Scientists Map How Human Cells Coordinate Across Tissues and How Cancer Breaks This Order
A landmark single-cell atlas of 35 human tissues reveals 12 multicellular coordination modules—and shows how cancer systematically dismantles them.
Summary
Researchers at Peking University assembled a 2.3-million-cell transcriptomic atlas spanning 35 human tissues and developed a computational tool called CoVarNet to identify 12 recurring 'cellular modules' (CMs)—groups of cell types that co-occur and communicate in coordinated ways across tissues. These modules have distinct spatial organizations, age-related dynamics, and tissue preferences. In the spleen, two immune CMs show opposing trajectories with aging. In breast tissue, a menopausal fibroblast-driven transition was mapped. Critically, in cancer, two simultaneous changes occur: healthy tissue-specific modules are lost, and a convergent pro-tumor ecosystem emerges across cancer types. The findings establish fundamental organizing principles of multicellular coordination in health and disease.
Detailed Summary
Understanding how trillions of cells self-organize into functional tissues—and how that organization collapses in disease—is one of biology's central challenges. This landmark study from Peking University addresses that challenge at unprecedented scale, assembling a pan-tissue single-cell transcriptomic atlas of 2,293,951 high-quality cells from 706 healthy human samples across 35 tissues. Using a custom computational framework called CoVarNet, the authors identified 12 cross-tissue cellular modules (CMs): recurring, coordinated groupings of non-epithelial cell subsets that co-vary in abundance across samples and exhibit structured intercellular communication.
CoVarNet works by combining non-negative matrix factorization (NMF)—to extract co-occurring cell subset signatures—with pairwise covariance analysis to define the edges (interaction pairs) of each module network. The resulting 12 CMs were validated against ~12,000 bulk RNA-seq profiles from the GTEx project, with strong concordance in cell-type marker expression. Epithelial cells were excluded due to their highly tissue-divergent nature. Most cell subsets (88%) participated in at least one CM, and 25% participated in multiple, indicating functional versatility without clear hub dominance.
The 12 CMs display striking tissue preferences: CM04–CM06 and CM09 concentrate in immune organs and blood; CM07 and CM12 in the reproductive system; CM08 in barrier tissues (skin, oral mucosa, trachea); CM10 forms a vascular unit with pericytes and smooth muscle cells; and CM01—featuring tissue-resident macrophages, universal fibroblasts, and lymphatic endothelial cells—is broadly distributed across nearly all body systems, suggesting a universal organizational scaffold. Spatial transcriptomics confirmed that CM components co-localize in tissue sections, and in vivo perturbation data supported coordinated intramodule intercellular signaling.
In the aging spleen, two immune CMs showed opposing chronological dynamics, with one expanding and one contracting with age—a finding with direct implications for immunosenescence research. In the breast, a menopausal trajectory was mapped through fibroblast dynamics, revealing how hormonal transitions reshape tissue-level multicellular organization. In cancer, analysis across tumor types uncovered a dual rewiring: the progressive loss of tissue-specific healthy CMs and the simultaneous emergence of a convergent 'cancerous ecosystem' shared across cancer types, suggesting that tumors may co-opt or impose a common pro-tumorigenic multicellular program regardless of tissue of origin.
This work establishes a foundational framework for understanding tissue-level multicellular coordination and its disruption in aging and cancer. The public atlas and CoVarNet tool are available at cm.cancer-pku.cn, enabling broad follow-up research into how multicellular ecosystems govern tissue homeostasis, aging, and disease progression.
Key Findings
- 12 cross-tissue cellular modules (CMs) identified from 2.3M cells across 35 human tissues, each with distinct tissue preferences and spatial organization.
- Two splenic immune CMs show opposing age-related dynamics, providing a multicellular map of immunosenescence.
- A menopausal breast tissue trajectory is driven by fibroblast-centered multicellular reorganization.
- Cancer simultaneously dismantles healthy tissue-specific CMs and installs a convergent pro-tumor multicellular ecosystem across cancer types.
- CoVarNet framework validated against 12,000 GTEx bulk RNA-seq profiles, confirming CM robustness across data modalities.
Methodology
Cross-sectional analysis of 2,293,951 cells from 706 healthy samples across 35 tissues using scRNA-seq, integrated via BBKNN and annotated hierarchically. CoVarNet combined NMF and pairwise covariance to define 12 cellular modules, validated by GTEx bulk RNA-seq and spatial transcriptomics (10x Visium).
Study Limitations
The atlas is cross-sectional, limiting causal inference about the directionality of multicellular changes. Epithelial cells were excluded from CM analysis due to tissue-divergent profiles, potentially missing epithelial-stromal coordination. Single-cell data representation may be biased by tissue dissociation and sorting protocols across the aggregated datasets.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.