A hard-to-treat colorectal tumor may lack a supply line, not soldiers: a Shenzhen, China team reports in Science that OX40 plus PD-1 rewired immunity in mice

The same class of drug can look like two different medicines in colorectal cancer. In some patients, a PD-1 antibody lifts the brakes on the immune system and the tumor starts to shrink. In others, almost nothing happens. Researcher Lei Zhang and colleagues at Shenzhen Bay Laboratory in Shenzhen, China, argue in Science that the difference is not simply whether immune cells are already inside the tumor. In tumors that respond, a set of helper T cells leaves the draining lymph node, enters the tumor and assembles a recognizable spatial niche. In a mouse model of the harder-to-treat subtype, pushing that supply line with an OX40 agonist plus PD-1 blockade slowed the cancer and extended survival.

PD-1 drugs are often described as taking the brakes off T cells that kill diseased cells. Colorectal cancers fall into two broad groups. Tumors that are mismatch-repair deficient and microsatellite instability–high (dMMR/MSI-H) tend to respond better to PD-1 blockade. Tumors that are mismatch-repair proficient and microsatellite stable (pMMR/MSS) usually do not. Much of the field has tried to loosen immunosuppression already present inside the tumor. Zhang’s team, working with physicians at Nanfang Hospital of Southern Medical University in Guangzhou, China, started instead with a comparison of the two kinds of patient tissue, then followed the same T-cell clones through lymph nodes, blood and tumors in people receiving PD-1 therapy.

Science published the study on Nov. 12, 2026. Using imaging mass cytometry, the researchers analyzed 295 multiparameter tissue images. In dMMR/MSI-H tumors they found a helper subset of CD4 T cells with TH1-like features, sitting in tight company with LAMP3-positive dendritic cells, CXCL9-positive macrophages and nearby CD8 T cells — a small, repeatable fighting unit. That spatial arrangement was more common in patients who mounted an active response after PD-1 treatment. pMMR/MSS tumors were instead dominated by regulatory T-cell interactions that suppress immunity. Single-cell and T-cell receptor analyses of patients on PD-1 therapy then showed a stem-like CD4 population in tumor-draining lymph nodes that shared clones with the TH1-like cells inside the tumor; that clonal link grew stronger after treatment. In a pMMR mouse model, PD-1 blockade alone could start a similar differentiation path but did not control tumor growth. Combining an OX40 agonist with PD-1 blockade significantly slowed the tumors, prolonged survival and expanded a wave of IL-7R–positive early effector CD4 T cells.

The image that helps is a war with a rear staging area, not just a fight at the front. Loosening the brakes at the tumor is not enough if the reserve units never get trained and sent in. OX40 is a costimulatory receptor on activated T cells, a booster switch. The combination first expanded stem-like CD4 T cells in the lymph node, turned some of them into early effectors, and then helped those cells migrate into the tumor and become TH1-like. Signals such as interferon-gamma also pushed macrophages into a more inflammatory, helpful state and expanded CD8 effector T cells. When the team blocked T cells from leaving the lymph node, neutralized interferon-gamma, or depleted CD4 T cells, CD8 T cells or macrophages, the benefit of the combination fell away. The effect depended on the supply line, not only on cells that were already on site.

For patients, that points to a different place to intervene in colorectal cancers that shrug off PD-1 drugs: the peripheral helper T-cell response, not just the cells already trapped inside the tumor. The evidence that the drug combination shrinks tumors still comes from mice. It has not been tested as a therapy in people with this mechanism in mind. OX40 agonists have already been tried in clinical studies in the United States and elsewhere, with uneven results. The next question is whether this path can be turned into a regimen that is safe and measurable in patients. In a companion Perspective, Matthew R. Olson of Purdue University in West Lafayette, Indiana, US, wrote that engineering the immune milieu around tumors could make more of them sensitive to immunotherapy.

Original source published: Science, Nov. 12, 2026, Beijing time / Nov. 12, 2026 EST; Shenzhen Bay Laboratory, Nov. 13, 2026, Beijing time