Related Experiment Video
Updated: Jul 3, 2026

In Vitro Assay to Study Tumor-macrophage Interaction
Published on: August 1, 2019
A transcriptome-defined TAM-rich phenotype identifies a macrophage-enriched, hypoxia-linked immune contexture in
Guanpeng Li1, Xiaohao Yan2,3, Shenying Fang1
1School of Public Health and Emergency Management, Southern University of Science and Technology, Shenzhen, China.
Background:
Tumor-associated macrophages (TAMs) are a major immune component of the glioblastoma microenvironment, but macrophage-oriented transcriptomic stratification remains difficult to reproduce across datasets and to validate at the tissue level.
Methods:
TCGA-GBM was used as the discovery cohort, and two independent CGGA cohorts (CGGA_693 and CGGA_325) were used for external validation. A continuous transcriptome-defined TAM score was constructed a priori as the mean log-scale expression of five macrophage-associated genes: CD163, MSR1, MRC1, CSF1R, and SPP1. Cohort-specific median dichotomization was used as a secondary categorical representation for visualization and group-level analyses. Differential expression, marker-level characterization, survival analysis, continuous-score sensitivity analyses, external immune benchmark analyses, and negative-control analyses were performed. Local histopathological correlation was assessed in a glioblastoma cohort from the Sichuan Cancer Hospital using immunohistochemistry and digital pathology for CD163, CD68, CD8, and CAIX.
Results:
In TCGA, the TAM-rich phenotype was associated with broad upregulation of macrophage- and myeloid-related genes. Across TCGA, CGGA_693, and CGGA_325, macrophage-associated markers, including CD163, MSR1, MRC1, CSF1R, SPP1, VSIG4, MARCO, and TREM1, were consistently increased, and the hypoxia-associated marker CA9 was also elevated. By contrast, CD8A, CXCL9, and CXCL10 showed only modest concordant increases. External immune benchmark analyses using xCell macrophage-related scores and published no-TAM5 macrophage/TAM signatures supported concordance with macrophage-related immune context, and random-panel and label-permutation analyses reduced concern that the associations reflected arbitrary five-gene scoring or random grouping. Survival association was modest and cohort-dependent. In the local cohort, CD163-defined macrophage-enriched tumors showed higher CD163-positive and CD68-positive macrophage density, greater CAIX-positive area, and a lower CD8/CD163 ratio, reflecting relative macrophage dominance rather than absolute loss of CD8-positive cells.
Conclusions:
A compact transcriptome-defined TAM-rich phenotype captures a biologically reproducible, macrophage-enriched and hypoxia-linked immune contexture in glioblastoma. It should be interpreted primarily as a quantitative immune-state classifier and tissue-correlated biological stratifier, rather than as a uniformly independent prognostic factor.