ResearchPod Summary
Acute myeloid leukemia (AML) remains difficult to treat due to the lack of highly specific tumor antigens and a suppressive bone marrow microenvironment. This study investigates whether targeting LILRB4—a receptor highly expressed on monocytic AML blasts but largely absent on normal hematopoietic stem cells—using a novel synthetic T-cell receptor and antigen receptor (STAR-T) platform can provide a safe and effective therapeutic option for relapsed/refractory (R/R) AML.
The researchers developed a dual-epitope STAR-T cell therapy (DE STAR-T) by fusing two high-affinity nanobodies targeting distinct LILRB4 epitopes to the constant regions of the T-cell receptor. They compared the efficacy of these DE STAR-T cells against single-epitope STAR-T cells and conventional CAR-T cells in vitro and in vivo. Following preclinical validation, they conducted a first-in-human phase I clinical trial (NCT05548088) in nine patients with LILRB4-positive R/R AML, using single-cell RNA sequencing to analyze mechanisms of response and treatment failure.
Preclinical results showed that DE STAR-T cells exhibited superior cytotoxicity, cytokine production, and tumor control compared to monovalent STAR-T or dual-epitope CAR-T cells, particularly against tumor cells with low LILRB4 expression. In the clinical trial, the therapy demonstrated an overall response rate (ORR) of 50% in the efficacy-evaluable set. While no severe cytokine release syndrome (CRS) or neurotoxicity (ICANS) was observed, all patients experienced grade 3 or higher adverse events, primarily pancytopenia and severe infections, which led to three deaths. Single-cell analysis suggested that monocyte-mediated suppression of autologous T-cell function is a primary driver of treatment failure in nonresponders.
This study provides proof-of-concept for LILRB4 as a viable target for AML immunotherapy. By utilizing the STAR-T platform, which mimics native TCR signaling, the researchers achieved potent antitumor effects even against low-antigen-expressing leukemia cells. While the safety profile—specifically the high rate of severe infections—necessitates caution and improved supportive care, the clinical responses observed in heavily pretreated patients suggest that this approach could be a transformative strategy for monocytic AML.
Sam: A phase I trial of a dual-epitope STAR-T cell therapy against LILRB4 reported a 50 percent objective response rate in heavily pretreated AML, with no grade 3 or higher neurotoxicity. The work was led by Xiangyu Zhao and published in Signal Transduction and Targeted Therapy.
Alex: Fifty percent in that population is notable, but I'd want the denominator before I react to it. And why a synthetic T-cell receptor rather than a conventional CAR?
Sam: The denominator is six evaluable patients, so three responders. I'll come back to what that can support. On the architecture, the problem in AML is antigen density. Conventional CARs need a fairly high density of target to trigger signaling. STAR-T cells instead fuse targeting nanobodies to the native TCR alpha and beta constant regions, so they use the native TCR signaling cascade, which is inherently more sensitive to sparse antigen than synthetic CAR domains.
Alex: So the receptor lowers the activation threshold. Where does the dual-epitope part come in?
Sam: Two nanobodies bind different contact points on LILRB4. That raises avidity, so the T cell stays engaged even when expression on the blasts is low. It also hedges against antigen escape, because losing one epitope doesn't remove the other.
Alex: Binding strength is one thing. What spares normal hematopoiesis?
Sam: Target selection. LILRB4 is highly expressed on monocytic AML blasts but largely absent from normal hematopoietic stem cells. That should limit the on-target, off-tumor toxicity that would otherwise drive myelosuppression. Together with the absence of severe neurotoxicity, that makes the safety profile look favorable on its face.
Alex: "On its face" is doing some work, though. These are people with advanced AML. You'd expect a lot of adverse events regardless of the product.
Sam: Yes. The patients were heavily pretreated and already profoundly neutropenic. There were deaths, but the study did not classify them as dose-limiting toxicities, because they were linked to infections in that neutropenic setting rather than to the cells. That's the authors' attribution, and with six patients it's hard to test independently. Persistent cytopenia and infection risk are the reason they argue future trials need stricter screening for fungal and bacterial colonization before infusion.
AI-generated third-party summary by ResearchPod. Not official content or an endorsement by the paper authors or affiliated organizations.
Alex: Then let's go to the non-responders. Why did half the patients not respond?
Sam: This is where the single-cell RNA sequencing comes in, and I'd call it the most informative part of the paper. In non-responders, the myeloid compartment looked immunosuppressive. Monocytic myeloid-derived suppressor cells were present, and the data pointed to suppression of T-cell function through pathways including LGALS9 and MIF. Their myeloid cells also showed higher reactive oxygen species activity, which correlated with inhibitory capacity.
Alex: Correlated is the operative word. With three patients per arm, this is hypothesis generation, not a mechanism demonstrated by perturbation.
Sam: That's a fair reading, and the authors describe the study as exploratory. The data suggest that the dual-epitope design addresses antigen density but not what the marrow does to the T cells afterward. Even a sensitive receptor may not help if the cells are dysfunctional before they can kill. So the question shifts from whether the product kills to why it stops killing.
Alex: Which implies the receptor isn't the ceiling so much as the microenvironment around it.
Sam: That's the interpretation the data support. Some responders also relapsed, so long-term persistence is unresolved. On the positive side, three patients achieved remission and some went on to successful transplants, which is a meaningful signal. But a careful referee would point to the sample size. With six evaluable patients there's no statistical power to claim efficacy. It is a signal, not a proof.
Alex: So the load-bearing evidence is the response and safety signal, read cautiously, plus the single-cell comparison that offers a explanation for the non-responders.
Sam: Yes, and the architecture is the thing the field can carry forward. Fusing dual nanobodies to the native TCR constant regions appears to get around the activation-threshold problem in low-antigen settings. Whether that yields durable remission in larger, randomized cohorts is the open question.
Alex: And the follow-up work would be about the niche as much as the receptor.
Sam: The authors point toward combining the platform with agents like venetoclax to upregulate LILRB4, along with better supportive care. They also raise off-the-shelf allogeneic T cells as a way to sidestep the functional impairment seen in these autologous products.
Alex: If you want the figures, method choices, and caveats we skipped, you can generate a deep dive of this paper. The paper has the rest either way.
Sam: Thanks for listening.