BATF is a major driver of NK cell epigenetic reprogramming and dysfunction in AML

Abstract

Myelodysplastic syndrome and acute myeloid leukemia (AML) belong to a continuous disease spectrum of myeloid malignancies with poor prognosis in the relapsed/refractory setting necessitating novel therapies. Natural killer (NK) cells from patients with myeloid malignancies display global dysfunction with impaired killing capacity, altered metabolism, and an exhausted phenotype at the single-cell transcriptomic and proteomic levels. In this study, we identified that this dysfunction was mediated through a cross-talk between NK cells and myeloid blasts necessitating cell-cell contact. NK cell dysfunction could be prevented by targeting the αvβ-integrin/TGF-β/SMAD pathway but, once established, was persistent because of profound epigenetic reprogramming. We identified BATF as a core transcription factor and the main mediator of this NK cell dysfunction in AML. Mechanistically, we found that BATF was directly regulated and induced by SMAD2/3 and, in turn, bound to key genes related to NK cell exhaustion, such as HAVCR2, LAG3, TIGIT, and CTLA4. BATF deletion enhanced NK cell function against AML in vitro and in vivo. Collectively, our findings reveal a previously unidentified mechanism of NK immune evasion in AML manifested by epigenetic rewiring and inactivation of NK cells by myeloid blasts. This work highlights the importance of using healthy allogeneic NK cells as an adoptive cell therapy to treat patients with myeloid malignancies combined with strategies aimed at preventing the dysfunction by targeting the TGF-β pathway or BATF.

Fig 1.. NK cells derived from patients with MDS and AML display reduced cytokine production, killing capacity and altered metabolism.

(A) Schematic diagram describing the experimental plan of flow cytometry-based sorting of NK cells from healthy donors, MDS and patients with AML followed by functional studies including degranulation and intracellular staining for cytokines and killing assays. (B), Graphs showing percentage of TNF-α production by NK cells from healthy controls (HC, n=6), patients with MDS (low-risk (n=6) or high-risk (n=5) by Revised International Prognostic Scoring System (IPSS-R) and patients with AML (n=6). (C), Graphs showing percentage of IFN-γ production by NK cells from HC (n=6), patients with MDS (low-risk (n=11) or high-risk (n=5) by IPSS-R) and patients with AML (n=5). (D), Graphs showing percentage of CD107a production by NK cells from from healthy HC (n=7), patients with MDS (low-risk (n=11) or high-risk (n=5) by IPSS-R) and patients with AML (n=5). (E-G), Incucyte assays showing the cytotoxicity of NK cells from HC (n=6), MDS (n=4) and AML patients (n=4) against various cell lines, K562 (E), THP-1 (F) and MOLM14 (G) at an effector to target ratio of 1:5. Graphs showing the percentage of tumor cell counts normalized to time 0hrs (H) Representative extracellular acidification rate (ECAR) measurements from a glycostress seahorse assay comparing HC and AML NK cells. (I) Summary bar graph showing the glycolysis and glycolytic capacity measurements comparing HC (n=4) and AML (n=4) NK cells; Each seahorse experiment has been performed using 4 biological donors. (J) Representative oxygen consumption rate (OCR) measurements from a mitostress seahorse assay comparing HC and AML NK cells. (K) Summary bar graph showing the basal and maximal OCR comparing HC (n=4) and AML (n=4) NK cells. Each seahorse experiment has been performed using 4 biological donors. Each symbol denotes an individual donor, data are shown as mean ± SEM. P values were determined by two-tailed paired or unpaired t-tests in panels B-D, or two-way ANOVA test (Dunnet correction to compare the different groups to HC) in panels E-G, or two-tailed paired t-test in panels I, K. * denotes P<0.05, **P<0.01, *** P<0.001.

Fig 2.. NK cells from patients with MDS and AML exhibit altered phenotype and increased exhaustion signature at the single cell transcriptomic and proteomic levels.

(A), SPADE analysis of CyTOF data showing the phenotype of NK cells in healthy controls (HC, n=4), patients with MDS (n=7) and patients with AML (n=8) that were available for the analysis. Samples were pooled and separated into the three categories: HC versus MDS versuss AML. Clustering by SPADE analysis revealed 6 main clusters (Cluster 1–6). Frequencies of each cluster are indicated; size and color of nodes within each cluster represent numbers of clustered cells. (B), Heatmap representing the expression of NK cell markers within the main sub-clusters of Clusters 1–6. Each column represents a major node in the clusters of the spade tree. The major nodes are those that are representative of the majority of cells from all corresponding conditions. The expression for each marker is represented on a color scale ranging from the color blue (low) to the color red (high). (C), Dimension reduction plot using NK cells profiled by scRNA-seq from both HCs (n=2) and patients with AML (n=8). (D), Heatmap showing significantly differentially expressed genes (DEGs) in NK cells between HC and AML NK cells. Each row is a gene andeach column is an NK cell. Biologically relevant DEGs are labeled on the right side of the heatmap. Gradient colors ranging from green to pink indicate gene expression (Z-scored) ranges from low to high. (E), Violin plots showing some DEGs related to NK inhibition/exhaustion in HC and AML NK cells. Significance is indicated as FDRs underneath each plot. (F), Violin plots showing module score for exhaustion markers from HC and AML NK cells in two independent scRNA-seq datasets (Abbas 2021, Van Galen 2019). Significance is calculated using the Wilcoxon test and labeled underneath each plot. (G), Violin plots showing differentially expressed TGF-β pathway related genes in AML NK cells compared to healthy NK cells (Abbas dataset). (H), Bar graph depicting the differentially enriched upregulated and downregulated pathways in AML NK cells compared to HC NK cells (Abbas dataset).

Fig 3.. Evidence of cross-talk between NK cells and myeloid blasts leading to cell-cell contact mediated dysfunction.

(A), Schematic diagram showing NK cells cultured alone or co-cultured with AML blasts for 72hrs then purified and used for cytotoxicity assays. (B), Incucyte cytotoxicity assay showing the mCherry+ MOLM14 cell count over time after culture with healthy NK cells (NK alone) versus healthy NK cells previously co-cultured with AML blasts for 72hs (NK:AML). Data is shown as mCherry+ object count normalized to time 0hr. n=3 biological replicates. (C), Schematic diagram showing how NK cells are co-cultured with AML blasts either in direct contact (magenta color) or indirect contact (separated by a transwell, blue color) for 72hrs before being purified and used for functional assays. (D), Graphs showing the percent secretion of TNF-α, IFN-γ and CD107a in response to K562 targets by NK cells previously cultured in direct or indirect (transwell) contact with AML blasts for 72hrs compared to NK cells cultured alone. n=6 biological replicates. (E), Incucyte cytotoxicity assay showing the mCherry+ MOLM14 object counts over time normalized to time 0hr (left panel) following multiple tumor rechallenges (black arrow indicate timing of tumor rechallenge). Bar graph showing the normalized mCherry+ MOLM14 object counts at day 10 (240hrs) normalized to time 0hr (right panel). n=3 biological donors. (F), Incucyte cytotoxicity assay showing the normalized mCherry+ THP-1 object counts over time normalized to time 0hr (left panel) following multiple tumor rechallenges (black arrow indicate timing of tumor rechallenge). Bar graph showing the normalized mCherry+ THP-1 object counts at day 10 (240hrs) normalized to time 0hr (right panel). n=3 biological donors. (G), Bar graphs showing the percent expression of exhaustion markers (TIM3, LAG3 and TIGIT) on the surface NK cells cultured alone (n=10 donors), or in direct (n=10 donors) or indirect (transwell, n=7 donors) contact with AML for 72hrs, as measured by flow cytometry. (H), Representative histograms showing the expression of LAG3, TIM3 and TIGIT on the surface NK cells cultured alone, or in direct or indirect (transwell) contact with AML. Each symbol represents an individual data point from a biological replicate. Data are shown as mean ± SEM. P values were determined by two-tailed paired t-test in panels D,G and E,F (right panels) or two-way ANOVA test in panels B, E (left panel), F (left panel). * Denotes P<0.05, **P<0.01 and *** P<0.001.

Fig 4.. Activated TGF-β/SMAD pathway in MDS/AML NK cells alters their function and targeting the TGF-β pathway prevents this dysfunction.

(A), Bar graph showing mean fluorescent intensity (MFI) of phosphorylated SMAD2/3 in NK cells from patients with MDS (n=6) and AML (n=15) compared to healthy controls (n=10). NK cells treated with exogenous TGF-β1 (10ng/ml) were used as positive control (n=6). (B), Representative histograms showing TGF-β-LAP expression on CD34+ cells from patients with AML, low IPSS-R and high IPSS-R MDS compared to healthy control CD34+cells. Fluorescence minus one (FMO) was used as negative control. (C), Graph showing the percentage of TGF-β-LAP expression on CD34+ cells from patients with AML (n=15), low IPSS-R and high IPSS-R MDS (n=5 and 4 biological donors respectively) compared to healthy control CD34+ cells (n=7). (D), Bar graph showing the concentration of TGF-β1 (pg/ml/million cells) in the supernatant of NK cells and AML blasts either cultured alone or co-cultured together in direct contact or indirectly through a transwell. n=5 biological replicates (E), Schematic diagram depicting the experimental plan for panel F where HC NK cells were cultured alone or with AML cells (NK:AML) for 72hrs in the presence or absence of TGF-β inhibitors and then purified and used in Incucyte cytotoxicity assays. (F), Graph of Incucyte assay showing the percentage of cytotoxicity (mCherry+NIR/mCherry) of the various groups of NK cells against MOLM14 cells normalized to time 0hr. Prior to initiating the Incucyte assay, NK cells were either cultured alone or co-cultured with AML (1:2 ratio) cells for 72hrs with or without the presence of TGF-β inhibitors (Galunisertib, luspatercept and Cilengitide), n=4 biological replicates. (G), Graph of Incucyte rechallenge assay showing the mCherry+ THP-1 object counts over time normalized to time 0hr, comparing the anti-tumor activity of Cas9 control NK cells versus TGFBR2 KO NK cells. Prior to initiating the Incucyte assay, Cas9 control NK cells or TGFBR2 KO NK cells were either cultured alone or co-cultured with AML cells (1:2 ratio) for 72hrs. (H), Schematic diagram depicting the experimental plan of the AML mouse model testing the in vivo efficacy of adoptively transferred NK cells (Cas9 control or TGFBR2 KO) or NK cells in combination with pharmacologic inhibitors of TGF-β (Galunisertib, luspatercept and Cilengitide). NSG Mice were injected with 1×10⁵ THP-1 tumor cells intravenously and either left untreated or treated with TGF-β inhibitors alone, or with NK cells (CTRL or TGFBR2 KO) or with NK cells in combination with TGF-β inhibitors. (I), Bioluminescent imaging (BLI) images of mice in the various groups described in Panel H. (J), Bar graph showing the BLI total flux [p/s] at day 30 comparing tumor burden in the various groups of mice shown in panel I. (K), Bar graphs showing the percentage of AML chimerism (GFP+/humanCD45+ THP-1 cells out of total live cells) in the bone marrow (BM), spleen and peripheral blood (PB) in the various groups of mice shown in Panel I. Each symbol represents an individual data point from a biological replicate. Data are shown as mean ± SEM. P values were determined by unpaired t-test in panels A, C, K, or paired t-test for panel D, or two-way ANOVA test in panels F, G, J. * Denotes P<0.05, **P<0.01 and *** P<0.001, ns not significant.

Fig. 5.. Direct co-culture with AML induces epigenetic regulation of important NK dysfunction cellular programs.

(A), Graphs of Incucyte live cell imaging assay showing the percentage of mCherry+ MOLM14 cells normalized to time 0 hr as a surrogate for NK cell killing. Prior to the Incucyte assay, NK cells were either cultured alone (control NK cells, blue lines) or co-cultured with AML for 72hrs (AML:NK, red lines). The left panel shows the results of the cytotoxicity assay at baseline (pre-rescue) and the right panel shows the results of the cytotoxicity assay following the 1 week-long activation and expansion (post-rescue). (B). Graphs of Incucyte live cell imaging assay showing the percentage of mCherry+ MOLM14 cells normalized to time 0 hr, as a surrogate for NK cell killing. NK cells were either derived from healthy controls (HC NK cells) or from paired AML patient samples at diagnosis during active disease and following complete remission (CR). (C), An UpSet plot highlighting the number of enriched accessible chromatin regions which are either unique or shared among the three experimental conditions, 1) NK cells cultured alone (Control) 2) NK cells co-cultured with AML indirectly through a Transwell (Transwell) or 3) NK cells co-culured with AML in direct contact (Direct). (D), Heatmap showing differential enrichment of accessible chromatin regions in all three conditions (Control n=2, Tranwell n=3, Direct n=3 biological replicates). (E), Volcano plot showing the differential expression of genes associated with/near the open chromatin regions in direct versus transwell cultures. (F), Integrative Genomics Viewer (IGV) plots of accessible chromatin regions from all three conditions showing gain of exclusive/differential enhancers in Direct co-culture of NK and AML cells in proximity of known inhibitory receptor genes HAVCR2 and TGFBR2. (G), Integration of differential accessible regions which are computationally linked to nearest genes from ATAC-seq and differential expression genes (DEGs) from RNA-seq in Transwell and Direct co-cultures. (H), Motif enrichment analyses highlighting presence of distinct sets of Transcription factors (TFs) at gained accessible chromatin regions in Transwell/control and Direct conditions. (I) Core TF circuitry derived by integration of ATAC-seq and RNA-seq informs distinct self-regulatory expressed set of TFs in innermost red circle driving major phenotypes in NK cells Direct and Transwell co-culture condition. (J), IGV plots showing gain of enhancers around BATF and BATF3 genes in Direct co-culture conditions in comparison to Transwell/Control. Data are shown as mean ± SEM. P values were determined one way ANOVA test in panels A and B. *** Denotes P<0.001.

Fig 6.. BATF mediated gene regulatory program is a main driver of myeloid blast induced NK epigenetic dysfunction.

(A), Venn diagram showing gained accessible chromatin peaks with BATF motifs annotated to their nearest gene and overlapped with differentially upregulated genes (B), Pathway enrichment analysis showing enriched pathways for epigenetically accessible genes that are overexpressed at the RNA-seq level (termed as epigenetically upregulated). (C), Western blots showing efficient overexpression (OE) of BATF in healthy peripheral blood NK cells from 2 representative donors; empty vector (EV) was used as negative control. (D), Western blots showing efficient knockout of BATF in healthy peripheral blood NK cells from 2 representative donors; Cas9 electroporated NK cells were used as negative control. (E), Graph of Incucyte assay showing the GFP+ MOLM14 object count over time normalized to time 0hrs, comparing the anti-tumor activity of empty vector (EV) control NK cells versus BATF OE NK cells following multiple tumor rechallenges. (F), Bar graph showing the percent expression of exhaustion markers in BATF OE NK cells compared to EV control NK cells. (G), Graph of Incucyte assay showing the GFP+ MOLM14 object count over time normalized to time 0hrs, comparing the anti-tumor activity of Cas9 control NK cells vs. BATF KO NK cells following multiple tumor rechallenges. (H), Bar graph showing the percent expression of exhaustion markers in BATF KO NK cells compared to Cas9 control NK cells. (I), BLI images of mice injected with THP-1 and either left untreated or treated with various NK cell preparations, EV control, BATF OE, Cas9 control or BATF KO. (J), Bar graph showing the average radiance (p/s/cm²/sr) at day 27 in the various mice groups shown in panel H. (K), Bar graphs showing the percentage of AML chimerism (GFP+/humanCD45+ THP-1 cells out of total live cells) in the bone marrow (BM), spleen and peripheral blood (PB) in the various groups of mice shown in Panel H. (L), IGV plots from CUT&RUN analysis showing that following treatment with TGF-β (10ng/ml) (used to mimic NK:AML direct co-culture), BATF binds to key genes related to NK cell exhaustion and dysfunction and binds to its own gene showing self-transcriptional regulation. IgG was used as negative control. Highlighted grey regions represent differentially open chromatin regions identified by ATAC-seq in the NK:AML direct co-culture condition, proving that BATF is binding to these regions. Each symbol represents an individual data point from a biological replicate. Data are shown as mean ± SEM. P values were determined by two-way ANOVA in panels E,G or paired t-test in panels F, H or unpaired t-tests in panels J,K. * Denotes P<0.05, **P<0.01 and *** P<0.001.

Fig 7.. TGFBR2 KO NK cells maintain potency in direct co-culture with AML through enhancer reprogramming.

(A), Western blot showing BATF protein expression in healthy NK cells without any treatment or following treatment with recombinant TGF-β1 treatment (10ng/ml) alone or in addition to Luspatercept (SMAD2/3 inhibitor, 2μg/ml). β-actin was used as a loading control. Left panel shows representative images from 2 donors, right panel shows a summary quantification graph of BATF expression in treated NK cells (TGF-β1 alone or TGF-β1+luspatercept) normalized to mean of untreated control NK cells (n=4 donors). (B), Bar graph from ChIP-qPCR experiment showing the percent input of pSMAD2 enrichment normalized to IgG at enhancer and intronic region of BATF. NC=negative control. (C), Venn diagram showing shared and exclusive accessible chromatin regions between NK ^WT and NK^TGBR2KO cells co-cultured with AML cells in Direct contact. (D), Heatmaps for open and closed chromatin regions in NK^WT and NK^TGBR2KO cells co-cultured with AML cells in Direct contact. The average intensity plots (right side) show the differences in accessibility in open or and closed in NK^TGBR2KO cells to NK ^WT. (E), PCA plot of gained accessible peaks in NK^TGBR2KO and NK ^WT cells cultured alone or in Direct or Transwell co-culture with AML cells. (F), Motif analyses for gained and lost accessible peaks in NK^TGBR2KO Cells in Direct co-culture indicate distinct TF regulations. (G), Pathway analyses reflecting enriched pathways in gained and lost accessible peaks in NK^TGBR2KO Cells in Direct co-culture compared to NK^WT NK cells in Direct co-culture. Each symbol represents an individual data point from a biological replicate. Data are shown as mean ± SEM. P values were determined by paired t-test in panels A,B and multiple unpaired t-tests in panel C. * Denotes P<0.05, **P<0.01, ns = non significant.