Identifying gene expression programs of cell-type identity and cellular activity with single-cell RNA-Seq

Programs, origins and immunomodulatory functions of myeloid cells in glioma

Gliomas are incurable malignancies notable for having an immunosuppressive microenvironment with abundant myeloid cells, the immunomodulatory phenotypes of which remain poorly defined1. Here we systematically investigate these phenotypes by integrating single-cell RNA sequencing, chromatin accessibility, spatial transcriptomics and glioma organoid explant systems. We discovered four immunomodulatory expression programs: microglial inflammatory and scavenger immunosuppressive programs, which are both unique to primary brain tumours, and systemic inflammatory and complement immunosuppressive programs, which are also expressed by non-brain tumours. The programs are not contingent on myeloid cell type, developmental origin or tumour mutational state, but instead are driven by microenvironmental cues, including tumour hypoxia, interleukin-1β, TGFβ and standard-of-care dexamethasone treatment. Their relative expression can predict immunotherapy response and overall survival. By associating the respective programs with mediating genomic elements, transcription factors and signalling pathways, we uncover strategies for manipulating myeloid-cell phenotypes. Our study provides a framework to understand immunomodulation by myeloid cells in glioma and a foundation for the development of more-effective immunotherapies.

Interpretable single-cell factor decomposition using sciRED

Single-cell RNA sequencing maps gene expression heterogeneity within a tissue. However, identifying biological signals in this data is challenging due to confounding technical factors, sparsity, and high dimensionality. Data factorization methods address this by separating and identifying signals in the data, such as gene expression programs, but the resulting factors must be manually interpreted. We developed Single-Cell Interpretable REsidual Decomposition (sciRED) to improve the interpretation of scRNA-seq factor analysis. sciRED removes known confounding effects, uses rotations to improve factor interpretability, maps factors to known covariates, identifies unexplained factors that may capture hidden biological phenomena, and determines the genes and biological processes represented by the resulting factors. We apply sciRED to multiple scRNA-seq datasets and identify sex-specific variation in a kidney map, discern strong and weak immune stimulation signals in a PBMC dataset, reduce ambient RNA contamination in a rat liver atlas to help identify strain variation and reveal rare cell type signatures and anatomical zonation gene programs in a healthy human liver map. These demonstrate that sciRED is useful in characterizing diverse biological signals within scRNA-seq data.

IL-27 elicits a cytotoxic CD8+ T cell program to enforce tumour control

Although cytotoxic CD8+ T lymphocytes (CTLs) are essential for anti-tumour immunity, they are frequently dysfunctional in tumours1. Cytokines that sustain CTL activity are attractive for cancer immunotherapy, but avoiding inflammatory toxicity remains a challenge for their clinical use2. Here we show that expression of a CTL signature is strongly associated with IL27 expression in human and mouse tumours. In mice, IL-27 acts directly on tumour-specific CTLs to promote their persistence and effector function in the tumour microenvironment. Moreover, treatment with inducible IL-27 overexpression or a half-life-extended IL-27 protein in vivo is well tolerated, induces regression of established tumours, drives an enhanced cytotoxic program in anti-tumour CTLs and synergizes with PD-L1 blockade. In patients with cancer who were treated with anti-PD-1/PD-L1 therapy, high expression of IL-27 correlates with a favourable clinical response, and IL-27 supports human CTL function during chronic antigen stimulation ex vivo. Our data demonstrate that endogenous IL-27 is essential for anti-tumour immunity and that IL-27 receptor agonism can safely improve anti-tumour T cell responses alone or in combination with PD-L1 blockade.

ChromaFactor: Deconvolution of single-molecule chromatin organization with non-negative matrix factorization

The investigation of chromatin organization in single cells holds great promise for identifying causal relationships between genome structure and function. However, analysis of single-molecule data is hampered by extreme yet inherent heterogeneity, making it challenging to determine the contributions of individual chromatin fibers to bulk trends. To address this challenge, we propose ChromaFactor, a novel computational approach based on non-negative matrix factorization that deconvolves single-molecule chromatin organization datasets into their most salient primary components. ChromaFactor provides the ability to identify trends accounting for the maximum variance in the dataset while simultaneously describing the contribution of individual molecules to each component. Applying our approach to two single-molecule imaging datasets across different genomic scales, we find that these primary components demonstrate significant correlation with key functional phenotypes, including active transcription, enhancer-promoter distance, and genomic compartment. Also, we find that some bulk trends exist at the single-cell level, but only in a small fraction of cells, suggesting that critical changes in genome organization may be driven by specific rare subpopulations rather than occurring uniformly across all cells. ChromaFactor offers a robust tool for understanding the complex interplay between chromatin structure and function on individual DNA molecules, pinpointing which subpopulations drive functional changes and fostering new insights into cellular heterogeneity and its implications for bulk genomic phenomena.

MorPhiC Consortium: towards functional characterization of all human genes

Recent advances in functional genomics and human cellular models have substantially enhanced our understanding of the structure and regulation of the human genome. However, our grasp of the molecular functions of human genes remains incomplete and biased towards specific gene classes. The Molecular Phenotypes of Null Alleles in Cells (MorPhiC) Consortium aims to address this gap by creating a comprehensive catalogue of the molecular and cellular phenotypes associated with null alleles of all human genes using in vitro multicellular systems. In this Perspective, we present the strategic vision of the MorPhiC Consortium and discuss various strategies for generating null alleles, as well as the challenges involved. We describe the cellular models and scalable phenotypic readouts that will be used in the consortium's initial phase, focusing on 1,000 protein-coding genes. The resulting molecular and cellular data will be compiled into a catalogue of null-allele phenotypes. The methodologies developed in this phase will establish best practices for extending these approaches to all human protein-coding genes. The resources generated-including engineered cell lines, plasmids, phenotypic data, genomic information and computational tools-will be made available to the broader research community to facilitate deeper insights into human gene functions.

High order expression dependencies finely resolve cryptic states and subtypes in single cell data

Single cells are typically typed by clustering into discrete locations in reduced dimensional transcriptome space. Here we introduce Stator, a data-driven method that identifies cell (sub)types and states without relying on cells' local proximity in transcriptome space. Stator labels the same single cell multiply, not just by type and subtype, but also by state such as activation, maturity or cell cycle sub-phase, through deriving higher-order gene expression dependencies from a sparse gene-by-cell expression matrix. Stator's finer resolution is clear from analyses of mouse embryonic brain, and human healthy or diseased liver. Rather than only coarse-scale labels of cell type, Stator further resolves cell types into subtypes, and these subtypes into stages of maturity and/or cell cycle phases, and yet further into portions of these phases. Among cryptically homogeneous embryonic cells, for example, Stator finds 34 distinct radial glia states whose gene expression forecasts their future GABAergic or glutamatergic neuronal fate. Further, Stator's fine resolution of liver cancer states reveals expression programmes that predict patient survival. We provide Stator as a Nextflow pipeline and Shiny App.

Variants and vaccines impact nasal immunity over three waves of SARS-CoV-2

Viral variant and host vaccination status impact infection with severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), yet how these factors shift cellular responses in the human nasal mucosa remains uncharacterized. We performed single-cell RNA sequencing (scRNA-seq) on nasopharyngeal swabs from vaccinated and unvaccinated adults with acute Delta and Omicron SARS-CoV-2 infections and integrated with data from acute infections with ancestral SARS-CoV-2. Patients with Delta and Omicron exhibited greater similarity in nasal cell composition driven by myeloid, T cell and SARS-CoV-2hi cell subsets, which was distinct from that of ancestral cases. Delta-infected samples had a marked increase in viral RNA, and a subset of PER2+EGR1+GDF15+ epithelial cells was enriched in SARS-CoV-2 RNA+ cells in all variants. Prior vaccination was associated with increased frequency and activation of nasal macrophages. Expression of interferon-stimulated genes negatively correlated with coronavirus disease 2019 (COVID-19) severity in patients with ancestral and Delta but not Omicron variants. Our study defines nasal cell responses and signatures of disease severity across SARS-CoV-2 variants and vaccination.

The dynamics of hematopoiesis over the human lifespan

Over a lifetime, hematopoietic stem cells (HSCs) adjust their lineage output to support age-aligned physiology. In model organisms, stereotypic waves of hematopoiesis have been observed corresponding to defined age-biased HSC hallmarks. However, how the properties of hematopoietic stem and progenitor cells change over the human lifespan remains unclear. To address this gap, we profiled individual transcriptome states of human hematopoietic stem and progenitor cells spanning gestation, maturation and aging. Here we define the gene expression networks dictating age-specific differentiation of HSCs and the dynamics of fate decisions and lineage priming throughout life. We additionally identifiy and functionally validate a fetal-specific HSC state with robust engraftment and multilineage capacity. Furthermore, we observe that classification of acute myeloid leukemia against defined transcriptional age states demonstrates that utilization of early life transcriptional programs associates with poor prognosis. Overall, we provide a disease-relevant framework for heterochronic orientation of stem cell ontogeny along the real time axis of the human lifespan.

A multi-omics spatial framework for host-microbiome dissection within the intestinal tissue microenvironment

The intricate interactions between the host immune system and its microbiome constituents undergo dynamic shifts in response to perturbations to the intestinal tissue environment. Our ability to study these events on the systems level is significantly limited by in situ approaches capable of generating simultaneous insights from both host and microbial communities. Here, we introduce Microbiome Cartography (MicroCart), a framework for simultaneous in situ probing of host and microbiome across multiple spatial modalities. We demonstrate MicroCart by investigating gut host and microbiome changes in a murine colitis model, using spatial proteomics, transcriptomics, and glycomics. Our findings reveal a global but systematic transformation in tissue immune responses, encompassing tissue-level remodeling in response to host immune and epithelial cell state perturbations, bacterial population shifts, localized inflammatory responses, and metabolic process alterations during colitis. MicroCart enables a deep investigation of the intricate interplay between the host tissue and its microbiome with spatial multi-omics.

Causal modeling of gene effects from regulators to programs to traits: integration of genetic associations and Perturb-seq

Genetic association studies provide a unique tool for identifying causal links from genes to human traits and diseases. However, it is challenging to determine the biological mechanisms underlying most associations, and we lack genome-scale approaches for inferring causal mechanistic pathways from genes to cellular functions to traits. Here we propose new approaches to bridge this gap by combining quantitative estimates of gene-trait relationships from loss-of-function burden tests with gene-regulatory connections inferred from Perturb-seq experiments in relevant cell types. By combining these two forms of data, we aim to build causal graphs in which the directional associations of genes with a trait can be explained by their regulatory effects on biological programs or direct effects on the trait. As a proof-of-concept, we constructed a causal graph of the gene regulatory hierarchy that jointly controls three partially co-regulated blood traits. We propose that perturbation studies in trait-relevant cell types, coupled with gene-level effect sizes for traits, can bridge the gap between genetics and biology.

Epigenetic dysregulation of metabolic programs mediates liposarcoma cell plasticity

Sarcomas are rare connective tissue cancers thought to arise from aberrant mesenchymal stem cell (MSC) differentiation. Liposarcoma (LPS) holds valuable insights into dysfunctional differentiation given its well- and dedifferentiated histologic subtypes (WDLPS, DDLPS). Despite well-established differences in histology and clinical behavior, the molecular pathways underlying each subtype are poorly understood. Here, we performed single-nucleus multiome sequencing and spatial profiling on carefully curated human LPS samples and found defects in adipocyte-specific differentiation within LPS. Loss of insulin-like growth factor 1 (IGF1) and gain of cellular programs related to early mesenchymal development and glucagon-like peptide-1 (GLP-1)-induced insulin secretion are primary features of DDLPS. IGF1 loss was associated with worse overall survival in LPS patients. Through in vitro stimulation of the IGF1 pathway, we identified that DDLPS cells are deficient in the adipose-specific PPARG isoform 2 (PPARG2). Defects in IGF1/PPARG2 signaling in DDLPS led to a block in differentiation that could not be fully overcome with the addition of exogenous IGF1 or the pro-adipogenic agonists to PPARG and GLP-1. However, we noted upregulation of the IGF1 receptor (IGF1R) in the setting of IGF1 deficiency, which promoted sensitivity to an IGF1R-targeted antibody-drug conjugate that may serve as a novel therapeutic strategy in LPS. In summary, lineage-specific defects in adipogenesis drive dedifferentiation in LPS and may translate into selective therapeutic targeting in this disease.

CSI-GEP: A GPU-based unsupervised machine learning approach for recovering gene expression programs in atlas-scale single-cell RNA-seq data

Exploratory analysis of single-cell RNA sequencing (scRNA-seq) typically relies on hard clustering over two-dimensional projections like uniform manifold approximation and projection (UMAP). However, such methods can severely distort the data and have many arbitrary parameter choices. Methods that can model scRNA-seq data as non-discrete "gene expression programs" (GEPs) can better preserve the data's structure, but currently, they are often not scalable, not consistent across repeated runs, and lack an established method for choosing key parameters. Here, we developed a GPU-based unsupervised learning approach, "consensus and scalable inference of gene expression programs" (CSI-GEP). We show that CSI-GEP can recover ground truth GEPs in real and simulated atlas-scale scRNA-seq datasets, significantly outperforming cutting-edge methods, including GPT-based neural networks. We applied CSI-GEP to a whole mouse brain atlas of 2.2 million cells, disentangling endothelial cell types missed by other methods, and to an integrated scRNA-seq atlas of human tumors and cell lines, discovering mesenchymal-like GEPs unique to cancer cells growing in culture.

Evolving cell states and oncogenic drivers during the progression of IDH-mutant gliomas

Isocitrate dehydrogenase (IDH) mutants define a class of gliomas that are initially slow-growing but inevitably progress to fatal disease. To characterize their malignant cell hierarchy, we profiled chromatin accessibility and gene expression across single cells from low-grade and high-grade IDH-mutant gliomas and ascertained their developmental states through a comparison to normal brain cells. We provide evidence that these tumors are initially fueled by slow-cycling oligodendrocyte progenitor cell-like cells. During progression, a more proliferative neural progenitor cell-like population expands, potentially through partial reprogramming of 'permissive' chromatin in progenitors. This transition is accompanied by a switch from methylation-based drivers to genetic ones. In low-grade IDH-mutant tumors or organoids, DNA hypermethylation appears to suppress interferon (IFN) signaling, which is induced by IDH or DNA methyltransferase 1 inhibitors. High-grade tumors frequently lose this hypermethylation and instead acquire genetic alterations that disrupt IFN and other tumor-suppressive programs. Our findings explain how these slow-growing tumors may progress to lethal malignancies and have implications for therapies that target their epigenetic underpinnings.

Liver X receptor unlinks intestinal regeneration and tumorigenesis

Uncontrolled regeneration leads to neoplastic transformation1-3. The intestinal epithelium requires precise regulation during continuous homeostatic and damage-induced tissue renewal to prevent neoplastic transformation, suggesting that pathways unlinking tumour growth from regenerative processes must exist. Here, by mining RNA-sequencing datasets from two intestinal damage models4,5 and using pharmacological, transcriptomics and genetic tools, we identified liver X receptor (LXR) pathway activation as a tissue adaptation to damage that reciprocally regulates intestinal regeneration and tumorigenesis. Using single-cell RNA sequencing, intestinal organoids, and gain- and loss-of-function experiments, we demonstrate that LXR activation in intestinal epithelial cells induces amphiregulin (Areg), enhancing regenerative responses. This response is coordinated by the LXR-ligand-producing enzyme CYP27A1, which was upregulated in damaged intestinal crypt niches. Deletion of Cyp27a1 impaired intestinal regeneration, which was rescued by exogenous LXR agonists. Notably, in tumour models, Cyp27a1 deficiency led to increased tumour growth, whereas LXR activation elicited anti-tumour responses dependent on adaptive immunity. Consistently, human colorectal cancer specimens exhibited reduced levels of CYP27A1, LXR target genes, and B and CD8 T cell gene signatures. We therefore identify an epithelial adaptation mechanism to damage, whereby LXR functions as a rheostat, promoting tissue repair while limiting tumorigenesis.

Dissecting tumor transcriptional heterogeneity from single-cell RNA-seq data by generalized binary covariance decomposition

Profiling tumors with single-cell RNA sequencing has the potential to identify recurrent patterns of transcription variation related to cancer progression, and to produce therapeutically relevant insights. However, strong intertumor heterogeneity can obscure more subtle patterns that are shared across tumors. Here we introduce a statistical method, generalized binary covariance decomposition (GBCD), to address this problem. We show that GBCD can decompose transcriptional heterogeneity into interpretable components-including patient-specific, dataset-specific and shared components relevant to disease subtypes-and that, in the presence of strong intertumor heterogeneity, it can produce more interpretable results than existing methods. Applied to data on pancreatic ductal adenocarcinoma, GBCD produced a refined characterization of existing tumor subtypes, and identified a gene expression program prognostic of poor survival independent of tumor stage and subtype. The gene expression program is enriched for genes involved in stress responses, and suggests a role for the integrated stress response in pancreatic ductal adenocarcinoma.

Same-Slide Spatial Multi-Omics Integration Reveals Tumor Virus-Linked Spatial Reorganization of the Tumor Microenvironment

The advent of spatial transcriptomics and spatial proteomics have enabled profound insights into tissue organization to provide systems-level understanding of diseases. Both technologies currently remain largely independent, and emerging same slide spatial multi-omics approaches are generally limited in plex, spatial resolution, and analytical approaches. We introduce IN-situ DEtailed Phenotyping To High-resolution transcriptomics (IN-DEPTH), a streamlined and resource-effective approach compatible with various spatial platforms. This iterative approach first entails single-cell spatial proteomics and rapid analysis to guide subsequent spatial transcriptomics capture on the same slide without loss in RNA signal. To enable multi-modal insights not possible with current approaches, we introduce k-bandlimited Spectral Graph Cross-Correlation (SGCC) for integrative spatial multi-omics analysis. Application of IN-DEPTH and SGCC on lymphoid tissues demonstrated precise single-cell phenotyping and cell-type specific transcriptome capture, and accurately resolved the local and global transcriptome changes associated with the cellular organization of germinal centers. We then implemented IN-DEPTH and SGCC to dissect the tumor microenvironment (TME) of Epstein-Barr Virus (EBV)-positive and EBV-negative diffuse large B-cell lymphoma (DLBCL). Our results identified a key tumor-macrophage-CD4 T-cell immunomodulatory axis differently regulated between EBV-positive and EBV-negative DLBCL, and its central role in coordinating immune dysfunction and suppression. IN-DEPTH enables scalable, resource-efficient, and comprehensive spatial multi-omics dissection of tissues to advance clinically relevant discoveries.

Single-cell transcriptome analysis reveals reciprocal epithelial and endothelial cell evolution in ovarian cancer

Tumor neovascularization mediated by endothelial cells (ECs) is essential for ovarian cancer (OC) progression, but interactions between epithelial cells and ECs are not well understood. Here, we analyze single-cell transcriptome of 87,847 epithelial cells and 11,696 ECs from fallopian tubes, primary and metastatic ovarian tumors. Cell differentiation trajectory analysis reveals that fallopian tube cells exhibit a potential development trend toward primary OC epithelial cells. We identify a sub-population of fallopian tube epithelial cells (FTSEC3), which highly express tumor cell markers and are enriched in vascular endothelial growth factor production. Two neovascularization-related EC phenotypes (MKI67+ proliferating ECs and ESM1+ tip cells) are specially found in ovarium tumors, which exhibit strong interactions with FTSEC3. We validate that genetic disruption of LAMININ and TGF-β with CRISPR in ECs inhibits sprouting angiogenesis. In summary, this study reveals a reciprocal evolution and interaction between epithelial and ECs in OC development and progression.

IFN-γ-producing TH1 cells and dysfunctional regulatory T cells contribute to the pathogenesis of Sjögren's disease

Sjögren's disease (SjD) is an autoimmune disorder characterized by progressive salivary and lacrimal gland dysfunction, inflammation, and destruction, as well as extraglandular manifestations. SjD is associated with autoreactive B and T cells, but its pathophysiology remains incompletely understood. Abnormalities in regulatory T (Treg) cells occur in several autoimmune diseases, but their role in SjD is ambiguous. We had previously shown that the function and development of Treg cells depend on store-operated Ca2+ entry (SOCE), which is mediated by ORAI1 Ca2+ channels and stromal interaction protein 1 (STIM1) and STIM2. Here, we show that mice with a Foxp3+ Treg cell-specific deletion of Stim1 and Stim2 develop a phenotype that fulfills all classification criteria of human SjD. Mutant mice have salivary and lacrimal gland inflammation characterized by strong lymphocyte infiltration and transcriptional signatures dominated by T helper 1 (TH1) and interferon (IFN) signaling. CD4+ T cells from mutant mice are sufficient to induce SjD-like disease in an IFN-γ-dependent manner. Inhibition of IFN signaling with the JAK1/2 inhibitor baricitinib alleviated CD4+ T cell-induced SjD in mice. These findings are consistent with the transcriptional profiles of CD4+ T cells from patients with SjD, which indicate enhanced TH1 but reduced memory Treg cell function. Together, our study provides evidence for a critical role of dysfunctional Treg cells and IFN-γ-producing TH1 cells in the pathogenesis of SjD.

Interpretable single-cell factor decomposition using sciRED

Single-cell RNA sequencing (scRNA-seq) maps gene expression heterogeneity within a tissue. However, identifying biological signals in this data is challenging due to confounding technical factors, sparsity, and high dimensionality. Data factorization methods address this by separating and identifying signals in the data, such as gene expression programs, but the resulting factors must be manually interpreted. We developed Single-Cell Interpretable REsidual Decomposition (sciRED) to improve the interpretation of scRNA-seq factor analysis. sciRED removes known confounding effects, uses rotations to improve factor interpretability, maps factors to known covariates, identifies unexplained factors that may capture hidden biological phenomena and determines the genes and biological processes represented by the resulting factors. We apply sciRED to multiple scRNA-seq datasets and identify sex-specific variation in a kidney map, discern strong and weak immune stimulation signals in a PBMC dataset, reduce ambient RNA contamination in a rat liver atlas to help identify strain variation, and reveal rare cell type signatures and anatomical zonation gene programs in a healthy human liver map. These demonstrate that sciRED is useful in characterizing diverse biological signals within scRNA-seq data.

An integrated single-nucleus and spatial transcriptomics atlas reveals the molecular landscape of the human hippocampus

The hippocampus contains many unique cell types, which serve the structure's specialized functions, including learning, memory and cognition. These cells have distinct spatial organization, morphology, physiology, and connectivity, highlighting the importance of transcriptome-wide profiling strategies that retain cytoarchitectural organization. Here, we generated spatially-resolved transcriptomics (SRT) and single-nucleus RNA-sequencing (snRNA-seq) data from adjacent tissue sections of the anterior human hippocampus in ten adult neurotypical donors to define molecular profiles for hippocampal cell types and spatial domains. Using non-negative matrix factorization (NMF) and label transfer, we integrated these data by defining gene expression patterns within the snRNA-seq data and inferring their expression in the SRT data. We identified NMF patterns that captured transcriptional variation across neuronal cell types and indicated that the response of excitatory and inhibitory postsynaptic specializations were prioritized in different SRT spatial domains. We used the NMF and label transfer approach to leverage existing rodent datasets, identifying patterns of activity-dependent transcription and subpopulations of dentate gyrus granule cells in our SRT dataset that may be predisposed to participate in learning and memory ensembles. Finally, we characterized the spatial organization of NMF patterns corresponding to non-cornu ammonis pyramidal neurons and identified snRNA-seq clusters mapping to distinct regions of the retrohippocampus, to three subiculum layers, and to a population of presubiculum neurons. To make this comprehensive molecular atlas accessible to the scientific community, both raw and processed data are freely available, including through interactive web applications.

Computer-aided analysis reveals metallothionein-positive cancer-associated fibroblasts promote angiogenesis in gastric adenocarcinoma

Gastric adenocarcinoma (GC), along with its tumor microenvironment (TME), poses great challenges for clinical treatment strategies. Single-cell sequencing has become an important tool for analyzing TME heterogeneity, cell subpopulation, and gene expression patterns. 56 GC single-cell sequencing samples were analyzed, focusing on TME by delineating cancer cells, cancer-associated fibroblasts (CAFs), and macrophages. The spatial transcriptome was used to clarify the distribution characteristics of each cellular component in the tissue slice. Despite the widespread genetic mutations observed in cancer cells, certain recurrent alterations were identified in specific chromosomal regions. The heterogeneity among GC cells is profound, four cancer cell subpopulations were identified through drug sensitivity profiling. Subtype 4, although only present in some samples, demonstrates the strongest stemness and metabolic activity, possibly indicative of an early-stage cancer subpopulation. Their drug sensitivity profiles may hold promise for guiding clinical intervention. In addition, robust spatial co-localization patterns were observed between CAFs, M2 macrophages, and endothelial cells. CAFs were further categorized into six subgroups, among which a novel subgroup termed metallothionein(mt)-positive CAF (mtCAF), characterized by elevated expression of metallothionein 1X (MT1X) and subsequent vascular endothelial growth factor A (VEGFA) secretion, was identified. Immunohistochemistry preliminary confirmed the presence of this unique CAF subgroup. Additionally, M2d macrophages, besides exhibiting high VEGFA expression, also demonstrated various growth factors such as Aamphiregulin (AREG). The M2d-mtCAF axis may play an important role in GC angiogenesis. This study not only enhances our understanding of the TME heterogeneity in GC but also sheds light on the interaction between CAFs and tumor-associated macrophages (TAMs) in tumor angiogenesis.

NiCo identifies extrinsic drivers of cell state modulation by niche covariation analysis

Cell states are modulated by intrinsic driving forces such as gene expression noise and extrinsic signals from the tissue microenvironment. The distinction between intrinsic and extrinsic cell state determinants is essential for understanding the regulation of cell fate in tissues during development, homeostasis and disease. The rapidly growing availability of single-cell resolution spatial transcriptomics makes it possible to meet this challenge. However, available computational methods to infer topological tissue domains, spatially variable genes, or ligand-receptor interactions are limited in their capacity to capture cell state changes driven by crosstalk between individual cell types within the same niche. We present NiCo, a computational framework for integrating single-cell resolution spatial transcriptomics with matched single-cell RNA-sequencing reference data to infer the influence of the spatial niche on the cell state. By applying NiCo to mouse embryogenesis, adult small intestine and liver data, we demonstrate the ability to predict novel niche interactions that govern cell state variation underlying tissue development and homeostasis. In particular, NiCo predicts a feedback mechanism between Kupffer cells and neighboring stellate cells dampening stellate cell activation in the normal liver. NiCo provides a powerful tool to elucidate tissue architecture and to identify drivers of cellular states in local niches.

Cell-to-cell distance that combines gene expression and gene embeddings

The application of large-language models (LLMs) to single-cell gene-expression data has introduced a new type of data that includes a gene-embedding matrix, in addition to the experimentally obtained gene-expression matrix. This paper addresses a fundamental problem in analyzing such data: how to effectively combine the information from both matrices to better define cell-to-cell distance. We identify a computationally feasible solution that demonstrates superior ability to cluster cells of the same type across all six real datasets we tested, underscoring its advantage as a measure of cell-to-cell distance.

Mapping the gene space at single-cell resolution with gene signal pattern analysis

In single-cell sequencing analysis, several computational methods have been developed to map the cellular state space, but little has been done to map or create embeddings of the gene space. Here we formulate the gene embedding problem, design tasks with simulated single-cell data to evaluate representations, and establish ten relevant baselines. We then present a graph signal processing approach, called gene signal pattern analysis (GSPA), that learns rich gene representations from single-cell data using a dictionary of diffusion wavelets on the cell-cell graph. GSPA enables characterization of genes based on their patterning and localization on the cellular manifold. We motivate and demonstrate the efficacy of GSPA as a framework for diverse biological tasks, such as capturing gene co-expression modules, condition-specific enrichment and perturbation-specific gene-gene interactions. Then we showcase the broad utility of gene representations derived from GSPA, including for cell-cell communication (GSPA-LR), spatial transcriptomics (GSPA-multimodal) and patient response (GSPA-Pt) analysis.

Identifying somatic fingerprints of cancers defined by germline and environmental risk factors

Numerous studies over the past generation have identified germline variants that increase specific cancer risks. Simultaneously, a revolution in sequencing technology has permitted high-throughput annotations of somatic genomes characterizing individual tumors. However, examining the relationship between germline variants and somatic alteration patterns is hugely challenged by the large numbers of variants in a typical tumor, the rarity of most individual variants, and the heterogeneity of tumor somatic fingerprints. In this article, we propose statistical methodology that frames the investigation of germline-somatic relationships in an interpretable manner. The method uses meta-features embodying biological contexts of individual somatic alterations to implicitly group rare mutations. Our team has used this technique previously through a multilevel regression model to diagnose with high accuracy tumor site of origin. Herein, we further leverage topic models from computational linguistics to achieve interpretable lower-dimensional embeddings of the meta-features. We demonstrate how the method can identify distinctive somatic profiles linked to specific germline variants or environmental risk factors. We illustrate the method using The Cancer Genome Atlas whole-exome sequencing data to characterize somatic tumor fingerprints in breast cancer patients with germline BRCA1/2 mutations and in head and neck cancer patients exposed to human papillomavirus.

Coordinated protein modules define DNA damage responses to carboplatin at single cell resolution in human ovarian carcinoma models

Tubo-ovarian high-grade serous carcinoma (HGSC) is the most lethal gynecological malignancy and frequently responds to platinum-based chemotherapy because of common genetic and somatic impairment of DNA damage repair (DDR) pathways. The mechanisms of clinical platinum resistance are diverse and poorly molecularly defined. Consequently, there are no biomarkers or medicines that improve patient outcomes. Herein we use single cell mass cytometry (CyTOF) to systematically evaluate the phosphorylation and abundance of proteins known to participate in the DNA damage response (DDR). Single cell analyses of highly characterized HGSC cell lines that phenocopy human patients show that cells with comparable levels of intranuclear platinum, a proxy for carboplatin uptake, undergo different cell fates. Unsupervised analyses revealed a continuum of DDR responses. Decompositional methods were used to identify eight distinct protein modules of carboplatin resistance and sensitivity at single cell resolution. CyTOF profiling of primary and secondary platinum-resistance patient models shows that a complex DDR sensitivity module is strongly associated with response, suggesting it as a potential tool to clinically characterize complex drug resistance phenotypes.

scGO: interpretable deep neural network for cell status annotation and disease diagnosis

Machine learning has emerged as a transformative tool for elucidating cellular heterogeneity in single-cell RNA sequencing. However, a significant challenge lies in the "black box" nature of deep learning models, which obscures the decision-making process and limits interpretability in cell status annotation. In this study, we introduced scGO, a Gene Ontology (GO)-inspired deep learning framework designed to provide interpretable cell status annotation for scRNA-seq data. scGO employs sparse neural networks to leverage the intrinsic biological relationships among genes, transcription factors, and GO terms, significantly augmenting interpretability and reducing computational cost. scGO outperforms state-of-the-art methods in the precise characterization of cell subtypes across diverse datasets. Our extensive experimentation across a spectrum of scRNA-seq datasets underscored the remarkable efficacy of scGO in disease diagnosis, prediction of developmental stages, and evaluation of disease severity and cellular senescence status. Furthermore, we incorporated in silico individual gene manipulations into the scGO model, introducing an additional layer for discovering therapeutic targets. Our results provide an interpretable model for accurately annotating cell status, capturing latent biological knowledge, and informing clinical practice.

A multi-regional human brain atlas of chromatin accessibility and gene expression facilitates promoter-isoform resolution genetic fine-mapping

Brain region- and cell-specific transcriptomic and epigenomic features are associated with heritability for neuropsychiatric traits, but a systematic view, considering cortical and subcortical regions, is lacking. Here, we provide an atlas of chromatin accessibility and gene expression profiles in neuronal and non-neuronal nuclei across 25 distinct human cortical and subcortical brain regions from 6 neurotypical controls. We identified extensive gene expression and chromatin accessibility differences across brain regions, including variation in alternative promoter-isoform usage and enhancer-promoter interactions. Genes with distinct promoter-isoform usage across brain regions were strongly enriched for neuropsychiatric disease risk variants. Moreover, we built enhancer-promoter interactions at promoter-isoform resolution across different brain regions and highlighted the contribution of brain region-specific and promoter-isoform-specific regulation to neuropsychiatric disorders. Including promoter-isoform resolution uncovers additional distal elements implicated in the heritability of diseases, thereby increasing the power to fine-map risk genes. Our results provide a valuable resource for studying molecular regulation across multiple regions of the human brain and underscore the importance of considering isoform information in gene regulation.

Molecular convergence of risk variants for congenital heart defects leveraging a regulatory map of the human fetal heart

Congenital heart defects (CHD) arise in part due to inherited genetic variants that alter genes and noncoding regulatory elements in the human genome. These variants are thought to act during fetal development to influence the formation of different heart structures. However, identifying the genes, pathways, and cell types that mediate these effects has been challenging due to the immense diversity of cell types involved in heart development as well as the superimposed complexities of interpreting noncoding sequences. As such, understanding the molecular functions of both noncoding and coding variants remains paramount to our fundamental understanding of cardiac development and CHD. Here, we created a gene regulation map of the healthy human fetal heart across developmental time, and applied it to interpret the functions of variants associated with CHD and quantitative cardiac traits. We collected single-cell multiomic data from 734,000 single cells sampled from 41 fetal hearts spanning post-conception weeks 6 to 22, enabling the construction of gene regulation maps in 90 cardiac cell types and states, including rare populations of cardiac conduction cells. Through an unbiased analysis of all 90 cell types, we find that both rare coding variants associated with CHD and common noncoding variants associated with valve traits converge to affect valvular interstitial cells (VICs). VICs are enriched for high expression of known CHD genes previously identified through mapping of rare coding variants. Eight CHD genes, as well as other genes in similar molecular pathways, are linked to common noncoding variants associated with other valve diseases or traits via enhancers in VICs. In addition, certain common noncoding variants impact enhancers with activities highly specific to particular subanatomic structures in the heart, illuminating how such variants can impact specific aspects of heart structure and function. Together, these results implicate new enhancers, genes, and cell types in the genetic etiology of CHD, identify molecular convergence of common noncoding and rare coding variants on VICs, and suggest a more expansive view of the cell types instrumental in genetic risk for CHD, beyond the working cardiomyocyte. This regulatory map of the human fetal heart will provide a foundational resource for understanding cardiac development, interpreting genetic variants associated with heart disease, and discovering targets for cell-type specific therapies.

Single cell and spatial transcriptomics highlight the interaction of club-like cells with immunosuppressive myeloid cells in prostate cancer

Prostate cancer treatment resistance is a significant challenge facing the field. Genomic and transcriptomic profiling have partially elucidated the mechanisms through which cancer cells escape treatment, but their relation toward the tumor microenvironment (TME) remains elusive. Here we present a comprehensive transcriptomic landscape of the prostate TME at multiple points in the standard treatment timeline employing single-cell RNA-sequencing and spatial transcriptomics data from 120 patients. We identify club-like cells as a key epithelial cell subtype that acts as an interface between the prostate and the immune system. Tissue areas enriched with club-like cells have depleted androgen signaling and upregulated expression of luminal progenitor cell markers. Club-like cells display a senescence-associated secretory phenotype and their presence is linked to increased polymorphonuclear myeloid-derived suppressor cell (PMN-MDSC) activity. Our results indicate that club-like cells are associated with myeloid inflammation previously linked to androgen deprivation therapy resistance, providing a rationale for their therapeutic targeting.

Multiplexed perturbation of yew reveals cryptic proteins that enable a total biosynthesis of baccatin III and Taxol precursors

Plants make complex and potent therapeutic molecules, but difficulties in sourcing from natural producers or chemical synthesis can challenge their use in the clinic. A prominent example is the anti-cancer therapeutic paclitaxel (Taxol ® ). Identification of the full paclitaxel biosynthetic pathway would enable heterologous drug production, but it has eluded discovery despite a half century of intensive research. Within the search space of Taxus' large, enzyme-rich genome, we suspected the complex paclitaxel pathway would be difficult to resolve using conventional gene co-expression analysis and small sample sets. To improve the resolution of gene set identification, we developed a multiplexed perturbation strategy to transcriptionally profile cell states spanning tissues, cell types, developmental stages, and elicitation conditions. This approach revealed a set of paclitaxel biosynthetic genes that segregate into expression modules that suggest consecutive biosynthetic sub-pathways. These modules resolved seven new genes that, when combined with previously known enzymes, are sufficient for the de novo biosynthesis and isolation of baccatin III, an industrial precursor for Taxol, in Nicotiana benthamiana leaves at levels comparable to the natural abundance in Taxus needles. Included are taxane 1β-hydroxylase (T1βH), taxane 9α-hydroxylase (T9αH), taxane 7β- O -acyltransferase (T7ΑΤ), taxane 7β- O -deacetylase (T7dA), taxane 9α- O -deacetylase (T9dA), and taxane 9-oxidase (T9ox). Importantly, the T9αH we discovered is distinct and independently evolved from those recently reported, which failed to yield baccatin III with downstream enzymes. Unexpectedly, we also found a nuclear transport factor 2 (NTF2)-like protein (FoTO1) crucial for high yields of taxanes; this gene promotes the formation of the desired product during the first taxane oxidation step, resolving a longstanding bottleneck in paclitaxel pathway reconstitution. Together with a new β-phenylalanine-CoA-ligase, the eight genes discovered in this study enables the complete reconstitution of 3'- N -debenzoyl-2'-deoxy-paclitaxel with a 20-enzyme pathway in Nicotiana plants. More broadly, we establish a generalizable approach for pathway discovery that scales the power of co-expression studies to match the complexity of specialized metabolism, enabling discovery of gene sets responsible for high-value biological functions.

INTERSPECIES ORGANOIDS REVEAL HUMAN-SPECIFIC MOLECULAR FEATURES OF DOPAMINERGIC NEURON DEVELOPMENT AND VULNERABILITY

The disproportionate expansion of telencephalic structures during human evolution involved tradeoffs that imposed greater connectivity and metabolic demands on midbrain dopaminergic neurons. Despite the central role of dopaminergic neurons in human-enriched disorders, molecular specializations associated with human-specific features and vulnerabilities of the dopaminergic system remain unexplored. Here, we establish a phylogeny-in-a-dish approach to examine gene regulatory evolution by differentiating pools of human, chimpanzee, orangutan, and macaque pluripotent stem cells into ventral midbrain organoids capable of forming long-range projections, spontaneous activity, and dopamine release. We identify human-specific gene expression changes related to axonal transport of mitochondria and reactive oxygen species buffering and candidate cis- and trans-regulatory mechanisms underlying gene expression divergence. Our findings are consistent with a model of evolved neuroprotection in response to tradeoffs related to brain expansion and could contribute to the discovery of therapeutic targets and strategies for treating disorders involving the dopaminergic system.

Hallmarks of uterine receptivity predate placental mammals

Embryo implantation requires tightly coordinated signaling between the blastocyst and the endometrium, and is crucial for the establishment of a uteroplacental unit that persists until term in eutherian mammals. In contrast, marsupials, with a unique life cycle and short gestation, make only brief fetal-maternal contact and lack implantation. To better understand the evolutionary link between eutherian implantation and its ancestral equivalent in marsupials, we compare single-cell transcriptomes from the receptive and non-receptive endometrium of the mouse and guinea pig with that of the opossum, a marsupial. We identify substantial differences between rodent peri-implantation endometrium and opossum placental attachment, including differences in the diversity and abundance of stromal and epithelial cells which parallel the difference between histotrophic and hemotrophic provisioning strategies. We also identify a window of conserved epithelial gene expression between the opossum shelled blastocyst stage and rodent peri-implantation, including IHH and LIF . We find strong conservation of blastocyst proteases, steroid synthetases, Wnt and BMP signals between eutherians and the opossum despite its lack of implantation. Finally, we show that the signaling repertoire of the maternal uterine epithelium during implantation displays substantial overlap with that of the post-implantation placental trophoblast, suggesting that the fetal trophoblast can compensate for the loss of endometrial epithelium in eutherian invasive placentation. Together, our results suggest that eutherian implantation primarily involved the re-wiring of maternal signaling networks, some of which were already present in the therian ancestor, and points towards an essential role of maternal innovations in the evolution of invasive placentation.

HiDDEN: a machine learning method for detection of disease-relevant populations in case-control single-cell transcriptomics data

In case-control single-cell RNA-seq studies, sample-level labels are transferred onto individual cells, labeling all case cells as affected, when in reality only a small fraction of them may actually be perturbed. Here, using simulations, we demonstrate that the standard approach to single cell analysis fails to isolate the subset of affected case cells and their markers when either the affected subset is small, or when the strength of the perturbation is mild. To address this fundamental limitation, we introduce HiDDEN, a computational method that refines the case-control labels to accurately reflect the perturbation status of each cell. We show HiDDEN's superior ability to recover biological signals missed by the standard analysis workflow in simulated ground truth datasets of cell type mixtures. When applied to a dataset of human multiple myeloma precursor conditions, HiDDEN recapitulates the expert manual annotation and discovers malignancy in early stage samples missed in the original analysis. When applied to a mouse model of demyelination, HiDDEN identifies an endothelial subpopulation playing a role in early stage blood-brain barrier dysfunction. We anticipate that HiDDEN should find wide usage in contexts that require the detection of subtle transcriptional changes in cell types across conditions.

Disentanglement of single-cell data with biolord

Biolord is a deep generative method for disentangling single-cell multi-omic data to known and unknown attributes, including spatial, temporal and disease states, used to reveal the decoupled biological signatures over diverse single-cell modalities and biological systems. By virtually shifting cells across states, biolord generates experimentally inaccessible samples, outperforming state-of-the-art methods in predictions of cellular response to unseen drugs and genetic perturbations. Biolord is available at https://github.com/nitzanlab/biolord .

Single-cell integration reveals metaplasia in inflammatory gut diseases

The gastrointestinal tract is a multi-organ system crucial for efficient nutrient uptake and barrier immunity. Advances in genomics and a surge in gastrointestinal diseases1,2 has fuelled efforts to catalogue cells constituting gastrointestinal tissues in health and disease3. Here we present systematic integration of 25 single-cell RNA sequencing datasets spanning the entire healthy gastrointestinal tract in development and in adulthood. We uniformly processed 385 samples from 189 healthy controls using a newly developed automated quality control approach (scAutoQC), leading to a healthy reference atlas with approximately 1.1 million cells and 136 fine-grained cell states. We anchor 12 gastrointestinal disease datasets spanning gastrointestinal cancers, coeliac disease, ulcerative colitis and Crohn's disease to this reference. Utilizing this 1.6 million cell resource (gutcellatlas.org), we discover epithelial cell metaplasia originating from stem cells in intestinal inflammatory diseases with transcriptional similarity to cells found in pyloric and Brunner's glands. Although previously linked to mucosal healing4, we now implicate pyloric gland metaplastic cells in inflammation through recruitment of immune cells including T cells and neutrophils. Overall, we describe inflammation-induced changes in stem cells that alter mucosal tissue architecture and promote further inflammation, a concept applicable to other tissues and diseases.

SPP1+ TAM Regulates the Metastatic Colonization of CXCR4+ Metastasis-Associated Tumor Cells by Remodeling the Lymph Node Microenvironment

Lymph node metastasis, the initial step in distant metastasis, represents a primary contributor to mortality in patients diagnosed with oral squamous cell carcinoma (OSCC). However, the underlying mechanisms of lymph node metastasis in OSCC remain incompletely understood. Here, the transcriptomes of 56 383 single cells derived from paired tissues of six OSCC patients are analyzed. This study founds that CXCR4+ epithelial cells, identified as highly malignant disseminated tumor cells (DTCs), exhibited a propensity for lymph node metastasis. Importantly, a distinct subset of tumor-associated macrophages (TAMs) characterized by exclusive expression of phosphoprotein 1 (SPP1) is discovered. These TAMs may remodel the metastatic lymph node microenvironment by potentially activating fibroblasts and promoting T cell exhaustion through SPP1-CD44 and CD155-CD226 ligand-receptor interactions, thereby facilitating colonization and proliferation of disseminated tumor cells. The research advanced the mechanistic understanding of metastatic tumor microenvironment (TME) and provided a foundation for the development of personalized treatments for OSCC patients with metastasis.

Single-cell transcriptional mapping reveals genetic and non-genetic determinants of aberrant differentiation in AML

In acute myeloid leukemia (AML), genetic mutations distort hematopoietic differentiation, resulting in the accumulation of leukemic blasts. Yet, it remains unclear how these mutations intersect with cellular origins and whether they converge upon similar differentiation patterns. Single-cell RNA sequencing (scRNA-seq) has enabled high-resolution mapping of the relationship between leukemia and normal cell states, yet this application is hampered by imprecise reference maps of normal hematopoiesis and small sample sizes among patient cohorts. As a first step we constructed a reference atlas of human bone marrow hematopoiesis from 263,519 single-cell transcriptomes spanning 55 cellular states, that was benchmarked against independent datasets of immunophenotypically pure hematopoietic stem and progenitor cells. Using this reference atlas, we mapped over 1.2 million single-cell transcriptomes spanning 318 AML, mixed phenotype acute leukemia (MPAL), and acute erythroid leukemia (AEL) samples. This large-scale analysis, together with systematic mapping of genotype-to-phenotype associations between driver mutations and differentiation landscapes, revealed convergence of diverse genetic alterations on twelve recurrent patterns of aberrant differentiation in AML. This included unconventional lymphoid and erythroid priming linked to RUNX1 and TP53 mutations, respectively. We also identified non-genetic determinants of AML differentiation such as two subgroups of KMT2A-rearranged AML that differ in the identity of their leukemic stem cells (LSCs), likely reflecting distinct cellular origins. Furthermore, distinct LSC-driven hierarchies can co-exist within individual patients, providing insights into AML evolution. Together, precise mapping of normal and malignant cell states provides a framework for advancing the study and disease classification of hematologic malignancies thereby informing therapy development.

A longitudinal single-cell atlas of anti-tumour necrosis factor treatment in inflammatory bowel disease

Precision medicine in immune-mediated inflammatory diseases (IMIDs) requires a cellular understanding of treatment response. We describe a therapeutic atlas for Crohn's disease (CD) and ulcerative colitis (UC) following adalimumab, an anti-tumour necrosis factor (anti-TNF) treatment. We generated ~1 million single-cell transcriptomes, organised into 109 cell states, from 216 gut biopsies (41 subjects), revealing disease-specific differences. A systems biology-spatial analysis identified granuloma signatures in CD and interferon (IFN)-response signatures localising to T cell aggregates and epithelial damage in CD and UC. Pretreatment differences in epithelial and myeloid compartments were associated with remission outcomes in both diseases. Longitudinal comparisons demonstrated disease progression in nonremission: myeloid and T cell perturbations in CD and increased multi-cellular IFN signalling in UC. IFN signalling was also observed in rheumatoid arthritis (RA) synovium with a lymphoid pathotype. Our therapeutic atlas represents the largest cellular census of perturbation with the most common biologic treatment, anti-TNF, across multiple inflammatory diseases.

Consensus tissue domain detection in spatial omics data using multiplex image labeling with regional morphology (MILWRM)

Spatially resolved molecular assays provide high dimensional genetic, transcriptomic, proteomic, and epigenetic information in situ and at various resolutions. Pairing these data across modalities with histological features enables powerful studies of tissue pathology in the context of an intact microenvironment and tissue structure. Increasing dimensions across molecular analytes and samples require new data science approaches to functionally annotate spatially resolved molecular data. A specific challenge is data-driven cross-sample domain detection that allows for analysis within and between consensus tissue compartments across high volumes of multiplex datasets stemming from tissue atlasing efforts. Here, we present MILWRM (multiplex image labeling with regional morphology)-a Python package for rapid, multi-scale tissue domain detection and annotation at the image- or spot-level. We demonstrate MILWRM's utility in identifying histologically distinct compartments in human colonic polyps, lymph nodes, mouse kidney, and mouse brain slices through spatially-informed clustering in two different spatial data modalities from different platforms. We used tissue domains detected in human colonic polyps to elucidate the molecular distinction between polyp subtypes, and explored the ability of MILWRM to identify anatomical regions of the brain tissue and their respective distinct molecular profiles.

Gene regulatory network structure informs the distribution of perturbation effects

Gene regulatory networks (GRNs) govern many core developmental and biological processes underlying human complex traits. Even with broad-scale efforts to characterize the effects of molecular perturbations and interpret gene coexpression, it remains challenging to infer the architecture of gene regulation in a precise and efficient manner. Key properties of GRNs, like hierarchical structure, modular organization, and sparsity, provide both challenges and opportunities for this objective. Here, we seek to better understand properties of GRNs using a new approach to simulate their structure and model their function. We produce realistic network structures with a novel generating algorithm based on insights from small-world network theory, and we model gene expression regulation using stochastic differential equations formulated to accommodate modeling molecular perturbations. With these tools, we systematically describe the effects of gene knockouts within and across GRNs, finding a subset of networks that recapitulate features of a recent genome-scale perturbation study. With deeper analysis of these exemplar networks, we consider future avenues to map the architecture of gene expression regulation using data from cells in perturbed and unperturbed states, finding that while perturbation data are critical to discover specific regulatory interactions, data from unperturbed cells may be sufficient to reveal regulatory programs.

Spatial oncology: Translating contextual biology to the clinic

Microscopic examination of cells in their tissue context has been the driving force behind diagnostic histopathology over the past two centuries. Recently, the rise of advanced molecular biomarkers identified through single cell profiling has increased our understanding of cellular heterogeneity in cancer but have yet to significantly impact clinical care. Spatial technologies integrating molecular profiling with microenvironmental features are poised to bridge this translational gap by providing critical in situ context for understanding cellular interactions and organization. Here, we review how spatial tools have been used to study tumor ecosystems and their clinical applications. We detail findings in cell-cell interactions, microenvironment composition, and tissue remodeling for immune evasion and therapeutic resistance. Additionally, we highlight the emerging role of multi-omic spatial profiling for characterizing clinically relevant features including perineural invasion, tertiary lymphoid structures, and the tumor-stroma interface. Finally, we explore strategies for clinical integration and their augmentation of therapeutic and diagnostic approaches.

The conserved genetic program of male germ cells uncovers ancient regulators of human spermatogenesis

Male germ cells share a common origin across animal species, therefore they likely retain a conserved genetic program that defines their cellular identity. However, the unique evolutionary dynamics of male germ cells coupled with their widespread leaky transcription pose significant obstacles to the identification of the core spermatogenic program. Through network analysis of the spermatocyte transcriptome of vertebrate and invertebrate species, we describe the conserved evolutionary origin of metazoan male germ cells at the molecular level. We estimate the average functional requirement of a metazoan male germ cell to correspond to the expression of approximately 10,000 protein-coding genes, a third of which defines a genetic scaffold of deeply conserved genes that has been retained throughout evolution. Such scaffold contains a set of 79 functional associations between 104 gene expression regulators that represent a core component of the conserved genetic program of metazoan spermatogenesis. By genetically interfering with the acquisition and maintenance of male germ cell identity, we uncover 161 previously unknown spermatogenesis genes and three new potential genetic causes of human infertility. These findings emphasize the importance of evolutionary history on human reproductive disease and establish a cross-species analytical pipeline that can be repurposed to other cell types and pathologies.

SIMVI reveals intrinsic and spatial-induced states in spatial omics data

Spatial omics technologies enable the analysis of gene expression and interaction dynamics in relation to tissue structure and function. However, existing computational methods may not properly distinguish cellular intrinsic variability and intercellular interactions, and may thus fail to capture spatial regulations for further biological discoveries. Here, we present Spatial Interaction Modeling using Variational Inference (SIMVI), an annotation-free framework that disentangles cell intrinsic and spatial-induced latent variables for modeling gene expression in spatial omics data. We derive theoretical support for SIMVI in disentangling intrinsic and spatial-induced variations. By this disentanglement, SIMVI enables estimation of spatial effects (SE) at a single-cell resolution, and opens up various opportunities for novel downstream analyses. To demonstrate the potential of SIMVI, we applied SIMVI to spatial omics data from diverse platforms and tissues (MERFISH human cortex, Slide-seqv2 mouse hippocampus, Slide-tags human tonsil, spatial multiome human melanoma, cohort-level CosMx melanoma). In all tested datasets, SIMVI effectively disentangles variations and infers accurate spatial effects compared with alternative methods. Moreover, on these datasets, SIMVI uniquely uncovers complex spatial regulations and dynamics of biological significance. In the human tonsil data, SIMVI illuminates the cyclical spatial dynamics of germinal center B cells during maturation. Applying SIMVI to both RNA and ATAC modalities of the multiome melanoma data reveals potential tumor epigenetic reprogramming states. Application of SIMVI on our newly-collected cohort-level CosMx melanoma dataset uncovers space-and-outcome-dependent macrophage states and the underlying cellular communication machinery in the tumor microenvironments.

Spatiotemporal modeling reveals high-resolution invasion states in glioblastoma

BACKGROUND

Diffuse invasion of glioblastoma cells through normal brain tissue is a key contributor to tumor aggressiveness, resistance to conventional therapies, and dismal prognosis in patients. A deeper understanding of how components of the tumor microenvironment (TME) contribute to overall tumor organization and to programs of invasion may reveal opportunities for improved therapeutic strategies.

RESULTS

Towards this goal, we apply a novel computational workflow to a spatiotemporally profiled GBM xenograft cohort, leveraging the ability to distinguish human tumor from mouse TME to overcome previous limitations in the analysis of diffuse invasion. Our analytic approach, based on unsupervised deconvolution, performs reference-free discovery of cell types and cell activities within the complete GBM ecosystem. We present a comprehensive catalogue of 15 tumor cell programs set within the spatiotemporal context of 90 mouse brain and TME cell types, cell activities, and anatomic structures. Distinct tumor programs related to invasion align with routes of perivascular, white matter, and parenchymal invasion. Furthermore, sub-modules of genes serving as program network hubs are highly prognostic in GBM patients.

CONCLUSION

The compendium of programs presented here provides a basis for rational targeting of tumor and/or TME components. We anticipate that our approach will facilitate an ecosystem-level understanding of the immediate and long-term consequences of such perturbations, including the identification of compensatory programs that will inform improved combinatorial therapies.

Scalable, compressed phenotypic screening using pooled perturbations

High-throughput phenotypic screens using biochemical perturbations and high-content readouts are constrained by limitations of scale. To address this, we establish a method of pooling exogenous perturbations followed by computational deconvolution to reduce required sample size, labor and cost. We demonstrate the increased efficiency of compressed experimental designs compared to conventional approaches through benchmarking with a bioactive small-molecule library and a high-content imaging readout. We then apply compressed screening in two biological discovery campaigns. In the first, we use early-passage pancreatic cancer organoids to map transcriptional responses to a library of recombinant tumor microenvironment protein ligands, uncovering reproducible phenotypic shifts induced by specific ligands distinct from canonical reference signatures and correlated with clinical outcome. In the second, we identify the pleotropic modulatory effects of a chemical compound library with known mechanisms of action on primary human peripheral blood mononuclear cell immune responses. In sum, our approach empowers phenotypic screens with information-rich readouts to advance drug discovery efforts and basic biological inquiry.

C-ziptf: stable tensor factorization for zero-inflated multi-dimensional genomics data

In the past two decades, genomics has advanced significantly, with single-cell RNA-sequencing (scRNA-seq) marking a pivotal milestone. ScRNA-seq provides unparalleled insights into cellular diversity and has spurred diverse studies across multiple conditions and samples, resulting in an influx of complex multidimensional genomics data. This highlights the need for robust methodologies capable of handling the complexity and multidimensionality of such genomics data. Furthermore, single-cell data grapples with sparsity due to issues like low capture efficiency and dropout effects. Tensor factorizations (TF) have emerged as powerful tools to unravel the complex patterns from multi-dimensional genomics data. Classic TF methods, based on maximum likelihood estimation, struggle with zero-inflated count data, while the inherent stochasticity in TFs further complicates result interpretation and reproducibility. Our paper introduces Zero Inflated Poisson Tensor Factorization (ZIPTF), a novel method for high-dimensional zero-inflated count data factorization. We also present Consensus-ZIPTF (C-ZIPTF), merging ZIPTF with a consensus-based approach to address stochasticity. We evaluate our proposed methods on synthetic zero-inflated count data, simulated scRNA-seq data, and real multi-sample multi-condition scRNA-seq datasets. ZIPTF consistently outperforms baseline matrix and tensor factorization methods, displaying enhanced reconstruction accuracy for zero-inflated data. When dealing with high probabilities of excess zeros, ZIPTF achieves up to 2.4 × better accuracy. Moreover, C-ZIPTF notably enhances the factorization's consistency. When tested on synthetic and real scRNA-seq data, ZIPTF and C-ZIPTF consistently uncover known and biologically meaningful gene expression programs. Access our data and code at: https://github.com/klarman-cell-observatory/scBTF and https://github.com/klarman-cell-observatory/scbtf_experiments .

Integrating genomics and AI to uncover molecular targets for mRNA vaccine development in lupus nephritis

Lupus nephritis (LN), a complex complication of systemic lupus erythematosus, requires in-depth cellular and molecular analysis for advanced treatment strategies, including mRNA vaccine development. In this study, we analyzed single-cell RNA sequencing data from 24 LN patients and 10 healthy controls, supplemented by bulk RNA-seq data from additional LN patients and controls. By applying non-negative matrix factorization (NMF), we identified four distinct leukocyte meta-programs in LN, highlighting diverse immune functions and potential mRNA vaccine targets. Utilizing 12 machine learning algorithms, we developed 417 predictive models incorporating gene sets linked to key biological pathways, such as MTOR signaling, autophagy, Toll-like receptor, and adaptive immunity pathways. These models were instrumental in identifying potential targets for mRNA vaccine development. Our functional network analysis further revealed intricate gene interactions, providing novel insights into the molecular basis of LN. Additionally, we validated the mRNA expression levels of potential vaccine targets across multiple cohorts and correlated them with clinical parameters such as the glomerular filtration rate (GFR) and pathological stage. This study represents a significant advance in LN research by merging single-cell genomics with the precision of NMF and machine learning, broadening our understanding of LN at the cellular and molecular levels. More importantly, our findings shed light on the development of targeted mRNA vaccines, offering new possibilities for diagnostics and therapeutics for this complex autoimmune disease.

Network modeling links kidney developmental programs and the cancer type-specificity of VHL mutations

Elucidating the molecular dependencies behind the cancer-type specificity of driver mutations may reveal new therapeutic opportunities. We hypothesized that developmental programs would impact the transduction of oncogenic signaling activated by a driver mutation and shape its cancer-type specificity. Therefore, we designed a computational analysis framework by combining single-cell gene expression profiles during fetal organ development, latent factor discovery, and information theory-based differential network analysis to systematically identify transcription factors that selectively respond to driver mutations under the influence of organ-specific developmental programs. After applying this approach to VHL mutations, which are highly specific to clear cell renal cell carcinoma (ccRCC), we revealed important regulators downstream of VHL mutations in ccRCC and used their activities to cluster patients with ccRCC into three subtypes. This classification revealed a more significant difference in prognosis than the previous mRNA profile-based method and was validated in an independent cohort. Moreover, we found that EP300, a key epigenetic factor maintaining the regulatory network of the subtype with the worst prognosis, can be targeted by a small inhibitor, suggesting a potential treatment option for a subset of patients with ccRCC. This work demonstrated an intimate relationship between organ development and oncogenesis from the perspective of systems biology, and the method can be generalized to study the influence of other biological processes on cancer driver mutations.