ASP1570 in advanced solid tumors: a DGK-zeta brake that read clean in mice and flat in patients

What was tried

Astellas ran a Phase 1/2 study (NCT05083481, protocol 1570-CL-0101) of ASP1570, an oral small molecule inhibitor of diacylglycerol kinase zeta (DGK-zeta, gene DGKZ), as monotherapy and in combination with pembrolizumab and standard chemotherapy in locally advanced or metastatic solid tumors. Merck Sharp and Dohme was a collaborator and supplied pembrolizumab (ClinicalTrials.gov). The trial enrolled 226 participants (actual) and distributed them across eleven arms, including single-agent dose escalation, monotherapy expansion in microsatellite stable colorectal cancer and in non-small cell lung cancer, and combination expansions that added docetaxel, bevacizumab, FOLFOX or FOLFIRI, and pembrolizumab plus pemetrexed and carboplatin (ClinicalTrials.gov). Astellas terminated the study with the posted reason "lack of clinical benefit" and an actual primary completion date of 2026-05-11 (ClinicalTrials.gov).

The biological hypothesis

DGK-zeta phosphorylates diacylglycerol (DAG) into phosphatidic acid, which terminates DAG-driven signaling downstream of the T cell receptor and NK cell activating receptors. Blocking that enzyme was designed to raise intracellular DAG, sustain RAS/ERK and PKC-theta signaling, and lower the activation threshold of cytotoxic lymphocytes. The framing treats DGK-zeta as an intracellular immune checkpoint that operates where surface PD-1 blockade cannot reach. Preclinically the compound enhanced human T cell activation, released anergic T cells from a hyporesponsive state, and restored functions suppressed by PD-1 and other inhibitory signals (Int Immunopharmacol 2023). It also augmented NK cell IFN-gamma production and degranulation and increased NK-mediated tumor clearance in vivo, with antitumor activity in models resistant to anti-PD-1 (Mol Cancer Ther 2025). The rationale rested on a well-characterized node of lymphocyte signaling (Front Cell Dev Biol 2016).

What actually happened

Astellas stopped the program for lack of clinical benefit after enrolling 226 patients across dose escalation and the full set of monotherapy and combination expansions (ClinicalTrials.gov). The posted primary outcomes were safety measures, dose-limiting toxicities, adverse events, and laboratory and ECG changes, rather than a powered efficacy endpoint (ClinicalTrials.gov), consistent with an early-phase program whose go decision rested on response signals in the expansion cohorts (Ann Oncol 2024, abstract 1004P). The breadth of the design is the most informative fact on the record. By termination the sponsor had tested ASP1570 as a single agent in two immunologically distinct settings, microsatellite stable colorectal cancer, a cold tumor, and non-small cell lung cancer, and had layered it onto chemotherapy and anti-PD-1 backbones. None of these combinations produced benefit sufficient to continue (ClinicalTrials.gov).

Failure mechanism, best guess

The Open Targets association between DGKZ and neoplasm is 0.0986, built almost entirely from literature mining (datatype score 0.81) with no genetic association evidence (Open Targets Platform v25). The target was nominated from mechanism and mouse pharmacology, not from human genetics linking DGKZ loss to lower cancer risk or better outcomes. That gap matters because the DAG brake is redundant. DGK-alpha and DGK-zeta both constrain DAG in T cells, and the tumor and its suppressive microenvironment can compensate through one isoform when the other is inhibited, which is the stated rationale for dual DGK-alpha/zeta inhibitors such as BMS-986408 (Cancer Immunol Res 2025). Single-isoform inhibition raises DAG only partially, and the residual phosphatidic acid route plus parallel checkpoints can hold effector cells below the activation threshold (Front Immunol 2022). A monotherapy that depends on releasing a redundant brake, in tumors that were not selected for DAG-pathway dependence, reads clean in engineered models and flat in unselected patients. The most likely failure mode is translational mismatch compounded by isoform redundancy, surfacing as absent efficacy rather than toxicity.

How to prevent this next time

Two quantitative levers would have changed the decision calculus. First, a Bayesian predictive probability of success computed at each expansion interim, in place of a fixed enrollment plan, would have stopped the weakest of eleven arms early and concentrated patients. The predictive probability integrates the chance of a positive final readout over the current posterior on the response rate:

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

With a Beta(1,1) prior on response rate and a go threshold near a 20 percent objective response rate, the absence of three to four responses in the first 15 to 20 evaluable patients per arm drives the predictive probability below 10 percent, a defensible stop. Second, a translational validation sequence with biomarker enrichment would have front-loaded the question that the genetics never answered. Pairing a pharmacodynamic readout of DAG-pathway release, phospho-ERK or a DAG-responsive transcriptional signature in paired on-treatment biopsies, with a DGKA-low expression cut converts eleven empirical arms into one or two enrichment hypotheses anchored to a measurable engagement endpoint. Historical base rates reinforce the point, since oncology assets that enter the clinic without human genetic support have advanced at roughly half the rate of genetically supported ones. The single highest leverage change would have been gating expansion-cohort entry on a prospective pharmacodynamic biomarker of DAG-pathway release together with a DGK-alpha-low enrichment, so that a redundant-brake mechanism was tested only where it could plausibly work.

What this means for similar programs

Intracellular checkpoint inhibitors aimed at redundant nodes inherit the redundancy as clinical risk. The lesson extends to DGK-alpha-selective and dual DGK programs, to CISH, and to other release-the-brake targets where a paralog or a parallel pathway can compensate. The Claidex Mechanism Risk Score for DGKZ is 37 of 100 (yellow), and the genetic deficit component contributes 13.5 of its 15 available points (Claidex MRS), a direct quantification of the missing human anchor. A second program against this target should state, before first patient, which compensating isoform it neutralizes and which patients are selected for dependence.

Open questions

Did ASP1570 reach sustained target engagement and DAG-pathway release at tolerated doses in humans, and was that engagement simply insufficient. Would dual DGK-alpha/zeta inhibition, which addresses the redundancy head on, clear the bar that single-isoform inhibition did not. Is there a DGK-alpha-low or DAG-pathway-high subset, for example in NK-cell-rich tumors, where single-isoform inhibition retains activity. The terminated record cannot answer these, because the program closed before posted results.

Sources

1. • ClinicalTrials.gov, NCT05083481 (terminated, lack of clinical benefit, n=226, eleven arms, primary completion 2026-05-11): https://clinicaltrials.gov/study/NCT05083481 - Open Targets Platform v25, DGKZ (ENSG00000149091) to neoplasm (EFO_0000616), overall association 0.0986, literature datatype 0.81: https://platform.opentargets.org/target/ENSG00000149091 - ASP1570 augments NK cell function, Int Immunopharmacol 2023.- Enhanced antitumor immunity by ASP1570, Mol Cancer Ther 2025.- BMS-986408, first-in-class dual DGK-alpha/zeta inhibitor, Cancer Immunol Res 2025.- Diacylglycerol kinases in T cell tolerance and effector function, Front Cell Dev Biol 2016.- DAG and ERK-mediated control of CD8 T cell responses, Front Immunol 2022.- Phase 1/2 trial of ASP1570 in advanced solid tumors, Ann Oncol 2024, abstract 1004P: https://www.annalsofoncology.org/article/S0923-7534(24)02582-1/fulltext.