Single-cell profiling of human gliomas reveals macrophage ontogeny as a basis for regional differences in macrophage activation in the tumor microenvironment

Immunotherapy has transformed cancer treatment across the board — checkpoint inhibitors, CAR-T cells, and immune modulators. Yet, in glioma, the most common and deadly primary brain tumor, those same strategies keep failing. The bitter irony is that gliomas are not immunologically cold; they are packed with immune cells. So, why isn't the immune system doing anything? Müller and colleagues at the University of California, San Francisco, and the University of Pittsburgh believe they have found a significant part of that answer. It comes down not to how many immune cells are in the tumor, but which kind, and where those cells originally came from. The immune cells in question are called tumor-associated macrophages, or TAMs. They are abundant in gliomas, often making up a large fraction of the tumor mass. However, TAMs aren't a single population. They come from two completely different sources. One group is brain-resident microglia, which are cells whose progenitors colonized the central nervous system early in embryonic development and have lived there ever since. The other group is macrophages that originated in the bone marrow, traveled through the bloodstream, and crossed the blood-brain barrier into the tumor. The critical unknown going into this study was whether those blood-derived invaders simply adopt the identity of the local microglia once they're inside the brain, becoming functionally indistinguishable from the cells that were already there. If they do, the distinction doesn't matter much for therapy. If they don't, it changes everything. To answer that question, Müller and colleagues conducted single-cell RNA sequencing, a technique that reads the gene activity of individual cells rather than averaging across millions at once. They physically isolated TAMs from patient biopsies using CD11b microbeads, a surface marker that labels macrophages, achieving purities above 96 percent. They then sequenced those cells on two platforms: the Fluidigm C1, which provides full-transcript coverage, and the higher-throughput 10X Genomics Chromium. One particularly elegant quality-control move involved performing exome sequencing on each patient's tumor and blood, identifying somatic mutations unique to the cancer cells, and checking whether those mutations appeared in the single-cell transcriptomes. Cells expressing clonal tumor mutations were flagged as neoplastic. The TAMs were clean, showing no somatic mutations and robust macrophage markers. This combination of physical purification and genetic verification gave the team high confidence in what they were actually examining. The final dataset comprised 5,455 TAMs from 19 patients, combining 4,181 newly profiled cells with 1,274 from published sources. The key analytical achievement was a sixty-six-gene signature that distinguishes blood-derived TAMs from microglial TAMs. The team started with lineage-tracing RNA sequencing data from murine glioma models, which identified 836 genes that were differentially expressed between the two TAM populations in mice. They mapped those to human equivalents, found 237 that were expressed in human TAMs, and ran principal component analysis on those 237 genes. The scores along the first principal component separated the TAMs into two distinct clusters, confirmed by Gaussian mixture modeling, with less than five percent overlap between populations. From the strongest loadings along that first component, they distilled the final sixty-six-gene set. This signature was reproducible across both sequencing platforms, correlated across 558 bulk-RNA glioma cases from The Cancer Genome Atlas, and was validated at the protein level by multicolor flow cytometry using the surface markers P2RY12, CX3CR1, CD49D, and HLA-DR. The biology was also spatially coherent. Mapping the signature onto the Ivy Glioblastoma Atlas Project, which provides anatomically dissected RNA sequencing from defined tumor regions, showed that microglial TAM genes were enriched at the tumor's infiltrating edge and in adjacent white matter. In contrast, blood-derived TAM genes were concentrated in hyperplastic blood vessels, microvascular proliferation zones, and peri-necrotic regions. In situ hybridization confirmed this directly: a blood-derived marker called TGFBI lit up near blood vessels, while a microglial marker called BIN1 was enriched in infiltrated white matter and diminished in the tumor core. So, the two populations do not blend. They stay spatially and molecularly distinct. Now, the question becomes: what are the blood-derived TAMs actually doing in those perivascular and necrotic niches? The answer, in short, is immunosuppression — along with a metabolic signature to match. Blood-derived TAMs upregulate two canonical immunosuppressive cytokines, interleukin ten and transforming growth factor beta two. They also show elevated class II human leukocyte antigen expression and enrichment of phagocytic machinery, consistent with an actively scavenging, suppressive phenotype. Metabolically, they display a pattern characteristic of alternatively activated macrophages: elevated expression of genes at precisely the rate-limiting steps of the tricarboxylic acid cycle — the steps typically stalled in classically activated, pro-inflammatory macrophages. In classically activated cells, the tricarboxylic acid cycle breaks after citrate and again after succinate. In blood-derived TAMs, those breaks don't appear. The cycle runs, supporting an oxidative metabolic program that is the hallmark of immune-suppressing macrophages. Microglial TAMs do not show this signature. They are metabolically distinct, less immunosuppressive, and concentrated at a completely different part of the tumor. Then comes the finding that genuinely scrambles the standard model of macrophage biology. The field has traditionally divided macrophages into two functional states: M1, the pro-inflammatory killers, and M2, the anti-inflammatory suppressors. The assumption is that a given cell is one or the other. But in the single-cell data, sixty-six percent of TAMs expressing the M2 marker interleukin ten also expressed the M1 marker tumor necrosis factor alpha — in the same cell. Flow cytometry at the protein level confirmed this: substantial fractions of both P2RY12-positive microglial TAMs and CD49D-positive blood-derived TAMs co-expressed the M1 marker CD86 and the M2 marker CD206 simultaneously. Individual cells are not neatly M1 or M2; they hold both programs at once. This means any therapeutic strategy built around converting M2 macrophages into M1 macrophages — a popular concept in immuno-oncology — may be targeting a distinction that doesn't cleanly exist at the single-cell level. The survival data sharpens why the blood-derived population matters clinically. Using the sixty-six-gene ontogeny signature mapped to 558 Cancer Genome Atlas glioma cases, the team found that high expression of the blood-derived TAM signature correlated with significantly shorter overall survival in low-grade glioma, which includes grade two and three tumors. The hazard ratio was 3.44, corrected for age and gender, with a p-value of 0.016. The microglial signature showed no such correlation. In glioblastoma, the same directional trend appeared for blood-derived TAMs, with a hazard ratio of 1.61, though it didn't reach statistical significance at a p-value of 0.109. TAM composition also shifts by tumor subtype: glioblastoma shows significantly more blood-derived infiltration than low-grade glioma. Within low-grade tumors, astrocytomas have a higher degree of microglial infiltration than oligodendrogliomas — a difference confirmed at a p-value below 0.01. The balance of these two populations varies, and that variation tracks with outcomes. That brings everything back to therapy. The current approach, targeting TAMs as a single population primarily through colony-stimulating factor one receptor inhibition to block macrophage survival signaling, has worked impressively in mouse models. Murine gliomas can regress with that blockade. However, clinical trials of the same approach have failed to improve overall survival. The data from Müller and colleagues suggest one reason why: you may be hitting both populations, disrupting microglia that may not be the problem while leaving the blood-derived immunosuppressive fraction to do its work. Their results argue directly against indiscriminate TAM depletion and for strategies that specifically target the blood-derived, immunosuppressive subset, whether through selective depletion or metabolic reprogramming. This study also provides the first single-cell baseline of myeloid cells in untreated, pre-therapy glioblastoma. That's a necessary foundation before any therapy-focused follow-up can be interpreted. What did the tumor's immune compartment look like before treatment changed it? Now there’s an answer. That answer reframes the immunotherapy failure problem in brain cancer: the barrier isn't the immune system's absence. It's the presence of the wrong macrophages, in the wrong places, doing the wrong things — and doing them in ways that current blunt-force strategies weren't designed to stop. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.