Orthogonal proteogenomic analysis identifies the druggable PA2G4-MYC axis in 3q26 AML

Abstract

The overexpression of the ecotropic viral integration site-1 gene (EVI1/MECOM) marks the most lethal acute myeloid leukemia (AML) subgroup carrying chromosome 3q26 abnormalities. By taking advantage of the intersectionality of high-throughput cell-based and gene expression screens selective and pan-histone deacetylase inhibitors (HDACis) emerge as potent repressors of EVI1. To understand the mechanism driving on-target anti-leukemia activity of this compound class, here we dissect the expression dynamics of the bone marrow leukemia cells of patients treated with HDACi and reconstitute the EVI1 chromatin-associated co-transcriptional complex merging on the role of proliferation-associated 2G4 (PA2G4) protein. PA2G4 overexpression rescues AML cells from the inhibitory effects of HDACis, while genetic and small molecule inhibition of PA2G4 abrogates EVI1 in 3q26 AML cells, including in patient-derived leukemia xenografts. This study positions PA2G4 at the crosstalk of the EVI1 leukemogenic signal for developing new therapeutics and urges the use of HDACis-based combination therapies in patients with 3q26 AML.

Fig. 1. Identification of EVI1 modulators from the intersection of phenotypic and genome-based approaches.

A Identification of EVI1 “Off” modulators by the integration of phenotypic and in silico gene-expression-based screen approaches. B Circos plot summarizing primary small molecule screening in EVI1^High AML cell lines. The heatmap at the external circle shows the normalized percentage of cell death (ΔPOC) for the 95th percentile for the compounds tested. The middle circos shows the compound libraries: i) Spectrum Collection (n = 1914); ii) the anti-cancer compound library (n = 343); and iii) the NDL-3000 library (n = 3035). The inner plot shows a hierarchical cluster of top candidates’ scalar fingerprints (SMILE), based on the Tanimoto similarity metric. C Dot plot showing the effects of the compound tested (n = 5292), ranked based on ΔPOC. The yellow area indicates the 95th percentile. HDAC inhibitors (HDACis) are labeled in red. D Rank-score log-scale plots for HDACis (n = 32), histamine (n = 52), glucocorticoid (n = 47), and acetylcholine receptor antagonists (n = 66). In black, the number of perturbagenes inferred in the connectivity map of which 2836 are small molecules. Compounds of interest are color-coded and expressed as a percentage relative to the entire library. E Bar plot showing the HDACis (in red) ranking in the ConnectivityMap dataset. The tau score represents the level of similarity to the EVI1 “On” or “Off” status. F L1000FWD fireworks visualization of drug-induced signatures mimicking and reversing the differential gene expression signature generated from EVI1-silenced HNT34 (left) and from EVI1^High (n = 5) or EVI1^Low (n = 2) AML cell lines contained in the E-MTAB-2225 dataset (right). Aquamarine circles indicate small molecules causing a reverse signature EVI1 “Off”. Light violet circles indicate small molecules mimicking a signature EVI1 “On”. HDACis drug-induced signatures are indicated in green overlapping with aquamarine circles. See also Supplementary Fig. 1-2 and Supplementary Data 1.

Fig. 2. Validation of HDACis as EVI1 modulators.

A EVI1 and ∆EVI1 expression in EVI1^High AML cell lines (n = 3 biological replicates). B Nuclear localization expression of EVI1 (in red) detected by immunofluorescence (IF) in EVI1^High AML cell lines (n = 3 biological replicates). Cell nuclei were stained with DAPI (blue), scale bar: 10 µm. C EVI1 and ∆EVI1 expression in TF1 cells 6 days after shRNA transduction. D Percentage of EVI1 mRNA relative to the control gene RPL13A (ΔΔCT) in TF1 cells 6 days after shRNA transduction. E Effect of EVI1 loss in TF1 cells at three or six days after shRNA selection on cell proliferation. F Effect of shRNA-directed EVI1 loss in the ability of TF1 to form colonies in methylcellulose compared to NT (n = 2 biological replicates). Scale bar: 1000 µm. The histogram on the right shows the mean number of colonies per field 20 days after plating cells. NT = non-targeting, sh#16 and sh#87 = shRNA directed against EVI1 (C–F). G Effect of AR-42, belinostat, and entinostat on cell viability in EVI1^High cell lines following 72 hours (hr) of treatment. H Annexin V/propidium iodide (PI) staining after 72 hr of HDACis treatment. Events ≥ 20,000. I Apoptotic fold increase, expressed as a percentage of annexin V-positive cells relative to vehicle control. J Effect of HDACis on the ability of 3q26 AML cell lines to form colonies in methylcellulose. Representative images of UCSD/AML1 treated with vehicle (DMSO), 0.5 µM AR-42, 0.5 µM belinostat, and 0.5 µM entinostat (n = 3 biological replicates). Scale bar: 1000 µm. K The histogram shows the mean number of colonies per field 20 days after plating cells (n = 3 biological replicates). Statistical significance among groups was determined by one-way (D, K) or two-way (E) ANOVA using Tukey’s correction for multiple comparison testing. Data are presented as mean ± standard deviation (SD) in D (n = 3), E (n = 4), F (n = 10 fields per condition), G (n = 2), and K (n = 10 fields per condition). Source data are provided as a Source Data file. See also Supplementary Fig. 3.

Fig. 3. EVI1 sensitizes AML to HDACis.

A EVI1, ∆EVI1, and cleaved caspase 3 expression in 3q26 EVI1^High cell lines after 24 hr of treatment with DMSO or HDACis (n = 2 biological replicates). B Representative IF images of EVI1 (red) in 3q26 EVI1^High cell lines (left, n = 3 biological replicates). Cell nuclei were stained with DAPI (blue), scale bar: 100 µm. Right: fluorescence intensity of nuclear EVI1 content in 3q26 EVI1^High AML cell lines after 24 hr of treatment. C IF images of U937T_E10 cells cultured in the presence (EVI1^Low, top panel) or absence (EVI1^High, bottom panel) of tetracycline, incubated with an anti-EVI1 (red) antibody (n = 2 biological replicates). Nuclei in blue (DAPI). Scale bar: 100 µm. D Effect of HDACis and ara-C on cell viability after 72 hr of treatment in U937T_E10 cell cultured in the presence (EVI1^Low, black line) or absence (EVI1^High, red line) of tetracycline. E Representative IF showing HL-60 cells transduced with an empty (top panels) or an ORF-EVI1 cDNA (bottom panels) vector and stained with an anti-EVI1 antibody (in red, n = 2 biological replicates). Nuclei in blue (DAPI). Scale bar: 100 μm. F Effect of HDACis and ara-C on cell viability after 72 h of treatment in HL-60 ± ORF-EVI1 cDNA. G Abnormal 3q26 pattern on fluorescence in situ hybridization (FISH) in MOLM1 sorted by cell size. The break-apart hybridization pattern 1F1G1O (one fusion and two separated signals, one green and one orange) indicates the break and split of the EVI1 locus. The abnormal pattern was observed in cells with large nuclei. At least 100 nuclei/cells were analyzed (n = 2 biological replicates). Scale bar: 20 μm. H Expression of EVI1 in MOLM1 sorted based on cell size. Large cells express EVI1 compared to small cells. I Effects of HDACis and chemotherapy treatment on viability in MOLM1 EVI1^High and MOLM1 EVI^Low after 72 h of drug exposure. J Effect of AR-42, belinostat, entinostat, ara-C, vincristine, methotrexate, and daunorubicin on EVI1^High (MOLM1, UCSD/AML1, HNT34, TF1, and MUTZ-3) and EVI1^Low (NOMO1, MOLM13, OCI/AML1, OCI/AML2, GDM1, SKM1, and IMS-M2) AML cell lines calculated using the area under the curve (AUC) model of the log-transformed dose-response data. A lower AUC corresponds to greater sensitivity. Statistical significance was determined by two-tailed non-parametric t-test (Mann-Whitney) (J), one-way ANOVA with Tukey’s correction for multiple comparison testing (B). Data are presented as mean ± SD in B (UCSD/AML1 n = 828, MOLM1 n = 1028, HNT34 n = 571), D (n = 2), F (n = 2), I (n = 2) and J (EVI1^High n = 5, EVI1^Low n = 7). Source data are provided as a Source Data file. See also Supplementary Fig. 4.

Fig. 4. Effect of HDACis in human 3q26 EVI1^High AML.

A H&E histological sections of PR#003 bone marrow cells grown in collagen type I scaffolds with vehicle (DMSO) or HDACis (AR-42, belinostat, and entinostat) for 72 hr. Scale bar: 40 µm. B Percentage of cell viability of primary 3q26 AML grown in collagen type I scaffolds and treated with HDACis. Scale bar: 100 µm. Drug concentration (IC₅₀) used in (A) and (B) were established following a dose-response titration assay presented in Supplementary Fig. 6A. C AUC effect of HDACis and chemotherapy agents in EVI1^High (samples PR#002-008, and PR#023-024) and EVI1^Low (samples PR#010-022, and PR#025-039) AML blasts. D EVI1, ∆EVI1, and cleaved caspase 3 protein expression of primary 3q26 AML samples after 24 h of treatment with the indicated concentrations of HDACis. E Representative images of IF staining of primary 3q26 EVI1^High AML cell showing EVI1 nuclear content (in red) following 24 hr of treatment with DMSO or HDACis at the indicated concentrations (n = 3 biological replicates). Nuclei in blue (DAPI), scale bar: 100 µm. F Quantitative IF analysis of nuclear EVI1 content in 3q26 EVI1^High primary AML cells after 24 h of treatment with indicated compounds. G Effect of entinostat (left) or ara-C (right) on EVI1 nuclear localization (in red) following 6 hr of treatment in PR#003 or PR#004, respectively. The nuclei were stained with DAPI (blue). Scale bar: 100 µm. IF quantification of EVI1 nuclear content is presented at the bottom. H EVI1 expression (brownish) in bone marrow leukemia cells at diagnosis and following two cycles of azacitidine and entinostat (PR#002, left) or three cycles of ara-C and daunorubicin (3 + 7) (PR#004, right). Top, formalin-fixed, paraffin-embedded (FFPE) tissue sections were stained with anti-EVI1 antibody revealed by immunoperoxidase. Scale bar: 100 µm. Statistical significance was determined by a two-tailed non-parametric t-test (Mann-Whitney) (C, G, H) or one-way ANOVA with Tukey’s correction for multiple comparison testing (B, F). Data are presented as mean ± SD in B (PR#003 n = 3, PR#004 n = 2), C (EVI1^High n = 9, EVI1^Low n = 28), F (PR#002 n = 318, PR#003 n = 573), G (PR#003 n = 235, PR#004 n = 214), and H (n = 10 fields per condition). Source data are provided as a Source Data file. See also Supplementary Fig. 5-6 and Supplementary Data 2.

Fig. 5. HDAC-mediated suppression of EVI1 modulates Myc signaling.

A Heatmap showing ssGSEA enrichment of MSigDB of gene signature in MOLM1 and UCSD/AML1 HDACi-treated cells. Hot or cold colors indicate correlation or anticorrelation of the top enriched gene sets from each functional group (P ≤ 0.05 in bold, calcolated according to) among cell lines and treatments (bottom). B Uniform manifold approximation and projection (UMAP) plot of clustering results of PR#002 bone marrow cells before and after therapy. Colors indicate the cell populations on the basis of the reference mapping approach. C UMAP plot of clustering results of PR#002 bone marrow cells before (left) and after (right) therapy, colored according to the UCell score of leukemic markers. D GSEA running score plot of the top enriched MYC targets V1 pathway in HSC, LMPP, and LP (Adj.P = 0.0011) cell populations of PR#002. Each graph indicates the running enrichment score (ES) of the pathway (top), the location of single genes of the gene set in the ranking (central), and the distribution of the ranking metric (bottom). E MYC expression (brownish) in bone marrow leukemia cells of PR#002 at diagnosis and following two cycles of azacitidine and entinostat (left). Scale bar: 100 µm. Right, the scatter dot blot indicates the mean ± SD of the percentage of the MYC-positive cells in FFPE tissue sections (n = 10 fields per condition). F EVI1, ∆EVI1, and MYC expression in EVI1^High AML cell lines, 6 days after shRNA transduction. NT = non-targeting, sh#87 = shRNA directed against EVI1 (n = 3 biological replicates). G In vivo antileukemic effect of entinostat and azacitidine. Mice (PDLX_PR#008) were treated with vehicle (DMSO), azacitidine (1 mg/kg) for 5 days, entinostat (10 mg/kg) (5 days/week) or the combination of both for 3 weeks. Histograms show the percentage hCD45⁺ in bone marrow (top) and the percentage of EVI1 mRNA relative to the control gene RPL13A (ΔΔCT) in CD45+ cells (bottom) at the end of treatment. H UMAP plot of clustering results of PDLX_PR#008 bone marrow LP before (i) and after entinostat (ii), azacytidine (iii) or the combination of both (iv), colored according to the UCell score of leukemic markers. I GSEA running score plot of the top enriched MYC targets V1 pathway in PDLX_PR#008 LP after the treatment with azacitidine/entinostat (Adj.P = 3.40 × 10^-8) or entinostat as a single agent (Adj.P = 2.17 × 10^-12). Each graph indicates the running enrichment score (ES) of the pathway (top), the location of single genes of the gene set in the ranking (central), and the distribution of the ranking metric (bottom). J Dot plot of GSEA results illustrating Molecular Signatures Database (MsigDB) biological processes associated with the indicated treatments compared to the vehicle in PDLX_PR#008 LP. Set size refers to the number of genes associated with each (MsigDB) biological process. Dot color indicates the range of Adj.P for each pathway. K EVI1, MYC, and Ki67 expression (brownish) in bone marrow leukemia cells (PDLX_PR#008) following three weeks treatment of azacitidine and entinostat as described in panel G. FFPE tissue were stained with anti-EVI1, anti-MYC and anti-Ki67 antibodies revealed by immunoperoxidase. Scale bar: 100 μm. L Histograms display the mean ± SD of the percentage of the EVI1, MYC or Ki67-positive cells in FFPE tissue sections. Statistical significance was determined by a two-tailed non-parametric t-test (Mann-Whitney) (E), and one-way ANOVA with Tukey’s correction for multiple comparison testing (G, L). GSEA enrichment score significance was based on a weighted Kolmogorov Smirnov (WKS) test corrected for multiple hypotheses testing: Benjamini & Hochberg (BH or FDR) (D, I, J). Data are presented as mean ± SD in E (n = 10 fields per condition), G (vehicle n = 7, treated n = 5 per group; n = 9 for the bottom panel), K (vehicle n = 7, treated n = 5 per group), L (n = 10 fields per condition). Source data are provided as a Source Data file. See also Supplementary Fig. 7-8 and Supplementary Data 3.

Fig. 6. PA2G4 mediates the effect of HDACis on EVI1 and MYC.

A Circle plot showing chromatin-associated proteins immunoprecipitated with an anti-EVI1 antibody by RIME. The bar length indicates the mean of mass spectrometry spectral count (SPEC) of uniquely identified proteins of two biological replicates per condition. Bar colors indicate protein ontology of the EVI1 interactors. B Histograms show the normalized gene expression levels of SERBP1, RPL14, RPL18, RPS2, PA2G4, RPS10, FBL, PABPC1, RACK1, RSL1D1, and DDX21 following 16 hr of 0.5 µM AR-42, 2 µM entinostat in UCSD/AML1, or 0.8 µM AR-42 and 4 µM entinostat in MOLM1 compared to vehicle. Statistical significance was determined using DESeq2. Whiskers show median values (central black lines) and 25^th and 75^th percentiles (bottom and top bounds), respectively. The bars represent values that exceed 1.5 times the interquartile range (IQR) from the edge of each box: (vehicle n = 6, treated n = 12). C Tracks showing EVI1 binding and H3K27ac enrichment across the DDX21, FBL, and PA2G4 locus in HNT34 cells treated with vehicle (DMSO), AR-42 and entinostat. The bottom bar represents the genes (hg19), and the y-axis represents normalized read density scaled to 1 million. D PA2G4, EVI1, ∆EVI1, and MYC expression in TF1, HNT34, and UCSD/AML1 cells 6 days after shRNA transduction (n = 3 biological replicates). E Effect of PA2G4 overexpression on HNT34 cells treated with HDACis. Western blot analysis showing PA2G4, EVI1, ∆EVI1, and MYC in wild-type or PA2G4-overexpressing HNT34 cells after 24 h of treatment with HDACi at the indicated doses (n = 2 biological replicates). (F) Histogram indicates the percentage of viable (white), dead (black), or rescued (red) HNT34 cells (n = 20.000) by the overexpression of PA2G4 on the basis of positivity for annexin V/PI staining after 72 h of HDACis treatment (n = 2 biological replicates). G Percentage of PA2G4, EVI1, and MYC mRNA relative to the control gene ACTB (ΔΔCT) in HNT34 and UCSD/AML1 cells 6 days after shRNA transduction. Data are presented as mean ± SD (n = 3). Statistical significance was determined by a two-tailed unpaired, parametric t-test. H Proteasome inhibition rescues WS6-induced EVI1 and MYC protein degradation. EVI1, ∆EVI1, MYC expression, and PA2G4 in HNT34 and UCSD/AML1 cell lines after treatment with DMSO and WS6 (24 hr) or MG132 (4 hr) at the indicated concentrations (n = 2 biological replicates). NT = non-targeting, sh#65 = shRNA directed against PA2G4 in C and F. Source data are provided as a Source Data file. See also Supplementary Fig. 9-10 and Supplementary Data 4.

Fig. 7. PA2G4 suppression alters 3q26 AML proliferation in vivo.

A Effects of WS6 treatment on viability in primary 3q26 AML samples following 72 hr of drug treatment. B Effect of WS6 in EVI1^High (samples PR#002 - 006, PR#008 and PR#023 - 024) and EVI1^Low (samples PR#025-039) primary AML samples. The AUC model of the log-transformed dose-response data is depicted. C Effects of WS6 treatment compared to control (DMSO) on viability in primary 3q26 AML samples following 72 hr of drug treatment. Left: H&E of PR#005 bone marrow cells in 3D cell culture after 72 hr of treatment with vehicle (DMSO) or WS6 at 1/2 IC₅₀ and IC₅₀ concentrations. Scale bar: 40 µm. Right: histograms indicate the fraction of viable cells expressed as a percentage relative to control. D PA2G4, EVI1, ∆EVI1, and MYC and cleaved caspase 3 protein expression in primary 3q26 AML samples following 24 h of treatment with the vehicle (DMSO) or WS6 at the indicated concentrations. E Effect of WS6 (50 mg/kg) on EVI1 nuclear localization (in red) following 6 hr of treatment in PDLX_PR#009 (n = 3 biological replicates). Nuclei in blue (DAPI). Scale bar: 100 µm. Histograms indicate the fluorescence intensity of EVI1 nuclear content before and after treatment. F In vivo antileukemic effect of WS6. On the top the draft of the experiments. Mice were treated with vehicle (DMSO) or 25 mg/Kg WS6 for 5 days/week for 15 days. On the bottom-left dot plot showing the number of hCD45+ cells expressed as the percentage difference of BM leukemic cells at T1 (endpoint-day 15) vs. T0 (pre-treatment-day 0) normalized for T0 in control (n = 7) and WS6-treated PDLX_PR#003 (n = 6). On the bottom-right the dot plot shows the percentage of hCD45+ cells in the bone marrow between vehicle (n = 4) or WS6-treated PDLX_PR#008 (n = 5) at the endpoint. Events ≥ 20.000. G UMAP plot of clustering results of PDLX_PR#003 hCD45 bone marrow–positive cells before (left panel) and after (right panel) 15 days of WS6 treatment (25 mg/kg/IP/5 days a week), color-coded according to the UCell score of leukemic markers (MECOM, MYC, CD45, CD34, KIT, CD33, ANPEP, CD38, CD2, TFRC, HLA-DRA, HLA-DRB1, and HLA-DRB5). H EVI1, MYC, and Ki67 expression in PDLX_PR#003 after 15 days of WS6 treatment (25 mg/kg/IP/5 days a week) or vehicle (DMSO). Scale bar: 100 µm. Histograms to the right indicate the mean ± SD of the percentage of the EVI1, MYC, or Ki67-positive cells in FFPE tissue sections. Statistical significance was determined by a two-tailed non-parametric t-test (Mann-Whitney) (B, E, F, H). Data are presented as mean ± SD in A (n = 2), B (EVI1^High n = 8, EVI1^Low n = 15), C (n = 2), in E (n = 125), F (vehicle PDLX_PR#003 n = 7, treated PDLX_PR#003 n = 6, vehicle PDLX_PR#008 n = 4, treated PDLX_PR#008 n = 5) and H (vehicle n = 7, treated n = 6; n = 10 fields per condition). Source data are provided as a Source Data file. See also Supplementary Fig. 11.