High-throughput drug screening identifies SMAC mimetics as enhancers of NK-cell cytotoxicity in chronic myeloid leukemia

Abstract

Natural killer (NK) cells have proven to be safe and effective immunotherapies, associated with favorable treatment responses in chronic myeloid leukemia (CML). Augmenting NK-cell function with oncological drugs could improve NK-cell-based immunotherapies. Here, we used a high-throughput drug screen consisting of >500 small-molecule compounds, to systematically evaluate the effects of oncological drugs on primary NK cells against CML cells. We identified second mitochondrially derived activator of caspases (SMAC) mimetics as potent enhancers of NK-cell cytotoxicity in both cell lines and primary patient samples. In contrast, several drug classes, including glucocorticoids and tyrosine kinase inhibitors such as dasatinib, inhibited NK-cell cytotoxicity. Single-cell RNA sequencing revealed drug-induced transcriptomic changes in both NK and target CML cells. SMAC mimetics upregulated NF-κB target genes in NK cells, potentially contributing to their enhanced cytotoxicity. Inhibitory drugs dexamethasone, dasatinib, and sotrastaurin prevented NK-cell transition to an activated state and suppressed the expression of interferon gamma (IFN-γ) by NK cells, thus preventing IFN-γ-mediated target cell transcriptomic response. In conclusion, we discovered that SMAC mimetics sensitize cancer cells to NK-cell-mediated killing, with potential clinical applications especially in patients with advanced phase CML.

Graphical abstract

Figure 1.

High-throughput drug screen identifies drugs modulating NK-cell cytotoxicity. (A) Schematic of the high-throughput coculture DSRT. (B) K562-luc cell viability with different E:T ratios of NK cells cocultured for 4 hours or 24 hours, in which bars indicate standard deviations and dots indicate the 6 technical replicates for each condition. (C) Overview of the main drug classes included in the compound library. Individual drugs are found in supplemental Table 1. (D) Overview of the landscape of drug responses in the 24-hour NK-cell cytotoxicity screen. A positive dDSS between NK-cell–treated and control drug screens indicates that the drug enhances NK-cell cytotoxicity, whereas a negative score indicates inhibition. The dotted line indicates cutoff values of 7.5 and −7.5. Drugs falling between the dotted lines are considered either only modestly effective or have no effect, whereas drugs falling outside the dotted lines are considered to have an effect.

Figure 2.

Enhancers and inhibitors of NK-cell cytotoxicity in K562 CML cells and primary CML sample. (A) Top 15 drugs most potently enhancing NK-cell cytotoxicity ordered by the dDSS). (B) Top 15 drugs most potently inhibiting NK-cell cytotoxicity, ordered by the dDSS. (C) Top 15 drugs most potently inhibiting NK-cell viability. (D) Dose-response curves showing the effects of the 3 SMAC mimetics included in the drug library on NK-cell viability. (E) Heat map showing the variability of drug responses across 3 healthy NK-cell donors from a validation drug screen. Red boxes represent enhancement of NK-cell cytotoxicity, whereas purple represents inhibition of NK-cell cytotoxicity. (F) Dose-response curves for the top drugs when validated in 3 different NK-cell donors. Orange curve and points indicate drug responses with coculture of NK cells and target cells, and the blue curve and points indicate drug responses with the target cells only. Curves are drawn using the median percent inhibition values across NK-cell donors shown as darker orange dots, with lighter orange dots indicate individual NK-cell donor percent inhibition values. BTK, Bruton tyrosine kinase; CDK, cyclin-dependant kinase; CSC, cancer stem cell; ERK, extracellular signal-regulated kinase; HDAC, histone deacetylase inhibitor; HER, human epidermal growth factor receptor; IMPDH, inosine-5'-monophosphate dehydrogenase; VDAC, voltage dependant anion channel.

Figure 2.

Enhancers and inhibitors of NK-cell cytotoxicity in K562 CML cells and primary CML sample. (A) Top 15 drugs most potently enhancing NK-cell cytotoxicity ordered by the dDSS). (B) Top 15 drugs most potently inhibiting NK-cell cytotoxicity, ordered by the dDSS. (C) Top 15 drugs most potently inhibiting NK-cell viability. (D) Dose-response curves showing the effects of the 3 SMAC mimetics included in the drug library on NK-cell viability. (E) Heat map showing the variability of drug responses across 3 healthy NK-cell donors from a validation drug screen. Red boxes represent enhancement of NK-cell cytotoxicity, whereas purple represents inhibition of NK-cell cytotoxicity. (F) Dose-response curves for the top drugs when validated in 3 different NK-cell donors. Orange curve and points indicate drug responses with coculture of NK cells and target cells, and the blue curve and points indicate drug responses with the target cells only. Curves are drawn using the median percent inhibition values across NK-cell donors shown as darker orange dots, with lighter orange dots indicate individual NK-cell donor percent inhibition values. BTK, Bruton tyrosine kinase; CDK, cyclin-dependant kinase; CSC, cancer stem cell; ERK, extracellular signal-regulated kinase; HDAC, histone deacetylase inhibitor; HER, human epidermal growth factor receptor; IMPDH, inosine-5'-monophosphate dehydrogenase; VDAC, voltage dependant anion channel.

Figure 3.

Effects of birinapant on NK cells in primary samples from patients with CML. (A) Lineplot of NK-cell cytotoxicity at different E:T ratios in 5 CP and 6 BP patient samples. E:T ratio of 0 represents a condition containing no NK cells, only target CD34⁺ CML cells. CD34⁺ counts have been normalized to this no-NK control. (B) Box plot showing drug sensitivity profiling of primary CML CD34⁺ cells’ sensitivity to NK cells at different E:T ratios with and without drug treatment ex vivo. Dots represent normalized CML CD34⁺ cells of all CP and BP patients after 24-hour coculture with NK cells. Median, interquartile ranges, minima, and maxima are shown. Cells are normalized by dividing average of the 3 technical replicates for each condition from each patient by the corresponding average of CML cells–only condition for each patient. Control CD34⁺ patient cells are represented by the dashed line. P values have been calculated with a paired t test. (C) Lineplot of CD34⁺ cell viability of 2 patients at different E:T ratios, 1 (left) being highly sensitive to birinapant treatment alone, and the other (right) being resistant to birinapant treatment alone. 0 represents control, no NK cells. (D) Box plot showing drug sensitivity profiling of 2 healthy donor bone marrow CD34⁺ cells’ sensitivity to NK cells at different E:T ratios with and without drug treatment ex vivo. Dots represent normalized healthy CD34⁺ cells of the 2 donors after 24-hour coculture with NK cells. Median, interquartile ranges, minima, and maxima are shown. Cells are normalized by dividing the average of the 3 technical replicates for each condition from each donor by the corresponding average of CD34⁺ cells–only condition for each patient. Control CD34⁺ cells are represented by the dashed line. (E) Bar plot of normalized total colony counts from a colony forming assay done on 2 healthy samples and 3 CML samples. Total colony counts from each condition have been normalized to the CD34-DMSO condition within each sample. (F) Box plot showing effect of drug treatment on the efficacy of unexpanded NK cells at an E:T ratio of 2:1 from 3 patients with CML, receiving imatinib (right panel), healthy control NK cells from 3 healthy donors (middle panel), and control K562-luc cells exposed only to drug and not NK cells. Median, interquartile ranges, minima, and maxima are shown.

Figure 4.

Phenotypes of NK cells identified by single-cell RNA sequencing with and without drug treatment. (A) Schematic of the single-cell sequencing experiment. (B) Uniform manifold approximation and projection (UMAP) of primary expanded NK cells from all conditions. The genes in boxes indicate marker genes for each cluster. Genes are selected from a list of differentially expressed genes based on significant P value, fold change, and biological relevance. (C) UMAPs of expanded NK cells from selected drug conditions. Dots colored light red (TRUE) represent NK cells from the respective condition, and gray dots (FALSE) represent NK cells from all other conditions. (D) Bar plot showing percentages of NK cells in each cluster by experimental/drug condition.

Figure 5.

Transcriptional changes identified in expanded NK cells caused by drug and target cell coculture. (A) Dot plot of selected genes upregulated and downregulated in NK cells exposed to K562 and drugs. Transcriptional data used for comparison are NK cells from coculture alone (NK-K562-DMSO). Dots encircled by gray indicate false discovery rate (FDR) <0.1. Magnitudes of gene fold changes are depicted by dot color, and the magnitude of FDR is depicted by dot size. FDR was calculated using the Benjamini-Hochberg (BH) test. (B-E) Volcano plots of differentially expressed genes from selected conditions of NK-target cell coculture and drug. Genes with an adjusted P value (BH) < .1 are colored red or blue. (F) Violin plot of IFNG expression in the NK coculture conditions.

Figure 6.

Phenotypes of K562 CML cells identified by single-cell RNA sequencing with and without drug treatment. (A) UMAP of K562 cells from all conditions. The genes in boxes indicate marker genes for each cluster. Genes are selected from a list of differentially expressed genes based on significant P value, fold change, and biological relevance. (B) UMAPs of K562 cells from selected drug conditions. Dots colored red (TRUE) represent K562 cells from the respective condition, and gray dots (FALSE) represent K562 cells from all other conditions. (C) Bar plot showing percentages of K562 cells in each cluster by experimental/drug condition.

Figure 7.

Transcriptional changes identified in K562 target cells caused by drug and NK-cell coculture. (A) Dot plot of selected genes upregulated and downregulated in K562 cells exposed to NK cells and drugs. Transcriptional data used for comparison are K562 cells from coculture alone (NK-K562-DMSO). Dots encircled by gray indicate FDR <0.1. Magnitudes of gene fold changes are depicted by dot color, and the magnitude of FDR is depicted by dot size. FDR was calculated using the BH test. (B-C) Volcano plots of differentially expressed genes from selected conditions of NK-target cell coculture and drug. Genes with an adjusted P value (BH) < .1 are colored red. (D) Top upregulated and downregulated HALLMARK pathways in K562 cells in birinapant-K562-NK condition as compared with K562 cells from K562-NK-DMSO control condition (E-F) Volcano plots of differentially expressed genes from selected conditions of NK-target cell coculture and drug. Genes with an adjusted P value (BH) < .1 are colored red or blue. (G) Top upregulated and downregulated HALLMARK pathways in K562 cells in dasatinib-K562-NK condition (left) and imatinib-K562-NK condition (right).

Figure 7.

Transcriptional changes identified in K562 target cells caused by drug and NK-cell coculture. (A) Dot plot of selected genes upregulated and downregulated in K562 cells exposed to NK cells and drugs. Transcriptional data used for comparison are K562 cells from coculture alone (NK-K562-DMSO). Dots encircled by gray indicate FDR <0.1. Magnitudes of gene fold changes are depicted by dot color, and the magnitude of FDR is depicted by dot size. FDR was calculated using the BH test. (B-C) Volcano plots of differentially expressed genes from selected conditions of NK-target cell coculture and drug. Genes with an adjusted P value (BH) < .1 are colored red. (D) Top upregulated and downregulated HALLMARK pathways in K562 cells in birinapant-K562-NK condition as compared with K562 cells from K562-NK-DMSO control condition (E-F) Volcano plots of differentially expressed genes from selected conditions of NK-target cell coculture and drug. Genes with an adjusted P value (BH) < .1 are colored red or blue. (G) Top upregulated and downregulated HALLMARK pathways in K562 cells in dasatinib-K562-NK condition (left) and imatinib-K562-NK condition (right).