Faculty Bibliography

Cancer diagnosis and management depend upon the extraction of complex information from microscopy images by pathologists, which requires time-consuming expert interpretation prone to human bias. Supervised deep learning approaches have proven powerful, but are inherently limited by the cost and quality of annotations used for training. Therefore, we present Histomorphological Phenotype Learning, a self-supervised methodology requiring no labels and operating via the automatic discovery of discriminatory features in image tiles. Tiles are grouped into morphologically similar clusters which constitute an atlas of histomorphological phenotypes (HP-Atlas), revealing trajectories from benign to malignant tissue via inflammatory and reactive phenotypes. These clusters have distinct features which can be identified using orthogonal methods, linking histologic, molecular and clinical phenotypes. Applied to lung cancer, we show that they align closely with patient survival, with histopathologically recognised tumor types and growth patterns, and with transcriptomic measures of immunophenotype. These properties are maintained in a multi-cancer study.

Cancer is a disease that stems from a fundamental liability inherent to multicellular life forms in which an individual cell is capable of reneging on the interests of the collective organism. Although cancer is commonly described as an evolutionary process, a less appreciated aspect of tumorigenesis may be the constraints imposed by the organism's developmental programs. Recent work from single-cell transcriptomic analyses across a range of cancer types has revealed the recurrence, plasticity, and co-option of distinct cellular states among cancer cell populations. Here, we note that across diverse cancer types, the observed cell states are proximate within the developmental hierarchy of the cell of origin. We thus posit a model by which cancer cell states are directly constrained by the organism's "developmental map." According to this model, a population of cancer cells traverses the developmental map, thereby generating a heterogeneous set of states whose interactions underpin emergent tumor behavior.

Microsporidia are single-celled intracellular parasites that cause opportunistic diseases in humans. Encephalitozoon intestinalis is a prevalent human-infecting species that invades the small intestine. Dissemination to other organ systems is also observed, and is potentially facilitated by macrophages. The macrophage response to infection and the developmental trajectory of the parasite are not well studied. Here we use single cell RNA sequencing to investigate transcriptional changes in both the host and parasite during infection. While a small population of infected macrophages mount a response, most remain transcriptionally unchanged, suggesting that the majority of parasites may avoid host detection. The parasite transcriptome reveals large transcriptional changes throughout the life cycle, providing a blueprint for parasite development. The stealthy microsporidian lifestyle likely allows these parasites to harness macrophages for replication and dissemination. Together, our data provide insights into the host response in primary human macrophages and the E. intestinalis developmental program.

BACKGROUND:Xenotransplantation of genetically engineered porcine organs has the potential to address the challenge of organ donor shortage. Two cases of porcine-to-human kidney xenotransplantation were performed, yet the physiological effects on the xenografts and the recipients' immune responses remain largely uncharacterized. METHODS:We performed single-cell RNA sequencing (scRNA-seq) and longitudinal RNA-seq analyses of the porcine kidneys to dissect xenotransplantation-associated cellular dynamics and xenograft-recipient interactions. We additionally performed longitudinal scRNA-seq of the peripheral blood mononuclear cells (PBMCs) to detect recipient immune responses across time. FINDINGS/RESULTS:Although no hyperacute rejection signals were detected, scRNA-seq analyses of the xenografts found evidence of endothelial cell and immune response activation, indicating early signs of antibody-mediated rejection. Tracing the cells' species origin, we found human immune cell infiltration in both xenografts. Human transcripts in the longitudinal bulk RNA-seq revealed that human immune cell infiltration and the activation of interferon-gamma-induced chemokine expression occurred by 12 and 48 h post-xenotransplantation, respectively. Concordantly, longitudinal scRNA-seq of PBMCs also revealed two phases of the recipients' immune responses at 12 and 48-53 h. Lastly, we observed global expression signatures of xenotransplantation-associated kidney tissue damage in the xenografts. Surprisingly, we detected a rapid increase of proliferative cells in both xenografts, indicating the activation of the porcine tissue repair program. CONCLUSIONS:Longitudinal and single-cell transcriptomic analyses of porcine kidneys and the recipient's PBMCs revealed time-resolved cellular dynamics of xenograft-recipient interactions during xenotransplantation. These cues can be leveraged for designing gene edits and immunosuppression regimens to optimize xenotransplantation outcomes. FUNDING/BACKGROUND:This work was supported by NIH RM1HG009491 and DP5OD033430.

Inflammatory epithelial diseases are spurred by the concomitant dysregulation of immune and epithelial cells. How these two dysregulated cellular compartments simultaneously sustain their heightened metabolic demands is unclear. Single-cell and spatial transcriptomics (ST), along with immunofluorescence, revealed that hypoxia-inducible factor 1α (HIF1α), downstream of IL-17 signaling, drove psoriatic epithelial remodeling. Blocking HIF1α in human psoriatic lesions ex vivo impaired glycolysis and phenocopied anti-IL-17 therapy. In a murine model of skin inflammation, epidermal-specific loss of HIF1α or its target gene, glucose transporter 1, ameliorated epidermal, immune, vascular, and neuronal pathology. Mechanistically, glycolysis autonomously fueled epithelial pathology and enhanced lactate production, which augmented the γδ T17 cell response. RORγt-driven genetic deletion or pharmacological inhibition of either lactate-producing enzymes or lactate transporters attenuated epithelial pathology and IL-17A expression in vivo. Our findings identify a metabolic hierarchy between epithelial and immune compartments and the consequent coordination of metabolic processes that sustain inflammatory disease.

The development of dendritic cells (DCs), including antigen-presenting conventional DCs (cDCs) and cytokine-producing plasmacytoid DCs (pDCs), is controlled by the growth factor Flt3 ligand (Flt3L) and its receptor Flt3. We genetically dissected Flt3L-driven DC differentiation using CRISPR-Cas9-based screening. Genome-wide screening identified multiple regulators of DC differentiation including subunits of TSC and GATOR1 complexes, which restricted progenitor growth but enabled DC differentiation by inhibiting mTOR signaling. An orthogonal screen identified the transcriptional repressor Trim33 (TIF-1γ) as a regulator of DC differentiation. Conditional targeting in vivo revealed an essential role of Trim33 in the development of all DCs, but not of monocytes or granulocytes. In particular, deletion of Trim33 caused rapid loss of DC progenitors, pDCs, and the cross-presenting cDC1 subset. Trim33-deficient Flt3⁺ progenitors up-regulated pro-inflammatory and macrophage-specific genes but failed to induce the DC differentiation program. Collectively, these data elucidate mechanisms that control Flt3L-driven differentiation of the entire DC lineage and identify Trim33 as its essential regulator.

Loss-of-function mutations in KEAP1 frequently occur in lung cancer and are associated with poor prognosis and resistance to standard of care treatment, highlighting the need for the development of targeted therapies. We previously showed that KEAP1 mutant tumors consume glutamine to support the metabolic rewiring associated with NRF2-dependent antioxidant production. Here, using preclinical patient-derived xenograft models and antigenic orthotopic lung cancer models, we show that the glutamine antagonist prodrug DRP-104 impairs the growth of KEAP1 mutant tumors. We find that DRP-104 suppresses KEAP1 mutant tumors by inhibiting glutamine-dependent nucleotide synthesis and promoting antitumor T cell responses. Using multimodal single-cell sequencing and ex vivo functional assays, we demonstrate that DRP-104 reverses T cell exhaustion, decreases T_regs, and enhances the function of CD4 and CD8 T cells, culminating in an improved response to anti-PD1 therapy. Our preclinical findings provide compelling evidence that DRP-104, currently in clinical trials, offers a promising therapeutic approach for treating patients with KEAP1 mutant lung cancer.

Self-supervised learning (SSL) automates the extraction and interpretation of histopathology features on unannotated hematoxylin-and-eosin-stained whole-slide images (WSIs). We trained an SSL Barlow Twins-encoder on 435 TCGA colon adenocarcinoma WSIs to extract features from small image patches. Leiden community detection then grouped tiles into histomorphological phenotype clusters (HPCs). HPC reproducibility and predictive ability for overall survival was confirmed in an independent clinical trial cohort (N=1213 WSIs). This unbiased atlas resulted in 47 HPCs displaying unique and sharing clinically significant histomorphological traits, highlighting tissue type, quantity, and architecture, especially in the context of tumor stroma. Through in-depth analysis of these HPCs, including immune landscape and gene set enrichment analysis, and association to clinical outcomes, we shed light on the factors influencing survival and responses to treatments like standard adjuvant chemotherapy and experimental therapies. Further exploration of HPCs may unveil new insights and aid decision-making and personalized treatments for colon cancer patients.

Organisms determine the transcription rates of thousands of genes through a few modes of regulation that recur across the genome¹. In bacteria, the relationship between the regulatory architecture of a gene and its expression is well understood for individual model gene circuits^2,3. However, a broader perspective of these dynamics at the genome scale is lacking, in part because bacterial transcriptomics has hitherto captured only a static snapshot of expression averaged across millions of cells⁴. As a result, the full diversity of gene expression dynamics and their relation to regulatory architecture remains unknown. Here we present a novel genome-wide classification of regulatory modes based on the transcriptional response of each gene to its own replication, which we term the transcription-replication interaction profile (TRIP). Analysing single-bacterium RNA-sequencing data, we found that the response to the universal perturbation of chromosomal replication integrates biological regulatory factors with biophysical molecular events on the chromosome to reveal the local regulatory context of a gene. Whereas the TRIPs of many genes conform to a gene dosage-dependent pattern, others diverge in distinct ways, and this is shaped by factors such as intra-operon position and repression state. By revealing the underlying mechanistic drivers of gene expression heterogeneity, this work provides a quantitative, biophysical framework for modelling replication-dependent expression dynamics.

Pattern formation originates during embryogenesis by a series of symmetry-breaking steps throughout an expanding cell lineage. In Drosophila, classic work has shown that segmentation in the embryo is established by morphogens within a syncytium, and the subsequent action of the gap, pair-rule, and segment polarity genes. This classic model however does not translate directly to species that lack a syncytium - such as Caenorhabditis elegans - where cell fate is specified by cell-autonomous cell lineage programs and their inter-signaling. Previous single-cell RNA-Seq studies in C. elegans have analyzed cells from a mixed suspension of cells from many embryos to study late differentiation stages, or individual early stage embryos to study early gene expression in the embryo. To study the intermediate stages of early and late gastrulation (28- to 102-cells stages) missed by these approaches, here we determine the transcriptomes of the 1- to 102-cell stage to identify 119 embryonic cell states during cell fate specification, including 'equivalence-group' cell identities. We find that gene expression programs are modular according to the sub-cell lineages, each establishing a set of stripes by combinations of transcription factor gene expression across the anterior-posterior axis. In particular, expression of the homeodomain genes establishes a comprehensive lineage-specific positioning system throughout the embryo beginning at the 28-cell stage. Moreover, we find that genes that segment the entire embryo in Drosophila have orthologs in C. elegans that exhibit sub-lineage-specific expression. These results suggest that the C. elegans embryo is patterned by a juxtaposition of distinct lineage-specific gene regulatory programs each with a unique encoding of cell location and fate. This use of homologous gene regulatory patterning codes suggests a deep homology of cell fate specification programs across diverse modes of development.