Genetic and Phenotypic Correlates of Clinical Outcomes with Venetoclax in Acute Myeloid Leukemia: The GEN-PHEN-VEN Study

Abstract

Resistance to venetoclax (VEN)-based therapy in acute myeloid leukemia (AML) includes genetic (i.e., mutations in N/KRAS, FLT3-ITD, TP53) and phenotypic (i.e., monocytic differentiation) features. Whether monocytic differentiation contributes to clinical VEN resistance secondary to a genetic bias remains unknown. This multimodal, multicenter, international analysis, inclusive of 678 patients, comprehensively characterized the prognostic role of monocytic differentiation in patients with AML treated with hypomethylating agents combined with VEN. AML genetics and monocytic differentiation (HR = 1.89; 95% confidence interval, 1.35-2.66; P < 0.001) in NPM1 wild-type cases correlated with an increased risk of death, clustering of centralized quantitative multiparameter flow cytometry data, evaluation of RNA sequencing-derived AML maturation stage, and single-cell proteogenomics linked driver mutations with AML phenotype and antiapoptotic gene expression. This comprehensive analysis of AML genetics, phenotype, and antiapoptotic protein expression highlights the complementary role these factors impart following VEN-based therapy.

Significance: AML with monocytic differentiation often occurs in the context of co-occurring mutations within signaling pathways. In certain AML subgroups (such as NPM1 wild-type and signaling pathway gene-mutated), a monocytic phenotype is associated with decreased overall survival following VEN-based therapy. See related commentary by Renders, p. 403.

Figure 1.

OS stratified by (A) the four-gene mPRS classifier (based on mutations in K/NRAS, FLT3-ITD, and TP53) proposed by Döhner and colleagues and (B) monocytic differentiation status determined by morphology and/or MFC. C, OS censored at HCT of patients with mutated K/NRAS, PTPN11, or FLT3-ITD with wild-type TP53, NPM1, and IDH1/2, stratified by monocytic differentiation. D, OS based on the presence of monocytic differentiation and mutated vs. wild-type NPM1. E, Forest plot depicting the adjusted effects of prognostic factors on survival from multivariable Cox analysis applied to the clinical patient cohort. Multiple rows for monocytic phenotype and NPM1 mutation status reflect the included two-way interaction term involving these factors. Reported P values utilized the Cox model Wald test. karyo, karyotype; mono, monocytic; mut, mutated; nonmono, nonmonocytic; wt, wild-type.

Figure 2.

A, Uniform Manifold Approximation and Projection (UMAP) representation of AML cases with available centralized flow cytometry (N = 144) clustered by MFC-assayed surface immunophenotype (left) and overlaid with FAB M5 classification (right). B, Correlation between normalized MFI of surface marker expression within each UMAP cluster, with depicted values representing the PBCC consistent with unique surface marker expression dependent upon AML phenotype. C, Distribution of clusters within each genetic subgroup indicated on the left. Asterisks represent mutation enrichment (measured using Fisher’s exact test) in cases with a monocytic immunophenotype measured using MFC (cluster 3) vs. nonmonocytic clusters. D, PC1 value (derived from MFI expression for CD38, CD14, CD36, CD64) for each cluster displayed in (A). E, Multivariable analysis of monocytic differentiation as a continuous variable in the N = 144 patient subgroup. ****, Wilcoxon P value < 0.0001. karyo, karyotype; mono, monocytic; mut, mutated; nonmono, nonmonocytic; wt, wild-type.

Figure 3.

A, Density plots demonstrating the relative frequency of cell state scores [derived from the transcriptional cell state signatures using the Beat AML cohort (n = 228)] for monocyte- or progenitor-like scores within co-mutated subgroups of AML. B and C, Uniform Manifold Approximation and Projection (UMAP) representation of single-cell proteogenomics (DAb-seq; single-cell DNA + surface immunophenotyping) of AML cells clustered based on DAb-seq measured surface immunophenotype in a patient with NPM1/ETV6/TET2/NF1–mutated AML. Increased monocytic/mature myeloid cell bias was observed within the clone harboring mutations in NF1. D and E, UMAP representation of DAb-seq of AML cells clustered based on DAb-seq measured surface immunophenotype in a patient with IDH1-mutated AML with a relative increase in primitive myeloid progenitors observed in the IDH1-mutated clone. F–I, DAb-seq in a patient with IDH2-mutated AML treated with HMA + VEN. F and G, UMAP representation of the cell populations at both diagnosis and at the end of treatment cycle 1 (EOC1), clustered by cell type, demonstrating the IDH2-mutated clone. H, Bar plot demonstrating the relative frequency of DAb-seq measured cell populations between diagnosis and EOC1, demonstrating an immunophenotypic shift to a more mature blast population. I, Violin plots depicting changes to surface CD34, CD14, CD64, and CD11b between diagnosis and remission with differences represented using the Wilcoxon test. DC, dendritic cell; mono, monocytic.

Figure 4.

A, For the Beat AML cohort (n = 228), correlation plots demonstrating differences in ex vivo VEN sensitivity measured by area under the cell viability curve (AUC; in which larger values denote VEN resistance) with respect to transcriptional cell state signature–derived scores. Individual jitter points represent individual patients. B, BCL2 and MCL1 expression [log₂ reads per kilobase per million mapped reads (RPKM)] correlation with transcriptional cell state signature–derived scores from the Beat AML cohort. C, Antiapoptotic gene expression ratios (computed on the single-gene log₂ RPKM values) within monocytic vs. nonmonocytic cases defined using available MFC (with a minimum of three positive surface markers) or using transcriptional cell state signature–derived scores. D, BCL2:BCL2A1 expression ratios across mutational subgroups of samples from patients with AML. *, Wilcoxon P value < 0.05; ***, Wilcoxon P value < 0.001; ****, Wilcoxon P value < 0.0001. Nonmono, nonmonocytic; RNA-seq, RNA sequencing.