CAR T Cells Accelerate Disease Progression in AMLs Harboring an Inflammatory Transcriptional Signature

Background: Chimeric antigen receptor T cells (CART) for acute myeloid leukemia (AML) have yielded disappointing responses in most reported clinical studies, with high rates of treatment failure or progressive disease (PD). We recently identified an unexpected mechanism of resistance arising from CART-associated cytokines that suppress apoptosis in AML blasts via JAK/STAT signaling, and showed this resistance mechanism was active in our clinical study of CART targeting CD123 (CART123) on AML blasts (NCT03766126) (Bhagwat et al., 2024). Here we sought to deepen our understanding of CART treatment failure in AML in order to advance cellular immunotherapies for patients with otherwise untreatable myeloid leukemias. Methods and Results: Of 12 patients treated on our clinical study, 4 patients had expansion of blasts within 14 days of CART123 infusion and were assessed as having PD on study. We have identified primary human AML samples that appear to behave similarly ex vivo. Unlike typical AMLs (typAML) which are killed by CART, these progressive AMLs (pAMLs) are instead stimulated to proliferate. The proliferative phenotype is observed even when pAML are physically separated from activated CART123 cells using a solute-permeable membrane, implicating CART-secreted factors in this phenotype. Surprisingly, proliferation is observed in pAML even when exposed to anti-CD19 CART cells that had been activated by CD19+ leukemia, suggesting that soluble factors from other clinically used CART products can trigger proliferation of myeloid leukemia cells. In support of our hypothesis that secreted cytokines mediate this proliferative response in AML, blockade of cytokine signaling using the JAK1/2 inhibitor ruxolitinib prevents CART-stimulated AML proliferation. To understand why some AML proliferates in response to CART stimulation, we performed phospho-proteomic profiling of 39 phospho-peptides spanning several independent signaling pathways in pAML and typAML samples. CART exposure stimulates phospho-STAT5 (pSTAT5) equally in both types of AML samples, consistent with our prior work. However in pAML, levels of pSTAT5 are dramatically elevated at baseline when compared to typAML, suggesting that this pathway is primed in pAML. Phospho-flow cytometry confirmed elevated baseline pSTAT5 in pAML compared to typAML. Single cell RNA sequencing (scRNA-seq) of primary pAMLs and typAMLs mixed with either CART cells (“stimulated” condition) or donor-matched normal T cells (“baseline” condition) supports our proteomic findings. When comparing stimulated pAML to stimulated typAML, overrepresented transcriptional pathways include “Inflammatory Response” and numerous cytokine-stimulated JAK-STAT pathways. Importantly, these inflammatory signaling pathways are already overrepresented in baseline pAML vs typAML. Thus, baseline inflammatory signaling identifies a subgroup of AML with exceptionally poor response to CART therapy. To test the clinical relevance of this finding, we applied our pAML gene expression signature to baseline bone marrow aspirate scRNA-seq samples from patients (n = 7) with AML who were enrolled on our phase I clinical study of CART-123. Patients with clear increase in blast count by 14 days after CART-123 infusion (n = 3) had higher scores than patients who had stable or decreased blast counts after CART-123 (mean signature score 4.302 vs 1.321, P = 0.0046), suggesting this score has clinical utility in predicting responses to CART123. Conclusion: Disease progression after CART therapy is typically attributed to the natural history of aggressive cancer. Our data suggest that CART cells paradoxically have a direct role in accelerating progression of AML via cytokine release. In particular, we find evidence of inflammatory signaling in AML progresses upon exposure to CART, and we have developed a tool that can identify such AMLs in pre-CART samples. This may be used to refine patient selection, or to highlight those patients who will require additional combination therapy, such as ruxolitinib, to improve outcomes with AML-directed CART.