Mouse models remain the principal preclinical systems for evaluating cancer immunotherapy, but the degree to which they reproduce the human tumor microenvironment has been incompletely defined. A recent systematic study performed cross-species immune profiling to directly compare immune landscapes between commonly used mouse tumor models and corresponding human cohorts. The aim was to map where mouse and human tumor immunity diverge and where they converge, to guide interpretation of preclinical findings and improve alignment of models to clinical biology.
The referenced analysis surveyed immune composition and transcriptional programs across 15 widely used mouse tumor models and matched patient cohorts. By comparing cellular composition, chemokine networks, cell–cell interactions, and consensus gene expression modules, the study assessed both shared and species‑specific features of tumor immunity. This approach combined cellular profiling with gene expression to identify patterns that either translate between species or remain distinct.
A central finding was that commonly used mouse tumor models predominantly capture a subset of human tumor immune phenotypes—specifically those that are macrophage-rich and T cell–poor. This bias means many mouse models reflect immune microenvironments dominated by myeloid populations rather than those characterized by strong adaptive T‑cell infiltration. As a result, preclinical studies using such models may underrepresent mechanisms and responses that depend on abundant intratumoral T cells.
By contrast, the profiling revealed that a class of human tumors organized around CXCL13 and rich in immune cells—features associated with better responsiveness to immune checkpoint blockade in the clinic—were largely absent from the commonly used mouse model set. These CXCL13-organized, immune-rich human tumor types have clinical relevance because they tend to show greater benefit from checkpoint inhibitors. The lack of corresponding mouse models implies a gap in preclinical tools for studying therapies whose activity depends on that immune context.
Beyond differences in cellular makeup, the study identified species-specific chemokine networks and distinct patterns of cell–cell communication that likely contribute to disparate immune architectures in mouse versus human tumors. These species differences in signaling and interactions offer mechanistic explanations for why certain immune phenotypes emerge in one species but not the other, and they caution against simple extrapolation of intercellular regulatory relationships from mice to patients.
Despite the compositional and signaling differences, consensus gene expression profiling uncovered conserved transcriptional modules shared between mouse and human tumors. A notable conserved program links interferon-responsive myeloid cells with T‑cell cytotoxicity; this transcriptional axis was strongly associated with measures of T‑cell effector function and had predictive value for patient survival. This convergence identifies a translationally relevant biological signature that can be interrogated in both species.
The authors packaged these cross-species comparisons into a resource described as a queryable atlas. The atlas is intended to help researchers precisely match preclinical mouse models to the human tumor immune states they best emulate, rather than treating mouse models as universally representative. Such alignment supports more targeted selection of models for mechanistic studies and preclinical testing of immunotherapies.
The work serves as both a cautionary guide and a practical tool. It cautions that many widely used mouse tumor models do not recapitulate immune-rich, CXCL13-organized human tumor types that respond to checkpoint blockade, and it highlights species-specific chemokine and interaction differences that may confound translation. At the same time, the identification of conserved transcriptional modules—particularly the interferon-responsive myeloid and T‑cell cytotoxicity program—points to shared biology that can inform translational hypotheses.
Practically, researchers should consider the immune phenotype they aim to study and use the atlas to select mouse models that reflect that phenotype where possible. Where no mouse model captures a given human immune context, investigators should interpret preclinical results cautiously and may need to develop or adopt alternative models. The source text does not provide details on specific model-by-model mappings or the atlas interface; those details were not reported in the abstract and would require consulting the full study.