Systematic Dissection of the Metabolic-Apoptotic Interface in AML Reveals Heme Biosynthesis to Be a Regulator of Drug Sensitivity

Abstract

Crosstalk between metabolic and survival pathways is critical for cellular homeostasis, but the connectivity between these processes remains poorly defined. We used loss-of-function CRISPR/Cas9 knockout screening to identify metabolic genes capable of influencing cellular commitment to apoptosis, using sensitization to the BCL-2 inhibitor ABT-199 in BCL-2-dependent acute myeloid leukemia (AML) cell lines as a proxy for apoptotic disposition. This analysis revealed metabolic pathways that specifically cooperate with BCL-2 to sustain survival. In particular, our analysis singled out heme biosynthesis as an unappreciated apoptosis-modifying pathway. Although heme is broadly incorporated into the proteome, reduction of heme biosynthesis potentiates apoptosis through the loss of ETC activity, resulting in baseline depolarization of the mitochondrial membrane and an increased propensity to undergo apoptosis. Collectively, our findings chart the first apoptotic map of metabolism, motivating the design of metabolically engaged combination chemotherapies and nominating heme biosynthesis as an apoptotic modulator in AML.

Trial registration: ClinicalTrials.gov NCT01371981.

Keywords: CRISPR; acute myeloid leukemia; apoptosis; cancer metabolism; genetic screens; heme biosynthesis; mitochondria.

Conflict of interest statement

Declaration of interests:

The remaining authors have no competing interests to declare.

Copyright © 2019 Elsevier Inc. All rights reserved.

Figures

Figure 1:. Validation of CRISPR/Cas9-based screens for the identification of metabolic sensitizers to ABT-199-induced apoptosis.

(A) CRISPR/Cas9 loss-of-function screening strategy. (B) Replicate-to-replicate comparison of gene-level essentiality phenotypes in MOLM-13. Essential controls are shown in red, non-essential controls in yellow, and non-targeting controls in blue. (C) Gene-level representation of ABT-199 sensitization phenotypes from both cell lines, ranked by their log2-transformed three score (TS) (ABT-199-treated / DMSO-treated). Apoptotically reactive genes are denoted by the red box. (D) Top-scoring genes (red), ranked by their log2(TS) values, displayed against each gene’s corresponding essentiality score (yellow). (E) Results of pre-ranked GSEA performed with gene ontology gene sets, ordered by the negative log2 of gene set q-values.

Figure 2:. Metabolic-apoptotic interaction map

Scoring library genes are highlighted in red; non-scoring library genes in black; pathway-relevant non-library genes in gray. Pathways with > 80% scoring genes are shaded dark red; pathways with 20 – 80% scoring genes are light red; pathways with

Figure 3:. Interaction map predicts apoptotic sensitization through targeting of nucleotide synthesis, TCA-associated nodes, and redox regulation.

Pathway genes that scored as sensitizers are highlighted in red. (A, E, I) Gene-level representation of ABT-199-sensitization, ranked by log2(TS) values. Highlighted genes correspond to: (A) purine and pyrimidine biosynthesis, (E) TCA cycle, and (I) redox regulation. (B, F, J) Annotated pathway schematics for: (B) purine/pyrimidine biosynthesis, (F) TCA cycle, and (J) redox regulation. Pharmacologic interrogations are shown in blue. (C, G, K) Validation of pharmacologic sensitization of: (C) purine/pyrimidine biosynthesis, (G) TCA nodes, and (K) redox regulation to ABT-199-mediated cell death using 72hr growth inhibition (GI50) assays. Shown are relative 72hr GI50 values derived from ABT-199 dose-response curves in the presence and absence of background metabolic drugs, used at sublethal doses based on single-drug GI50 curves (Figure S2A). (D, H, L) Fold change in 72hr GI50 values derived from ABT-199 dose-response curves for OCI-AML2 cells transduced with 2 sgRNAs targeting (D) TYMS or PPAT; (H) SLC25A1, GLS, OGDH, DLST, or DLD; and (L) GPX4, GCLC, HMGCR, or SEPHS1, relative to a non-targeting sgRNA. *** p ≤ 0.001, **** p ≤ 0.0001 by student’s t-test; n=3. Data are mean ± SD.

Figure 4:. Heme biosynthesis is downregulated in clinical AML.

(A) Heatmap depicting expression of metabolic pathway genes in samples from healthy bone marrow (HBM) n=74, myelodysplastic syndrome (MDS) n=206, or acute myeloid leukemia (AML) n=542. Inset: heme biosynthetic genes highlighted in red. (B) Expression pattern of heme biosynthetic pathway genes in samples from HBM (n=74), MDS (n=206), or AML (n=542). **** p ≤ 0.0001 by Welch’s t-test. (C) Comparison of 5-aminolevulinic acid levels from pediatric AML patients taken at diagnosis (n = 20) vs. upon remission (n = 7). ** p ≤ 0.01 by Welch’s t-test.

Figure 5:. Modulation of intracellular heme potentiates ABT-199-induced apoptosis.

(A) Gene-level representation of ABT-199-sensitization, ranked by log2(TS) values. Heme biosynthesis genes and iron transporters are highlighted. (B) Annotated pathway schematic of heme biosynthesis. Pharmacologic interrogations are shown in blue. (C) Fold change in 72hr GI50 values, derived from ABT-199 dose-response curves for OCI-AML2 cells transduced with 2 sgRNAs targeting ALAS1, ALAD, or FECH, relative to a non-targeting sgRNA. (D) Relative heme content after 48hr treatment with SA in six AML cell lines. (E) Fold change in 72hr GI50 values derived from ABT-199 dose-response curves treated in the presence and absence of SA. Absolute GI50 values normalized to viability in SA alone. (F) ABT-199 dose-response curves in OCI-AML2 cells pre-treated with background doses of 250µM SA or 250µM SA + 5µM PPIX for 48hr, then treated with an ABT-199 dilution series for 72h. Dose response curves normalized to the viability of SA alone or SA and PPIX alone. (G) Immunoblots of cleaved PARP, cleaved caspase 3, and β-actin in cells treated for 48hr with SA (OCI-AML2: 100µM, MOLM-13: 750µM, THP-1: 2mM, MV4–11: 1mM, KG-1α: 1.5mM, HL-60: 2mM), 8hr treatment with ABT-199 (OCI-AML2: 50nM, MOLM-13: 30nM, THP-1: 200nM, MV4–11: 100nM, KG-1α: 25nM, HL-60: 50nM), or the combination. (H) Percentage annexin V+ in cells treated with ABT-199 for 24hr (OCI-AML2: 75nM, MOLM-13: 25nM), SA for 72hr (OCI-AML2: 250µM, MOLM-13: 2mM), or the combination. Percentages of annexin V+ drug-treated cells normalized to signal from vehicle-treated cells. (I) Immunoblots of Bax, Bak, and VDAC1 in OCI-AML2 cells treated with SA 100µM for 48hr, ABT-199 50nM for 8h, or the combination. Shown are mitochondrial fractions only; VDAC1 is shown as a mitochondrial loading control. (J) ABT-199 dose-response curves in OCI-AML2 and MOLM-13 cells expressing shGFP or one of two shRNAs targeting Bax, treated in the presence and absence of SA (OCI-AML2: 300µM, MOLM-13: 600µM). Immunoblots of Bax. (K) Quantification of OCI-AML2 cell populations cultured continuously in vehicle, SA 100µM, ABT-199 1.0µM, PPIX 1.0µM, or the stated combinations. Total cell number is an extrapolation from weekly growth rates. (L) OCI-AML2 cells cultured continuously in vehicle, SA 100µM, cytarabine 100nM, PPIX 1.0µM, or the stated combinations. (M) Percentage annexin V+ in primary AML patient cells treated with ABT-199 and/or SA for 24hr. n=3 for all samples except DP0285 (n=1). ** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001 by student’s t-tests; n=3. Data are mean ± SD.

Figure 6:. Targeting heme biosynthesis primes the cell for apoptosis by disrupting electron transport chain integrity.

(A) 83 library genes that associate with heme, ranked by log2(TS) score. (B) Schematic showing incorporation of heme species into electron transport chain complexes. Reactive genes are highlighted in red. (C) Fold change in 72hr GI50 values derived from ABT-199 dose-response curves for OCI-AML2 cells transduced with 2 sgRNAs targeting COX10 or COX15, relative to a non-targeting sgRNA. (D) Immunoblot of subunits of ETC complexes I – V in cells treated with SA or SA + PPIX 2.5µM for 48hr (OCI-AML2: 250µM, MOLM-13: 750µM). (E) Basal OCR as measured by Seahorse XF Analyzer’s Mito Stress Test. Cells were treated with SA (OCI-AML2: 100µM, MOLM-13: 750µM) or SA and PPIX (OCI-AML2: 2.5µM, MOLM-13: 1.0µM). (F) Enzymatic activity of complex IV in cells treated with SA (OCI-AML2: 250µM, MOLM-13: 750µM) or complex IV inhibitor sodium azide (OCI-AML2: 1mM, MOLM-13: 2mM) for 48hr. Enzymatic activity normalized to vehicle treatment. (G) TMPD-driven oxygen consumption rate as measured by Seahorse XF Analyzer. Cells were treated for 48hr with SA (OCI-AML2: 250µM, MOLM-13: 750µM), resuspended in minimal media, and introduced to TMPD/ascorbate. Shown is the maximal OCR. (H) Immunoblots of cleaved PARP, cleaved caspase 3, and β-actin in cells treated for 48h with SA (OCI-AML2: 100µM, MOLM-13: 750µM), 48h with sodium azide (OCI-AML2: 1mM, MOLM-13 2mM), 8h with ABT-199 (OCI-AML2: 50nM, MOLM-13: 30nM), or the stated combinations. (I) JC-1 signal reported as ratios of red/green fluorescence normalized to vehicle, following 48hr treatment of cells with H2O (vehicle), SA (OCI-AML2: 250µM, MOLM-13: 750µM), or sodium azide (OCI-AML2: 500µM, MOLM-13: 2mM). (J) JC-1 signal following 24hr treatment of primary AML patient cells with SA. All data reported are n =1. (K) Immunoblots of cytochrome c, AIF, Smac, HtrA2, XIAP, cleaved PARP, cleaved caspase 3, and β-actin. OCI-AML2 cells were treated with DMSO, 100µM SA, 20nM ABT-199, or 100µM SA + 20nM ABT-199 for 24hr, collected, and fractionated. Cytoplasmic fractions are shown for drug-treatments. Cytoplasmic versus membrane/organellular distribution in untreated cells shown for reference. (L) Model of heme-dependent apoptotic sensitization in AML cells. Heme depletion (right) drives baseline MMP depolarization through destabilization of heme-containing ETC complexes, sensitizing cells to MOMP and consequent release of pro-apoptotic factors. * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001, **** p ≤ 0.0001 by student’s t-tests; n=3. Data are mean ± SD.