Revealing the underlying cell-state landscape from single-cell data requires overcoming the critical obstacles of batch integration, denoising and dimensionality reduction. Here we present CONCORD, a unified framework that simultaneously addresses these challenges within a single self-supervised model. At its core, CONCORD implements a probabilistic sampling strategy that corrects batch effects through dataset-aware sampling and enhances biological resolution through hard-negative sampling. Using only a minimalist neural network with a single hidden layer and contrastive learning, CONCORD surpasses state-of-the-art performance without relying on deep architectures, auxiliary losses or external supervision. It seamlessly integrates data across batches, technologies and even species to generate high-resolution cell atlases. The resulting latent representations are denoised and biologically meaningful, capturing gene coexpression programs, revealing detailed lineage trajectories and preserving both local geometric relationships and global topological structures. We demonstrate CONCORD's broad applicability across diverse datasets, establishing it as a general-purpose framework for learning unified, high-fidelity representations of cellular identity and dynamics.

Testosterone is the most studied androgen in hormone replacement therapy in postmenopausal women. Many transgender men also receive long-term testosterone replacement therapy (TRT) and have high serum testosterone levels compared to the low levels seen in cisgender women. Compared to other sex hormones such as estrogen and progesterone, the effects of testosterone on the mammary gland have been relatively understudied and there is little data regarding the long term safety of this treatment. Comparison of mammary glands from transgender men on TRT and cisgender women can reveal the effects of testosterone treatment on mammary gland biology and provide critical information regarding the long-term effects of TRT on patient health and disease outcomes. In this study, we performed single-cell RNA sequencing of breast tissues from a demographics-matched cohort of cisgender women and transgender men on TRT. Surprisingly, participants on TRT had unchanged serum levels of estradiol compared to controls. Among the observed transcriptional differences for participants on TRT were a dramatically reduced expression of genes downstream of estrogen signaling pathways in hormone receptor positive (HR+) luminal epithelial cells, as well as a decreased overall menstrual cycle-related hormone signaling. We confirmed this finding experimentally by showing reduced expression of progesterone receptor a/b, a prominent marker of estrogen signaling, in donors on TRT. Our results support the hypothesis that high levels of testosterone in transgender men on TRT suppress sex hormone signaling in the breast as seen on their impact on HR+ mammary epithelial cells, with implications for TRT as an antagonist of estrogen signaling and protection against breast cancer.

Large-scale scATAC-seq experiments are challenging because of their costs, lengthy protocols, and confounding batch effects. Several sample multiplexing technologies aim to address these challenges, but do not remove batch effects introduced when performing transposition reactions in parallel. We demonstrate that sample-to-sample variability in nuclei-to-Tn5 ratios is a major cause of batch effects and develop MULTI-ATAC, a multiplexing method that pools samples prior to transposition, as a solution. MULTI-ATAC provides high accuracy in sample classification and doublet detection while eliminating batch effects associated with variable nucleus-to-Tn5 ratio. We illustrate the power of MULTI-ATAC by performing a 96-plex multiomic drug assay targeting epigenetic remodelers in a model of primary immune cell activation, uncovering tens of thousands of drug-responsive chromatin regions, cell-type specific effects, and potent differences between matched inhibitors and degraders. MULTI-ATAC therefore enables batch-free and scalable scATAC-seq workflows, providing deeper insights into complex biological processes and potential therapeutic targets.

Gene regulatory networks within cells modulate the expression of the genome in response to signals and changing environmental conditions. Reconstructions of gene regulatory networks can reveal the information processing and control principles used by cells to maintain homeostasis and execute cell-state transitions. Here, we introduce a computational framework, D-SPIN, that generates quantitative models of generegulatory networks from single-cell mRNA-seq data sets collected across thousands of distinct perturbation conditions. D-SPIN models the cell as a collection of interacting gene-expression programs, and constructs a probabilistic model to infer regulatory interactions between gene-expression programs and external perturbations. Using large Perturb-seq and drug-response datasets, we demonstrate that D-SPIN models reveal the organization of cellular pathways, sub-functions of macromolecular complexes, and the logic of cellular regulation of transcription, translation, metabolism, and protein degradation in response to gene knockdown perturbations. D-SPIN can also be applied to dissect drug response mechanisms in heterogeneous cell populations, elucidating how combinations of immunomodulatory drugs can induce novel cell states through additive recruitment of gene expression programs. D-SPIN provides a computational framework for constructing interpretable models of gene-regulatory networks to reveal principles of cellular information processing and physiological control.

Single-cell studies on breast tissue have contributed to a change in our understanding of breast epithelial diversity that has, in turn, precipitated a lack of consensus on breast cell types. The confusion surrounding this issue highlights a possible challenge for advancing breast atlas efforts. In this perspective, we present our consensus on the identities, properties, and naming conventions for breast epithelial cell types and propose goals for future atlas endeavors. Our proposals and their underlying thought processes aim to catalyze the adoption of a shared model for this tissue and to serve as guidance for other investigators facing similar challenges.

Three-dimensional (3D) printing can be beneficial to tissue engineers and the regenerative medicine community because of its potential to rapidly build elaborate 3D structures from cellular and material inks. However, predicting changes to the structure and pattern of printed tissues arising from the mechanical activity of constituent cells is technically and conceptually challenging. This perspective is targeted to scientists and engineers interested in 3D bioprinting, but from the point of view of cells and tissues as mechanically active living materials. The dynamic forces generated by cells present unique challenges compared to conventional manufacturing modalities but also offer profound opportunities through their capacity to self-organize. Consideration of self-organization following 3D printing takes the design and execution of bioprinting into the fourth dimension of cellular activity. We therefore propose a framework for dynamic bioprinting that spatiotemporally guides the underlying biology through reconfigurable material interfaces controlled by 3D printers.

In breast cancer (BC), radial alignment of collagen fibers at the tumor-matrix interface facilitates collective invasion of cancer cells into the surrounding stromal matrix, a critical step toward metastasis. Collagen remodeling is driven by proteases and cellular forces, mediated by matrix mechanical plasticity, or irreversible matrix deformation in response to force. However, the specific mechanisms causing collagen radial alignment remain unclear. Here, we study collective invasion of BC tumor spheroids in collagen-rich matrices. Increasing plasticity to BC-relevant ranges facilitates invasion, with increasing stiffness potentiating a transition from single cell to collective invasion. At enhanced plasticity, cells radially align collagen at the tumor-matrix interface prior to invasion. Surprisingly, cells migrate tangentially to the tumor-matrix interface in a swirling-like motion, perpendicular to the direction of alignment. Mechanistically, swirling generates local shear stresses, leading to distally propagating contractile radial stresses due to negative normal stress, an underappreciated property of collagen-rich matrices. These contractile stresses align collagen fibers radially, facilitating collective invasion. The basement membrane (BM), which separates epithelia from stroma in healthy tissues, acts as a mechanical insulator by preventing swirling cells from aligning collagen. Thus, after breaching the BM, swirling of BC cells at the tumor-stroma interface radially aligns collagen to facilitate invasion.

Complex and robust self-organization requires defined initial conditions and dynamic boundaries - neighboring tissues and extracellular matrix (ECM) that actively evolve to guide morphogenesis. A major challenge in tissue engineering is identifying material properties that mimic dynamic tissue boundaries but that are compatible with the engineering tools necessary for controlling the initial conditions of culture. Here we describe a highly tunable granular biomaterial, MAGIC matrix, that supports long-term bioprinting and gold-standard tissue self-organization. MAGIC matrix is designed for two temperature regimes: at 4 °C it exhibits reversible yield-stress behavior to support hours-long high-fidelity 3D printing without compromising cell viability; when transferred to cell culture at 37 °C, the material cross-links and exhibits viscoelasticity and stress relaxation that can be tuned to match numerous conditions, including that of reconstituted basement membrane matrices like Matrigel. We demonstrate that the timescale of stress relaxation and loss tangent are decoupled in MAGIC matrices, allowing us to test the role of stress relaxation rate and strain-dependence across formulations with identical storage and loss moduli. We find that fast absolute stress relaxation rates and large relative deformation magnitudes are required to optimize for morphogenesis. We apply optimized MAGIC matrices toward precise extrusion bioprinting of saturated cell suspensions directly into 3D culture. The ability to carefully control initial conditions for tissue growth yields dramatic increases in organoid reproducibility and complexity across multiple tissue types. We also fabricate perfusable 3D microphysiological systems that experience large strains in response to pressurization due to the compliant and dynamic tissue boundaries. Combined, our results both identify key parameters for optimal organoid morphogenesis in an engineered material and lay the foundation for fabricating more complex and reproducible tissue morphologies by canalizing their self-organization in both space and time.

Mutations in mitochondrial DNA (mtDNA) accumulate during aging and contribute to age-related conditions. High mtDNA copy number masks newly emerged recessive mutations; however, phenotypes develop when cellular levels of a mutant mtDNA rise above a critical threshold. The process driving this increase is unknown. Single-cell DNA sequencing of mouse and human hepatocytes detected increases in abundance of mutant alleles in sequences governing mtDNA replication. These alleles provided a replication advantage (drive) leading to accumulation of the affected genome along with a wide variety of associated passenger mutations, some of which are detrimental. The most prevalent human mtDNA disease variant, the 3243A>G allele, behaved as a driver, suggesting that drive underlies prevalence. We conclude that replicative drive amplifies linked mtDNA mutations to a threshold at which phenotypes are seen thereby promoting age-associated erosion of the mtDNA and influencing the transmission and progression of mitochondrial diseases.

Single-cell sample multiplexing technologies function by associating sample-specific barcode tags with cell-specific barcode tags, thereby increasing sample throughput, reducing batch effects, and decreasing reagent costs. Computational methods must then correctly associate cell-tags with sample-tags, but their performance deteriorates rapidly when working with datasets that are large, have imbalanced cell numbers across samples, or are noisy due to cross-contamination among sample tags - unavoidable features of many real-world experiments. Here we introduce deMULTIplex2, a mechanism-guided classification algorithm for multiplexed scRNA-seq data that successfully recovers many more cells across a spectrum of challenging datasets compared to existing methods. deMULTIplex2 is built on a statistical model of tag read counts derived from the physical mechanism of tag cross-contamination. Using generalized linear models and expectation-maximization, deMULTIplex2 probabilistically infers the sample identity of each cell and classifies singlets with high accuracy. Using Randomized Quantile Residuals, we show the model fits both simulated and real datasets. Benchmarking analysis suggests that deMULTIplex2 outperforms existing algorithms, especially when handling large and noisy single-cell datasets or those with unbalanced sample compositions.

Tissue folds are structural motifs critical to organ function. In the intestine, bending of a flat epithelium into a periodic pattern of folds gives rise to villi, finger -like protrusions that enable nutrient absorption. However, the molecular and mechanical processes driving villus morphogenesis remain unclear. Here, we identify an active mechanical mechanism that simultaneously patterns and folds the intestinal epithelium to initiate villus formation. At the cellular level, we find that PDGFRA+ subepithelial mesenchymal cells generate myosin IIdependent forces sufficient to produce patterned curvature in neighboring tissue interfaces. This symmetry -breaking process requires altered cell and extracellular matrix interactions that are enabled by matrix metalloproteinase-mediated tissue fluidization. Computational models, together with in vitro and in vivo experiments, revealed that these cellular features manifest at the tissue level as differences in interfacial tensions that promote mesenchymal aggregation and interface bending through a process analogous to the active dewetting of a thin liquid film.

Somatic mutations in mitochondrial genomes (mtDNA) accumulate exponentially during aging. Using single cell sequencing, we characterize the spectrum of age-accumulated mtDNA mutations in mouse and human liver and identify directional forces that accelerate the accumulation of mutations beyond the rate predicted by a neutral model. "Driver" mutations that give genomes a replicative advantage rose to high cellular abundance and carried along "passenger" mutations, some of which are deleterious. In addition, alleles that alter mtDNA-encoded proteins selectively increased in abundance overtime, strongly supporting the idea of a "destructive" selection that favors genomes lacking function. Overall, this combination of selective forces acting in hepatocytes promotes somatic accumulation of mutations in coding regions of mtDNA that are otherwise conserved in evolution. We propose that these selective processes could contribute to the population prevalence of mtDNA mutations, accelerate the course of heteroplasmic mitochondrial diseases and promote age-associated erosion of the mitochondrial genome.

Juvenile Dermatomyositis (JDM) is one of several childhood-onset autoimmune disorders characterized by a type I interferon response and autoantibodies. Treatment options are limited due to incomplete understanding of how the disease emerges from dysregulated cell states across the immune system. We therefore investigated the blood of JDM patients at different stages of disease activity using single-cell transcriptomics paired with surface protein expression. By immunophenotyping peripheral blood mononuclear cells, we observed skewing of the B cell compartment towards an immature naive state as a hallmark of JDM. Furthermore, we find that these changes in B cells are paralleled by signatures of Th2-mediated inflammation. Additionally, our work identified SIGLEC-1 expression in monocytes as a composite measure of heterogeneous type I interferon activity in disease. We applied network analysis to reveal that hyperactivation of the type I interferon response in all immune populations is coordinated with dysfunctional protein processing and regulation of cell death programming. This analysis separated the ubiquitously expressed type I interferon response into a central hub and revealed previously masked cell states. Together, these findings reveal the coordinated immune dysregulation underpinning JDM and provide novel insight into strategies for restoring balance in immune function.

Metastasis is the leading cause of cancer-related deaths. It is unclear how intratumor heterogeneity (ITH) contributes to metastasis and how metastatic cells adapt to distant tissue environments. The study of these adaptations is challenged by the limited access to patient material and a lack of experimental models that appropriately recapitulate ITH. To investigate metastatic cell adaptations and the contribution of ITH to metastasis, we analyzed single-cell transcriptomes of matched primary tumors and metastases from patient-derived xenograft models of breast cancer. We found profound transcriptional differences between the primary tumor and metastatic cells. Primary tumors upregulated several metabolic genes, whereas motility pathway genes were upregulated in micrometastases, and stress response signaling was upregulated during progression. Additionally, we identified primary tumor gene signatures that were associated with increased metastatic potential and correlated with patient outcomes. Immune-regulatory control pathways were enriched in poorly metastatic primary tumors, whereas genes involved in epithelial-mesenchymal transition were upregulated in highly metastatic tumors. We found that ITH was dominated by epithelial-mesenchymal plasticity (EMP), which presented as a dynamic continuum with intermediate EMP cell states characterized by specific genes such as CRYAB and S100A2. . Elevated expression of an intermediate EMP signature correlated with worse patient outcomes. Our findings identified inhibition of the intermediate EMP cell state as a potential therapeutic target to block metastasis.

A key aspect of nutrient absorption is the exquisite division of labor across the length of the small intestine, with individual classes of micronutrients taken up at different positions. For millennia, the small intestine was thought to comprise three segments with indefinite borders: the duodenum, jejunum, and ileum. By examining fine-scale longitudinal segmentation of the mouse and human small intestines, we identified transcriptional signatures and upstream regulatory factors that define five domains of nutrient absorption, distinct from the three traditional sections. Spatially restricted expression programs were most prominent in nutrient-absorbing enterocytes but initially arose in intestinal stem cells residing in three regional populations. While a core signature was maintained across mice and humans with different diets and environments, domain properties were influenced by dietary changes. We established the functions of Ppar-ẟ and Cdx1 in patterning lipid metabolism in distal domains and generated a predictive model of additional transcription factors that direct domain identity. Molecular domain identity can be detected with machine learning, representing the first systematic method to computationally identify specific intestinal regions in mice. These findings provide a foundational framework for the identity and control of longitudinal zonation of absorption along the proximal:distal small intestinal axis.

The gastrointestinal tract relies on the production, maturation, and transit of mucin to protect against pathogens and to lubricate the epithelial lining. Although the molecular and cellular mechanisms that regulate mucin production and movement are beginning to be understood, the upstream epithelial signals that contribute to mucin regulation remain unclear. Here, we report that the inflammatory cytokine tumor necrosis factor (TNF), generated by the epithelium, contributes to mucin homeostasis by regulating both cell differentiation and cystic fibrosis transmembrane conductance regulator (CFTR) activity. We used genetic mouse models and noninflamed samples from patients with inflammatory bowel disease (IBD) undergoing anti-TNF therapy to assess the effect of in vivo perturbation of TNF. We found that inhibition of epithelial TNF promotes the differentiation of secretory progenitor cells into mucus-producing goblet cells. Furthermore, TNF treatment and CFTR inhibition in intestinal organoids demonstrated that TNF promotes ion transport and luminal flow via CFTR. The absence of TNF led to slower gut transit times, which we propose results from increased mucus accumulation coupled with decreased luminal fluid pumping. These findings point to a TNF/CFTR signaling axis in the adult intestine and identify epithelial cell-derived TNF as an upstream regulator of mucin homeostasis.

Tissues comprise ordered arrangements of cells that can be surprisingly disordered in their details. How the properties of single cells and their microenvironment contribute to the balance between order and disorder at the tissue-scale remains poorly understood. Here, we address this question using the self-organization of human mammary organoids as a model. We find that organoids behave like a dynamic structural ensemble at the steady state. We apply a maximum entropy formalism to derive the ensemble distribution from three measurable parameters - the degeneracy of structural states, interfacial energy, and tissue activity (the energy associated with positional fluctuations). We link these parameters with the molecular and microenvironmental factors that control them to precisely engineer the ensemble across multiple conditions. Our analysis reveals that the entropy associated with structural degeneracy sets a theoretical limit to tissue order and provides new insight for tissue engineering, development, and our understanding of disease progression.