XMT-2056, a HER2-directed STING-agonist ADC, ends in Phase 1 after a fatal dose-escalation signal and an acquisition

What was tried

The trial was a Phase 1, first-in-human, dose-escalation and dose-expansion study of XMT-2056 in adults with advanced or recurrent solid tumors that express HER2 (NCT05514717). The primary objectives were dose-limiting toxicities during the first cycle, the overall incidence of adverse events, and objective response rate in the expansion phase. Actual enrollment reached 57 participants. The study started in January 2023 and recorded a primary completion date of May 2026. ClinicalTrials.gov lists it as terminated with the posted reason "Development discontinued."

XMT-2056 is an Immunosynthen STING-agonist antibody-drug conjugate. It pairs an antibody against a novel HER2 epitope with a STING-agonist payload, so the construct is designed to deliver an innate-immune trigger to HER2-positive tumor cells and the immune cells around them (Bukhalid et al., 2025). The asset originated at Mersana Therapeutics, in a collaboration that involved GSK. The current ClinicalTrials.gov sponsor of record is Day One Biopharmaceuticals, which acquired Mersana in a deal announced in November 2025 and closed in January 2026.

The biological hypothesis

HER2 is one of the best-validated targets in oncology. The Open Targets Platform scores the HER2 to neoplasm association at 0.9245 overall, with a clinical evidence component of 0.998 and a somatic mutation component of 0.976, and the platform catalogs 47 drugs or clinical candidates against the target. HER2 antibody-drug conjugates are an established modality, with trastuzumab deruxtecan showing activity in HER2-low breast cancer and HER2-mutant non-small cell lung cancer (Modi et al., 2022, and Li et al., 2022).

XMT-2056 tried to change the payload rather than the target. Instead of a cytotoxic warhead, it carried a STING agonist, with the goal of converting HER2 expression into a homing signal for local innate-immune activation. STING agonism stimulates type I interferon and can turn immunologically cold tumors hot, and several groups have pursued the same antibody-delivered STING concept to widen the narrow therapeutic window of free STING agonists (Long et al., 2025, and Wu et al., 2025).

What actually happened

In March 2023, Mersana voluntarily paused the trial and the FDA placed it on clinical hold after a Grade 5, meaning fatal, serious adverse event that was assessed as related to XMT-2056. The event occurred in the second patient dosed at the initial dose level of the escalation. In October 2023, the FDA lifted the hold after the investigators lowered the starting dose for dose escalation.

The study then continued for more than two years and reached a primary completion date in May 2026 before being terminated as "Development discontinued." The termination coincided with Day One's acquisition of Mersana, after which Day One emphasized other pipeline assets, including its adenoid cystic carcinoma program. XMT-2056 has no marketed adverse event record, so FAERS returns no signal for the molecule. For class context, the approved HER2 ADC trastuzumab deruxtecan carries a prominent interstitial lung disease signal in FAERS, which underlines that even validated HER2 payload delivery brings serious organ toxicity.

Failure mechanism, best guess

The posted reason is a development discontinuation, and the timing points to a strategic reprioritization after an acquirer inherited the asset. That is the proximate cause. The mechanistic cause that shaped the asset's value was the therapeutic window of a systemically administered innate-immune agonist.

STING activation is potent and pleiotropic. A circulating conjugate that can trip STING signaling carries the risk of activating the pathway in normal tissue and in the vasculature, which can produce cytokine-driven, sometimes catastrophic toxicity. The fatal event at the first dose level, before any meaningful exposure-response had been established, is the kind of early signal that compresses the dose range a program can explore. Lowering the starting dose reopened the trial but also lowered the ceiling on the deliverable STING stimulus, which squeezes the gap between a dose that activates antitumor immunity and a dose that harms the patient. A target as validated as HER2 does not rescue a payload whose on-mechanism toxicity appears at or near the lowest tested dose. When an acquirer then weighed a Phase 1 asset with a fatal dose-escalation history against a competitive landscape dominated by approved HER2 ADCs, the expected value did not justify continued spend. The best guess is a viable target paired with a payload whose systemic safety margin was too thin to clear, made terminal by a strategic decision.

How to prevent this next time

Two quantitative tools would have framed the decision more cleanly.

First, a Bayesian predictive probability of success that conditions on the early safety data and the absence of an efficacy readout:

PP(success)=∫θ​P(final success∣θ,interim data)⋅π(θ∣interim data)dθ

A fatal dose-limiting event at the first dose level shifts the posterior on the tolerable dose downward and widens uncertainty, so the predictive probability of reaching a registrational endpoint falls well before any response data arrive.

Second, explicit interval estimation on the realized response experience. With no confirmed responses reported and a small dosed cohort, the binomial uncertainty is large. A Wilson 95 percent interval for 0 of 20 evaluable patients spans 0 to about 16 percent, and for 0 of 57 it spans 0 to about 6 percent, so a Phase 1 that has not produced a clear response signal cannot statistically exclude a clinically meaningful rate, which is exactly when continuation should hinge on the safety ceiling rather than hope of efficacy. A historical base-rate adjustment reinforces this, since oncology assets advancing from Phase 1 to approval succeed about 3.4 percent of the time historically, and trials with biomarker-based selection do better (Wong, Siah, and Lo, 2019). Finally, a competitive red-team would have scored XMT-2056 against approved HER2 ADCs early, recognizing that a novel payload must beat, not match, an entrenched standard of care. The single highest leverage change would have been a pre-specified safety-window stopping rule that treated an on-mechanism Grade 5 event at the lowest dose as a near-fatal blow to the program's viability, rather than a tolerated restart at a lower dose.

What this means for similar programs

Immune-stimulating antibody conjugates inherit two risk layers: the target and the payload pharmacology. When the payload is a systemic innate-immune agonist, the binding antigen, no matter how validated, does not control where the agonist acts once it is in circulation. Programs in this class should define the therapeutic window before scaling investment, ideally with biomarkers of on-target versus off-tissue STING activation, and should treat early high-grade toxicity at low doses as information about the mechanism, not just the molecule.

Open questions

Did any confirmed responses occur across the 57 dosed patients before the program stopped, and will those data be published? Was the fatal event driven by tumor-bound or by systemic STING activation, since that distinction decides whether a better antibody or a different payload could rescue the concept? And does any HER2-directed STING approach have a window wide enough to matter, or does the modality need tumor-restricted activation chemistry to be viable at all?

Sources

1. • ClinicalTrials.gov, NCT05514717, Phase 1 first-in-human design, dose-limiting toxicity and objective response endpoints, enrollment of 57, primary completion May 2026, and termination reason "Development discontinued." https://clinicaltrials.gov/study/NCT05514717 - Bukhalid RA, Duvall JR, Lancaster K, et al. XMT-2056, a HER2-Directed STING Agonist Antibody-Drug Conjugate, Induces Innate Antitumor Immune Responses by Acting on Cancer Cells and Tumor-Resident Immune Cells. Clinical Cancer Research.

2. - Long Y, Tang B, Xie F, et al. Novel Quaternary Ammonium Salt-Linked STING Agonist Antibody-Drug Conjugate: Synergistic Activation of Tumor Immunity with Mitigated Off-Target Toxicity. Advanced Science.

3. - Wu G, Yu C, Ding C, et al. Next-Generation HER-2 Tumor-Targeted Delivery of the STING Agonist Immune-Stimulating Antibody Conjugate Improves Anticancer Efficacy. MedComm.

4. - Mersana Therapeutics. Announces Clinical Hold on XMT-2056 Phase 1 Clinical Trial (fatal serious adverse event, second patient at the initial dose level). March 13, 2023. https://www.globenewswire.com/news-release/2023/03/13/2625598/0/en/Mersana-Therapeutics-Announces-Clinical-Hold-on-XMT-2056-Phase-1-Clinical-Trial.html - Mersana Therapeutics. Announces FDA has Lifted Clinical Hold on Phase 1 Clinical Trial of XMT-2056 (after lowering the dose-escalation starting dose). October 31, 2023. https://www.globenewswire.com/news-release/2023/10/31/2770089/0/en/Mersana-Therapeutics-Announces-FDA-has-Lifted-Clinical-Hold-on-Phase-1-Clinical-Trial-of-XMT-2056.html - Day One Biopharmaceuticals to Acquire Mersana Therapeutics. November 13, 2025. https://www.globenewswire.com/news-release/2025/11/13/3187105/0/en/Day-One-Biopharmaceuticals-to-Acquire-Mersana-Therapeutics.html - Open Targets Platform, ERBB2 (ENSG00000141736) to neoplasm (EFO_0000616) association, overall 0.9245, clinical 0.998, somatic mutation 0.976, with 47 drugs or clinical candidates. https://platform.opentargets.org/target/ENSG00000141736 - Modi S, Jacot W, Yamashita T, et al. Trastuzumab Deruxtecan in Previously Treated HER2-Low Advanced Breast Cancer. New England Journal of Medicine.

5. - Li BT, Smit EF, Goto Y, et al. Trastuzumab Deruxtecan in HER2-Mutant Non-Small-Cell Lung Cancer. New England Journal of Medicine.

6. - openFDA FAERS, trastuzumab deruxtecan (ENHERTU) class adverse event counts, interstitial lung disease as a leading reported term. https://api.fda.gov/drug/event.json - Wong CH, Siah KW, Lo AW. Estimation of clinical trial success rates and related parameters. Biostatistics. 2019;20(2):273-286.