Proposed KLEOS arm: CD40 agonist plus anti-PD-1 in biomarker-selected MMR-proficient colorectal

Cohort: Neoadjuvant CRC (one of Treatments A–C). Endpoint: pCR/cCR. Prepared by Encounter (encounter.bio) for the Project Cure CRC / KLEOS consortium · June 2026.

The arm, in one line. A biomarker-enriched arm testing a CD40 agonist plus anti-PD-1 in MMR-proficient (MSS) resectable colorectal, selected for a pre-treatment CD8 signature that identifies tumours where effector T cells were built but never trafficked to the tumour.

The problem it addresses. MSS colorectal does not respond to checkpoint blockade, where MMR-deficient disease does, roughly 50% versus under 5%. The field reads MSS as immunologically cold and treats cold as empty, on the assumption that no anti-tumour T cells were made. In a defined subset that assumption is wrong. In the Pelka 2021 colorectal atlas (76,965 CD8 T cells across 110 tumours), about a third of MMR-proficient tumours carry fully built CD8 effectors that are held in the draining lymph node and never reach the tumour bed. Anti-PD-1 acts at the synapse inside the tumour, so for these patients there is no substrate there for it to release. The limiting step is trafficking, upstream of the checkpoint.

The intervention. A CD40 agonist (sotigalimab, selicrelumab, or 2141-V11) supplies the cDC1–CD40–IL-12 licensing signal that releases primed CD8 from the node; anti-PD-1 then unblocks those cells once they reach the tumour. Concurrent neoadjuvant dosing puts the licensing signal in the same window as the developing response. Both drug classes, and the combination, are already in clinical use, so the arm uses existing agents rather than a new molecule.

The enrichment biomarker. A pre-treatment CD8 signature read on the diagnostic biopsy: effector pool full (memory plus resident CD8 above 65% of CD8), engaged and exhausted compartment minimal (under 5%), IL7R and TCF7 retained, PD-1 and TOX low, CXCL13 absent. Specified on Pelka 2021, it appears in 32% of MMR-proficient versus 9% of MMR-deficient tumours, Fisher odds ratio 4.64, p = 0.0028. Combined with the upstream licensing step it separates 58.9% versus 14.8%, odds ratio 8.25. (Encounter's internal name for this read is the Garrison signature.)

Fit to the cohort. It drops into the Neoadjuvant CRC cohort without changing the schema. Screening adds the signature assay to the diagnostic biopsy. Randomisation, neoadjuvant treatment, and surgery proceed as drawn, with pCR (and cCR for rectal) as the cohort's primary endpoint. The signature enters as a pre-specified enrichment or stratification factor.

The pre-specified prediction, and how it fails. The pCR benefit of the CD40-agonist arm concentrates in signature-positive patients, with little or no gain in signature-negative patients. If pCR does not separate by the signature, the hypothesis is wrong, and the arm has cost a surgical window rather than years. The neoadjuvant endpoint reads this in weeks, which is the reason this belongs in the neoadjuvant cohort rather than the metastatic or refractory ones.

Honest limits. The signature is colorectal-grounded, specified on the cohort it was derived from, so its reach to other indications is open. Read on the effector alone it cannot fully separate a tumour holding cells back from healthy gut tissue waiting correctly; reading the cDC1 licensing partner alongside tightens it, and inside tumours the confounder is small. Proving the held-in-node mechanism would need paired node and tumour tissue, but the arm does not depend on that, since enrolment runs on the tumour-side signature.

What Encounter contributes. The signature, the matching rule, and the structural read that selected this combination for this subset, supplied as a dated, pre-registered, falsifiable call that can be scored against the cohort's pCR readout.

Contact: Raimo van der Klein, founder of Encounter · raimo@encounter.bio