Clinical and molecular characteristics of AML patients with respect to sex

Our analysis of pretreatment characteristics in the CALGB/Alliance cohort revealed that female patients tended to be younger (median age, 51 vs 54 years, P = 0.06), and had higher white blood cell counts (median, 26.0 vs 22.2 × 109/L, P = 0.009) and percentages of bone marrow blasts (median, 68% vs 65%, P = 0.04). Females had also more often cytogenetically normal AML (CN-AML; 52% vs 42%, P < 0.001), and less often complex karyotype (8% vs 12%, P = 0.005; Supplementary Table 1). Mutational analysis revealed that female patients harbored more often DNMT3A (P < 0.001), NPM1 (P < 0.001) and WT1 (P = 0.02) mutations as well as FLT3-ITD (P = 0.03), and less often ASXL1 (P < 0.001), SRSF2 (P < 0.001), U2AF1 (P = 0.001), RUNX1 (P = 0.04), or KIT (P = 0.05) mutations (Table 1). Notably, all aforementioned genes are located in autosomal chromosomes. We also observed sex-associated differences in the frequencies of mutations in genes categorized into the major AML-associated functional groups, that included female patients having a higher frequency of mutations in methylation-related genes (51% vs 46%, P = 0.02) and male patients having a higher frequency of mutations in genes involving chromatin remodeling (23% vs 18%, P = 0.02), spliceosome (22% vs 11%, P < 0.001), and, by trend, transcription factors (28% vs 24%, P = 0.06); men also harbored myelodysplasia-related gene mutations more often than women (37% vs 27%, P < 0.001; Table 2).

Table 1 Frequencies of mutations in single genes in female and male patients with AML treated on the CALGB/Alliance frontline protocols and those treated on the German AML Cooperative Group frontline protocols.Table 2 Frequencies of gene mutations arranged in functional groups in female and male patients with AML treated on the CALGB/Alliance frontline protocols and those treated on the German AML Cooperative Group frontline protocols.

To validate the observed sex-associated differences, we compared the frequencies of cytogenetic findings (Supplementary Table 1) and gene mutations in patients from AMLCG (Tables 1 and 2). The results were largely concordant, with CN-AML and all mutations, except KIT and WT1 mutations, which differed between males and females among CALGB/Alliance patients being also significantly different in the AMLCG cohort. However, in the latter patient population, females also less often carried EZH2 (P = 0.005), SMC1A (P = 0.003) and STAG2 (P < 0.001) mutations and mutations in the cohesion complex genes (P < 0.001) than males.

The proportions of patients assigned to genetic-risk groups in the 2022 ELN classification [1] also differed between sexes. In the CALGB/Alliance cohort, a higher percentage of female patients was categorized in the intermediate-risk group (30% vs 20%) and a lower percentage of women was included in the adverse-risk group (31% vs 44%, Supplementary Table 1).

Sex-associated differences in co-occurring molecular alterations

We next compared patterns of co-occurring recurrent mutations in male and female patients and differences between sexes (Pearson correlation <0.2 and P < 0.0002 for one sex and Pearson correlation >0.03 and P > 0.5). We identified four gene mutation pairs, including NRAS/BCOR, U2AF1/CBL, NOTCH1/BCORL1, and SF1/SRSF2 whose presence was strongly associated with female patients. Conversely, ASXL1/TET2, NF1/SF3B1, RUNX1/PHF6, RUNX1/U2AF1 and RUNX1/STAG2 were strongly positively associated with male patients. Conversely, WT1/SRSF2 tended to be mutually exclusive in males but less so in females (Fig. 1).

Fig. 1: Differences in co-existing molecular features between female and male patients with AML.

Triangle correlation plot depicting co-occurring mutations in A female and B male patients with AML treated on CALGB/Alliance frontline protocols. All squares denoting gene mutation pairs that differ between the sexes are marked using stars.

Treatment outcome of adults with AML aged <60 years with respect to patient sex

We analyzed clinical outcomes of 381 female and 463 male AML patients aged 17–59 years in the CALGB/Alliance cohort, for whom frontline treatment with intensive induction chemotherapy is currently still a standard of care. We found no significant differences in CR and early death rates, DFS or OS between female and male AML patients (Supplementary Table 2). There were also no differences between sexes in CR and early death rates or DFS in the AMLCG cohort. However, male German patients aged 18–59 years had a shorter OS than female patients (5-year rates, 42% vs 51%, P = 0.005).

Sex-associated prognostic impact of genetic alterations in patients aged < 60 years

Since cytogenetic and molecular genetic findings are routinely used for risk stratification of AML patients, we assessed sex-specific outcomes based on the 2022 ELN genetic-risk categories [1]. Despite differences in proportions of females and males in the intermediate- and adverse-risk groups, DFS and OS was essentially equal within each genetic-risk group, except for longer OS of female adverse-risk patients (P < 0.001; Supplementary Fig. 1A, B).

Next, we assessed associations of recurrent genetic alterations with outcome of CALGB/Alliance patients. In univariable analysis (UVA) for CR, presence of FLT3-ITD, PTPN11 and RUNX1 mutations affected outcome of only female patients (Fig. 2A), of which FLT3-ITD and PTPN11 mutations remained significantly associated with CR achievement also in subsequent multivariable analysis (MVA, Supplementary Table 3). In contrast, a normal karyotype and SF3B1 mutations were adverse predictors only in male patients in both UVA and MVA.

Fig. 2: Outcomes of patients with AML aged <60 years who were treated on the CALGB/Alliance study protocols.

Forest plot illustrating univariable analyses of A complete remission, B disease-free survival and C overall survival. Depicted in the plot are all gene mutations with sex-specific survival impact for either male or female patients in univariable analyses for the respective outcome endpoint. Results of the corresponding multivariable analyses and markers with significance are shown in Supplementary Table 3.

For DFS in female patients, DNMT3A and WT1 mutations associated with shorter, and FLT3-TKD mutations with longer DFS in both UVA and MVA (Fig. 2B, Supplementary Fig. 2A). While for male patients detection of ASXL1, CEBPAbZIP, NRAS, SRSF2 and TP53 mutations had sex-specific survival impact in UVA, only CEBPAbZIP mutations were significant in multivariable models.

In univariable analysis of OS, FLT3-TKD, DNMT3A and WT1 mutations affected outcome of only female patients (Fig. 2C) of which, again, WT1 mutations held their significance also in multivariable analyses (Fig. 3A). Notably, the adverse outcome impact of female WT1-mutated patients seemed to be driven by WT1/NPM1 co-mutated female patients (Fig. 3B, C). OS of men only was influenced by CEBPAbZIP, GATA2, KIT, NRAS, SF3B1, SRSF2, STAG2 and U2AF1 mutations in UVA, of which SF3B1 mutations had again sex-specific survival association in MVA, too (Fig. 3D, Supplementary Table 3).

Fig. 3: Overall survival of female and male patients with AML according to their WT1, NPM1 and SF3B1 mutation status.

A Overall survival of female and male patients with and without WT1 mutations and overall survival of B female and C male patients according to their NPM1 and WT1 mutation status. D Overall survival of female and male patients with and without SF3B1 mutations. E Overall survival of female and male patients classified in the 2022 ELN adverse-risk group because of the presence of SF3B1 mutation (and lack of favorable-risk genetic markers) and of the remaining female and male patients classified in the 2022 ELN adverse-risk group who did not have SF3B1 mutation.

Focusing on those sex-specific molecular alterations that held their significance in both UV and MV models and at least 2 survival endpoints (female: WT1 mutations [DFS, OS]; male: SF3B1 mutations [CR, OS]), we tested their outcome impacts in the AMLCG cohort.

While WT1 mutations associated with a trend for inferior DFS in female AMLCG patients, there was no association with inferior OS, possibly suggestive of a rescue effect of more frequently used intensive consolidation including allogeneic HSCT which is more commonly administered in Germany (Supplementary Fig. 2B, C). Although limited by relatively small sample sizes, SF3B1-mutated male patients in the AMLCG cohort also had a lower CR rate (38% vs 72%, P = 0.04) and tended to have inferior OS (Supplementary Fig. 3).

As SF3B1 mutations demonstrated distinct prognostic differences and were recently added to the 2022 ELN classification as adverse-risk outcome prognosticator (in the absence of favorable-risk genetic markers), we compared OS of adverse-risk female and male patients with SF3B1 mutations and OS of the remaining adverse-risk female and male patients (i.e., without SF3B1 mutations). While male SF3B1-mutated 2022 ELN adverse-risk patients and both female and male adverse-risk patients without SF3B1 mutations had similarly poor overall survival, female SF3B1-mutated adverse-risk patients had significantly longer OS (5-year rates, 43% vs 8%, P = 0.01) than other 2022 ELN adverse-risk patients (Fig. 3E).

Treatment outcome of adults with AML aged ≥60 years with respect to patient sex and the sex-associated prognostic impact of genetic alterations

Given the generally poor survival of older patients with AML, we performed a subset outcome analyses of the AML patients aged 60 years and older included in our study. We did not find significant differences between sexes among 524 older patients treated on the CALGB/Alliance protocols nor among 414 older German patients (Supplementary Table 4). There were no significant differences in DFS or OS between female and male patients in any of the 2022 ELN genetic-risk groups, with patients in the intermediate-risk and adverse-risk groups performing similarly poorly (Supplementary Fig. 4A, B). Although drawing definitive conclusions from the multivariable analyses is difficult because of the generally poor treatment outcome of older patients with AML, we found that such mutations included in the 2022 ELN genetic-risk classification as FLT3-ITD, NPM1 and TP53 mutations were among main factors affecting CR rates and survival (Supplementary Table 5). Only in men, DFS was also negatively affected by PTPN11 mutations and OS by KRAS mutations.

Sex-specific alternative splicing events and lineage-associated gene expression patterns

To investigate the potential role of sex bias on gene and splicing programs, we next analyzed patients with existing RNAseq data (n = 848) to quantify alternative splicing events (ASE) and differentially expressed genes (DEGs) associated with recurrent gene mutations or cytogenetic findings using the AltAnalyze workflow, including those affecting spliceosome genes and other recurrent AML-associated genes or cytogenetic features (n = 24) and select clinical/patient demographic variables.

To set the stage, we first determined the spectrum of mutations, which result in common splicing impacts via supervised and unsupervised analyses irrespective of sex, which could be verified in independent AML cohorts (Supplementary Methods and Supplementary Fig. 5A, B). Most AML-associated mutations and chromosome rearrangements resulted in splicing impacts, most significantly associated with splicing-factor mutations (SRSF2-P95*, U2AF1-S34*, SF3B1, SF3A1, U2AF1-Q157*, SF1), common mutations (NPM1, TP53), CN-AML, complex karyotype, and chromosome rearrangements resulting in gene fusions involving CBFB and KMT2A (Fig. 4A, B).

Fig. 4: Alternative splicing events (ASEs) segregated by mechanism of predicted regulation for recurrent mutations, gene fusions and clinical covariates in the CALGB/Alliance RNA-Seq cohort.

The extent of exon/intron A inclusion or B exclusion ASEs. The extent of unique sex-associated ASEs C or differentially expressed genes (DEGs) D associated with each patient subtype, unique regulated events or genes compared with all other AMLs and sex/subtype overlapping events/genes, derived from the software AltAnalyze. Green color denotes the number of unique ASEs or DEGs in all patients with the indicated molecular genetic or cytogenetic subtype versus all other AML patients in the cohort. Blue indicates the number of unique ASEs or DEGs in males versus females in the indicated AML subtype and red illustrates overlap between subtype-associated ASEs or DEGs and male versus female (subtype regulated events/genes that vary according to patient sex). E Quantification of SF3B1-mutated sex-associated alternative splicing events as percent spliced in (PSI) values for the indicated parental exon-exon junction. F = female, M = Male. F Heatmap of gene set enrichments (GO-Elite) for known AML subtypes against human bone marrow cell-type markers to identify lineage skewing, BM cell-type markers were derived from extensive prior human single-cell analyses [43], to identify markers of rare hemopoietic stem cells, progenitors, and immune cell-types. Only AML subtype sex-associated gene sets with significant enrichments are shown. G Sex-associated differentially expressed genes in SF3B1-mutated males versus females corresponding to human bone marrow progenitor populations.

Next, we compared female and male patients in each of these defined subtypes. We considered subtypes with >17 samples and an unadjusted P-value for these comparisons. Interestingly, we found different magnitudes of splicing and gene-expression impacts associated with sex in AML subtypes, with generally greater impacts associated with ASEs versus DEGs (Fig. 4C, D). In general, splicing-factor mutations (SRSF2-P95*, U2AF1-S34*, SF3B1, SF3A1, SF3A1, SF1) resulted in large sex-associated differences for both ASEs and DEGs. Notably, while it has been extensively documented that SF3B1-K700E and functionally similar mutations result in alternative splice sites only selected for within the presence of these mutations, examination of SF3B1-K700E and related mutation ASEs predicted sex bias for several of previously annotated (USP25, MYO15B) and novel SF3B1-K700E selected cryptic splice sites (AL133352.1, CNTRL, DOCK10), in addition to non-cryptic splice sites with clear sex bias (Fig. 4E). At the pathway level, we found that SF3B1 mutations in male patients transcriptionally induce genes associated with integrin cell-surface receptor-linked signaling pathways (MAP kinase, Netrin, PECAM1, Ephrin B, interferon alpha/beta signaling, L1CAM), whereas RUNX1- and TP53-mutated male patients induce inflammatory response pathways (IL12, FGF, CXCR4 and TCR signaling). Further, complex karyotype and TP53 mutations in female patients induce a distinct set of inflammatory and immune signaling programs (Supplementary Fig. 5C, D). When we considered marker genes for hematopoietic human progenitor and differentiated cell populations, we noted distinct lineage enrichments shared by different sets of mutations. In particular, SF3B1 (females) and KRAS (males) were enriched in hematopoietic stem cells, MultiLin, common lymphoid precursor and eosinophil/mast programs, whereas SF3B1 (males) and WT1 (females) induced a dominant megakaryocytic progenitors/erythroid progenitors (Fig. 4F). We note that in male SF3B1-mutated patients, genes associated with these programs represent well-defined lineage determining factors, such as HOPX and PROM1 for hematopoietic stem cells, ITGA2B, VWF and PF4 for megakaryocytic progenitors and EPOR and GATA1 for erythroid progenitors (Fig. 4G). Together, these data indicate that sex contributes to both alternative splicing and transcriptional differences among patients with splicing factor and other dominant AML mutations.