Adoptively Infused Memory-like Natural Killer Cells Impact Adaptive Immune Responses in Patients with Acute Myeloid Leukemia

Natural killer (NK) therapies have been challenging owing to antigen escape, restricted trafficking, and limited tumor infiltration, translating into modest clinical benefit. We have shown that memory-like (ML)-NK cells acquire an in vivo differentiated phenotype distinct from conventional NK cells and are well-tolerated without cytokine release syndrome, GVHD, or neurotoxicity. WU-NK-101 (W-NK) is an allogeneic cytokine-reprogrammed, expanded, cryopreserved ML-NK cell product derived from healthy human peripheral blood mononuclear cells, with scalable manufacturability, that is currently in clinical development in relapsed/refractory (R/R) AML (NCT# 05470140).

We conducted a multi-modal analysis of the immune tumor microenvironment (TME) in 13 patients with R/R AML receiving ML-NK cells (NCT#01898793) with the aim to determine whether TME modifications induced by ML-NK cells correlate with clinical benefit.

NK cells were purified from leukapheresate, and activated with IL-12, IL-15, and IL-18 for 12-16 hours. Patients were lymphodepleted with fludarabine and cyclophosphamide. NK cells were infused in escalating doses: 0.5 × 10⁶/kg, 1.0 × 10⁶/kg, and all NK cells from a single leukapheresis, capped at 10 × 10⁶/kg. Response was defined as <5% BM blasts on d+28. BM trephines for immune gene expression profiling (nCounter IO360®) and spatial proteomics (GeoMx™ DSP; n=53 immuno-oncology [IO] targets; NanoString Technologies, Seattle) were collected at baseline (BL) and on treatment (OT). Immune infiltration, NK and T cell number from BM samples was quantified manually on immunofluorescent imaging. Oncomine™ TCR Pan-Clonality assay (Thermo Fisher) was used to profile TCRb and g repertoires by NGS. The Shannon index was used to describe TCR repertoire diversity, and clonal evenness was used as a measure of the similarity of clone sizes.

BM infiltration with CD8⁺ effectors, enhanced glycolysis and cell cycling was higher in BM samples from responders (CR) to ML-NK cells. By contrast, OXPHOS RNA scores and gene signatures of immune senescence and terminal T-cell exhaustion were higher in non-responders (NR). We employed machine learning to construct an 8-gene predictor of response which encompassed T/NK markers, such as CD7, and IFN regulatory factors and TLRs. This score was significantly higher at baseline in CR and positively correlated with response duration. Principal component analysis (PCA) of spatially resolved IO proteins separated CR from NR, particularly in OT specimens, where T-cell markers CD3/CD8 and inhibitory receptors CTLA4/LAG3 were among the top contributing features. CD34, type I IFN signaling molecule STING1 and PARP1 were significantly higher at baseline compared with OT, with OT showing higher expression of cytotoxicity markers and inhibitory receptors (e.g., Tim-3 and CD276). This observation strongly suggests modulation of the immune TME after the infusion of ML-NK cells. Response to ML-NK cells was associated with coordinated up-regulation of T cell, NK cell and B cell protein markers. To further substantiate the co-localization of T cells and NK cells in the BM FFPE sections (Fig. A), and its correlation with clinical efficacy we annotated the 93 ROIs in our spatial proteomics dataset based on a median split of CD3 and CD56 protein barcodes. ROIs from CR were largely CD3^highCD56^high, both at BL and OT. Conversely, ROIs from BM sections of NR to ML-NK cells had a predominantly CD3^lowCD56^low phenotype. Quantitative analysis of T-cell infiltration post ML-MK dosing showed a significant association with response, 0.9 (PD) vs. 5.3 (CR)-fold-increase (P=0.01). Interestingly, data demonstrate selected intra-BM clonal expansion as evidenced by decreased TCR diversity (mean ± SEM, 4.6±0.3 vs. 3.3±0.4, pre vs. post;P=0.013) and increased clonality (Fig. B).

Thus, pre-existing T-cell responses appear to be co-opted in the AML TME of CR, suggesting the potential for durable effectiveness of ML-NK cells. Multi-modal analyses of WU-NK-101 highlight unique transcriptional and functional states, which could underpin in vivo potency and persistence. WU-NK-101 are currently being evaluated in a phase 1 study of R/R AML (#NCT05470140).