Phosphoproteomics of primary AML patient samples reveals rationale for AKT combination therapy and p53 context to overcome selinexor resistance

Abstract

Acute myeloid leukemia (AML) is a heterogeneous disease with variable patient responses to therapy. Selinexor, an inhibitor of nuclear export, has shown promising clinical activity for AML. To identify the molecular context for monotherapy sensitivity as well as rational drug combinations, we profile selinexor signaling responses using phosphoproteomics in primary AML patient samples and cell lines. Functional phosphosite scoring reveals that p53 function is required for selinexor sensitivity consistent with enhanced efficacy of selinexor in combination with the MDM2 inhibitor nutlin-3a. Moreover, combining selinexor with the AKT inhibitor MK-2206 overcomes dysregulated AKT-FOXO3 signaling in resistant cells, resulting in synergistic anti-proliferative effects. Using high-throughput spatial proteomics to profile subcellular compartments, we measure global proteome and phospho-proteome dynamics, providing direct evidence of nuclear translocation of FOXO3 upon combination treatment. Our data demonstrate the potential of phosphoproteomics and functional phosphorylation site scoring to successfully pinpoint key targetable signaling hubs for rational drug combinations.

Keywords: CP: Cancer; CP: Molecular biology; MK-2206; acute myeloid leukemia; combination therapy; drug resistance; functional scoring; mass spectrometry; nutlin-3a; phosphoproteomics; selinexor; subcellular proteomics.

Graphical abstract

Figure 1

Phosphoproteomics analysis of the selinexor response in AML ex vivo samples (A) Schematic representation of the workflow for characterization of the selinexor response in ex vivo-treated AML patient samples (n = 44). (B and C) Patient distribution in cohort according to the French-American-British (FAB) subtype classification of AML morphology (B) and the European LeukemiaNet 2017 risk stratification model by genetic abnormality (C). (D) Selinexor EC50 values for the 20 ex vivo AML patient samples in the phosphoproteome analysis with grouping into selinexor responders (EC50 < 1,000 nM; n = 9) and non-responders (EC50 > 1,000 nM; n = 11). See also Figure S1. (E) Distribution of selinexor non-responders and responders according to FAB classification (left) and risk stratification by genetics (right). (F) Western blot analysis (upper) and quantification (lower) of phospho-Rb (S780) and Rb levels in MV-4-11 cells after treatment with selinexor for 0.5–8 h. Data in bar graph represent mean ± SD (n = 3 independent experiments). ^∗p < 0.05 by two-sample Student’s t test with Bonferroni correction. (G) Experimental workflow of the MS-based quantitative phosphoproteome analysis of the selinexor response in AML cells isolated from 20 patients presented in (D). (H) Summary of phosphoproteome data with the number of identified phosphorylation sites and phosphoproteins. See also Figures S1–S3; Tables S1 and S2.

Figure 2

Functional scoring of phosphorylation sites identifies key signaling rewiring in response to selinexor (A) Overview of the functional sites assigned to phosphorylation sites identified to be down- (left) or upregulated (right) among responders and non-responders in response to selinexor treatment. Phosphorylation sites with a functional score >0.6 are represented. Known regulatory site functions and their involvement in a biological process as indicated by colored circles are functionally annotated based on PhosphoSitePlus (Hornbeck et al., 2015) and retrieved from Perseus software (Tyanova et al., 2016). (B) Workflow for single-shot MS-based proteome analysis of 30 ex vivo AML patient samples. (C and D) Volcano plots showing differentially regulated proteins from the single-shot proteome MS analysis. Fold change represents selinexor treatment versus DMSO control of ex vivo-grown AML cells for responders (n = 8) (C) and non-responders (n = 22) (D). Significance was deemed based on a two-sided t test (FDR < 0.05, s0 = 0.1) using Perseus software and highlighted in orange coloring. Red and blue coloring indicate the proteins with regulated phosphosites shown in (A). See also Figures S4 and S5; Tables S2 and S3.

Figure 3

Phosphoproteomics analysis of the selinexor response in sensitive and resistant AML cell lines (A) Dose-response cell viability curves for a panel of AML cell lines treated with selinexor or DMSO as control and estimated relative and absolute EC50 values for each of the cell lines (n = 2–3 independent experiments; n = 4 technical replicates each dose). (B) Lysates from AML cells treated with selinexor or DMSO for 24 h, immunoblotted for cleaved PARP and actin (for reference) (upper) and quantified (lower). Data in bar graph represent mean ± SD (n = 3 independent experiments). (C) Overview of the experimental workflow for quantitative MS-based phosphoproteome analysis of four selected AML cell lines. Each cell line was treated with selinexor or DMSO (n = 4 independent experiments). (D) Summary of phosphoproteome data including the number of identified phosphorylation sites and phosphoproteins derived from the analysis. (E) Overlap between phosphoproteins identified in the AML cell lines; selinexor-sensitive (blue coloring) and -resistant cells (red coloring). (F) Volcano plots showing differentially regulated phosphorylation sites. Fold change represents selinexor treatment versus DMSO control for each AML cell line. Significance was deemed by two-sided t test (n = 4 independent experiments, FDR < 0.05, s0 = 0.1) using Perseus software. (G) Kinase-substrate enrichment analysis (KSEA). Dots in orange with labels are the significantly enriched kinases corresponding to the selinexor-treated versus untreated (DMSO control). Significance was considered for p values <0.05. (H) Overlap in enriched kinases based on KSEA in (G). See also Figure S6; Table S4.

Figure 4

The selinexor responses in the AML ex vivo model and cell lines show common and unique signaling responses (A) List of phosphorylation sites with a functional score >0.6 commonly regulated between ex vivo analysis and AML cell lines. The list highlights known regulatory site functions based on information obtained from Perseus software. Asterisk (^∗) indicates shared phosphorylation sites between response groups. (B) Hierarchical clustering of significantly regulated proteins (n = 504; FDR = 0.01, s0 = 0.1) between selinexor-sensitive (blue color) and -resistant (red color) cells. KEGG pathway overrepresentation with Benjamini-Hochberg FDR-corrected p values for cluster 1 and 2. (C) Heatmap of median log2-transformed label-free quantification (LFQ) intensities for a selection of regulated proteins belonging to FOXO3/mTOR signaling from (B) and include the top five most regulated proteins in (A). Arrows indicate the phosphorylation site (p-site) regulation directionality in (A) and blue/red colored text refers to the selinexor response group. Asterisk (^∗) indicates housekeeping genes. (D) Western blot analysis (upper) and quantification (lower) of FOXO3, AKT, and p53 in AML cells. Data in bar graph represent mean ± SD (n = 3). See also Figure S7; Tables S2, S3, and S4.

Figure 5

Selinexor combination therapy with MDM2 or AKT inhibition shows synergistic effects (A) 3D plots showing the mean relative cell viability (color scale, z axis) against selinexor and nutlin-3a doses (x and y axes) for AML cells treated with the drugs alone or in combination (n = 2–3 independent experiments; n = 4 technical replicates each dose). (B) Heatmaps showing the synergy scores derived for experiments in (A). (C) 3D plots showing mean relative cell viability (color scale, z axis) against selinexor and MK-2206 doses (x and y axes) for NOMO-1 and PL-21 cells treated with the drugs alone or in combination (n = 2–3 independent experiments; n = 4 technical replicates each dose). (D) Heatmaps showing the synergy scores for the different cell lines and drug combinations. (E) Calculated EC50 values for selinexor plotted for each dose of MK-2206 for each AML cell line as indicated. (F) Western blot (left) and quantification (right) of lysates from PL-21 cells treated with selinexor, MK-2206 alone, or the combination for 24 h and immunoblotted for cleaved PARP and β-actin (for reference). Quantification of blots (right) shows mean ± SD (n = 5). ^∗p < 0.05 by two-sample Student’s t test with Bonferroni correction. n.s., non-significant. (G) Model of the selinexor response in ex vivo AML samples and cell lines. See also Figure S8.

Figure 6

Subcellular proteome analysis of selinexor synergy combination with MK-2206 confirms FOXO3 nuclear translocation in PL-21 cells (A) Brief overview of the spatial-proteomics workflow and summary of results. (B) Heatmap of scaled intensities per fraction of the subcellular proteome of MV-4-11 (left) and PL-21 (right) cells (DMSO-treated cells; n = 4 independent experiments), showing protein and sample clustering. (C and D) Volcano plots showing differentially regulated proteins in the subcellular proteome analysis of fraction 2 (C) and 1 (D) of MV-4-11 cells. Fold change represents selinexor treatment versus DMSO control. Significance was deemed by a two-sided t test (FDR < 0.05 [dotted line], FDR < 0.01 [solid line], s0 = 0.1) using Perseus software. (E) Overlap of proteins identified in MV-4-11 cells to be significantly downregulated in fraction 1 and upregulated in fraction 2 by selinexor and upregulated by MK-2206 treatment in fraction 2. (F and G) Volcano plots showing differentially regulated proteins in the subcellular proteome analysis of fraction 2 (F) and 1 (G) of PL-21 cells. Fold change represents the combination treatment (selinexor and MK-2206) versus DMSO control. Significance was deemed by a two-sided t test (FDR < 0.05 [dotted line], FDR < 0.01 [solid line], s0 = 0.1) using Perseus software. (H) Overlap of proteins identified in PL-21 cells to be significantly downregulated in fraction 1 and upregulated in fraction 2 by the synergy combination and upregulated by selinexor in fraction 2. (I and J) KEGG pathway enrichment analysis for proteins significantly upregulated in fraction 2 by selinexor versus DMSO in MV-4-11 cells (I) and the synergy combination versus DMSO in PL-21 cells (J). See also Figures S9–S11; Tables S5 and S6.