarios where labeled data is scarce or unavailable. This dissertation addresses these challenges by exploring few-shot and zero-shot learning within the graph domain. We first propose COLA, a self-supervised few-shot node classification method that exploits unlabeled graph structures. Next, we introduce OneForAll (OFA), a unified graph foundation model capable of training across multiple datasets to improve few-shot generalization. Finally, we propose Generative OneForAll (GOFA), a generative foundation model integrating GNNs with large language model (LLM) architectures, specifically designed for zero-shot scenarios. Together, these contributions provide a principled path toward versatile and broadly applicable graph learning systems capable of operating effectively under minimal or no supervision
Special-Purpose Governments
Full text linkWhen one thinks of government, what comes to mind are familiar general-purpose entities like states, counties, and cities. But more than half of the 90,000 governments in the United States are strikingly different: They are “special-purpose” governments that do one thing, such as supply water, fight fire, or pick up the trash. These entities have expanded far more rapidly than any other form of government. Yet they remain understudied, and they present at least two puzzles. First, special-purpose governments are difficult to distinguish from entities that are typically regarded as business organizations—such as consumer cooperatives—and thus underscore the nebulous border between “public” and “private” enterprise. Where does that border lie? Second, special-purpose governments almost always provide exactly one service, in sharp contrast to general-purpose governments. There is little in between the two poles—such as two-, three-, or four-purpose governments. Why?
This Article answers those questions—and, in so doing, offers a general theory of special purpose government. The fundamental difference between public and private enterprise exists not in the services provided or even the governance structure, but in how they are created: Governments can compel membership and financial contributions in a given territory from the moment they are formed—in contrast to “private” cooperatives, which contract for membership and accept funds in exchange for the provision of services. Moreover, special-purpose governments typically offer one service because of the efficiency benefits of having “owners” with homogenous interests. Just as private agricultural cooperatives tend to involve a single crop—ensuring that controlling members will have convergent interests—special-purpose governments work best when they provide a single service for which all the members share common incentives. As a result, special-purpose governments have much in common with some private firms. And yet, despite these broad similarities, the patchwork of laws governing special-purpose governments has not kept pace with the evolution of organizational law in “private” contexts. Private law may thus suggest reforms to let special-purpose governments achieve their unfulfilled potential
Empirically Exploring the Physical Realizability of Adversarial Examples
Full text linkThe development of autonomous vehicles (AVs) has been accelerated by advancements in deep neural networks (DNNs), which power the complex perception systems necessary for safe and efficient real-world navigation. However, as AVs increasingly integrate into public transportation networks, the robustness of their perception systems against potential vulnerabilities is critical. Among these threats, adversarial attacks—particularly through the use of adversarial patches—pose significant risks. These patches are carefully crafted perturbations designed to mislead DNNs, potentially compromising AV safety by causing incorrect object recognition or misclassification.
While extensive research has demonstrated high attack success rates for adversarial patches in controlled digital environments, their performance under practical conditions remains underexplored. This gap is noteworthy because real-world environments introduce variability such as changes in lighting, object angles, and material textures, which could influence the practical applicability of these attacks. Furthermore, most existing defense mechanisms have been evaluated primarily in digital domains, leaving their robustness in real-world conditions largely unexplored. To address these gaps, we empirically evaluated the performance of adversarial patches through extensive physical experiments, following a systematic approach across diverse real-world environments.
To achieve this, we began by training adversarial patches using methodologies from previous works to establish a strong baseline and validate their effectiveness under idealized digital conditions. Following this, we printed the patches and applied them to real-world objects to assess their performance under varying physical conditions. Experiments were conducted with different types of vehicles, including SUVs and sedans, in diverse settings: outdoor environments during the day, outdoor environments at night, and indoor parking lots. This approach enabled us to evaluate the robustness of adversarial patches under conditions resembling real-world AV environments.
Our findings indicate that while adversarial patches achieve high attack success rates in controlled digital settings, their effectiveness is notably reduced in real-world environments due to environmental variability. This highlights the critical challenge of bridging the gap between theoretical vulnerability studies and practical adversarial threats in AV systems. By identifying key factors such as lighting conditions, object angles, and material textures that influence the success of adversarial attacks, we offer empirical insights that can guide efforts to improve AV perception system resilience. Furthermore, our results underscore the pressing need for robust defense mechanisms that account for real-world complexities, as current defenses largely focus on digital domains and may not adequately address real-world vulnerabilities. This study contributes to the growing body of knowledge on adversarial machine learning and lays a foundation for future research to enhance the security and robustness of AV technologies in practical settings
Barriers and Breakthroughs in the Evolution of Aggregative Multicellularity
Full text linkThe evolution of multicellular organization represents a key innovation in the history of life, yet fundamental questions remain about the evolutionary mechanisms that allowed some lineages to achieve remarkable complexity, while constrained others in their elaboration. The widely accepted single-cell bottleneck model provides a compelling explanation for the success of clonal multicellularity by reducing internal conflict and facilitating developmental integration. However, this perspective does not fully account for the stability and repeated evolution of aggregative multicellularity, where cooperation emerges among previously independent cells that may not be genetic relatives. Aggregative multicellularity demonstrates that the evolution of multicellularity is not a singular process and challenges our understanding of how multicellular complexity evolves and persists. These systems provide unique opportunities to examine how evolutionary and developmental processes intersect to enable major transitions in biological organization. My research focuses on an integrative analysis of aggregative multicellularity combining social evolution theory, experimental manipulation, comparative phylogenomics and transcriptomics. Using the social amoeba Dictyostelium discoideum as a model system, I examine three key aspects of aggregative multicellularity: mechanisms maintaining group integrity, developmental coordination between cells, and evolutionary constraints on developmental innovation. I demonstrate how collective behavior can emerge and persist despite potential genetic conflicts in the evolution of aggregative systems. First, through a comprehensive synthesis of empirical evidence from D. discoideum and the bacterium Myxococcus xanthus, I demonstrate how aggregative organisms employ multiple complementary mechanisms to maintain cooperation, including population structure, kin discrimination, and pleiotropic constraints on cheating. This analysis reveals that while aggregative multicellularity faces unique challenges compared to clonal development, robust solutions have evolved to maintain group cohesion. Second, I experimentally tested whether mixing genetically distinct cells impairs developmental coordination independent of social conflicts. Using experimentally evolved lines of D. discoideum that were not selected for social traits, I found that while basic coordination of slug migration remained intact in chimeric mixtures, there was a significant failure in coordinating migration with spore production. This provides the first direct evidence that developmental incompatibilities, separate from social conflicts, can emerge between divergent cell lines. Third, through phylogenomic and transcriptomic analyses spanning the developmental and evolutionary history of all protein coding genes in D. discoideum, I showed that early developmental stages are dominated by ancient genes while novel genes are predominantly expressed later in development. This pattern supports von Baer\u27s law of early developmental conservation and reveals how evolutionary innovations become integrated into aggregative development. Together, these findings demonstrate that aggregative multicellularity represents a distinct but viable evolutionary strategy that has solved fundamental challenges of collective organization through multiple mechanisms operating at different scales. This work advances our understanding of major evolutionary transitions and highlights how different routes to multicellularity navigate universal constraints while finding unique solutions
JARID2-mediated Reprogramming of Human Hematopoietic Progenitors for Stem Cell Transplantation
Full text linkHematopoietic stem cell transplantation (HSCT) is a well-established treatment for various hematological disorders. Umbilical cord blood (UCB) has emerged as an alternative source of hematopoietic stem and progenitor cells (HSPCs). However, the limited number of functional repopulating cells in UCB units has encouraged researchers to develop methods to expand them ex vivo. Hematopoiesis is strictly regulated by epigenetic modifiers. JARID2 is a co-factor of the Polycomb Repressive Complex 2 (PRC2), which recruits the complex to specific genomic loci and promotes the deposition of Histone 3 lysine 27 trimethylation (H3K27me3), a repressive epigenetic mark that mediates condensed chromatin packing to silence key developmental genes. In our previous studies, we found that mouse multipotent progenitor (MPP) cells with genetic loss of Jarid2 regained self-renewal capacity, unlike typical transient MPPs. These reprogrammed MPPs displayed transcriptional and epigenetic profiles more similar to hematopoietic stem cells (HSCs) than control MPPs, suggesting a role for Jarid2 in the developmental transition from HSCs to MPPs. Based on this, we hypothesized that human HSPCs may follow a similar mechanism and could be expanded ex vivo through JARID2 inhibition to improve outcomes in bone marrow transplantation. To investigate whether inhibition of JARID2 can enhance the functional potential of human HSPCs, we designed lentiviral vectors to achieve JARID2 knockdown using shRNAs in both constitutive and inducible systems. Our results showed that constitutive loss of JARID2 in human CD34+ HSPCs led to improved overall engraftment outcomes and an increased number of HSPCs in a xenograft NSG mouse model. JARID2-knockdown (JARID2-KD) HSCs also demonstrated enhanced functionality, reflected by increased self-renewal and multilineage reconstitution capacity in both primary and, more prominently, secondary transplantation experiments. We further found that transient JARID2 inhibition ex vivo exerted long-term positive effects on transplantation outcomes, suggesting the clinical potential of short-term JARID2 inhibition to improve bone marrow transplantation. Notably, one experiment showed that engraftment was enhanced only after in vivo doxycycline administration to induce JARID2 knockdown, indicating that JARID2-KD-mediated HSPC expansion and transplantation benefits can be controlled in vivo—further supporting its clinical relevance. Safety is a critical concern when considering the transplantation of expanded human HSPCs into patients. To assess the safety of JARID2 loss of function in the context of improved bone marrow transplantation, we conducted a long-term study in NSG mice. Human HSPCs with permanent JARID2 knockout via CRISPR-Cas9 were transplanted into NSG mice and monitored for one year. Histological and pathological analyses revealed no signs of abnormal development or fibrosis in the blood, spleen, or bone marrow. Additionally, we performed single-cell RNA sequencing and flow cytometry-based clustering on hCD34+ cells isolated from 20-week post-transplant bone marrow. UMAP and t-SNE analyses showed comparable cell population compositions between control and JARID2-KD groups. These results indicate that long-term JARID2 loss of function does not induce pathological changes in human HSPCs, further supporting its clinical applicability. To elucidate the cellular mechanisms underlying JARID2-mediated enhancement of bone marrow transplantation, we sorted HSCs, MPPs, and MLPs and cultured them in MethoCult assays. As expected, JARID2-KD HSCs exhibited increased colony-forming capacity compared to control HSCs. Interestingly, JARID2-KD MPPs also demonstrated enhanced colony-forming ability, with this effect persisting through two additional rounds of replating—indicating a gain in self-renewal capacity comparable to that of HSCs. These findings suggest that JARID2 knockdown enhances the functional potential of both HSCs and MPPs. Since ex vivo culture significantly alters cell surface marker expression, we aimed to identify the appropriate cell population for downstream sequencing and epigenetic analyses. Using CITE-seq, we identified CD34⁺CD90⁺EPCR⁺CD49f⁻CD71⁻ cells as the functional repopulating population after 8 days of ex vivo culture. Transplantation of 1,000 cells with this marker profile into NSG mice resulted in robust engraftment and full multilineage differentiation in JARID2-KD groups. These results suggest that JARID2 knockdown not only enhances functional capacity but also helps preserve HSC identity during ex vivo culture. While JARID2 functions as a co-factor of PRC2 by guiding the complex to specific genomic loci, EZH2 serves as the core enzymatic subunit responsible for catalyzing H3K27 trimethylation. To further explore the molecular mechanism underlying JARID2-mediated enhancement of HSPC function—and to determine whether its effects are dependent on PRC2—we conducted parallel experiments using EZH2 knockdown (KD). EZH2 shRNAs were cloned into the same lentiviral vectors used for JARID2 inhibition. In contrast to JARID2-tKD UCBs, EZH2-tKD significantly impaired transplantation outcomes, as evidenced by reduced peripheral blood engraftment and decreased HSPC numbers in the bone marrow at 20 weeks post-transplant. Notably, EZH2-KD engraftment started to decline only after in vivo doxycycline-induced shRNA expression, indicating a direct detrimental effect. Moreover, EZH2-KD HSPCs exhibited abnormal lineage distribution in xenograft models, suggesting potential pathogenic alterations. At the epigenetic level, EZH2-KD led to a marked reduction in global H3K27me3 levels, while JARID2-KD had minimal impact, highlighting distinct and non-redundant roles for JARID2 and EZH2 within the PRC2 complex. In conclusion, our findings suggest that JARID2, rather than EZH2, is a promising target for enhancing hematopoietic stem cell expansion and improving bone marrow transplantation outcomes. In the future, the development of small-molecule inhibitors targeting JARID2 may offer a novel strategy to optimize current ex vivo expansion and transplantation protocols for clinical applications
Nanoengineered Constructs for Immunomodulatory Therapeutic Delivery
Full text linkImmune processes are central to aspects of human health ranging from wound healing to pathogen recognition, yet even within the field of drug delivery, immunomodulatory therapeutics remain relatively niche. In this work we apply nanoscale peptide, lipid and polymer engineering principles to address present challenges in immunomodulatory therapeutics delivery, particularly in the treatment of non-healing wounds and vaccination against flaviviruses and other pathogens that exhibit antibody dependent enhancement. In both cases, nanoengineered delivery systems allowed us to overcome challenges associated with immune hyperactivation using targeted immune stimuli rather than nonspecific signals. A frequent driver of non-healing wounds is failure of local macrophages to “turn over” from a pro-inflammatory to a pro-proliferative phenotype. Although alginate based wound dressings have many favorable properties for wound healing, they are canonically crosslinked by Ca2+ ions, which can be potently pro-inflammatory if they leach into the extracellular wound environment. By developing two alginate sols, each composed of alginate polymer conjugated to several peptides of one member of a heterodimeric coiled coil pair, we were able to synthesize a self-supporting, self-healing hydrogel by mixing those sols. This coiled coil crosslinked gelinduced significantly lower macrophage driven inflammation than Ca2+ crosslinked alginate gels both in ex vivo culture of mouse peritoneal macrophages and after in vivo intraperitoneal implantation. Flaviviruses and other diseases that exhibit antibody-dependent enhancement are difficult to vaccinate against because immunization with the whole inactivated or attenuated pathogen can generate cross reactive and non-neutralizing antibodies to other pathogens which can make infections with those other pathogens far more lethal. To avoid this outcome, vaccines for such pathogens must contain only specific, highly-immunodominant, poorly-conserved epitopes so as to avoid antibody cross-reactivity. This presents difficulties with adjuvation as isolated peptides or epitopes have generally poor intrinsic immunogenicity. To address this concern, we developed nanocarriers that can localize specifically to plasmacytoid dendritic cells (pDCs), a cell population specialized in production of type-I interferons. To do this we capitalized on the homogenously high expression of CD71, a ubiquitous endocytosing iron transporter, by pDCs. By synthesizing nanoscale liposomes surface coated with poly(ethyleneglycol) and either gambogic acid or T7 as small molecule or peptide noncompetitive CD71 ligands respectively we were able to selectively deliver liposomes to pDCs over other leukocytes both ex vivo and in vivo. Further, we demonstrated that these liposomes effectively home to secondary lymphoid organs after subcutaneous administration and elicit type-I interferon stimulation through and IRF-7 dependent pathway when used to deliver ODN 2216 as a TLR9 agonist. Collectively, these developments contribute to a growing corpus of nanoengineered constructs for immunomodulation and may pave the way for enhanced wound healing technology, treatments for autoimmune disorders, vaccine and drug delivery and treatments for cancer
Investigating neural, neurovascular, and vascular function: insights from imaging 5-HT2AR compounds, optogenetic motor mapping, and cross-modal functional mapping
Full text linkUnderstanding how brain activity relates to behavior and cognition remains a central challenge in neuroscience. Brain function emerges from interactions between electrical, chemical, and blood dynamics across multiple spatial and temporal scales. Capturing this complexity requires methods that assess both large-scale neurophysiological activity and precise, cell-type-specific processes. Widefield optical imaging (WFOI) enables high-resolution monitoring of cortical function by simultaneously measuring hemodynamic and calcium dynamics. WFOI is particularly valuable for studying psychedelics, which show therapeutic potential but may alter neurovascular coupling, complicating the interpretation of hemodynamic signals. Using WFOI, we examined the hallucinogenic 5-HT2A receptor agonist DOI in awake mice, revealing dissociations between calcium and hemodynamic signals that were reversible with 5-HT2A receptor antagonism. Optogenetics enables precise, cell-specific neuronal activation, offering a bottom-up approach to functional mapping. We developed a system to systematically photostimulate the cortex in awake mice while tracking movement in 3D, revealing a more spatially distributed motor representation than previously recognized. Importantly, we demonstrate a strong correspondence between resting-state functional connectivity, assessed via spontaneous activity with WFOI, and action mapping via systematic whole-hemisphere photostimulation. Integrating these approaches advances our understanding of brain function and its modulation by pharmacological and behavioral interventions
Pituitary Adenylate Cyclase Activating Polypeptide and Its Receptor as Novel Therapeutic Targets for Opioid Induced Hyperalgesia
No full textOpioids are commonly prescribed for the treatment of pain and headache disorders. However, chronic opioid use can result in a paradoxical increased sensitivity to pain as well as an extension of pain area known as opioid-induced hyperalgesia (OIH). One field OIH has been clearly identified is migraine, known as medication overuse headache (MOH). While opioids may provide acute relief, chronic use results in increased migraine pain severity and progression of migraine from an episodic to chronic state. The primary treatment is opioid cessation, which is difficult to implement as patients are reluctant to stop treatment they believe is helping them, and they may also experience opioid withdrawal. Opioid cessation has low retention rates, and one study reports a 50% relapse rate within the first year, which can ultimately feed into a pattern of opioid use disorder. There is thus a desperate need to find effective therapies for OIH, and this disorder would benefit greatly from a focused drug development strategy. In this thesis, I provide a greater understanding about the pathophysiology of OIH and MOH and characterize novel treatment options. Using an unbiased proteomic screen, our lab had previously identified the pituitary adenylate cyclase activating polypeptide (PACAP) as a possible target for both OIH and migraine. In Chapter 2 I showed that PACAP alone could induce allodynia at translationally significant doses. I further demonstrated efficacy of a delta opioid receptor (DOR) agonist, SNC80, in reducing PACAP-induced allodynia. I also elucidated the expression of DOR with PACAP and its receptor PAC1 using in situ hybridization and found high co-expression of DOR with the PAC1 receptor. Together, these results suggest that PACAP can cause allodynia, that this allodynia is blocked by DOR, and that DOR may act by inhibiting PAC1 signaling. Previous studies from our lab have demonstrated the effectiveness of a peptide PAC1 antagonist M65 in an OIH model. PACAPergic blockade was shown to decrease migraine clinically. Small molecule PAC1 receptor antagonists and antibodies have different properties to peptides and are worth investigating in our OIH/MOH models. In Chapter 3, I demonstrate that a small molecule PAC1 inhibitor could block established and prevent development of OIH. Additionally, a PACAP targeting antibody also blocked established OIH and prevented development of OIH and MOH. PACAP and the related peptide, vasoactive intestinal peptide (VIP), can both bind to VPAC1, VPAC2, and PAC1. PAC1 has a lower affinity for VIP than PACAP. PACAP may also activate the MAS-related GPR, member B2 (Mrgprb2) receptor. I determined the receptor and endogenous ligand expression of the PACAPergic, and mu and delta opioid receptors, with cell type resolution using in situ hybridization in the trigeminal nucleus caudalis (TNC). The results indicate that of all PACAP receptors, PAC1 had the highest total expression (~90% cells), followed by VPAC2 (~52%), VPAC1 (~45%), and MRGPRB2 (10%). Considering this high expression, approximately 97% of all mu opioid receptor positive cells co-expressed PAC1 in this region. These results suggest that the PAC1 receptor plays an important role in head pain processing regions. Previously our lab had shown upregulation of PACAP in the periaqueductal gray (PAG) in models of OIH and chronic migraine. In Chapter 4, I focused on the role of the PAG in OIH specifically. I determined that among PACAP receptors, PAC1 exhibited the highest overall expression at 66% of all cells, followed by VPAC2 (~44%), VPAC1 (~27%), and MRGPRB2 (15%). Notably, approximately 86% of all mu opioid receptor positive cells were found to also express PAC1. I used excitatory chemogenetic and shRNA knockdown approaches to characterize whether PAC1 in the ventrolateral periaqueductal gray (vlPAG) is necessary and sufficient to produce mechanical hypersensitivity, similar to OIH. Acute excitation of PAC1-expressing vlPAG neurons resulted in freezing behavior and analgesia, but chronic excitation produced sustained mechanical allodynia. We also used a shRNA approach to knockdown PAC1 receptors in the vlPAG. Compared to the control-OIH group, chronic morphine in animals with knockdown of PAC1 in the vlPAG developed OIH to a lesser extent and recovered more quickly. Together, our data suggests that cells expressing PAC1 in the vlPAG are sufficient to induce hyperalgesia and that the chronicity of OIH is partially dependent on activation of vlPAG PAC1 receptors. The results presented in this thesis provide a deeper understanding of the pathophysiology of OIH and MOH and identify DOR agonists, PAC1 antagonists, and PACAP antibodies as potential therapies for this disorder
Private Space, Public Self: Studios of the Avant-Garde in Fin-de-Siècle France
No full textFor many artists of late nineteenth-century France, particularly those of the avant-garde, the studio served as the physical environment over which they had the most agency, relative to public venues. Studios could function alternatively as a workshop, domestic refuge, social gathering spot, salesroom, and exhibition hall, thereby providing artists with a hybrid space where they could engage their audiences. Affiliated with the artist’s character and creative process, the studio’s polysemic nature granted its inhabitants the opportunity to shape its meaning. From the locale and architecture, to the use of light and color, to the objects on view or hidden, the elements of the studio offered each artist a medium to express their aesthetics, their values, and their identities. Such communication could occur by means of the direct encounters of visitors as well as indirectly by painted representations, graphic illustrations, photographs, and written accounts. Through a series of paired case studies of avant-garde artists’ studios in France circa 1880–1910, this dissertation examines how such environments were not merely sites of artistic production, but also artistic products in and of themselves. The case studies are the colorful decorative programs of James McNeill Whistler (American, 1834–1903) and Claude Monet (French, 1840–1926); the domestic salon-studios of Berthe Morisot (French, 1841–1895) and Odilon Redon (French, 1840–1916); and the multimedia, multicultural displays of Edgar Degas (French, 1834–1917) and Paul Gauguin (French, 1848–1903). Each chapter probes questions of artistic reputation, strategies of representation, and the particular creative milieu of the avant-garde. This project sheds light on these overlooked but crucial spaces in avant-garde production—in terms of both the material creation of art and the symbolic creation of persona. At stake is a richer understanding of how these artists navigated their careers and of the networks formed by these endeavors
Elucidating Pathways in Lung Monocytes, Macrophages, and Dendritic Cells that Impact Mtb Pathogenesis
No full textMycobacterium tuberculosis (Mtb) remains the top infectious cause of death globally, with an estimated 1.6 million deaths in 2021. About 10% of infected, immunocompetent adults will develop active disease, and the host factors that determine whether the immune response to infection is protective remain incompletely understood. Early events in innate immunity carried out by lung monocytes, macrophages, and dendritic cells are critical for the establishment of a protective immune response, and the pathways in these cells that contribute to protection are of interest to the field. This dissertation covers investigations of three such pathways, sphingolipid biosynthesis in alveolar macrophages and their precursors, BHLHE40-mediated transcriptional regulation of innate immune differentiation and polarization, and autophagy in lung macrophages and dendritic cells during innate-adaptive immune transition. We identified that SPTLC2, a subunit of the enzyme that catalyzes the first and rate-limiting step of sphingolipid synthesis, is required in innate immune cells during Mtb infection. We infected mice with an innate immune cell specific deletion of the gene that encodes SPTLC2 (Sptlc2����/����-Lysm- Cre) and found that they succumb to Mtb infection with a median survival of 93 days, much earlier than Sptlc2����/���� control mice. Infection of the Sptlc2����/����-Lysm-Cre mice is characterized by higher bacterial burden, the development of a striking glossy lung gross pathology, and transcriptional signatures of hypoxia relative to control mice, all of which were observed later than 28 days-post-infection (dpi). In contrast to many other mice with susceptibility to Mtb infection, there are no changes in cellular inflammation of lymphocytes, granulocytes, or dendritic cells that precede the loss of control of bacterial growth. However, beginning by 14 dpi, there is a significant reduction in the abundance of alveolar macrophages (AMs) in the lungs of Sptlc2����/����-Lysm-Cre mice relative to control mice. To investigate the mechanism of the loss of AMs, we performed single-cell RNA sequencing and noted no difference in markers of AM self-renewal or cell death between Sptlc2����/����- Lysm-Cre mice and controls. Instead, we observed a defect in the generation of AM precursors from recruited monocytes in the absence of SPTLC2. The glossy gross pathology, presence of hypoxia, and specific loss of AMs are reminiscent of pulmonary alveolar proteinosis, a disease in which surfactant proteins and lipids that are homeostatically digested by AMs accumulate in the alveolar space. We detected elevated levels of surfactant proteins A and D in the lungs of Sptlc2����/����-Lysm- Cre mice by 56 dpi, confirming the presence of proteinosis. In experiments to confirm that the role for SPTLC2 is intrinsic to the CD11c-expressing AMs, we found that Sptlc2����/����-Cd11c-Cre mice are similarly susceptible to Mtb infection and phenocopy the loss of AMs, glossy lung pathology, and proteinosis observed in the Sptlc2����/����-Lysm-Cre system. These studies establish that SPTLC2 is required in CD11c-expressing cells to maintain an AM population, avoid proteinosis, and survive during Mtb infection in mice. Mtb has been shown to use neutrophils as a replicative niche, and dendritic cells and macrophages both play an important role in initiating adaptive immunity and controlling bacterial replication upon adaptive immune help. We have previously shown that loss of BHLHE40, a transcription factor that is known to regulate pro- and anti-inflammatory immune responses, in CD11c+ cells leads to a neutrophil driven innate immune response characterized by a loss of dendritic cells during Mtb infection and results in early death in mice. Furthermore, BHLHE40 expression is driven by GM-CSF, a cytokine abundant in the lungs during infection, and in vitro studies using GM- CSF cultures revealed that loss of BHLHE40 skewed cytokine responses toward an alternatively activated phenotype upon stimulation with Mtb antigen. The cell types and pathways leading to these phenotypes remained unclear. We turned to single-cell RNA sequencing (scRNAseq) of the GM-CSF culture system to better understand the transcriptional pathways leading to these phenotypes. Using GM-CSF cultured bone marrow from BHLHE40 wild-type or knockout mice, we performed scRNAseq and revealed a previously unappreciated complexity of macrophages, dendritic cells, neutrophils, and a mix of common and cell-type specific precursors. Analysis of the merged data demonstrated a striking loss of DCs and DC precursors in the absence of BHLHE40. RNA velocity analysis revealed a disruption in the ability of DC precursors to differentiate into DCs associated with upregulation of C/EBP�� transcripts. C/EBP�� is a transcription factor that must be downregulated for differentiation into DCs. BHLHE40 binds directly upstream of the Cebpb locus, suggesting that BHLHE40 may directly repress C/EBP�� expression during DC differentiation. Additionally, there is a shift in the macrophage cluster in the absence of BHLHE40. Use of a scoring module for classically vs alternatively activated macrophages confirmed previous results that loss of BHLHE40 results in an enrichment for alternatively activated macrophages. The single-cell transcriptomic data revealed that this enrichment for alternative polarization is associated with decreased HIF-1�� expression. This study demonstrated the strength of scRNAseq as a tool to understand the relationship between developing cells in a complex mixture and identified the critical role for BHLHE40 in DC maturation and macrophage polarization. Autophagy, the pathway that allows cells to target unwanted intracellular content to the lysosome for nutrient recycling, has long been implicated in macrophages during infection. It was previously thought that the primary role for autophagy in control of mycobacterial infections was through targeting the pathogen to the lysosome for degradation. We previously showed that Atg5����/����-Lysm- Cre mice, which lack ATG5 and autophagy in LysM-expressing cells, succumb to Mtb infection early via an ATG5-dependent but autophagy-independent mechanism, suggesting that autophagy machinery plays an important role in lung innate immune cells outside of the previously reported role in direct pathogen killing. The susceptibility follows a hyper-neutrophilic inflammatory response, and neutrophil depletion rescues the mice. I determined that there are fewer CD4+ and CD8+ T cells in the lungs of these mice at 19 dpi, when the adaptive immune response is initiating in Atg5����/���� mice. Because these mice have markedly elevated neutrophils in the lung, I tested whether the delay in T cell response could be rescued by depletion of neutrophils. Depletion of neutrophils resulted in a rescue of T cell response to the level of undepleted Atg5����/���� mice. However, there was an additional increase in T cells in the neutrophil depleted Atg5����/���� mice despite similar neutrophil levels after depletion, suggesting that there may be an additional neutrophil-independent role for ATG5 in the initiation of adaptive immunity. Preliminary results suggest that this phenotype is common to mice that lack essential genes of the autophagy pathway in innate immune cells. This study determined that autophagy is required in lung innate immune cells for the efficient initiation of adaptive immunity to Mtb infection. Together, these findings advance our understanding of lung macrophage, monocyte, and dendritic cell pathways required during Mtb infection