For decades, the immune protein complement C3 has been studied as a sentinel of the bloodstream, the first responder that marks pathogens for destruction. A team at Nagoya University has now shown that its most consequential role in cancer therapy may be entirely local. The C3 that determines whether checkpoint immunotherapy works is not the C3 circulating in the blood and made by the liver, but the C3 manufactured by fibroblasts inside the tumor itself. The distinction is stark: in mice, eliminating 89 percent of circulating C3 changed nothing, while a genetic deletion that barely dented blood levels destroyed the treatment’s power. The study, led by Yuki Miyai with Atsushi Enomoto as corresponding author, was published in Nature Communications on July 15, based on a preprint posted in May 2025.
C3 predates the circulatory system: sponges and jellyfish, animals with no blood at all, carry the gene. It is the convergence point of all three complement pathways, and in mammals roughly 90 percent of it is produced by hepatocytes and circulates at about 100 milligrams per deciliter of plasma. Researchers assumed the blood pool was the molecule’s stage. The existence of tissue-made C3 was known but its function was not. The new work argues that the local, tissue-made version has been quietly running an entirely different program.
The mouse experiments established the dissociation cleanly. Using MC-38 colon tumors treated with anti-PD-1 antibodies, the team compared three genetic situations. Mice whose hepatocytes could not make C3 had 89 percent less circulating C3, yet responded to immunotherapy exactly as well as controls (interaction P = 0.762). Mice whose cancer-associated fibroblasts could not make C3, a deletion that lowered circulating C3 by only about 9 percent, lost most of the response (P < 0.0001), had a complete-response odds ratio of 0.219, and survived shorter. Injecting AMY-101, an inhibitor of the C3 convertase enzyme, next to the tumor reproduced the failure (P = 0.0093), confirming that local complement activation, not systemic, is what the treatment needs.
The mechanism runs through the tumor’s immune environment. C3 made by fibroblasts is cleaved into a fragment called iC3b, which engages complement receptor 3, the CD11b/CD18 integrin, on myeloid cells. That engagement triggers Syk-dependent signaling that stops monocytes from adhering and migrating. With fewer monocytes entering the tumor, fewer become immunosuppressive M2-like macrophages, the cells that shield tumors from T cells. Remove fibroblast C3 and the gate opens: single-cell profiling of tumors in the knockout mice showed more M2 macrophages and exhausted T cells, and resistance to therapy.
The human data point the same way. Among 136 patients with non-small cell lung cancer treated with checkpoint inhibitors at Nagoya University Hospital, serum C3 levels were statistically indistinguishable between responders and non-responders. But in 35 tumor specimens, the amount of C3 in the stroma, the supporting tissue around the cancer cells, correlated negatively with M2 macrophage infiltration (Spearman’s r = -0.610, P = 0.0001). Patients whose tumors were rich in C3-producing fibroblasts responded far more often, about half versus none in the low-C3 group according to the university’s announcement, and had longer progression-free and overall survival. A public database of immunotherapy-treated patients confirmed the survival association (hazard ratio 0.75).
The therapeutic implication is that blocking myeloid infiltration can rescue resistant tumors: in mice with intrinsically resistant tumors, depleting CD11b-positive myeloid cells with an antibody, or treating with the partial agonist ADH-503, restored sensitivity to anti-PD-1 and prolonged survival. But the authors flag the caution themselves: ADH-503 failed in a clinical trial (NCT04060342) in patients with established, advanced tumors. Timing may be everything, and boosting local C3 in the tumor, rather than chasing myeloid cells systemically, may be the more faithful reproduction of what the biology does.
Several limitations frame the finding. The human cohort is single-center and the biomarker classification, a 20 percent positivity threshold for C3-producing fibroblasts, will need standardization across cancer types before it can guide treatment decisions. The Nature Communications version is an early-access manuscript; the complete analysis was first available as a preprint. And while the mouse genetics are causal, the human association, however consistent, remains correlational.
What the study offers is a new way to read an old molecule. C3 evolved before blood existed, which suggests its original job was local, done in place, tissue by tissue. In the tumor, that ancient local function turns out to be the switch that decides whether checkpoint drugs work. The blood test for C3, the one doctors have used for decades, was measuring the wrong pool. The next step, the team says, is learning how to raise C3 inside tumors, and when to do it.
Sources: Miyai, Y. et al. Local, but not circulating, complement C3 shapes immune checkpoint blockade efficacy by controlling myeloid cell infiltration. Nature Communications (2026). DOI: 10.1038/s41467-026-75542-3. Preprint: bioRxiv 10.1101/2025.05.12.653608. Nagoya University press release, July 22, 2026.