Overexpression of the signaling coordinator GAB2 can play an important role in acute myeloid leukemia progression

Abstract

Mutations that initiate acute myeloid leukemia (AML) can cause clonal expansion without transformation (clonal hematopoiesis). Cooperating mutations, usually in signaling genes, are needed to cause overt disease, but these may require a specific fitness state to be tolerated. Here, we show that nearly all AMLs arising in a mouse model expressing 2 common AML-initiating mutations (Dnmt3aR878H and Npm1cA) acquired a single copy amplification of chromosome 7 (chr7), followed by activating mutations in signaling genes. We show that overexpression of a single gene on chr7 (Gab2, which coordinates signaling pathways) was tolerated in the presence of the Npm1cA mutation, could accelerate the development of AML, and was important for the survival of fully transformed AML cells. GAB2 is likewise overexpressed in many human AMLs with mutations in NPM1 and/or signaling genes, and also in acute promyelocytic leukemia initiated by PML::RARA; the PML::RARA fusion protein may activate GAB2 by directly binding to its 5' flanking region. A similar pattern of GAB2 overexpression preceding mutations in signaling genes has been described in other human malignancies. GAB2 overexpression may represent an oncogene-driven adaptation that facilitates the action of signaling mutations, suggesting an important (and potentially targetable) missing link between the initiating and progression mutations associated with AML.

Keywords: Genetics; Leukemias; Oncology.

Figure 1. Mice with Dnmt3a^R878H and Npm1^cA mutations spontaneously develop AML with amplification of chr7 and cooperating signaling mutations.

(A) Whole-genome sequencing of 11 AMLs arising in Dnmt3a^R878H/+ × Npm1^cA/+ mice (mAMLs) revealed mutations in cancer driver genes and amplification of chr7 in 10 of 11 tumors. amp, amplification. (B) RNA expression (z scores) is shown for DEGs from mAML versus preleukemic Dnmt3a^R878H/+ × Npm1^cA/+ samples (FDR < 0.01). Shown are mAML samples (n = 9), pre-AML murine bone marrow samples (nontransplantable clonal expansions with +7 in Dnmt3a^R878H/+ × Npm1^cA/+ mice; n = 2), and preleukemic murine bone marrow samples of the indicated genotypes. Genes are split by genomic location (chr7 vs. the rest of the genome) and hierarchically clustered. (C) Left: Copy number changes in mAMLs (this study), panel sequencing data from a similar mouse model (24), and aCGH of an Npm1^cA-driven tumor model (28). Right: The minimally amplified region identified on chr7. WGS, whole-genome sequencing. (D) mRNA expression of genes on the chr7 minimally amplified region and their human orthologs. The x axis represents mouse mAML tumors compared with Dnmt3a^R878H/+ × Npm1^cA/+ preleukemic samples from this study; the y axis represents human AML tumors with DNMT3A^R882C/H and NPM1^c mutations compared with healthy CD34⁺ cells (75). (E–G) mRNA expression of Gab2 in mAML versus controls in this mouse model (E) and in patients from the BEAT-AML study (F) (y axis: log₁₀ scale to show the range in the presence of high Gab2 AML samples) (75) or TCGA (G) (7). (H) Protein abundance in AMLs versus healthy lineage-depleted bone marrow samples, measured by tandem-mass-tag mass spectrometry, scaled to between 0 and 1 for display (43). *P < 0.05, by 2-tailed t test. Boxes show median (line) and extend from the 25th to 75th inter-quartile range (IQR), with whiskers showing 1.5X the IQR, and outlier points as shown outside the whiskers (E–H), adjusted for multiple hypothesis testing in E using the Benjamini-Hochberg method.

Figure 2. Retroviral overexpression of Gab2 causes the expansion of HSPCs and accelerates the development of AML in the presence of Npm1^cA and Dnmt3a^R878H mutations.

(A) Retroviral vectors were transduced into lineage-depleted primary murine CD45.2⁺ bone marrow cells to create a mixed population of transduced (GFP⁺) and untransduced (GFP^–) cells. Cells were split into 2 groups, with 1 maintained in vitro and 1 transplanted into CD45.1⁺ recipient mice. The fraction of GFP⁺ cells was measured over time using flow cytometry. This illustration was created in BioRender. (B) Overexpression of Gab2 in Dnmt3a^R878H/+ × Npm1^cA/+ marrow shows a growth advantage in vitro, with GFP⁺ (Gab2-overexpressing) cells expanding over time. No effect was observed in cells transduced with the IRES-eGFP or Pak1-IRES-eGFP vectors. Representative data from 1 of 4 biological replicates are shown. (C and D) Overexpression of Gab2 in (C) WT or (D) Dnmt3a^R878H/+ bone marrow caused a selective disadvantage compared with IRES-eGFP vector–only cells or untransduced cells. Data shown in C and D are representative of 1 of 2 biological replicates. (E) Overexpression of Gab2 in Npm1^cA/+ bone marrow did not show a marked advantage or disadvantage compared with IRES-eGFP vector or untransduced cells. Data shown are representative of 1 of 2 biological replicates. (F) GFP positivity of CD45.2⁺ peripheral WBCs was measured by flow cytometry in CD45.1 mice transplanted with retrovirally transduced cells of the indicated genotypes. Dots show mean values at each time point, and error bars show SD. P values were calculated at each time point for the Gab2-IRES-eGFP curve using a 2-sided, 1-sample t test for change from baseline compared with a null hypothesis of a mean change of 0, with Benjamini-Hochberg multiple hypothesis correction. (G) Mice of the indicated genotypes were monitored for AML development. Retroviral overexpression of Gab2 in the Dnmt3a^R878H/+ × Npm1^cA/+ background significantly accelerated AML development. P values were determined by pairwise log-rank test.

Figure 3. Increased Gab2 expression does not cause major transcriptional or immunophenotypic changes in AML.

(A) Preleukemic Dnmt3a^R878H/+ × Npm1^cA/+, lineage-depleted murine bone marrow cells were transduced with retroviruses expressing Gab2-IRES-eGFP or IRES-eGFP only. Cells were maintained in vitro and harvested 1 or 4 weeks after transduction. GFP⁺ cells were then enriched by flow cytometry, and bulk RNA-Seq was performed. Two biological replicates were performed for each condition and time point (n = 4 per group). Each dot represents 1 gene. Gab2 (red) is the only DEG (FDR < 0.05). (B) Flow cytometry was performed on bone marrow samples from leukemic mice and WT bone marrow samples. The percentage of CD45.2⁺ cells positive for each of the markers indicated on the x axis is shown. mAML denotes murine AMLs developing spontaneously in Dnmt3a^R878H/+ × Npm1^cA/+ mice; GAB2 mAML denotes murine AMLs developing with Gab2 retroviral overexpression in Dnmt3a^R878H/+ × Npm1^cA/+ mice. (C and D) UMAP of cells from leukemic mice with mAML or GAB2 mAML, characterized using a 23-color hematopoietic flow cytometry panel. Note that the majority of cells in all samples are Cd11b⁺ myeloid cells. Cells are colored by cell type in C and by sample type in D. (E) UMAP created from bulk mRNA-Seq data collected from murine bone marrow samples of the indicated types (annotated by color). Preleukemic and leukemic cells formed mostly distinct clusters, whereas mAMLs and GAB2 mAMLs colocalized. (F) Mean expression levels of all genes as determined by bulk mRNA-Seq of GAB2 mAML tumors (n = 5) and mAML tumors (n = 9). DEGs (FDR < 0.01) are colored red. See Supplemental Table 6 for the list of 57 DEGs.

Figure 4. The GAB2 protein interactome is shaped by Dnmt3a^R878H and Npm1^cA mutations.

(A–C) Protein proximity data derived from TurboID (TID) expression performed in primary, lineage-depleted murine bone marrow cells of the indicated genotypes. Each box represents a single replicate, colored according to the log₂ fold change (FC) of spectral counts detected compared with the mean spectral counts detected in control samples using TurboID alone as the bait. As expected, each bait was self-labeling, and known interactions were detected, including for GAB2 (with GRB2, PTPN11, and CBL) and for DNMT3A (with DNMT3B). Yellow box highlights interactions with GAB2 and NPM1^cA observed in Dnmt3a^R878H/+ cells. (D and E) Normalized GRB2 (D) and GAB2 (E) spectral counts detected in TurboID proximity labeling experiments. The x axis indicates the bait used and the genotype of the primary murine bone marrow cells transduced. Boxes show the median (line) and extend from the 25th to 75th IQR, with whiskers showing 1.5 times the IQR and outlier points as shown outside the whiskers. P values were calculated by 2-sided, 2-sample t test and adjusted for multiple hypothesis correction using the Benjamini-Hochberg method. Note that NPM1^cA (but not WT NPM1) shows an interaction with GRB2 in WT and Dnmt3a^R878H/+ cells but interacts with GAB2 only in Dnmt3a^R878H/+ cells.

Figure 5. Some RTK/RAS signaling proteins display posttranscriptionally increased protein abundance and NPM1^cA physical interactions.

(A) Mean expression values of protein and mRNA as measured in leukemic mAML murine bone marrow. Protein expression was measured using LFQ mass spectrometry. The R² value was calculated on proteins and genes detected using both methods. S100a9 and S100a8 are known to be highly abundant genes/proteins in myeloid cells. Gab2 is indicated in red. (B) Hierarchical clustering of TMT proteomics data from preleukemic versus leukemic samples of the indicated genotypes using the unweighted pair group with arithmetic mean (UPGMA) method (76) and Pearson’s correlation of protein abundance profiles as the distance metric. A yellow box highlights preleukemic samples, and a blue box highlights AML samples. (C) t-SNE created from TMT proteomics data collected from murine bone marrow samples of the indicated types (annotated by color). Preleukemic and leukemic cells formed distinct clusters. mAML samples for B–F represent 2 biological replicates each of mAML 1 and mAML 3, transplanted into secondary recipients. (D–F) TMT protein abundance for the indicated genes from bone marrow of preleukemic and leukemic (mAML) mice from this model. Each dot represents a measurement from an individual mouse. TMT protein expression values were median centered at 0 and log₂ transformed. For mAML samples, adjusted **P < 0.05, by t test with Benjamini-Hochberg multiple hypothesis correction for differences compared with Dnmt3a^R878H/+ × Npm1^cA/+ preleukemic bone marrow. Red lines indicate the mean of each group. (G–I) mRNA expression of the indicated genes from whole bone marrow of preleukemic and leukemic (mAML) mice from this model. NS indicates no significant differential expression in the mAML groups compared with Dnmt3a^R878H/+ × Npm1^cA/+ preleukemic bone marrow. Red lines indicate the mean of each group. Data for PIK3CD are also presented in Supplemental Figure 5C (RNA) and Supplemental Figure 5H (protein).

Figure 6. Retroviral overexpression of GAB2 increases phosphorylation of activating sites on AKT and ERK1/2.

(A and B) Preleukemic Dnmt3a^R878H/+ × Npm1^cA/+ bone marrow cells were transduced with MSCV-based retroviruses expressing IRES-eGFP only or Gab2-IRES-eGFP and maintained in vitro until marked selection for Gab2-overexpressing cells was observed (>1 month, see Figure 2). GFP⁺ cells were purified and then cultured in serum-reduced media for 24 hours prior to lysis. ProteinSimple Jess Western Blotting performed using anti-phosphorylated AKT (anti-pAKT) (Ser473) (A, top), anti-AKT (A, bottom), anti-pERK1/2 (Thr202/Tyr204) (B, top), or anti-ERK1/2 (B, bottom) antibodies. Phosphorylated and total protein blotting was performed on the same blot after stripping using the RePlex system. Note that samples with Gab2-IRES-eGFP transduction show increased phosphorylation at the activating sites on AKT and ERK1/2 compared with the IRES-eGFP–only samples, despite no difference in total AKT or ERK1/2 protein abundance. Both blots are representative of 2 biological and technical replicates.

Figure 7. Inactivation of Gab2 mitigates the growth of some fully transformed mAML cells.

(A) Schematic of experimental design. Dnmt3a^R878H/+ × Npm1^cA/+ mAML cells were harvested from mouse bone marrow and transfected with Cas9 protein and guide RNA targeting Gab2 (2 independent guides) or Rosa26 (negative control). Cells were then transplanted into sublethally irradiated (6 Gy) CD45.1⁺ recipient mice. An aliquot of cells was maintained in vitro for 24–48 hours prior to harvesting to measure baseline editing efficiency. Bone marrow from transplanted mice was harvested when mice were moribund. All mice had more than 90% CD45.2⁺ leukemia cells in the bone marrow at the time of harvesting. DNA from bone marrow was purified, and targeted PCR-based sequencing of edited loci was performed. The figure in A was created with BioRender. (B) Results from the experiment performed as described in A, using mAML 1 (with +7 and Ptpn11^E69K mutation). n = 5 recipient mice per guide. Red line indicates the percentage of frameshift mutations detected at baseline. Boxes show the median (line) and extend from the 25th to 75th IQR, with whiskers showing 1.5 times the IQR; no outlier points were detected beyond the whiskers. P values were determined by 2-sided, 1-sample t test for differences from the baseline value, adjusted for multiple hypothesis correction using the Benjamini-Hochberg method. (C) Results from an experiment performed as described in A using 5 independent mAML tumors. Each point represents an individual mouse and shows the percentage of change in frameshift mutations detected at the time of harvesting compared with baseline. Red line indicates the median of each group. P values were calculated by Wilcoxon rank-sum test between the Rosa26 and Gab2 groups. Gab2 guides were combined for this plot. See also Supplemental Figure 7 for these data, separated by mAML tumor and individual guide.

Figure 8. Inactivation of GAB2 mitigates the growth of some human AML samples.

(A) GAB2 mRNA expression in primary human AML samples from TCGA (7), grouped by mutations. PML::RARA is the driver of acute promyelocytic leukemia (n = 16). Signaling-mutant AML includes 83 patients with FLT3, NRAS, KRAS, CBL, PTPN11, or KIT mutations; NPM1^c AML includes 48 patients; other AML includes 72 patients. Purified CD34⁺ cells from healthy donors (CD34⁺ healthy) served as a control (n = 3). P values were determined by 2-sided, 2-sample t test, adjusted with Benjamini-Hochberg multiple hypothesis correction. Boxes show the median (line) and extend from the 25th to 75th IQR, with whiskers 1.5 times the IQR. Numbers indicate the group median. (B) GAB2 CRISPR essentiality from the DepMap CHRONOS dataset in human cancer cell lines. Zero (0) indicates no effect. Black line indicates the group mean. Dotted red line shows the median value of pan-essential genes. P value was determined by Wilcoxon rank-sum test. (C–E) Cells were electroporated with Cas9 mRNA and an sgRNA targeting GAB2 (2 independent guides) or AAVS1 (negative control) and maintained in vitro. PCR-based sequencing determined the frequency of frameshift mutations at each targeted site. Cells with frameshift mutations in GAB2 (but not AAVS1) decreased in frequency in (C) human K562 cells and (D) a cryopreserved, primary human AML sample with NPM1^c and FLT3-ITD mutations, but not in (E) CD34⁺ HSPCs from human cord blood (representative of 2 biological replicates). (F) Cryopreserved primary human AML samples (each with an NPM1^c mutation and either FLT3-ITD, FLT3^D835Y or high GAB2 mRNA expression) were CRISPR edited and transplanted into NSG-SGM3 mice. After engraftment, bone marrow was sorted for human AML cells prior to targeted sequencing; sequencing results were compared with baseline editing. P values were determined by 2-sided, 1-sample t test for differences from 0 in each group. Boxes are as in A, with points outside the IQR range shown individually.

Figure 9. GAB2 mRNA abundance is increased by PML::RARA in both mouse and human hematopoietic cells.

(A) Genome browser tracks for the Gab2 locus in mouse for anti-V5 ChIP-Seq or CUT&RUN performed in WT lineage–depleted mouse bone marrow cells transduced with PML::RARA^WT-V5, compared with empty vector–transduced cells or cells transduced with a PML::RARA^C88A-V5 mutant that does not bind to its consensus DNA target sites (50, 77, 78). Assay for transposase-accessible chromatin using sequencing (ATAC-Seq) was performed on promyelocytes that were flow-enriched from the bone marrow of Ctsg-PML::RARA mice or WT littermates. (B) Gab2 mRNA expression from bulk RNA-Seq on flow-sorted promyelocytes from three 8- to 12-week-old littermate-matched Ctsg-PML::RARA mice versus WT mice. **FDR = 1 × 10^–14. (C) t-SNE plots of scRNA-Seq data from whole bone marrow cells from young, nonleukemic WT mice or Ctsg-PML::RARA mice (51). Known hematopoietic cell types are labeled according to Haemopedia gene expression profiling (79, 80). A unique population of myeloid precursor cells that were only present in the bone marrow from Ctsg-PML::RARA mice is outlined in blue (PML::RARA specific). Cells are colored according to Gab2 expression. GMPs, granulocyte-macrophage progenitors. DCs, dendritic cells. (D) Relative Gab2 expression by scRNA-Seq in the various lineage populations from C in WT versus Ctsg-PML::RARA (PR) bone marrow. **FDR = 1 × 10^–15, ***FDR = 1 × 10^–55. (E) Genome browser tracks for the GAB2 locus in humans for anti-V5 ChIP-Seq performed in healthy donor–derived (HD-derived), human CD34–enriched cord blood cells transduced with PML::RARA^WT-V5, treated with DMSO or ATRA for 48 hours (50); ATRA degrades the PML::RARA protein (52). ATAC-Seq was performed in flow-sorted promyelocytes from healthy donors and in primary human APL samples.

Figure 10. Gab2 inactivation mitigates the growth of murine APL cells.

(A and B) Bone marrow cells from WT (A) or Ctsg-PML::RARA (B) mice were harvested and electroporated with Cas9 protein and sgRNAs targeting Gab2 (2 independent guides) or Rosa26 (as a negative control). Cells were cultured in methylcellulose and replated weekly. Targeted PCR-based sequencing was performed at the indicated time points to determine the fraction of edited cells. (C and D) Cryovials of cells from 2 independent murine APLs were thawed and transfected with Cas9 mRNA and an sgRNA targeting Gab2 (2 independent guides) or Rosa26 (as a negative control). Cells were cultured in methylcellulose and replated every 5–7 days. Targeted PCR-based sequencing was performed at the indicated time points to determine the fraction of edited cells. For APL tumor 2 (D), editing efficiency was maximal (Max) at day 3 for Rosa26 and Gab2 guide 1, and at day 10 for Gab2 guide 2 (other time points were not measured). The maximum value is displayed for each guide for baseline comparison.