Acute myeloid leukemia (AML) includes a heterogeneous group of aggressive, neoplastic proliferations of immature myeloid cells (myeloblasts or blast equivalents) that typically expand within the bone marrow space leading to eventual hematologic compromise and, if untreated, can lead to death. In the United States, in 2019, AML accounted for 1.0% of all new cancer cases at an incidence of ∼4.1/100,000 persons, with ∼20,050 new cases, 11,540 deaths, and a 5-year relative survival rate of 30.5%. The median age at diagnosis is 68, with 59.4% of new cases and 74.1% of deaths occurring in patients >65 years old.1

With increasing revelation of the genomic landscape of AML, categorization of AML has moved gradually from a morphologic and cytochemical-based system to one constructed on common underlying genetic alterations.2,3 Thanks to the early advances in cytogenetics, AML classification systems began incorporating recurrent cytogenetic abnormalities in the classification of AML (e.g., RUNX1::RUNX1T1, PML::RARA, MYH11::CBFB).4 Subsequently, identification of recurring mutations in key genes, NPM1, FLT3, and CEBPA were utilized in prognostic algorithms, including the European LeukemiaNet and the National Cancer Center Network Guidelines for AML.5, 6, 7 Indeed, among genetic, clinical, and diagnostic variables, genetic alterations have been shown to be the most predictive of overall and relapse free survival.8

In recent years, due to the advent of next generation sequencing techniques, the World Health Organization (WHO) introduced three subdivisions of AML with gene mutations in the 4th edition of the Classification of Tumours of Haematopoietic and Lymphoid Tissue (WHO HAEM) and its subsequent revision: AML with mutated NPM1, AML with biallelic mutation of CEBPA, and provisionally AML with RUNX1 (Table 1).4 The WHO HAEM5 and particularly the newly released system by the authors of the International Consensus Classification (ICC) System have expanded the role of genetics in defining these diseases.9, 10, 11 This review aims to discuss the recent advances in the understanding of the molecular underpinnings of AML, and, where appropriate, compare the new classification systems with respect to the incorporation of molecular genetics into the definition of AML.

Approximately 50% of cases of AML have a normal karyotype, and on the whole, as compared to other malignancies, AML is reported to have a small number of genetic mutations per case (∼13 mutations overall, ∼5 recurrent in AML) with over seventy unique driver mutations having been identified.8,12,13 Most (97%) of AML have at least 1 driver mutation.14 The majority of mutations found in AML are believed to have arisen randomly in hematopoietic stem cells prior to the acquisition of a driver mutation, explaining why these mutations are typically unanimously present in AML blasts at diagnosis.15

More than two decades ago, Gilliland et al. proposed a 2-hit model for the development of AML, involving "class I" mutations in signaling molecules which confer a proliferative or survival advantage and "class II" mutations involving transcription factors that interfere with differentiation and apoptosis of cells.16 To this end, some of the earliest acquired mutations in AML are found in epigenetic modifiers/chromatin remodelers, including DNMT3A, ASXL1, IDH1/2, and TET2, and RNA splicing genes, such as SF3B1 and SRSF2, which are often found in age-related clonal hematopoiesis of indeterminate potential (CHIP), and appear insufficient for the generation of overt leukemia.8,14,17, 18, 19, 20 Further supporting this model, IDH2, DNMT3A, ASXL1, IKZF1, and CBFB::MYH11 alterations have all been identified in the hematopoietic stem cell compartment and help mutant HSCs outcompete the normal counterparts.18 These “preleukemic” clones may survive induction chemotherapy and be identified on minimal residual disease assays.18 In contrast, the “class I” mutations, such as tyrosine kinases (FLT3, KIT) RAS pathway mutations, and PTPN11 appear to occur late in disease development.8,21,22,14 Nevertheless, some acute leukemias, such as those with KMT2A or PML-RARA translocations typically show relatively few cooperating mutations, and may require a shorter path to leukemogenesis.13

Much work has been done analyzing the ∼50% of patients with cytogenetically normal AML, with mutations in a handful of genes being recognized as conferring prognostic information. Roughly 31% of normal karyotype (NK) AML harbor internal tandem duplications (ITD) and 11% harbor tyrosine kinase domain (TKD) mutations of the fms-related tyrosine kinase 3 gene (FLT3), with these mutations conferring a shorter overall survival and indicating a need for therapy with FLT3 inhibitors.5,23, 24, 25 Nucleophosmin 1 (NPM1), a protein with nuclear to cytoplasmic shuttling properties that regulates the alternate-reading-frame protein (ARF)-p53 tumor suppressor pathway, is mutated in 50-60% of NK AMLs, and in the absence of FLT3 mutations, is associated with a more favorable response to therapy.5,26 Similarly, mutations in CCAAT enhancer binding protein A (CEBPA) gene are present in 13% of NK AML and are associated with a more favorable prognosis.5,27,28 Partial tandem duplications (PTD) of KMT2A (MLL) occur in ∼12% of NK AML and have an increased likelihood of relapse and poor survival.29, 30, 31

Within the past decade, and after the first iteration of the WHO HAEM4, multiple groups performed sequencing analysis of AML, gaining insight into the breadth of the disease's mutational background, and allowing construction of mutation-associated clustering of AML. The Cancer Genome Atlas Research Network sequenced (via whole genome or exome sequencing) 200 paired de novo adult AML cases, and grouped the unique, recurrent mutated genes into nine ontological categories. The categories include: transcription factor fusions (18%, PML-RARA, MYH11-CBFB, RUNX1-RUNX1T1, PICALM-MLLT10); mutated NPM1 (27%); tumor suppressor genes (16%, TP53, WT1, PHF6); DNA methylation genes (44%, DNMT3A, DNMT3B, DNMT1, TET1, TET2, IDH1, IDH2); activating signaling genes (59%, FLT3, KIT, other Tyr-kinases, Ser-Thr kinases, KRAS/NRAS, PTPs); myeloid transcription-factor genes (22%, RUNX1, CEBPA, other myeloid transcription factors); chromatin-modifying genes (30%, MLL-X fusions, MLL-PTD, NUP98-NSD1, ASXL1, EZH2, KDM6A, and other modifiers); cohesin-complex genes (13%), and spliceosome-complex genes (14%).13

Patel et al analyzed the mutation status of 18 genes in 398 AML patients younger than 60-years-old, and identified co-occurring and mutually exclusive mutation groups.25 Co-occurring mutation groups include: KIT mutations with core-binding-factor cytogenetic abnormalities, IDH1 and IDH2 mutations with NPM1 mutations, and DNMT3A mutations with NPM1, FLT3, and IDH1. Mutually exclusive mutations include IDH1 and IDH2 with TET2 and WT1 and DNMT3A with MLL translocations.25,32 With respect to mutational prognostic significance, FLT3 ITD, PHF6 and ASXL1 mutations correlated with reduced overall survival, TET2 mutations correlated with reduced overall survival for intermediate-risk AML, while IDH2 (R140Q) mutations displayed better overall survival. KIT mutations were associated with reduced overall survival with t(8;21) AML but not inv(16)/t(16;16).25

Later, using a cohort of 1540 patients, Papaemmanuil et al. found that 48% of AML patients did not meet the molecularly-defined AML categories of the WHO HAEM4, and therefore used a Bayesian statistical model to identify 14 AML sub-groups based on the ontological categories of the mutations.8 The categories included: AML with NPM1 mutation (27%); AML with mutated chromatin-spliceosome with mutations in chromatin genes (ASXL1, STAG2, BCOR, MLLPTD, EZH2, and PHF6), splicing genes (SRSF2, SF3B1, U2AF1, and ZRSR2), and/or transcription genes (RUNX1) (18%); AML with TP53 mutations, chromosomal aneuploidy, or both (13%); AML with CBFB-MYH11 (5%); AML with biallelic CEBPA mutations (4%); AML with PML-RARA (4%); AML with RUNX1-RUNX1T1 (4%); AML with MLL fusion genes (3%); AML with GATA2, MECOM (1%); AML with IDHR172 mutations and no other class-defining lesions (1%); AML with DEK-NUP214 (1%); AML with driver mutations but no detected class-defining lesions (11%); AML with no detected driver mutations (4%); and AML meeting criteria for ≥2 genomic subgroups (4%).8 Importantly, the number of driver mutations was found to correlate with overall survival, independent of age or white blood cell count, while in certain cases, the mutational background influenced the prognostic impact of subsequent mutations.8TP53, chromatin-spliceosome genes (e.g., SRSF2 and ASXL1), and BRAF were independently associated with worse outcome.8

In 2022, two classification systems were proposed as modifications of the prior WHO HAEM4 system. First, the WHO HAEM5 revised the diagnostic criteria for AML in certain instances, attempting to retreat from arbitrary blast cutoffs of the past by removing the blast threshold criteria for most AML subtypes with defining genetic abnormalities, under the condition that these genetic abnormalities must drive disease pathology (table 1).9 The WHO additionally adds new subtypes based on recurrent cytogenetic abnormalities, as well as a category for rarer translocations, AML with other defined genetic alterations, for provisional entities that may eventually become officially endorsed AML subtypes.

While the ICC does not go so far as to remove blast cutoffs altogether, they have revised the diagnosis of MDS with excess blasts, type II to a combined category of MDS/AML, underscoring the biologic spectrum of these diseases. However, a 10% lower blast threshold for diagnosis of AML is required for most AML subtypes with recurrent genetic and/or molecular defining abnormalities, potentially putting the WHO and ICC in opposition to each other for rare cases with <10% blasts.10 Similar to the WHO, the ICC adds additional recurrent genetic abnormalities, including "AML with other rare recurring translocation". The ICC classification regroups a number of AMLs based on the underlying molecular alterations, converting prior “therapy-related” and “myelodysplasia-related” categories into modifiers rather than AML subtypes (Tables 1 and 2).11 The use of diagnostic qualifiers is shared by the WHO, which also converts "therapy-related myeloid neoplasms" to a qualifier of "post cytotoxic therapy".9

As stated above, the WHO HAEM5 has removed the 20% blast count requirement for all AML types with defining genetic abnormalities, with two exceptions (AML with BCR::ABL1 and AML with CEBPA mutation).9 This move is in contrast to the ICC system which lowers the blast threshold for most AMLs with definitional genetic lesions to 10% (excluding AML with BCR::ABL1 and AML with mutated TP53).10 From the perspective of the WHO, the expectation is that future cutoffs may be better informed by the size of the clone which, in turn, would be estimated by the variant allele frequency (VAF) of the clone. With respect to the ICC, having a blast count of 10% provides some diagnostic guidance and overcomes most discrepancies in sensitivities of molecular assays until more data regarding VAFs is obtained.

Both the WHO and the ICC classification systems both use specific gene mutations to define AML. The WHO and ICC both defined AML categories by mutations in NPM1 and CEBPA. The ICC additionally adds AML with mutated TP53. Finally, both classification systems have added a list of genes that can be used to define a subset of AML with myelodysplasia-related abnormalities. A discussion of these specific genes and AML types follows.

NPM1 is a multi-domain containing nucleolar phosphoprotein that shuttles between the nucleus and cytoplasm to carry out its many cellular functions, including proliferation, growth suppression, regulation of the ARF-p53 tumor suppressor pathway/stress response, ribosome biogenesis, and genetic stability.33, 34, 35 The diverse array of functions carried out by NPM1 explains why the lack of NPM1 is embryonically lethal, and why both overexpression of and loss of NPM1 in cells is associated with a variety of cancers.34,36 Additionally, translocations that disrupt the normal cellular trafficking of NPM1 are also associated with a variety of hematologic malignancies, including ALK+ anaplastic large cell lymphoma with NPM1::ALK, acute promyelocytic leukemia with NPM1::RARA, and AML with NPM1::MLF1.37

Wild-type NPM1 possesses an N-terminal oligomerization core with two weak nuclear export signals, followed by an acidic central region (containing a nuclear localization domain), a C-terminal basic region that binds TP53 and nucleic acids, and a C-terminal aromatic domain, with two conserved tryptophan residues (W288 and W290) that mediate nucleolar localization.38 Upon nucleolar stress, NPM1 migrates from the nucleolus to the nucleoplasm, where it binds HDM2. HDM2 functions as an E3 ligase, that serves to degrade p53, and its inhibition results in increased p53 protein and subsequent promotion of apoptosis.38

In addition to 50-60% of NK AML, mutated NPM1 is present in over 30% of adult patients and 8% of children with de novo AML, and thus the WHO HAEM4 introduced the provisional entity, AML with mutated NPM1, and confirmed the entity in the 2017 revision.4,26,39,40 Heterozygous mutations generally affect exon 12 of NPM1, resulting in frame-shifts of the C-terminus.26,41 The C-terminal mutations tend to cause loss of W288 and W290, and the effective formation of a nuclear export signal gives rise to cytoplasmic localization of mutant NPM1.26,38 Because NPM1 functions as an oligomer, mutant NPM1 protein can oligomerize with wild-type protein, resulting in redistribution of the non-mutated protein to the cytoplasm as well, and thus mutant NPM1 potentially exerts dominant negative activity on the nuclear functions of NPM1.42

The morphology of blasts in AML with mutated NPM1 ranges from myeloblastic to myelomonocytic or monoblastic/monocytic, while background hematopoiesis can be dysplastic (Fig 1A-C).43, 44, 45 NPM1 mutated AML is weakly associated with a cup-like morphology, whereas co-mutation of NPM1 and FLT3-ITD is strongly associated with this trait (Figure 1A).46 Immunohistochemistry may assist in identifying NPM1-mutated cases, either using N-terminally directed antibodies that uncover aberrant cytoplasmic localization of NPM1, or C-terminal mutation-specific antibodies.47,48 With respect to flow cytometry, NPM1-mutated AML shows a variety of immunophenotype expression patterns, including an APL-like pattern (CD34-, HLA-DR-), with the most common immunophenotypic pattern being CD34-, CD15+, and HLA-DR+ (Figure 1D).49 Otherwise, blasts typically express CD117, CD13, CD33 (strong), MPO and CD38, with ∼30% cases showing co-expression of CD7 and CD117, and monocytes not uncommonly showing expression of CD56.49, 50, 51 The median VAF for NPM1 mutations is 0.38, with no correlation between initial VAF and peripheral blood blast count or overall survival.51

NPM1 mutations appear to be secondary events, being conspicuously absent in CHIP, and occurring after mutations in DNMT3A, IDH1, or NRAS on the precipice of the development of AML.8,20,52,53 Indeed, acquisition of novel NPM1 mutations in patients with prior MDS has been shown to be associated with swift development of AML.54,55 On the whole, NPM1 mutations occur less frequently in secondary AML arising from MDS and MDS/MPN (13.3%) than they do in de novo AML, possibly due to the short latency between acquisition and eventual development of AML.54 The mutational background seen in the setting of AML with NPM1 mutations frequently includes mutations in DNMT3A (50-60%), FLT3 (40-50%; FLT3 ITD occurs twice as frequently in AML with mutant NPM1 than in AML with wild-type NPM1), PTPN11 (20-30%), TET2 (20-30%), IDH1/IDH2 (10-20%), and NRASG12/13 (10-20%), and excludes mutations in RUNX1 and TP53 or recurrent translocations of PML::RARA, MYH11-CBFB, and KMT2A.8,13,14,56, 57, 58 Prognostic interactions include the particularly deleterious co-mutation of FLT3 ITD with the combination of DNMT3A and NPM1 co-mutations, whereas NRASG12/13 co-mutated with DNMT3A and NPM1 is more favorable.8 Patients older than 75-years tend to have increased co-mutations of TET2 and SRSF2 and reduced co-mutation of DNMT3A, a finding that is similar to NPM1+ secondary AML, a leukemia that is more likely to have TET2 mutations than DNMT3A mutations, and more frequently harbors mutations in ASXL1 (19%), and RUNX1 (9.5%).54,59

The NPM1 VAF of NPM1+ AML has been shown to positively correlate with WBC, peripheral blood blast percentage, and bone marrow blast percentage, and negatively correlate with platelet count.58,60 There is conflicting data, however, on the effect of a high NPM1 VAF on NPM1+ AML patient survival. Patel et al. showed NPM1 VAF correlated with decreased OS and EFS when stratified at the median (≥0.40 or <0.40) or by quartile (<0.36, 0.36-0.39, 0.40-0.43, and ≥0.44), regardless of patient age or presence of FLT3-ITD co-mutations, and correlated with worse outcomes in patients treated with stem cell transplant in first remission or with co-mutation of DNMT3A.58 Abbas et al., in contrast, showed no significant correlation between NPM1 VAF and overall survival, event free survival, DNMT3A co-mutation, or stem cell transplant in first remission.60 A study by the German AML Cooperative Group confirmed the finding of an association between high NPM1 VAF and overall survival, but attributed this finding largely to the increased marrow blast percentage, FLT3-ITD and/or DNTM3A mutations, with no independent contribution from the NPM1 VAF alone, suggesting high NPM1 VAF is a marker for more proliferative disease.61

Consistent with NPM1 mutation as a secondary event leading to AML, NPM1 mutations are rarer in non-acute myeloid neoplasms, being present in ∼2% of cases of MDS and ∼3% MDS/MPN and are primarily found in chronic myeloid neoplasms with higher blasts counts (median 10%) and aggressive clinical courses.62, 63, 64,50,51,65,66 In a study of 8 CMML patients with NPM1 mutations, 5 of 8 progressed to acute leukemia.66 Two larger studies of MDS or MDS/MPN cases harboring NPM1 mutations (76 cases altogether) revealed these neoplasms occur in comparatively younger patients (median age 62-63), and similar to AML with NPM1 mutations, display normal karyotypes, frequently possess mutations in DNMT3A and PTPN11, and infrequently possess mutations in ASXL1, RUNX1 and TP53 compared to non-acute myeloid neoplasms lacking NPM1 mutations.50,51 Moreover, chronic myeloid neoplasms harboring NPM1 mutations tend to lack mutations in IDH1/IDH2 and FLT3, whereas evidence suggests acquisition of a FLT3 mutation is associated with progression of NPM1+ MDS to AML.50,54,63

Interestingly, NPM1+ chronic myeloid neoplasms have a shorter overall survival compared not only to chronic myeloid neoplasms lacking NPM1 mutations, but also to NPM1+ AML.50 Patel et al. showed that 39% of patients with NPM1+ chronic myeloid neoplasms being treated with hypomethylating agents progressed to AML, whereas patients that received upfront induction chemotherapy showed no evidence of progression.50 Similarly, Montalban-Bravo et al. showed that up front induction chemotherapy ± transplant showed improved overall survival and progression free survival compared to those receiving hypomethylating agents alone.51 On a related note, for patients >60 years old, treating NPM1+ AML with hypomethylating agents alone blunts the survival advantage of this mutation, whereas induction therapy or the addition of venetoclax improves outcomes, indicating NPM1 mutations may serve as a marker for patients that benefit from induction chemotherapy.67, 68, 69, 70, 71

Because NPM1+ chronic myeloid neoplasms tend to have an aggressive clinical course and appear to respond better with induction chemotherapy, and because there is a similar mutational backdrop for these neoplasms compared to NPM1+ AML, both the WHO and ICC have converted the detection of an NPM1 mutation to a leukemia-defining event.9, 10, 11 The ICC has reduced the blast threshold for such AMLs to 10%, similar to the ICC's blast threshold for other genetically-defined AMLs. In contrast, the WHO HAEM5 has removed a blast threshold entirely, with the caveat that cases with a VAF <10% should be diagnosed with caution.9

At face value, the discrepancy in diagnostic criteria for AML with mutated NPM1 poses a potential problem for those cases with <10% blasts. While the WHO HAEM5′s removal of a blast count eliminates the subjective nature of morphologic blast assessment, it shifts the burden to molecular assays that can similarly be discrepant across platforms.48 Moreover, there is little data to show that a blast count <10% in myeloid neoplasms harboring NPM1 mutations behave similarly to AML with mutated NPM1. However, the paucity of NPM1 mutations in myeloid neoplasms with <20% blasts and the tendency for these cases to have increased blasts suggests that the diagnostic dilemma of what to do with a small NPM1 clone with an allele fraction near 10% will be a rare event. Nevertheless, further study of the behavior of such cases is necessary to resolve the discrepancy in these two classification systems.

CCAAT enhancer binding protein A gene (CEBPA), located on chromosome 19.q13.1, encodes the transcription factor, C/EBPα, and is required for granulocytic and adipocyte differentiation and glucose metabolism.72 C/EBPα contains two N-terminal transactivation domains (TAD) and a C-terminal basic amino acid-rich DNA-binding domain and a leucine zipper dimerization domain (collectively referred to as b-ZIP).73 C/EBPα is expressed in myelomonocytic cells, but its expression appears to induce granulocytic differentiation genes, and as such, it was a prime suspect for AML-associated mutations causing blocks in granulocytic maturation.27,28

As predicted, CEBPA is mutated in AML (∼5-14%), with nearly 200 identified mutations which are predominantly biallelic.27,74,75 There is no specific cytologic morphology associated with the blasts in AML with biallelic CEBPA mutations (Fig 2). With respect to flow cytometry, AML with biallelic CEBPA mutations tends to have more robust expression of CD34, CD117, and HLA-DR, with aberrant early expression of CD15, CD65, CD64, or MPO (Figure 2).76 Moreover, these AMLs tend to show aberrant bright expression of CD7 on the majority of blasts (Fig. 2C).76 These leukemias also tend to show a block in granulocytic differentiation accompanied by lower SSC, and expanded erythroid progenitors which are often dysplastic.76

Mutation of CEBPA tends to cluster in two regions, with the first mutational hot spot in the bZIP region, typically consisting of in-frame insertion/deletion mutations which presumably result in altered DNA binding or dimerization.77 The second mutation bed is closer to the N-terminus, where frame-shift mutations result in protein truncation, yielding a shorter protein, p30, rather than the wild-type, p42, protein.72,73 The most common co-mutated genes in AML with biallelic CEBPA mutations are CSF3R (19.75%), WT1 (18.52%), and GATA2 (16.05%), with co-mutation of CSF3R or WT1 correlated with decreased relapse-free survival.77 Of note, CEBPA mutations are rare in MDS, and are typically seen as secondary events in AML transformed from MDS; most AML with mutated CEBPA tend not to have evolved from MDS, suggesting CEBPA mutations are leukemogenic.78

AMLs with biallelic mutated CEBPA are predominantly FAB M1/M2, affect mostly younger patients, have a unique gene expression profile, and have a more favorable prognosis.75,77,79, 80, 81 AML with monoallelic mutated CEBPA is more heterogeneous in clinical presentation, except for those with mutations in the bZIP region, which resemble biallelic AML.75 The WHO HAEM4 therefore introduced the category of acute myeloid leukemia with biallelic mutation of CEBPA to account for this more favorable form of AML.4 Nevertheless, a group for the Study Alliance Leukemia as well as Wakita et al. subsequently expanded on the findings that AMLs with monoallelic mutations of CEBPA within the bZIP region display similar features as the biallelic forms, including younger age, higher white blood cell concentration, co-mutations (GATA2 and WT1 and lacking NPM1), and overall survival.82,83 Moreover, Taube et al. further clarified that, regardless of allelic status, in-frame mutations in bZIP accounted for the constellation of clinical findings, while El-Sharkawi, et al. showed that non bZIP biallelic CEBPA AML may have poorer survival with a methylation profile that is different from bZIP biallelic CEBPA AML.82,84

In light of these studies, the ICC has revised the diagnosis of AML with biallelic mutation of CEBPA to AML with in-frame bZIP CEBPA mutations, removing the biallelic requirement, and lowering the blast threshold to 10%.10 In contrast, the WHO HAEM5 maintains the biallelic criteria for CEBPA mutations, with the exception of bZIP mutations, and maintains the blast threshold at 20%.9 As the majority of biallelic mutations are within the bZIP region, the significance of these separate allelic criteria may be minimal, nevertheless, there are potential issues with both classifications systems. With respect to the WHO criteria, there is an implication that biallelic CEBPA mutations outside of the bZIP region would be expected to show a similar clinical outcome; however, this conclusion is not currently supported by recent literature.79,82,84 For the ICC, there is little data showing that myeloid neoplasms with CEPBA mutations in the bZIP region and <20% blasts behave similarly to those with at least 20%. In light of the difference in blast criteria, it may be important to design clinical trials that navigate the differences in response criteria established by the ELN 2022 for AML and the International Working Group response criteria in myelodysplasia.6,85,86

The tumor suppressor gene, tumor protein 53 (TP53), is located on chromosome 17p13, and is widely mutated in a number of cancers.87(p53)TP53 encodes p53 which contains a DNA-binding domain, a transcription activation domain, a proline-rich domain, and a tetramerization domain, and functions to counteract cellular stressors.87,88 Under steady state conditions, p53 is rapidly degraded; however, in the setting of cellular stress, p53 may initiate cell-cycle checkpoints, DNA repair, and/or apoptosis.88,89 The majority of TP53 mutations result in single residue substitutions of the DNA-binding domain, leading to loss of activation of p53 target genes and associated tumor suppression, without loss of the mutant p53 protein.90 Persistence, rather than loss, of mutant p53 is unusual for a tumor suppressor, which may gain aberrant functional properties and/or dominant negative properties that allow mutant p53 to promote genetic instability.88

Given the propensity of TP53 mutations to foster genomic complexity, it is not surprising that while 10% of high grade myeloid neoplasms (AML and MDS with excess blasts) harbor TP53 mutations, 70% of AML with complex karyotypes demonstrate TP53 alterations (mutations and/or loss).91, 92, 93 Complex karyotype AML has a well-known association with poor outcomes, and in the WHO HAEM4, AML with a complex karyotype was grouped into the category of AML with myelodysplasia-related changes.4 Nevertheless, among complex karyotype-associated AML, TP53-altered AML tends to show a higher degree of genomic complexity, increased alterations of −5/5q−, −7/7q−, −16/16q−, −18/18q−, +1/+1p, and +11/+11q/amp11q13∼25, and monosomal karyotypes, with multivariable analysis showing TP53 alteration as the primary prognostic factor in complex karyotype AML.92 Additionally, TP53 mutation and the presence of a complex karyotype correlate independently and additively with survival.8 More recent reports show that, regardless of blast count, MDS and AML associated with a complex karyotype with multi-hit TP53 mutations are associated with an overall worse survival, particularly in therapy-related myeloid neoplasms.93,94

In light of the dismal outcomes associated with TP53 alterations and the designation of TP53 mutated AML as “adverse risk” in the 2017 European LeukemiaNet (ELN) risk classification, it was perhaps inevitable that myeloid neoplasms harboring mutated TP53 would be grouped together in hematopoietic classification systems.95 In 2022, the ICC classification introduced the category of myeloid neoplasm with mutated TP53, which is separated into MDS, MDS/AML, and AML with mutated TP53 based on blast counts of <10%, 10-19%, and ≥20% blasts, respectively (Fig 3).10,11 Relatedly, the WHO HAEM5 includes myelodysplastic neoplasm with biallelic TP53 inactivation, but a TP53-specific AML is not introduced.9

Both the ICC and WHO classifications require biallelic mutations of TP53 for a diagnosis of TP53-associated MDS; however, with respect to the ICC diagnosis of MDS/AML and AML with mutated TP53, a single somatic mutation of >10% VAF is sufficient. The lower allelic threshold for AML reflects the poor survival outcome of single allele mutations in AML versus MDS, where complex karyotype, high-risk presentation and poor outcomes were associated only with biallelic TP53 alterations 93,96. That being said, there is conflicting data on the significance of high allele burdens of mutated TP53 in AML, as better long-term outcomes for single allelic mutations of <40% VAF have been shown in patients treated with front-line cytarabine-based therapy compared to patients with >40% VAF in one study, whereas others have shown no association of VAF with survival.93,97

It remains to be determined if the conventional MDS vs AML distinctions are clinically meaningful in the setting of multi-hit TP53 alterations. Just as the ICC has blurred the lines between MDS with excess blasts in the new “MDS/AML” category, it is possible that future classification systems will lump myeloid neoplasms with complex karyotype and TP53 mutations into a broader category without distinctions. For example, TP53 alterations in the setting of MDS with excess/increased blasts or AML are not statistically different from molecular and outcome standpoints, regardless of allelic state.93 Relatedly, previous editions of the WHO have characterized a subset of AML as pure erythroleukemia (PEL) which is characterized by ≥80% erythroid precursors, with ≥30% proerythroblasts, and by definition, <20% myeloid blasts.4 This form of AML frequently presents in the setting of prior cytotoxic therapy or myelodysplastic syndrome and is associated with a complex karyotype, biallelic TP53 alterations, and poor prognosis.98,99 PEL are now termed acute erythroleukemia in the WHO HAEM5, whereas most, if not all of these would be termed AML with mutated TP53 in the ICC system. It may be the case that in the future these neoplasms will be grouped by the underlying presence of TP53 mutations and complex karyotype, rather than by subjective distinctions between proerythroblasts and basophilic erythroblasts, or by myeloid blast count thresholds.

Dysplasia in the setting of de novo acute myeloid leukemia, particularly with an unfavorable karyotype, is associated with a poor prognosis, inspiring the creation of the AML subtype, “AML with multilineage dysplasia”, in the WHO HAEM3 classification system.100,101 This form of leukemia required the demonstration of dysplasia in 50% or more of cells in at least two lineages. AML with multilineage dysplasia was then revised to include cases with a prior history of MDS or MDS/MPN and cases with myelodysplasia-related cytogenetic alterations, and renamed “AML with myelodysplasia related changes (AML with MRC) in the WHO HAEM4 (Table 2).4,40 This form of AML was meant to encompass leukemias from older patients with fewer blasts, and shorter overall survival. Nevertheless, dysplasia can be seen in AML with recurrent cytogenetic or molecular abnormalities or AML with a history of cytotoxic therapy, and as such, these classification subtypes take precedence over AML with MRC in the WHO HAEM4.

The myelodysplasia-related cytogenetic alterations of AML with MRC included complex karyotype (≥3 abnormalities), unbalanced abnormalities and balanced abnormalities. The unbalanced abnormalities included loss of chromosome 7, del(7q), del(5q), or t(5q), isochromosome 17q or t(17p), loss of chromosome 13 or del(13q), del(11q), del(12p) or t(12p), idic(X)(q13). Balanced abnormalities included t(11;16), t(3;21), t(1;3), t(2;11), t(5;12), t(5;7), t(5;17), t(5;10), and t(3;5). Among, the unbalanced abnormalities, a monosomal karyotype (one autosomal chromosomal monosomy coupled either with another autosomal monosomy or another structural abnormalities) is a particularly adverse prognostic indicator.102, 103, 104

Since the introduction of the AML with MRC subtype, sequencing of these AMLs has revealed alterations in a number of genes, including TP53, ASXL1, and U2AF1, with U2AF1 associated with trilineage dysplasia, and ASXL1 associated with dysgranulopoiesis and intermediate cytogenetics, and STAG2 mutations associated with megakaryocytic dysplasia (megakaryocytes with separated nuclear lobes).44,105, 106, 107 In contrast AML with MRC shows a decreased rate of mutation in FLT3, NPM1, and DNMT3A.105 With time it has become clear that morphologic dysplasia in the setting of de novo AML is not always associated with a poor prognosis, and sequencing data indicates cases with mutations of NPM1 or CEBPA may account for the discrepancies.108

Myelodysplastic syndromes in general frequently harbor mutations in TET2, SF3B1, ASXL1, SRSF2, DNMT3A, and RUNX1, and less frequently in U2AF1, ZrSR2, STAG2, TP53, EZH2, CBL, JAK2, BCOR, IDH2, NRAS, MPL, NF1, ATM, IDH1, KRAS, PHF6, BRCC3, ETV6, and LAMB4, whereas higher grade MDS typically shows increased overall mutations and enrichment in mutations in TP53, GATA2, KRAS, RUNX1, STAG2, ASXL1, ZRSR2, and TET2.65,109,110 Mutations in TET2, ASXL1, SRSF2, EZH2, CBL, and RUNX1 have been shown to correlate with worse leukemia-free survival for MDS patients.65 Makishima et al showed that secondary AML compared to high risk AML is enriched in class I mutations in FLT3, PTPN11, WT1, IDH1, NPM1, IDH2 and NRAS, Metzler et al showed secondary AML had more frequent mutations in RUNX1, SRSF2, ASXL1, STAG2, U2AF1, and PTPN11, while Lindsley et al showed that detection of a mutation of SRSF2, SF3B1, U2AF1, ZRSR2, ASXL1, EZH2, BCOR, or STAG2 was >95% specific for secondary AML.14,22,110

Interestingly, Papaemmanuil, et al. showed that the so-called, AML with mutated chromatin, RNA-splicing, or both (ASXL1, STAG2, BCOR, MLLPTD, EZH2, PHF6, SRSF2, SF3B1, U2AF1, ZRSR2, and RUNX1) harbored many of these secondary/MDS-associated mutations, and consisted of older patients with low blast counts, poorer response to chemotherapy, and poor prognosis; however, only 20% of patients had morphologic dysplasia and only 9% had an antecedent chronic myeloid neoplasm.8 Likewise, using machine learning, Baer et al. showed mutations in TP53, IDH2, SF3B1, ASXL1, TET2, EZH2, SETBP1, SRSF2, ETV6, RUNX1, BCOR, and STAG2, and U2AF1 showed the highest weight of 73 tested mutations for prediction of AML with MRC, yet these mutations were found in a substantial fraction of AML without MRC cases with comparably poor overall survival.111 Similarly, Gao et al. subsequently showed that ASXL1, TP53, U2AF1, SRSF2, and SETBP1 mutations were associated with worse survival in de novo AML patients.112

It is therefore becoming evident that a molecular signature seen in the setting of secondary AML is present in a subset of de novo AML and portends a worse outcome. In light of this data, both the ICC and the WHO HAEM5 have revised the classification of AML with MRC to include specific gene mutations as inclusion criteria. Both systems have eliminated morphologic dysplasia as an inclusion criterion. Moreover, in both systems, other AML-defining single gene mutations or cytogenetic abnormalities take precedence.

With respect to the WHO, the category has been renamed AML, myelodysplasia-related (MR), and the mutations identified by Lindsley et al., ASXL1, BCOR, EZH2, SF3B1, SRSF2, STAG2, U2AF1, and ZRSR2, are thus included as disease-defining. These mutations thus join complex karyotype and unbalanced chromosomal abnormalities as AML, MR-defining (Table 2).9 The WHO additionally removed balanced translocations as AML, MR defining, with many of these translocations now part of other recurrent translocation-defining neoplasms. Of note, as the WHO HAEM5 does not recognize AML with TP53 mutations as a unique category, mutation of TP53 does not exclude a case from the diagnosis of AML, MR.

The ICC has made more extensive classification alterations, eliminating AML-MRC as a category entirely, and replacing it with AML (≥20% blasts) and MDS/AML (≥ 10-19% blasts) with myelodysplasia-related gene mutations and AML and MDS/AML with myelodysplasia-related cytogenetic abnormalities.10 Similar to the approach with prior cytotoxic therapy, a history of MDS or MDS/MPN is used as a diagnostic modifier, "progressing from MDS" or "progressing from MDS/MPN", to denote those cases with prior MDS history. The myelodysplasia-related gene mutations are identical to those utilized by the WHO, with the exception that the ICC additionally includes RUNX1. The ICC thus eliminates the WHO HAEM4 provisional entity, AML with mutated RUNX1. The WHO HAEM5 similarly eliminates AML with mutated RUNX1, however mutated RUNX1 is not used as a disease-defining mutation in any AML category. With respect to the myelodysplasia-related cytogenetic abnormalities, the ICC also eliminates the group of balanced translocations, but adds +8 and del(20q).

In 2022 the ELN has revised the classification of genetics-based risk of adult AML (Table 3).6,95 One notable change is the removal of the FLT3-ITD allelic ratio as a consequence of the impact of minimal residual disease (MRD) testing and midostaurin-based therapy as well as the lack of standardization of assays for the allelic ratio measurement.6 In addition to RUNX1 and ASXL1, adverse risk now includes the remaining myelodysplasia-related genes (BCOR, EZH2, SF3B1, SRSF2, STAG2, U2AF1, and ZRSR2). Further, AML with mutated NPM1 is considered adverse risk if associated with adverse cytogenetic abnormalities.

With respect to up-front testing, the ELN recommends screening for FLT3, IDH1, IDH2, and NPM1 with results available within 3-5 days. They recommend screening for CEBPA, DDX41, TP53, ASXL1, BCOR, EZH2, RUNX1, SF3B1, SRSF2, STAG2, U2AF1, and ZRSR2 with results available within the first cycle. Additional genes recommended to be tested are ANKRD26, BCORL1, BRAF, CBL, CSF3R, DNMT3A, ETV6, GATA2, JAK2, KIT, KRAS, NRAS, NF1, PHF6, PPM1D, PTPN11, RAD21, SETBP1, TET2, and WT1.6

With respect to molecular MRD (Mol-MRD) testing, the ELN recommends a limit of detection of at least 10−3. For qPCR, the ELN endorses aberrations of NPM1; CBFB::MYH11, RUNX1::RUNX1T1, KMT2A::MLLT3, DEK::NUP214, and BCR::ABL1 gene fusions; and WT1 expression. For NPM1 and CBF-AML, qPCR showing MRD ≥2% or failure to reduce mutant transcript levels by 3 or 4 log after completion of consolidation chemotherapy may necessitate deviations in treatment.6 For 2022, the ELN expanded response terminology to include CRh (CR with partial hematologic recovery) and CRi (CR with incomplete hematologic recovery) without MRD. Because Mol-MRD at low levels (CRMRD-LL) may not have prognostic implications, these cases are included in the CR (complete remission), CRh and CRi response categories.6 MRD relapse is defined as MRD positivity after previous MRD negativity, or detection by qPCR of an increase in MRD copy number ≥ log10 between any two positive samples in patients with CRMRD-LL, CRhMRD-LL, or CRiMRD-LL.6