Triple-mutated AML samples are characterized by an aberrant GPR56^high CD34^low immunophenotype and high LSC frequency. (A) Mutational landscape of 65 AML samples at first diagnosis. Blue box indicates mutated, light gray box indicates nonmutated. Color coded bar on top shows grouping of samples according to their mutational status for FLT3-ITD, NPM1, and DNMT3A into single-mutated (n = 16, orange group), double-mutated (n = 14, green group), and triple-mutated samples (n = 22) split into those with a typical DNMT3A mutation at position R882 (n = 13, black) and those with other DNMT3A mutations (n = 9, dark gray). Samples not mutated for the 3 genes are included in the blue group (n = 13). Color-coded bar at the bottom indicates the immunophenotypic profile with regard to CD34 and GPR56 percentage determined by flow cytometry: profile 1 defined as CD34% > GPR56%, profile 2 as GPR56% > CD34%, and profile 3 as all CD34-negative samples (CD34 < 1%). (B) Representative examples of the 3 different immunophenotypic profiles as defined in panel A. (C) Numbers of triple mutations in patients with profile 2 vs other groups. Detection of the aberrant profile 2 at diagnosis implies an 84-fold greater chance to simultaneously harbor mutations in FLT3-ITD, NPM1, and DNMT3A (Fisher´s exact test, P < .0001). (D) GPR56 mRNA expression in AML samples at diagnosis in different genetic groups. Box and whiskers plot (Tukey) showing reads per kilobase per million mapped reads (RPKM, transformed as log10[RPKM + 0.001]) values for GPR56 mRNA based on RNA-Seq data in genetic groups (n = 388). Median RPKM for GPR56: 3.1 (WT), 6.6 (F), 12.47 (NF), 31.19 (DNF). P values provided in supplemental Table 2. (E, left) Box and whiskers plot (Tukey) showing GPR56 mRNA expression levels (transformed as log10(RPKM + 0.001) in FLT3-ITD-mutated patients with a mutant allele frequency > 0.5 (n = 34) and < 0.5 (n = 87). Medians were 9.25 vs 33.79 (P = .0005, Mann-Whitney U). (Right) FLT3-ITD mutant allele frequencies in samples with GPR56 mRNA expression above (n = 60) or below or equal to (n = 61) the median (RPKM 16). Allelic frequencies were 0.38 vs 0.55 (P < .0001, unpaired Student t test). (F) FLT3-ITD allelic frequencies in AML samples at diagnosis. Symbols represent individual samples, bars show average mutant allele frequencies in FLT3-ITD mutated patients with no mutation in NPM1 or DNMT3A (F, n = 34), with DNMT3A mutation and NPM1 WT (DF, n = 7), with NPM1 but no DNMT3A mutation (NF, n = 26), with NPM1 and DNMT3A mutation (DNF, n = 42). Average frequencies were 0.36 (F), 0.42 (DF), 0.41 (NF), and 0.48 (DNF; P = .018 for F vs DNF, unpaired Student t test). Of note, 24% of triple-mutated samples had FLT3-ITD allele frequencies above 0.5, suggesting that loss of heterozygosity had occurred vs only 8% in FLT3-ITD with nonmutated NPM1 and DNMT3A. (G) Numbers of samples with high LSC frequency in triple-mutated AML vs other groups. Five of 11 triple-mutated AML samples were categorized as “LSChigh,” defined by an LSC frequency greater than 1:30 000 cells compared with only 2 of 27 in the remaining samples (adapted from Pabst et al²⁶ ; P = .0077; odds ratio 6.67, χ² test). D, D mutated, NPM1wt, no FLT3-ITD; DF, D and F mutation, NPM1wt; F, FLT3-ITD, D and N not mutated; N, NPM1 mutated, no F or D mutation; ND, N and D mutation, no FLT3-ITD; NF, N and F mutation, D wt, DNF:triple-mutated; WT, not mutated for NPM1 (N) or DNMT3A (D) and no FLT3-ITD (F).