S129: adoptively infused memory-like (ml) natural killer (nk) cells elicit adaptive immune responses in patients with acute myeloid leukemia (aml)

Background: Natural killer (NK) cells possess a unique ability to kill tumor cells and are key effectors of immune surveillance. NK cellular therapy has shown marginal benefit in clinical trials due to antigen escape, restricted trafficking, and limited tumor infiltration. WU-NK-101 is derived from healthy donor NK cells, activated to induce a memory-like (ML) NK-cell program, expanded, and frozen to create an off-the-shelf cell therapy. We have previously shown that ML-NK cells, specifically WU-NK-101, have a unique in vivo differentiated phenotype (Berrien-Elliott et al. Cancer Discov 2020) and that they could overcome these limitations (Rutella et al. Blood 2022 & Muth et al. J Clin Oncol 2023). Aims: Herein, we aimed to determine whether ML-NK cells engage the adaptive immune response and whether modifications of the tumor microenvironment (TME) correlate with clinical benefit in 15 patients with relapsed/refractory (R/R) acute myeloid leukemia (AML) receiving ML-NK cells on study NCT#01898793. Methods: The transcriptome of ML-NKs was analyzed using bulk RNA-seq. GSVA was employed to derive an absolute enrichment score of KEGG pathways. Bone marrow (BM) trephine biopsies for targeted immune gene expression profiling (nCounter IO360®; n=740 targets) and spatially resolved proteomics (GeoMx™ DSP; n=53 targets; 93 circular regions of interest (ROIs); NanoString Technologies) were collected at baseline (BL) and on treatment (OT). Fluorescent antibodies against CD3/CD123 were used to resolve BM architecture (Fig.1A) and to select ROIs, which were dichotomized into CD3high/low and CD56high/low based on a median split of barcode counts. Differential expression analysis was performed using linear mixed models to account for subsampling, i.e., without treating ROIs from individual BM sections as independent observations. Response to ML-NK cells was defined as < 5% BM blasts on day +28.Results: Spatially resolved proteomics of BL and OT BM samples separated responders (CR) from non-responders (NR; Fig.1B). A significantly higher number of CD3highCD56high ROIs in OT BM samples associated with response (Fig.1C). Type I IFN signaling molecules STING and CD34 were the top downregulated proteins in OT samples (log2FC=-0.92, FDR=2.12×10-20 and log2FC=-0.84, FDR=9.05×10-35, respectively), in agreement with recent data on responses to checkpoint inhibitors after blockade of cancer cell type I IFN signaling (Qiu J et al, 2023). At the RNA level, NK cell infiltration positively and significantly correlated with the abundance of activated T cells (GZMB, TNFSF4, LAG3, CD27, HAVCR2, TNFSF9, PDCD1, CD44, TNFRSF18, CTLA4, IL2RA, ICOS) and with dendritic cell (DC) infiltration (CCL13, CD209, HSD11B1), suggesting coordinated changes in immune cell populations after the infusion of ML-NK cells (Fig.1D). Immune deconvolution of RNA expression corroborated these findings by showing a higher inferred abundance of macrophages, gd T cells and DCs in CR. ML-NK cells were significantly enriched in metabolic gene programs (Fig.1E), including glycolysis and OXPHOS, compared with conventional NK cells, suggesting the potential for overcoming restrictions imposed by the AML TME. Summary/Conclusion: ML-NK cell infusion was associated with TME modulation and engagement of the endogenous adaptive immune response. The high T-cell infiltration to the tumor site post-treatment points to the recruitment of the adaptive immune system and signals the potential for durable effectiveness. A phase 1 study of WU-NK-101 in R/R AML is in development (#NCT05470140). Keywords: Acute myeloid leukemia, Cellular therapy, NK cell