AML, NOS and AML-MRC as defined by multilineage dysplasia share a common mutation pattern which is distinct from AML-MRC as defined by MDS-related cytogenetics

Acute myeloid leukemia (AML) is a genetically complex disease. Many different driver mutations are known to cause development of AML, moreover genetic and genomic features are the most important prognostic factors for overall survival [1]. In addition to cytogenetics, which have been pivotal in AML diagnostics and classi ﬁ cation for many years, molecular genetics are becoming more important in everyday diagnostic routine. According to WHO 2017 guidelines, AML cases without prior history of cytotoxic treatment are classi ﬁ ed into subgroups with recurrent genetic abnormalities [2]. If these are not present, AML are either classi ﬁ ed into the subgroup of AML with myelodysplasia-related changes (AML-MRC) or AML, not otherwise speci ﬁ ed (AML, NOS). AML-MRC comprises patients with a history of myelodysplastic syndrome (MDS) or myelodysplastic/myelo- proliferative neoplasm (MDS/MPN), or with multilineage myelo-dysplasia (MLD), or speci ﬁ c cytogenetic abnormalities. AML, NOS is subclassi ﬁ ed based on morphological and cytochemical/immuno-phenotypic features of leukemic cells, indicating the lineages involved and their degree of maturation. Since genetic aberrations are causal for development of AML, the classi ﬁ cation on basis of speci ﬁ c gene mutations has the

were not classifiable according to WHO 2017 due to missing/ insufficient bone marrow material or missing NPM1 and/or CEBPA analyses in absence of recurrent genetic aberrations (Fig. 1A). Patients with fusion genes not defined by WHO were grouped in an additional "other rare rearrangements" group (details in Supplementary Table 1), since we aimed at analysing AML, NOS patients without any specific cytogenetic features. As RUNX1 mutated AML is a provisional entity, RUNX1 mutations were not used for classification and thus, according to the remaining classification criteria, all RUNX1-mutated cases were classified as AML, NOS. In summary, 273 cases (12%) were grouped into AML with recurrent genetic rearrangements, 476 cases (22%) into AML with gene mutations (NPM1 or biallelic CEBPA), and 90 cases (4%) into other rare rearrangements. The remaining 1240 cases (cohort "AML, NOS and AML-MRC", n = 1240, Fig. 1A) consisted of 497 (23%) cytogenetically-defined AML-MRC cases (AML-MRC-C), 78 (4%), AML-MRC defined by multilineage dysplasia (AML-MRC-MLD), and 665 patients (30%) with AML, NOS (detailed patient characteristics in Supplementary Table 1). The most prevalent karyotype in AML, NOS was a normal karyotype (70%), the most frequent chromosomal aberration was +8 (12%, Supplementary  Fig. 1A, B).
In AML, NOS, 67% of patients showed at least one MR mutation (Fig. 1B), in AML-MRC-MLD 71% of cases, respectively. In contrast, only 39% of patients in AML-MRC-C had one or more MR mutations (chi-square test, p < 0.001). Two or more MR mutations were found in 35% of AML, NOS, 40% of AML-MRC-MLD, but in only 8% of AML-MRC-C cases. The number of MR mutations per patient was significantly different between AML, NOS or AML-MRC-MLD on the one side and AML-MRC-C on the other side, but not between AML, NOS and AML-MRC-MLD (Welch ANOVA, Games-Howell post hoc test, p < 0.001): AML, NOS and AML-MRC-MLD cases showed a mean of 1.19 (median 1) and 1.27 (median 1) MR mutations per patient, whereas AML-MRC-C cases presented with a mean of 0.52 (median 0) MR mutations.
We found a substantial part of AML, NOS and AML-MRC-MLD harbouring MR mutations (67% and 71%). A differentiation between AML, NOS cases with and without MR mutations has not yet been included in diagnostic and prognostic routines. Following WHO 2017, patients in AML, NOS are further subclassified according to morphology. These subtypes as well as subclassification according to the French-American-British (FAB) system did not show prognostic relevance [2,6]. Likewise, the presence of MLD alone has no prognostic significance in AML-MRC [7].
Several studies have demonstrated that patients with de novo AML and MR mutations share the unfavourable clinical and prognostic features of AML progressed from MDS or MDS/MPN. Lindsley et al. showed these patients to be less responsive to chemotherapy [5]. Papaemmanuil et al. confirmed lower survival rates of patients with mutations of chromatin-spliceosome genes (e.g. SRSF2, ASXL1, STAG2) and found higher relapse rates compared to other subgroups [1]. A lower response rate and worse survival was corroborated for patients with mutations in SRSF2, U2AF1, and ASXL1 by Gao et al. [8]. In line with these   findings, Baer et al. identified a mutational pattern including SRSF2, U2AF1, SF3B1, ASXL1, EZH2, BCOR, and STAG2 that allowed to distinguish AML-MRC from non-AML-MRC patients. Overall survival (OS) was inferior in patients harbouring these mutations, irrespective of whether they were classified as AML-MRC or not [9]. No detailed survival data was available for our consecutive and unselected study cohort. We therefore evaluated a second retrospectively assembled cohort (Supplementary Results, Supplementary Fig. 4). AML-MRC harbouring MR gene mutations and AML, NOS harbouring MR gene mutations showed a comparable OS, hinting at a shared characteristic of AML biology.
We found a high prevalence of MR mutations in AML, NOS. These were described to be specific for AML arising from a previous myeloid malignancy and result in a similar prognosis as that of AML-MRC patients. We believe that the presence of MR mutations reflects an underlying biology of AML. Therefore, we suggest to include an additional subgroup in AML-MRC, which is defined by molecular genetics, harbouring MR mutations (AML-MRC-M). Since prognostic relevance of MLD, if it is the only criterion for inclusion into AML-MRC, has not been shown, we suggest omitting cases with only MLD from the AML-MRC group. These cases would instead be classified as AML-MRC-M in presence of MR mutations and otherwise as AML, NOS. In our cohort, 50 out of 70 (71%) AML-MRC-MLD cases had MR mutations and would be re-classified to AML-MRC-M. The remaining 20 cases (29%) would become AML, NOS. In AML, NOS 251 out of 372 (67%) cases had MR mutations and would be re-classified to AML-MRC-M. The remaining 121 cases (33%) would remain AML, NOS. This would reduce the group of AML, NOS from originally 372 patients to 141 (38%, Supplementary Fig. 5).
In conclusion, patients with mutations in SRSF2, SF3B1, U2AF1, ZRSR2, ASXL1, EZH2, BCOR, or STAG2 constitute a substantial proportion of patients diagnosed with AML, NOS or with AML-MRC-MLD. We suggest classifying cases with these mutations as a new subgroup of AML-MRC, likely reflecting the distinct biology of AML entities more accurately.