Genomic Subtypes of AML Define Sensitivity to NK Cell Cytotoxicity

Acute myeloid leukemia (AML) is characterized by poor prognosis and remains a challenging disease to treat. New therapeutic approaches are being introduced, and natural killer (NK) cell-based immunotherapy has been recognized as a promising option. However, which subgroups of AML patients that are most likely to respond to NK cell therapy, remain undefined. We explored the susceptibility of AML cells to NK cell cytotoxicity in ex vivo co-cultures.

We analyzed leukemic cell sensitivity to NK cells using vitally frozen bone marrow (BM) samples from 99 AML patients at diagnosis and expanded NK cells from one healthy donor. BM samples were exposed to NK cells at five effector-to-target (E:T) ratios 0:1, 1:1, 2:1, 4:1 and 8:1, followed by a flow cytometry-based killing assay. We also performed multiplexed single-cell RNA-sequencing (sc-RNAseq) to understand how NK and BM cells responded transcriptomically to their interaction. We characterized the clinical features of the AML cohort and integrated the information with sensitivity and transcriptomic data to reveal the clinical and molecular drivers behind leukemic cell susceptibility to NK cell killing.

The cohort displayed a diversity of molecular subtypes and a balanced distribution across the 2022 European LeukemiaNet (ELN) risk groups (favorable n=27, intermediate n=34, adverse n=35). The international consensus classification (ICC) of AML mapped the patients into all categories, ranging in size from 2.0% to 53.5%. The cohort also exhibited diversity in their French-American-British (FAB) subtypes, M0 (n=4), M1 (n=23), M2 (n=23), M3 (n=2), M4 (n=10), M5 (n=17), M6 (n=2), mixed subtype (n=14), undefined (n=4). The median age of patients was 63 (range 27-82) years.

The flow cytometry killing assay uncovered the AML blast populations. 60% of samples had CD34+ myeloblasts, 40% CD34- myeloblasts and 32% CD14+ monocytic-like blasts, with half of the CD14+ samples being CD34+. All blast populations displayed prominent heterogeneity to NK cell cytotoxicity, indicating the presence of both sensitive and resistant AML subgroups. The median killing of myeloblasts spanned from 13.3% (range 0%–75.7%) at 1:1 E:T ratio to 31.3% (range 0%–92.3%) at the highest ratio. We further measured sensitivity to NK cells by calculating the area under the dose-response curve (AUC) capturing the effect of all E:T ratios in each sample. The median AUC for myeloblasts was 81.8. Samples with mutated BCOR (n=6) and RAS oncogenes (n=27), including KRAS, NF1 and NRAS, showed increased sensitivity to NK cytotoxicity with significantly lower AUC values, 59.3 and 69.8 (Mann-Whitney p = .004 and p = .005) respectively, compared to cases lacking these mutations. NK cytotoxicity was especially predominant in CD34+ myeloblasts. The overall AUC median for CD34+ blasts was 76.3. BCOR (n=5) mutated cases had a median of 48.1 and RAS oncogenes (n=16) 63.4. CD14+ blasts were more prone to NK killing with a median of 21.8. Killing of CD34+ and CD14+ blasts correlated for samples exhibiting both myeloblasts and monocytic-like blasts (Spearman r_s= .66, p = .006).

Sc-RNAseq identified tumor-induced activation states in expanded NK cells upon interaction with AML cells. The percentage of NK cells displaying an activated phenotype correlated with increased killing of both CD34+ (Spearman r_s= -.56, p < .001) and CD14+ (Spearman r_s= -.51, p = .003) blasts. The median NK activation for CD34+ samples was 45.8%. Both BCOR and RAS mutated CD34+ samples triggered significantly higher responses in NK cells, with median activations of 80.2% and 67.2% (Mann-Whitney p = .01 and p = .02) respectively. NF1 mutated cases induced the highest NK activation median, 83.6%, among the RAS oncogenes.

Our study demonstrates substantial heterogeneity in sensitivity of AML cells to NK cytotoxicity. The findings reveal that AML patients with BCOR and RAS mutations, are more sensitive to NK killing, especially in CD34+ myeloblasts. The sc-RNAseq data showed that NK activation correlates with enhanced killing of CD34+ and CD14+ blasts. The variability in NK cell cytotoxicity against AML cells within our diverse cohort highlights the possibility of adoptive cell therapies in distinct patient subgroups. Our study improves the understanding of clinical and transcriptomic factors that affect immune clearance of AML cells, enhancing the therapeutic potential of NK cell-based approaches in AML treatment.