Over the past decade, significant advances in single-cell technologies and initiatives such as the Human Cell Atlas have enabled comprehensive profiling of various human tissues, revealing a substantial number of previously uncharacterized cell types. However, tissues are not random assemblages of cells. Rather, they rely on the coordinated interaction of multiple cell types to achieve specific structure and function. A notable illustration of multicellular coordination is evident in gut-associated mucosal immunity, where diverse cell types such as immune cells, stromal cells and epithelial cells collaborate to defend against pathogenic insults. Building upon these observations, it’s imperative to determine whether such coordinated multicellular ecosystems represent a recurrent theme across the human body, thereby addressing the fundamental question of how these ecosystems contribute to cohesive functional coordination among diverse cell types at the tissue level.

 

On May 28, 2025, the research team led by Professor Zemin Zhang from the Biomedical Pioneering Innovation Center (BIOPIC) at Peking University published a research article titled “Cross-tissue multicellular coordination and its rewiring in cancer” in Nature. The study introduced a novel concept of "cross-tissue cellular modules" and developed a computational framework called CoVarNet. By applying CoVarNet to single-cell transcriptomic data, the study systematically identified 12 cross-tissue coordinated cellular modules (CMs), uncovering their spatiotemporal dynamics, internal regulation, phenotypic associations and examining their rewiring in cancer.

 

To systematically characterize multicellular coordination, the study translated this concept into co-occurring cellular networks and developed a novel computational framework named CoVarNet (Figure 1). By analyzing the covariance of cell type frequencies within the samples, the study identified 12 CMs with distinct cellular compositions and tissue prevalences from the pan-tissue atlas. Among them, CM04, CM05, CM06 and CM09 were primarily composed of immune cells, enriched in immune organs and peripheral blood. CM02 and CM03 were mainly distributed in the urinary system and gastrointestinal tract. CM07 and CM12 demonstrated preferences for the reproductive system, containing reproductive-system-related fibroblasts. CM08 was enriched in barrier tissues, potentially representing multicellular ecosystems within mucosa-associated lymphoid tissues. CM10 appeared to function as a vascular unit while CM11 was involved in metabolic processes. CM01, characterized by tissue-resident macrophages, universal fibroblasts, and lymphatic endothelial cells, was broadly distributed across nearly all human body systems. These results demonstrate the tissue preferences and potential functional roles of CMs.

Figure 1. Systematic identification of cross-tissue CMs

 

Subsequently, the study mapped CMs onto spatial transcriptomics data to characterize the spatially resolved multicellular coordination (Figure 2a-e). By analyzing Visium data from the small intestine, the study found CM05 and CM02/CM03 were respectively located in the Peyer’s patch and intestinal mucosa, suggesting that they respectively recapitulate the inductive and effector modules of mucosal immunity. Utilizing high-resolution Xenium data, the study further distinguished CM02 and CM03: CM02, which contains CD8+ effector memory T cells and natural killer cells, exhibited a uniform distribution across the intestinal mucosa; while CM03, which contains IgA+ plasma cells and mast cells, was enriched in the lamina propria. These findings underscore the distinct functional roles of CMs within tissue ecosystems.

 

Additionally, the study explored the internal regulation within CMs (Figure 2f-h). Compared to CMs with a higher proportion of stromal and endothelial cells, lymphocyte-enriched CMs were more spatially concentrated and produced fewer types of ligands. This aligns with the notion that spatial proximity may enhance the specificity of intercellular communication. Furthermore, cytokine response analysis based on in vivo perturbation data demonstrated the crucial regulatory role of cytokines within CMs. The study observed that the same subsets exhibited different cytokine responses in different CMs. For example, CD8+ effector memory T cells exhibited TNF-α responses in CM02, CM08, and CM09, but absent in CM04 and CM06. This indicates that cellular phenotypes are collectively determined by intrinsic properties and local stimuli.

Figure 2. Spatially resolved multicellular coordination in CMs

 

The study revealed some fundamental insights by linking CMs with phenotypic data (Figure 3). Two immune CMs, CM05 and CM06, showed contrasting chronological dynamics in the spleen with aging. Four CM05 subsets and four CM06 subsets also exhibited significant age-related changes, which were more pronounced compared to the previously reported age-associated B cells. Further analysis revealed 17 overlapping regulons in CM05 subsets, with these convergent regulons displaying increased activity with aging. These results in the spleen underscored coordinated behavior at the molecular, cellular, and multicellular levels. Additionally, based on breast-enriched CM12, the study uncovered a fibroblast-engaged menopausal trajectory. It was found that the frequency of two fibroblast subsets (S10 and S06) gradually declined along this trajectory, accompanied by a consistent reduction in inflammatory characteristics.

Figure 3. Multicellular dynamics in context

 

Finally, the study explored the dynamic changes in CMs during the transition from health to tumor (Figure 4). Using public data, the study established a comprehensive pan-cancer single-cell transcriptomic atlas and focused on eight cancer types with matched healthy, tumor, and adjacent non-tumor samples (as a surrogate for precancerous samples). The study observed a notable reduction in these tissue-specific healthy CMs during tumor progression, along with a progressive increase of the tumor-associated cell module cCM02. Based on this finding, the study proposed a tumor progression model from a multicellular coordination perspective: as tumor progress, tissues gradually lose tissue-specific healthy organizations and acquire a convergent cancerous ecosystem.

Figure 4. Rewiring of multicellular ecosystems in cancer

 

In summary, the study systematically elucidated fundamental organizing principles of multicellular ecosystems in health and cancer, bridging the gap between well-characterized cellular diversity and the complex organization and function of tissues. It opens avenues for cellular regulation, tissue regeneration and disease intervention, as well as establishing a new computational framework for large-scale data integration and analysis. Furthermore, the pan-tissue and pan-cancer single-cell atlases curated in this study provide valuable resources for the community.

 

Academician Prof. Zemin Zhang from BIOPIC of Peking University / Chongqing Medical University and Dr. Qiang Shi from BIOPC / School of Life Sciences of Peking University were the co-corresponding authors of this paper. Dr. Qiang Shi and PhD student Yihan Chen in the 2024 cohort of the PTN program from Academy for Advanced Interdisciplinary Studies of Peking University were co-first authors. The research was supported by the National Key Research and Development Program of China, the National Natural Science Foundation of China, the China Postdoctoral Science Foundation and the Boehringer-Ingelheim Postdoctoral Fellowship Program.

 

Link: https://www.nature.com/articles/s41586-025-09053-4