Command Palette

Search for a command to run...

CLN-617 and the limits of anchoring two cytokines inside a tumor

OncologyEfficacyJune 23rd, 2026·5 min read·10.5281/zenodo.20479005

Cullinan built a single molecule to keep IL-2 and IL-12 inside an injected tumor, then stopped CLN-617 after a 23-patient Phase 1 that was too small to settle the efficacy question it was built to answer.

Mechanism Risk Score

ComponentPoints
Phase-weighted failure burden4.7 / 40
Archetype severity9.8 / 25
Temporal recency4.3 / 15
Genetic evidence deficit13.9 / 15
Programmatic saturation2.5 / 5

For IL12B in Advanced solid tumor, the Mechanism Risk Score is 35/100 (yellow band). The score is a failure-burden index derived from Claidex post-mortems on this target–disease pair, not a probability of approval.

MRS 35/100 (YELLOW). 1 programs across IL12B have been documented for IL12B in Advanced solid tumor: 0 Phase 3, 0 Phase 2, 1 Phase 1 — of which 1 were efficacy failures, 0 safety, 0 biomarker, and 0 operational (enrollment, sponsor, or funding). The most informative failure on file is CLN-617 and the limits of anchoring two cytokines inside a tumor. This score quantifies the documented failure burden; the Open Targets association score of 0.07 reflects weak genetic anchoring, compounding the documented failure record. The MRS is not a prediction of future trial outcomes — it is a structured summary of the empirical record, recomputed live from the Claidex claims table, and intended to flag mechanisms where any new program must explicitly resolve each prior failure mode before pursuit is justified.

This score does not predict whether the next trial will succeed. It flags how heavy the documented mechanistic failure record is before a new program is justified.

Primary figure supporting this claim (CLN-617 / IL12B / Advanced solid tumor): CLN-617 and the limits of anchoring two cytokines inside a tumor

What was tried

CLN-617 is a single-chain fusion protein that strings together four parts: human IL-2, the collagen-binding domain LAIR2, human serum albumin, and human IL-12, all joined by glycine and serine linkers (ClinicalTrials.gov, NCT06035744 intervention record). Cullinan Therapeutics ran it as a first-in-human Phase 1 study in advanced solid tumors, dosing CLN-617 alone and in combination with the anti-PD-1 antibody pembrolizumab. The trial opened in December 2023, used a single-group non-randomized design across dose escalation, optimization, and expansion, and enrolled 23 patients before it was terminated. The posted reason was "Company decision," with an actual primary completion date of 16 June 2026 (NCT06035744). In November 2025 Cullinan stated that it ended development of CLN-617 after a review of emerging clinical data in advanced solid tumors, and notified the Massachusetts Institute of Technology that it was terminating the underlying patent license (Cullinan Form 8-K, 18 November 2025).

The biological hypothesis

IL-2 and IL-12 are among the most potent T-cell and NK-cell stimulants known, and both have a long history of working in animals and failing in people because of systemic toxicity. The CLN-617 design tried to solve that with geography rather than potency. The LAIR2 domain binds collagen, which is abundant in tumor stroma, so an intratumoral injection should hold both cytokines inside the lesion and limit how much leaks into circulation. Albumin extends the local half-life. Preclinical work reported that intratumoral CLN-617 retained IL-2 and IL-12 in the injected tumor and still triggered systemic anti-tumor immunity against distant lesions (Cancer Immunol Res, 2024). The clinical bet was that tumor-restricted delivery would unlock the efficacy of IL-12 without reproducing the dose-limiting toxicity that ended earlier systemic programs.

What actually happened

The study reached 23 patients and stopped. No efficacy results are posted on the registry (NCT06035744, hasResults = false), and the company framed the decision around emerging clinical data rather than a safety halt or a slow-enrollment problem (Cullinan Form 8-K, 18 November 2025). The structural fact worth holding onto is the size. A 23-patient single-arm escalation is a tool for finding a tolerable dose, not for measuring response rate. If such a cohort had recorded zero objective responses, the Wilson 95 percent upper bound on the true response rate would still sit at 14.3 percent (computed from n = 23). The trial could not have distinguished an inactive drug from one with a clinically interesting response rate near 15 percent. The program was discontinued at exactly the point where the efficacy question becomes answerable, and before a dataset large enough to answer it existed.

Failure mechanism, best guess

The most likely failure mode is a translational gap between a delivery concept that works in mouse flank tumors and one that has to clear a much harder bar in human disease. Two pressures compound. First, intratumoral cytokine therapy only reaches lesions a needle can reach, so the systemic abscopal effect has to carry the clinical benefit, and that effect is far more fragile in patients than in syngeneic models. Second, the genetics do not point at this mechanism for cancer. In Open Targets, IL12B associates strongly with autoimmune disease, with scores of 0.79 for psoriasis and 0.74 for Crohn's disease, and weakly with the solid-tumor class at 0.070 (Open Targets Platform, IL12B, ENSG00000113302). The human evidence base for IL-12 biology is about restraining it in inflammation, which is the basis for approved p40 antibodies, not about deploying it against tumors. CLN-617 was an engineering answer to a toxicity problem, layered on a cancer hypothesis with thin human support.

How to prevent this next time

Two quantitative tools would have changed the decision structure. The first is an explicit power and precision plan tied to the efficacy question. A 23-patient cohort gives a 95 percent upper response bound of 14.3 percent at zero events, so a credible go decision needs a pre-specified expansion of roughly 40 to 60 evaluable patients to tighten that bound below a meaningful threshold. The second is a Bayesian interim framework that states, before dosing, what posterior on response would justify continued spend. The predictive probability of eventual success is

where the prior should be anchored to the historical base rate for oncology programs entering Phase 1, near 5.3 percent for the path to approval (Wong, Siah, and Lo, Biostatistics, 2019). A prior that honest forces the interim bar high and makes a low-information 23-patient readout a planned decision point rather than a surprise. A third lever, translational validation sequencing, would have demanded human pharmacodynamic evidence that injected cytokine actually drove distant-lesion immunity before committing to expansion, since the abscopal effect is the entire commercial thesis. The single highest leverage change would have been pre-registering the abscopal pharmacodynamic readout as the gate for expansion, so the program either generated human proof of its core mechanism or stopped with that proof in hand.

What this means for similar programs

Tumor-anchored cytokines are a crowded design space, and CLN-617 is a data point about the class, not just one molecule. Albumin-binding and matrix-binding IL-12 constructs such as SON-1010 are pursuing the same toxicity-limiting logic (Frontiers in Immunology, 2024). The lesson is that solving delivery does not create a target rationale. When the genetics of a cytokine point at autoimmune restraint rather than tumor control, the burden of proof shifts onto early human pharmacodynamics, and trials should be sized to clear that burden or not run at all.

Open questions

Did CLN-617 produce any measurable intratumoral or systemic immune activation in patients, and will those data be released. Would a biomarker-enriched population, selected for injectable lesions and high stromal collagen, have changed the readout. And does the abscopal mechanism that animates this entire class survive contact with human solid tumors at all, or is the mouse-to-human gap the real ceiling.

Sources

Related failure claims

Linked claims sharing target, indication, or failure mechanism.