SAR443579: a CD123 NK-cell engager handed back after modest single-agent remission rates in relapsed AML

What was tried

SAR443579, also called IPH6101, is a trifunctional NK-cell engager built on the Innate Pharma ANKET platform and developed by Sanofi. The molecule binds CD123 (the IL3RA chain, ENSG00000185291) on leukemic blasts, and it co-engages NKp46 and CD16 on natural killer cells to redirect NK cytotoxicity onto the tumor. NCT05086315 was an open-label, first-in-human, single-group dose-escalation and dose-expansion study in adults and children with relapsed or refractory acute myeloid leukemia, B-cell acute lymphoblastic leukemia, high-risk myelodysplastic syndromes, and blastic plasmacytoid dendritic cell neoplasm (ClinicalTrials.gov NCT05086315). It enrolled 101 participants and dosed by intravenous infusion across escalation cohorts from 10 micrograms per kilogram to 6000 micrograms per kilogram. Sanofi terminated the study with the posted reason "early discontinuation based on strategic sponsor decision not driven by any safety concerns," and returned the asset to Innate Pharma in April 2025 (Innate Pharma and FierceBiotech, 2025).

The biological hypothesis

CD123 is expressed on AML blasts and on leukemic stem cells at higher density than on normal hematopoietic progenitors, which is the standing rationale for targeting it. The NK-engager design was meant to avoid the cytokine release and neurotoxicity that limit CD3 T-cell engagers while recruiting an effector population that is often preserved in heavily pretreated AML. Open Targets scores the IL3RA to AML association at 0.4825, carried by literature and clinical evidence with no direct genetic association for the pair (Open Targets Platform, 2025). The target has approved-drug tractability, since tagraxofusp, a CD123 cytotoxin, is already marketed for BPDCN. The premise was mechanistically reasonable: high target expression, a validated antigen, and an effector arm chosen to widen the therapeutic window.

What actually happened

Safety behaved as designed. No dose-limiting toxicities were reported up to the highest dose of 6000 micrograms per kilogram weekly (Sanofi ASH 2023 presentation; Innate Pharma, 2024). The efficacy signal did not follow. In the adult dose-expansion Cohort A1 at 1000 micrograms per kilogram, which functioned as the recommended expansion dose, the composite complete remission rate was 5.0 percent, that is 1 of 20 evaluable participants, and the overall response rate was also 5.0 percent (ClinicalTrials.gov NCT05086315 results, Cohort A1, n=20). The Wilson 95 percent confidence interval for 1 of 20 runs from 0.9 percent to 23.6 percent, so the point estimate is low and the interval is wide. Escalation cohorts showed scattered responses of 20 to 50 percent, but each rested on single-digit denominators. Reported duration of composite complete remission in the expansion analysis was 29 days. Sanofi advanced to the Phase 2 expansion in April 2024, then discontinued and handed the program back within a year (Innate Pharma, 2024 and 2025).

Failure mechanism, best guess

This reads as a translational efficacy shortfall dressed as a portfolio decision. The stated reason was strategic, and no safety concern was cited, but a 5 percent single-agent remission rate at the expansion dose in relapsed AML is below the bar that would justify a competitive registration path. Two mechanistic factors are plausible. First, NK effector fitness in relapsed or refractory AML is frequently compromised, so an engager that depends on endogenous NK cytotoxicity may lack the effector supply that the design assumes. Second, CD123 density and the leukemic microenvironment vary widely across patients, and antigen-low or immune-excluded disease can blunt redirected killing even when the construct binds well. The clean safety record supports that the molecule engaged its target without excess toxicity, which points the failure toward insufficient effector-mediated cytotoxicity rather than an inability to reach the antigen.

How to prevent this next time

Two quantitative tools would have sharpened the decision. First, a Bayesian predictive-probability stopping rule at the expansion stage. With a beta prior on the remission rate and a go threshold set to a clinically meaningful floor, the predictive probability of eventual success can be evaluated at each interim:

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

At 1 of 20, a beta(1,1) prior updates to beta(2,19), whose mean is 0.095 and whose mass above a 20 percent remission floor is small, which would have triggered an early stop for futility on the monotherapy expansion. Second, an NK effector biomarker enrichment strategy. Screening for baseline NK count, CD16 genotype, and prior hypomethylating exposure, then enriching the expansion for effector-competent patients, would raise the number needed to screen but concentrate any true signal. Historical base rates reinforce the caution: oncology assets show roughly a 3.4 percent probability of success from Phase 1 to approval, against about 13.8 percent across all indications (Wong, Siah, and Lo, 2019). The single highest leverage change would have been ...

What this means for similar programs

CD123 remains one of the most crowded targets in myeloid oncology, with tagraxofusp approved and flotetuzumab, pivekimab sunirine, and vibecotamab among the programs that have struggled to convert target validity into durable single-agent remissions. The SAR443579 result suggests that clean safety for an NK engager is achievable but does not guarantee efficacy, and that effector-arm choice does not rescue an antigen whose therapeutic index is already narrow. Programs relying on endogenous NK cytotoxicity in relapsed AML should assume effector insufficiency until proven otherwise and should build effector-fitness readouts into first-in-human design.

Open questions

Did responders in the escalation cohorts share a measurable NK or CD123 phenotype that would define an enrichable subgroup. Would combination with a hypomethylating agent or venetoclax convert the 5 percent monotherapy rate into a workable regimen. Does the ANKET NK-engager format retain any advantage over CD3 engagers once effector fitness is accounted for. Was the decision to return the asset driven by this efficacy readout or by unrelated portfolio pressure at Sanofi.

Sources

1. • ClinicalTrials.gov, NCT05086315, full record and posted results, terminated status, whyStopped text, enrollment 101, Cohort A1 composite complete remission 1 of 20 and overall response 5.0 percent, duration 29 days: https://clinicaltrials.gov/study/NCT05086315 - Open Targets Platform, IL3RA (ENSG00000185291) to acute myeloid leukemia (MONDO_0018874) association 0.4825, tractability: https://platform.opentargets.org/target/ENSG00000185291 - Innate Pharma, advancement of SAR443579/IPH6101 to Phase 2, April 2024: https://www.innate-pharma.com/media/all-press-releases/innate-pharma-announces-advancement-sanofi-developed-nk-cell-engager-sar443579/iph6101-progressing-phase-2-blood-cancer-patients - FierceBiotech, Innate ends work on the NK-cell engager as part of pipeline refocus, 2025: https://www.fiercebiotech.com/biotech/innate-ends-work-nk-cell-engager-phase-12-trial-part-ongoing-pipeline-refocus - Wong CH, Siah KW, Lo AW. Estimation of clinical trial success rates and related parameters. Biostatistics. 2019;20(2):273-286.- Recent Advances in Immune-Based Therapies for Acute Myeloid Leukemia. Blood Cancer Discovery.

2. - Escape from T-cell-targeting immunotherapies in acute myeloid leukemia. Blood.

3. - Wilson score interval computed on the posted Cohort A1 result 1 of 20, 95 percent confidence interval 0.9 to 23.6 percent.