Augmenting NK cell-based immunotherapy by targeting mitochondrial apoptosis

Abstract

Interest in harnessing natural killer (NK) cells for cancer immunotherapy is rapidly growing. However, efficacy of NK cell-based immunotherapy remains limited in most trials. Strategies to augment the killing efficacy of NK cells are thus much needed. In the current study, we found that mitochondrial apoptosis (mtApoptosis) pathway is essential for efficient NK killing, especially at physiologically relevant effector-to-target ratios. Furthermore, NK cells can prime cancer cells for mtApoptosis and mitochondrial priming status affects cancer-cell susceptibility to NK-mediated killing. Interestingly, pre-activating NK cells confers on them resistance to BH3 mimetics. Combining BH3 mimetics with NK cells synergistically kills cancer cells in vitro and suppresses tumor growth in vivo. The ideal BH3 mimetic to use in such an approach can be predicted by BH3 profiling. We herein report a rational and precision strategy to augment NK-based immunotherapy, which may be adaptable to T cell-based immunotherapies as well.

Keywords: BCL-2; BH3 mimetics; BH3 profiling; MCL-1; T cells; immunotherapy; mitochondrial apoptosis; natural killer; synergy; venetoclax.

Figure 1.. Pre-activated NK cells induce mtApoptosis, but not necroptosis, ferroptosis, or pyroptosis of cancer cells.

Two representative cell lines OCI-AML3 (left panels) and HeLa (right panels) were tested in parallel. After CFSE labeling, they were co-cultured with pre-activated human NK cells at indicated E:T ratios and time points (see Methods for details). (A, M) Loss of mitochondrial membrane potential in cancer cells after NK treatment, as measured by CMXRos staining followed by flow cytometry analysis (See also Figure S1C for quantitative data at multiple time points and different E:T ratios). (B, N) Release of cytochrome c from cancer cell mitochondria after NK treatment, measured by BH3 profiling procedure but using DMSO instead of BH3 peptides (see Methods). See also Figure S1D. (C, O) NK cells induced activation of caspase-3 in target cells, as determined using NucView® 405 Blue Caspase-3 Dye. See also Figure S1E. (D, P) Exposure of phosphatidylserine after NK treatment at various E:T ratios, measured by annexin V staining. See also Figure S1F. (E, Q) Immunoblots showing cleavage of BID and caspases in target cells 2 h after NK treatment at E:T of 1:2. Immunoblotting was performed using cell lysis from FACS sorted CFSE⁺CD56^- cancer cells. β-Actin and GAPDH served as loading control. (F, R, and S) NK induced time-dependent apoptosis in OCI-AML3 cells and HeLa cells measured by annexin V staining via flow cytometry. (G and T) NK treatment induced membrane blebbing and formation of apoptotic bodies in cancer cells. E:T = 1:2, 6 h. (H and U) Effects of zDEVD-fmk (30 μM), zVAD-fmk (30 μM), necrostatin-1 (Nec-1, 20 μM), ferrostatin-1 (Fer-1, 2 μM) and α-tocopherol (Vit. E, 100 μM), and dimethyl fumarate (DMF, 25 μM) on NK-induced cell death. Data was normalized to control (without NK treatment). Statistical significance was calculated versus DMSO group. (I and V) Effects of EGTA (2 mM) on NK induced cell death. White bars: control, no NK cells added. (J and W) Effects of NK conditioned medium (CM, from three NK samples) on treated cancer cells. See methods for generation of conditioned medium. (K and X) Transwell assays. Gray bars: NK cells were added in direct contact with cancer cells growing on bottom of the wells as described in Methods. White bars: no NK cells were added. Cyan bars: NK cells were added into transwell inserts (0.4 μM pores). After 8 h, cell death of cancer cells was measured by flow cytometry (E:T = 1:2). (L and Y) Left panel: NK cells were incubated with non-targeting control or GZMB-targeting siRNA for 72 h before they were used for cytotoxicity assays. OCI-AML3 and HeLa cancer cells were then treated with the two NK samples for 5 h at E:T = 1:2, followed by flow cytometry analysis. Right panel: OCI-AML3 and HeLa cells were incubated with non-targeting control or BID-targeting siRNA for 72 h. Then the control and BID knockdown cells were treated with two primary NK samples for 5 h at E:T = 1:2, followed by flow cytometry analysis. Cell death % by flow cytometry was calculated as 100% − %AnxV⁻DAPI⁻ (H-L, and U-Y). Data were presented as mean ± SEM of triplicate experiments (F, H-L, and U-Y). Statistical significance was assessed by student’s t test. *p < 0.05, **p < 0.01, ***p < 0.001, ns, nonsignificant. See also Figure S1.

Figure 2.. Mitochondrial apoptosis machinery is essential for efficient NK killing, especially at E:T ratios ≤ 1.

(A) Representative flow plots of HeLa control (Ctrl) and BAX^−/−BAK^−/− (DKO) cells treated with primary human NK cells (E:T = 1:2, 8 h). Dead cells are in the P7 gate and live cells are in the P8 gate. (B and C) Percentage of apoptotic cells (B) and normalized live cell number (C) of HeLa control and DKO cells after NK treatment (E:T = 1:2). Cell number was enumerated by flow cytometry using absolute counting beads as described in Methods. (D and E) Apoptotic percentage (D) and morphology (E) of HeLa control and DKO cells after treatment with NK cells at different E:T ratios for 16 h. (F) Crystal violet staining of attached HeLa control and DKO cells after co-culture with two different NK samples for 16 h. (G) Normalized cell number and apoptotic percentage of HeLa control and DKO cells after treatment with human NK cell line KHYG-1 for 16 h, determined by flow cytometry as in A-C. (H) Induction of cell death of HeLa cells by NK samples with different capacity for killing (E:T = 1:2). These four samples were selected for killing assays in Figure 2I. (I) Percentage of dead SU-DHL6 (BAX^WTBAK^WT) and BBDL (BAX/BAK deficient) cells after treatment with different NK samples (6 h). Data in bar and line graphs were presented as mean ± SEM of triplicate experiments. Statistical significance by student’s t test. *p < 0.05, **p < 0.01, ***p < 0.001, ns, nonsignificant. See also Figure S2.

Figure 3.. NK cells effectively prime cancer cells for mtApoptosis

(A) Selectivity, specificity, and color code of synthetic BH3 peptides and ABT-199 used in BH3 profiling assays in this study. (B) Schematic showing cytochrome c release from a mitochondrion induced by BH3 peptides during BH3 profiling. (C) Workflow of BH3 profiling of cancer cells with or without NK treatment. See Methods for detailed BH3 profiling procedures. (D and E) Mitochondrial priming status of OCI-AML3 cells (D) and HeLa cells (E) before and after treatment with NK cells (2 h, E:T = 1:2 ). Values following peptide names are concentrations (µM). Mitochondrial priming is defined as a range from 0 – 100% cytochrome c release induced by BH3 peptides (see Methods for details). (F and G) NK treatment (2 h, E:T = 1:2) increased mitochondrial priming of 3/3 blood and 3/3 solid cancer cell lines, as determined using BIM peptide. (H) Representative histograms showing mitochondrial priming status of different cancer cells (determined using 2 nM BIM peptide) after NK treatment (E:T = 1:2, 2 h). BH3 profiling performed as in 3C. (I) Heatmap of mitochondrial priming of HeLa cells with or without NK treatment at multiple time points (E:T = 1:2). Each tile represents one of three experimental replicates of a certain treatment. Data in bar graphs and line graphs were presented as mean ± SEM of triplicate experiments. See also Figure S3.

Figure 4.. Reduced mitochondrial priming makes cancer cells less susceptible to NK-mediated killing

(A-C) Mitochondrial priming status of the paired control and BCL-2, BCL-XL or MCL-1 overexpressing cells as determined by BH3 profiling. Delta priming is the arithmetic difference between the control and overexpressing cells. Negative values measured by the BIM peptide indicate reduced overall priming in overexpressing cells (black bars in bottom panels). Statistics were calculated versus DMSO control. ## denotes shared control between panels A and B except for the BAD peptide datapoints because BAD at 1 μM or 0.1μM are most sensitive in determining delta priming for BCL-2 OE or BCL-XL OE cells respectfully. (D and E) Immunoblots of BCL-2, BCL-XL, and MCL-1 in paired control and overexpressing cancer cells. (F-H) Sensitivity of control and overexpressing cells to NK-mediated killing. Controls and overexpressors were treated with two NK samples at indicated E:T ratios and time points. Each dot is one of three technical replicates of a particular treatment. (I-K) Overexpression of BCL-2, BCL-XL, and MCL-1 altered the sensitivity of cancer cells to BH3 mimetics and to genotoxic cytarabine and etoposide (48 h treatment). Data were presented as mean ± SEM of triplicate experiments (A-C and I-K). Statistical significance by student’s t test. *p < 0.05, **p < 0.01, ***p < 0.001, ns, nonsignificant.

Figure 5.. Pre-activated NK cells are less primed for mtApoptosis and acquire resistance to BCL-2, MCL-1, and BCL-XL inhibitors

(A) Sensitivity of CD3⁻CD56⁺ NK cell subpopulation of PBMCs to BCL-2i ABT-199 (24 h treatment; PBMC, peripheral blood mononuclear cells), as measured by annexin V staining followed by flow cytometry analysis. (B) Immunoblots showing altered protein expression of NK cells after pre-activation for 3 days (4 biological replicates, see Methods for pre-activation method). (C) Reduced mitochondrial priming of NK cells after pre-activation as determined by BH3 profiling using the BIM peptide at 0.01 μM. Delta priming is the difference between pre-activated and resting cells, with a negative value indicating a decrease in priming. (D) Sensitivity of freshly isolated resting NK and pre-activated NK cells to BCL-2i (24 h treatment). (E) Normalized live cell number of pre-activated NK cells (measured by flow cytometry with counting beads) after treatment with BCL-2i for 24, 48, or 96 h. Numbers were normalized to DMSO control at each time point. (F and G) BCL-2i treatment (1 µM, 24 h) induced exposure of phosphotidylserine (as measured by annexin V staining), loss of mitochondrial membrane potential (MMP, as measured by CMXRos staining), and cytochrome c release in resting NK cells but not in pre-activated NK cells. Panel F are representative plots and panel G shows quantitative data. Note that matched pre-activated NK cells (day 5) and freshly-isolated resting NK cells (day 0) were used. Resting NK cells (day 5) were unsuitable for cell death studies because viability of resting NK cells quickly decreases in the absence of IL-2 pre-activation. (H-J) Sensitivity of resting and pre-activated NK cells to BCL-XL, MCL-1, and XIAP inhibitors (24 h), as measured by annexin V staining. For all bar and line graphs, data were presented as mean ± SEM of triplicate experiments. Statistical significance was assessed by student’s t test between matched NK samples. *p < 0.05, **p < 0.01, ***p < 0.001, ns, nonsignificant. See also Figure S4.

Figure 6.. Selective BH3 mimetics synergize with NK cells in killing cancer cells, which can be predicted by BH3 profiling.

(A and D) Basal level of mitochondrial priming of HeLa (A) and HL60-BCL-2 (D) cells as determined by BH3 profiling with the indicated peptides or BH3 mimetic (BCL-2i = ABT-199) at the indicated concentrations in μM. (B and E) Delta priming of HeLa (B) and HL60-BCL-2 (E) cells after 2 h treatment with 256 nM BCL-2i, BCL-XLi, and MCL-1i in the presence or absence of NK cells (E:T = 1:2). Delta priming = priming_treated – priming_untreated, as determined using BIM peptide. Each tile represents one data point (n = 3). (C) NK cells synergized with MCL-1i, but not with BCL-2i or BCL-XLi, in killing HeLa cells. HeLa cells were treated with escalating doses of different BH3 mimetics, in the presence or absence of pre-activated NK cells (E:T = 1:2, 6 h). Specific apoptosis in C, F, G and H was determined by annexin V staining. (F) NK cells synergized with BCL-2i, but not with BCL-XLi or MCL-1i, in killing HL60-BCL2 cells. HL60-BCL2 cells were treated with escalating doses of different BH3 mimetics, in the presence or absence of two different samples of pre-activated human NK cells (E:T = 1:2, 6 h). (G) Combined effects of NK cells and BCL-XLi on HL60-BCLXL cells (E:T = 1:2, 6 h). (H) Combined effects of NK cells and MCL-1i on OCI3-MCL1 cells (E:T = 1:2, 6 h). For all bar graphs, data were presented as mean ± SEM of triplicate experiments. Dashed lines in bar graphs denote apoptotic percentage induced by NK cells alone. See also Figures S5, S6, and S7.

Figure 7.. BH3 mimetics synergize with pre-activated NK cells in reducing tumor burden and prolonging mouse survival.

(A) Schematic outline of the HL60-BCL2 mouse model (6 groups, n = 7 per group). NK cells used for in vivo studies were generated and pre-activated following the same procedure as for in vitro studies. (B) ABT-199, but not S63845, synergized with NK cells in priming engrafted HL60-BCL2 cells treated as xenografts in vivo and then removed and subjected to BH3 profiling. Delta priming was determined using BH3 profiling with BIM peptide on dissociated cells from engrafted tumors, comparing tumors obtained from treated mice with those from untreated mice. (C) Mean tumor volume of NSG mice subcutaneously implanted with HL60-BCL2 cells. Statistical tests were performed as NK/ABT-199 group versus NK group, the second most efficacious treatment arm. (D) Kaplan-Meier survival curves of mice engrafted with HL60-BCL2 cells. Statistical significance was calculated using the log rank test. (E) Schematic outline of the HeLa-NSG mouse model. NK cells used for in vivo studies were generated and pre-activated following the same procedure as for in vitro studies as described in Methods. (F) S63845, but not ABT-199, synergized with NK cells in priming engrafted HeLa cells in vivo. Delta priming was determined as in 7B. (G) Photograph of representative HeLa tumors in different groups (day 49). (H) Mean tumor volume of NSG mice implanted with HeLa cells. Statistical tests were performed as NK/S63845 group vs. S63845 group, the second most efficacious treatment arm. Data were presented as mean ± SEM (B, C, F, and H). Statistical significance by student’s t test. *p < 0.05, **p < 0.01, ***p < 0.001, ns, nonsignificant.