Researchers at Peking University and Chongqing Medical University have published a sweeping review that maps the cellular landscape inside tumors with single-cell precision. The work synthesizes cutting-edge findings on immune cells, stromal tissue and neural cells — and introduces an AI-powered ‘virtual tumor’ framework that could reshape how oncologists design treatments.
A landmark review published June 5 in the journal Immunity & Inflammation offers one of the most comprehensive pictures yet of the complex cellular world inside tumors — and charts a course toward AI-guided, precision immunotherapy. Led by Linnan Zhu, an associate researcher at the Peking University Biomedical Pioneering Innovation Center (BIOPIC), and Zemin Zhang, an academician of the Chinese Academy of Sciences, president of Chongqing Medical University, and former director of BIOPIC, the paper synthesizes recent breakthroughs in single-cell sequencing and spatial omics to deliver, as the authors describe it, an integrated roadmap from fundamental tumor biology to next-generation cancer treatments.
What Is the Tumor Microenvironment — and Why Does It Matter?
Tumors are not simply clumps of rogue cells. They are dense, dynamic ecosystems populated by immune cells, structural support cells, blood vessels, and even neurons, all interacting with malignant cells in ways that can either fight cancer or help it spread. This surrounding ecosystem is called the tumor microenvironment, or TME, and understanding it has become one of oncology’s most important frontiers. Advances in single-cell RNA sequencing and spatial transcriptomics now allow scientists to identify and map individual cell types within this environment with unprecedented resolution — revealing specific subtypes that were previously invisible to researchers.
The review identifies several cell populations with strong implications for patient prognosis and treatment response. Among immune cells, CD8+ cytotoxic T lymphocytes are considered the primary soldiers of anti-tumor immunity, capable of recognizing and destroying cancer cells. However, these cells frequently become exhausted within the TME, losing their killing ability. A subtype called CXCL13+ T cells has emerged as particularly significant — pre-exhausted versions of these cells appear across multiple cancer types and are linked to better responses to immune checkpoint blockade, one of the most widely used immunotherapy approaches today.
On the suppressive side, TNFRSF9+ regulatory T cells mount a powerful immunosuppressive response that acts as a major barrier to effective treatment. Meanwhile, certain B cell and natural killer cell subtypes also play meaningful roles: tumor-associated B cells tagged with the marker FCRL4+ correlate with improved prognosis and checkpoint response, while DNAJB1+ NK cells show signs of dysfunction and are associated with resistance to PD-1 therapy.
Myeloid Cells, Fibroblasts and a Neural Surprise
The review dedicates considerable attention to myeloid cells — a broad category that includes macrophages, dendritic cells, neutrophils and mast cells. SPP1+ tumor-associated macrophages stand out as particularly harmful, driving blood vessel formation, supporting low-oxygen tumor environments, and restructuring surrounding tissue in ways that accelerate disease progression. The authors highlight that a polarity axis defined by mutually exclusive expression of CXCL9 and SPP1 in macrophages offers stronger predictive clinical value than the traditional, and now increasingly outdated, M1/M2 classification system.
Dendritic cells marked as LAMP3+ are described as mature and migratory, with a subset linked to enhanced CD8+ T cell infiltration and better immunotherapy outcomes. Even neutrophils and mast cells show antigen-presenting capabilities under certain conditions, underscoring the remarkable functional versatility of myeloid cells in the TME.
Perhaps the most striking section concerns cancer neuroscience. The review highlights that TGFBI+ Schwann cells — a type of nerve-supporting cell — can be induced by the signaling molecule TGF-β to actively promote tumor cell migration, correlating with worse patient outcomes. This finding adds a neurological dimension to cancer biology that researchers are only beginning to understand, reinforcing the idea that the TME is far more complex than its immune components alone.
From Individual Cells to Coordinated Networks
One of the review’s most conceptually significant contributions is its emphasis on cellular networks rather than isolated cell types.
“These cell subsets do not act in isolation but form complex multi-cellular networks through spatial organization and functional cooperation,” the authors noted.
Structures called immunity hubs — which cluster specialized dendritic cells, particular T cell populations, and supportive fibroblasts — function as integrated units whose spatial arrangement within a tumor strongly predicts whether a patient will respond to immunotherapy.
The research further describes how tumor progression involves the gradual breakdown of these healthy, coordinated networks and their replacement with aberrant, cancer-promoting modules that appear consistently across different cancer types. This shared remodeling pattern, the authors argue, lays the conceptual groundwork for developing broad-spectrum therapies that target TME organization rather than any single cell type.
The AI Virtual Tumor: A New Computational Frontier
Perhaps the most forward-looking element of the review is its introduction of the “AI virtual tumor” concept. Building on earlier work in AI-simulated single cells, this framework aims to integrate cellular composition, spatial tissue architecture, cell-cell communication, and perturbation response rules to extend single-cell behavior modeling to tumor-scale ecosystem dynamics. In practice, such a model could allow researchers to simulate how a tumor ecosystem responds to a given drug combination before any patient is treated — essentially running experiments in a computer rather than a lab or clinic.
“This could provide new computational frameworks for patient stratification, combination drug design, and efficacy prediction,” the authors suggested, opening up possibilities for truly personalized oncology at a scale that was not feasible even a few years ago.
Implications for Immunotherapy — and for Students Entering Science
The review also surveys the current state of immunotherapy through the lens of these TME insights. In immune checkpoint blockade, specific cell markers now help predict who will benefit and who will resist treatment. CAR-T cell therapy continues to show strong results in blood cancers, while a newer approach using engineered macrophages — CAR-M cells — is entering early clinical trials for solid tumors, where T cells struggle to penetrate. On the vaccine front, personalized mRNA neoantigen vaccines have demonstrated safety and immune activation in high-risk kidney cancer patients, pointing toward a future where treatments are tailored to each patient’s unique tumor biology.
For students pursuing careers in biology, medicine, bioinformatics or oncology, this review represents a useful synthesis of where the field stands. Single-cell technologies are reshaping cancer research, and the integration of AI modeling into tumor biology is rapidly creating new job categories and research directions. Understanding the TME is increasingly central to graduate-level cancer biology, clinical oncology training, and pharmaceutical development.
Source: Immunity & Inflammation