Introduction

Acute myeloid leukemia (AML) is a neoplasm of the clonal and irreversible expansion of myeloid blasts and most common among adult acute leukemia. It is heterogeneous regarding clinical feature, morphological and immunophenotypic features, and karyotypic and genetic abnormalities [1, 2]. Initial molecular biological studies started with cloning recurrent chromosomal translocations; RUNX1-RNNX1T1, CBFB-MYH11, PML-RARA, and MLL gene rearrangement from t(8;21), inv(16), t(15;17) and 11q23 translocations, respectively. These studies presented a paradigm of leukemogenesis. The translocations generate chimeric transcriptional factors, which alter expressions of genes critical for hematopoietic development and/or differentiation. In vitro and in vivo experiments support that these abnormalities block hematopoietic differentiation and/or promote self-renewal, while their transgenic mice developed AML with low frequency except MLL-AF4 and MLL-AF9 [3–7]. In addition, critical transcriptional factors for myeloid differentiation, CEBPA and RUNX1, are point-mutated in AML without above translocations [8–10]. Importantly, this class of mutations is closely associated with morphology, phenotypes and prognosis of AML and is mutually exclusive [11].

On the other hand, classical transformation assay revealed NRAS or KRAS mutation in AML as well as solid tumors [12, 13]. Thereafter, the receptor tyrosine kinase-encoding genes, FLT3 and KIT, were found to be frequently mutated in AML [14–20]. A receptor tyrosine kinase (RTK)-RAS-MAPK pathway is one of the most important intracellular signaling in myeloid malignancies including myeloproliferative neoplasm (MPN) and myelodysplastic syndrome (MDS). In vitro and in vivo studies suggest that this class of mutations augments proliferation and survival signals, but does not induce AML but MPN by itself in murine model [21]. Clinically, this type of mutations sometimes emerge or occasionally disappear during clinical course of AML and MDS [22–24]. These mutations (so-called class I mutation) are mutually exclusive but not so strictly as the above class of mutations (class II mutation), and not related with AML subtypes except for the preference of KIT mutation to CBF leukemia.

Simultaneous expression of two classes of mutations, i.e. PML-RARA and FLT3-ITD, in murine hematopoietic progenitor cells caused AML with high frequency and the following similar experiments suggested that two classes of mutations work synergistically to develop AML in vivo [6]. The cooperating model does not only fit with the murine experiments, but also with epidemiological findings in human AML, providing the backbone for the WHO classification in AML [25]. This model is also used to characterize the new gene mutation. NPM1 mutation which was found in 2005 and the functional significance not elucidated at that time are exclusive to class II mutations and frequently overlaps with class I mutations, suggesting that it belongs to class II mutation [26, 27]. However, it is often difficult to dissect functions of class I and class II mutations [7]. For example, a representative gene mutation of class I, FLT3-ITD, does not only mediate proliferation and anti-apoptosis signals, but also block differentiation through down-regulating C/EBPα and PU.1 [28–30].

Year 2009 is the first year of leukemia genome. TET2 located on 4q24 was identified as a target of deletions or mutations in myeloid malignancies [31, 32]. IDH1/2 mutation was found by whole-genome sequencing of a primary, cytogenetically normal AML (CN-AML) [33]. In the next year, DNMT3A and EZH2 were recognized to be mutated by genome-based methods [34, 35]. Either of these new mutations directly targets epigenetic control such as DNA/histone methylation status [36]. In addition to them, mutations were found in transcriptional corepressors, ASXL1, BCOR and BCORL1 [37–39]. In 2012, cohesion complex-encoding gene mutation was reported [40]. Most these mutations were generally identified in all types of myeloid malignancies, while discovery of frequent mutations of splicing machinery, which is highly associated with myelodysplasia and the formation of ring sideroblasts, is important to clarify disease type- and/or phenotype-specific pathological roles of gene mutations [41, 42].

Although more years are needed to draw the whole picture of AML gene abnormalities, we may have obtained a list of recurrently mutated genes in AML (Table 1). Although all the mutations are likely to be unrelated at first glance, if looking at its features, they are found to be related each other. In Fig. 1, we propose a pathway map, in which mutated gene products in AML are functionally connected. In this figure, we make a point not to draw uncontrolled cell-cycle and cell survival which are secondary operations from mutated gene products. In the case of acute lymphoblastic leukemia and lymphoma, genes associated with cell-cycle and cell survival are frequently targeted, suggesting that the mutational profiles depend on its cellular context. Accordingly, it may be possible to highlight that AML is a disease of transcription, in another word, differentiation.

Table 1 Recurrently mutated genes in AML
Full size table
Fig. 1
figure 1

A schematic diagram of intracellular signaling associated with AML. Gene mutations in AML are functionally classified into three general classes: constitutive signal transduction (class I), deregulated gene expression (class II), and epigenetics/chromatin change (class III). Three classes of mutations are functionally linked in this model, whereas according to the Gilliland’s model two classes of mutation work in a cooperative or complementary manner [26]. The molecular alterations are eventually summarized toward the deregulated gene expression in molecular pathogenesis

Full size image

Next, we summarize gene mutations found in AML during these 5 years.

Mutations in epigenetic modifiers

TET2

The TET family proteins (Tet1, Tet2, Tet3) are catalyzing enzymes 5-methylcytosine (5-mC) to hydroxymethylcytosine (5-hmC), which have been shown as an initial process of DNA demethylation [43–45]. Mutations of the TET2 gene have been found in various myeloid malignancies; 8–27 % of AML, 20–25 % of MDS, 4–13 % of MPN [31, 32, 46–49]. TET2 mutations are loss-of-function of mono allele in most cases, including missense, frameshift and nonsense mutations. Importantly, TET2 mutations are almost mutually exclusive with IDH1/2 mutations, suggesting a similar epigenetic defect as IDH1/2 mutations [36, 50].

The base 5-hmC is recently suspected as a new epigenetic marker in addition to being an intermediate metabolite during cytosine demethylation. It has been suggested that 5-hmC is associated with transcriptional regulation for promoter and enhancer functions [51–53]. In vivo, TET2 inactivation induced in hematopoietic abnormalities predisposing not only myeloid but also lymphoid malignancies [54]. However, the precise mechanisms and downstream effects of TET2 are yet unknown.

The prognostic relevance of TET2 mutations remains controversial. Some studies suggest an adverse impact of TET2 mutations on outcome in certain AML subgroups; in other studies, no prognostic significance was found [47–50].

DNMT3A

DNA methyltransferase (DNMT) family enzymes transfer a methyl radical to DNA, and in human, DNMT3A and 3B are methyltransferases that convert cytosine to 5-methylcytosine and generate de novo DNA methylation, whereas DNMT1 mediates maintenance methylation [55].

Mutations of DNMT3A are reported in 18–23 % of AML, including 20–35 % with normal karyotype, 8 % of MDS and 7–15 % of MPN [34, 36, 50, 56–58]. The locus of the mutations is clustered at R882 in a half of cases and distributed widely in other cases. The mutations are supposed to be loss-of-function, although it is not well characterized. Aberrant methylation patterns induced by the mutation might be associated with initiation and progression of leukemia, although the biological consequences of the mutations remain unclear [55–57, 59]. In vivo study revealed that Dnmt3a-null hematopoietic stem cells had an increased self-renewal capacity and lost their differentiation potential, which was accompanied by aberrant methylation pattern implicated in leukemogenesis [60, 61]. However, knockout of Dnmt3a alone was not sufficient to initiate leukemia.

DNMT3A mutations might be associated with poorer prognosis, and is frequently associated with NPM1 and FLT3 mutations [34, 50, 56–58]. Several different loss-of-function mutations have been found in all exons of DNMT3A, while a missense point mutation at amino acid R882 which decreases catalytic activity and DNA binding affinity is most frequently identified.

EZH2

EZH2 is a H3K27 methylatransferase which is one of the components of polycomb repressive complex 2 (PRC2) [62]. PRC2 has been supposed to be required for silencing target genes and maintaining “stemness” in stem cells [63]. Several studies established that EZH2 is overexpressed in solid tumors, and high expression is associated with tumor progression, indicating the role of EZH2 as an oncogene in these tumors [64]. Mutation at the tyrosine 641 residue within the catalytic domain of EZH2 is recurrently identified in germinal center-type diffuse large B-cell lymphoma and follicular lymphoma [65]. Biochemical analysis revealed that Y641 mutation increases di- and tri-methylation of H3K27 in spite of impairing H3K27 mono-methylation. Furthermore, in vivo study suggested that EZH2 augments leukemogenesis by reinforcing differentiation block in AML [66]. These results are consistent with the assumption that the increasing H3K27 methylation activity of PRC2 represses the expression of tumor suppressor genes. In contrast to these data, a wide variety of loss-of function mutation of the EZH2 gene has been found in myeloid malignancies. It is found in 6 % of MDS and 3–13 % of MPN, but extremely rare in AML [35, 36, 67]. Among them, mutations are prevalent in CMML and associated with adverse outcome [68]. The function of EZH2 remains unclear in hematopoiesis, because Ezh2 conditional knock-out (KO) mice had minimal myeloid differentiation defects [69]. However, loss-of-function rather than overexpression of EZH2 might be involved in pathogenesis of myeloid malignancies through the insufficient H3K27 methylation status.

IDH1/2

IDH1 and IDH2 are NADP-dependent isocitrate dehydrogenases that catalyze isocitrate to α-KG in TCA cycle, and the subcellular localizations are in cytosol and mitochondria, respectively [70]. IDH1/2 mutations are detected in 15–33 % of AML mostly in normal karyotype-AML, 3.5 % of MDS, 2–5 % of MPN, and also in glioma and multiple endochondromatosis [70–74]. The mutations have been shown to exhibit a gain-of-function leading to aberrant accumulation of 2-hydroxyglutarate (2-HG). 2-HG is an oncometabolite which inhibits an enzymatic activity of TET2 and stimulates HIF1-α, leading to initiation and promotion of cancer [75–77]. Therefore, the IDH1/2 mutations functionally overlap with TET2 mutation, resulting in hypermethylation of leukemia cells, disrupt TET2 function, and impair hematopoietic differentiation. Recently, IDH mutation was reported to expand murine hematopoietic progenitor pool as well as altering epigenetics [76].

The impact of IDH mutations on survival of AML patients is unclear like the mutation of TET2. Some studies have observed no difference in outcome with respect to the IDH mutation status, while others have demonstrated a poor prognostic impact in certain AML subgroups. Recent studies suggest that the impact of IDH2 mutation on prognosis depends on the mutation site, with the IDH2 R140 mutation being an independent favorable prognostic factor in AML patients [71].

Cohesin complex genes

Cohesin complex regulates the cohesion and separation of sister chromatids during cell division, and recently has been known to regulate gene transcription associated with cell development and differentiation [78, 79]. Cohesion complex is composed by Smc1, Smc3, Rad21 and SA1/2. SA2-encoding gene, STAG2, and RAD21 are reported to be deleted in AML genome, and recently SMC1A, SMC3, STAG2 and RAD21 are found to be mutated in a loss-of-function manner [40, 80, 81]. The mutations are found in a variety of AML except acute promyelocytic leukemia and are unrelated to chromosomal instability. The clinical and biological significance of cohesion complex-coding genes is to be further elucidated.

Transcriptional corepressor

BCOR

BCOR protein was identified as an interacting corepressor of BCL6, a POZ/zinc finger transcription repressor that is required for germinal center formation and may influence apoptosis [82]. Mutations in the BCOR gene have also been found to be associated with AML [38].

ASXL1

ASXL1 is involved in the maintenance of both activation and silencing of the HOX genes, which are involved in development of body axial structure as well as in chromatin remodeling [83]. ASXL1 also acts as a corepressor of nuclear receptors such as PPARγ and RARγ [84]. Although the KO mice of ASXL1 have no defect in hematopoiesis [83], it has been recently reported that ASXL1 mutants promote myeloid transformation through loss of PRC2-mediated H3K27 tri-methylation [85]. ASXL1 mutations have been found in 3–11 % of AML, 14 % of MDS and 2–23 % of MPN. The mutation frequency is high, from 27 to 49 % in MDS/MPN including CMML [50, 86–88]. The mutations consist of missense, nonsense and frame shift, suggesting a loss-of-function. It has been also reported that ASXL1 mutation status can change during disease evolution.

Interaction of genetic alterations in AML

Several somatic mutations are accumulated in cancer cells during the development and progression. These mutations are classified into two types of mutations: “driver” mutations that provide a selective advantage and “passenger” mutations that are simply happened without any role. In AML, it has been demonstrated both by clinical analysis and experimental models that the accumulation of two types of driver mutations (class I and class II mutations) is required for the clonal expansion of leukemia cells [7]. Therefore, it is important to evaluate whether each mutation is driver or passenger mutation and which combinations of overlap mutations are associated with the leukemogenesis and clinical significance in AML. To date, several combinations of class I and class II mutations have been apparent. Although FLT3-ITD and PML-RARA, KIT mutation and RUNX1-RUNX1T1 or CBFB-MYH11, and FLT3-ITD and NPM1 mutation are frequently identified, the other combinations are relatively infrequent. Recently, it has been suggested that a part of mutations, such as epigenetic modifiers, generates a new class (class III), because of their overlap mutations both with class I and class II mutations (Fig. 2) [36, 50].

Fig. 2
figure 2

Interaction of genetic alterations in AML frequently identified overlap mutations are shown. Mutations in epigenetic modifiers are overlapped both with class I and class II mutations

Full size image

Consistent with the biological evidence that TET2 and IDH1/2 mutations are functionally overlapped, these mutations are mutually exclusive in AML cells [36]. Likewise, DNMT3A, ASXL1 and MLL mutations, which directly regulate methylation state of DNA and/or the histone, are exclusively identified in AML cells. Importantly, overlap mutation of TET2 or IDH1/2 mutation with DNMT3A, ASXL1 or MLL mutation is identified in AML cells. This overlap is infrequently identified, but indicates the possible involvement of more than one molecules, which show the opposite methylation status by these epigenetic mutations, in leukemogenesis. In addition, these epigenetic mutations might contribute to pathogenesis of AML in concert with class I and/or class II mutations because of their frequent overlap mutations. Particularly, TET2, IDH1/2 and DNMT3A mutations are frequently overlapped with FLT3 and NPM1 mutations. In contrast, ASXL1 and MLL mutations, which mainly affect the histone methylation status, are exclusively mutated with class I mutations, except for the frequent overlap of MLL-partial tandem duplication (MLL-PTD) and FLT3 mutations. Furthermore, ASXL1 and MLL mutations are overlapped both with class II, such as RUNX1 and CEBPA, and TET2 or IDH1/2 mutations [37, 87]. On the other hand, ASXL1 mutation recurrently overlaps with JAK2, RAS and NF1 mutations as well as U2AF35 and RUNX1 mutations in MDS and MPN [89, 90]. In addition, it has been demonstrated that knock down of ASXL1 collaborates with NRAS G12D mutation to promote myeloid transformation in vivo [85]. Since ASXL1 mutation is preferentially identified in secondary AML (30.3 %) rather than de novo AML (6.5 %), further analysis is required to clarify whether the oncogenic role of ASXL1 in de novo AML is the same as that of MDS and MPN [91].

More recently, mutations in cohesion complex genes (STAG2, SMC3, SMC1A and RAD21) have been discovered [40]. These mutations are reportedly mutually exclusive, indicating the redundant potential for leukemogenesis. Although mutations in cohesion complex genes were so far not identified in AML with PML-RARA, CBFB-MYH11 and BCR-ABL1, their overlap mutations with class I, class II and epigenetic modifier mutations were identified. At present, it is unclear as to how these mutations are involved in the pathogenesis of AML, while their unique overlap pattern indicates the novel important pathway of leukemogenesis.

Several patterns of overlap mutations have been, to date, apparent as described above. However, the order of acquisition and stability of each mutation during the disease status is not fully elucidated. It has been well accepted that class II and epigenetic modifier mutations in stem/progenitor cells, which confer an advantage in self-renewal of leukemia progenitors, are earlier events than class I mutation during the development of AML. However, the interaction of class II and epigenetic modifier mutations both at the initiation and progression stages of leukemia remains unclear. Comparable analyses of mutation status between the diagnosis and the relapse of AML revealed the loss of class II mutations, such as NPM1 and CEBPA mutations, at the relapse. In contrast, epigenetic modifier mutations are stably identified during the disease progression of AML, while they are frequently acquired at the later stage of MPN, suggesting their multiple oncogenic potency not only for the initiation, but also for progression in myeloid malignancies [22–24].

Background mutation

It is notable that several background mutations occur in normal hematopoietic cells [40]. Whole-genome analysis using normal hematopoietic cells revealed that the mutated gene number is lowest in the cord blood and increased according to the aging, irrespective of synonymous or non-synonymous mutation. These results suggest that background mutations, which are randomly acquired in hematopoietic cells with age and stress, might be involved in the initiation of AML development. However, these results have raised further questions regarding how background mutations affect the acquisition of driver mutations and whether driver and background mutations cooperatively induce the clonal expansion. Furthermore, it is interesting that the number of background mutations is almost the same between AML with PML-RARA and normal karyotype, suggesting that most of the mutations found in AML are age-related background and leukemia-associated passenger mutations. In cytogenetically normal (CN)-AML, several kinds of driver mutations are involved in the initiation of AML, while those are not always stable during the progression of AML, and the number of background mutations is the same regardless the type of initiation driver mutations [22–24, 40, 92, 93]. These findings, therefore, suggest that pathological roles of back ground mutations might be different among the types of following driver mutations.

Prognostic relevance of genetic mutations in AML

Although prognostic risk classification based on cytogenetics has been well accepted in AML, there is a clinical heterogeneity of the intermediate risk group, particularly CN-AML [94]. Recent advance and accumulation of prognostic relevance of recurrent genetic alterations have made possible more detailed risk stratification in AML patients.

The European LeukemiaNet (ELN) has recommended a novel risk classification system based on the cytogenetic and genetic status [95]. In this system, CN-AML is stratified into two risk groups according to the mutation status of FLT3, NPM1 and CEBPA: patients with NPM1 mutation but not FLT3-ITD and those with CEBPA mutation are included in the favorable risk (FR) group, and patients with FLT3-ITD and those without NPM1 mutation and FLT3-ITD are categorized into intermediate-I risk group. Long-term prognosis according to the ELN classification system was retrospectively evaluated in two well-established cohorts, and both analyses demonstrated that the ELN system is useful for further risk stratification of younger adult patients with CN-AML. However, further studies are required to confirm prognostic relevance of epigenetic modifier mutations. In the patients who were treated on the Cancer and Leukemia Group B (CALGB) trials, TET2 mutation was identified to be a poor prognostic factor in the patients who were classified in the FR group of the ELN system [47]. In the patients who were treated on the Eastern Cooperative Oncology Group (ECOG) trials, only co-occurring NPM1 and IDH1/2 mutations were identified to be the favorable factor for the long-term prognosis [50]. Although both analyses demonstrated possible prognostic implication of TET2 and IDH1/2 genes in the FR group, it was not evident in the patients registered to Munich Leukemia Laboratory in Germany [88]. Some groups reported adverse effect of DNMT3A, MLL and ASXL1 mutations on long-term prognosis in AML patients, their prognostic relevance is still controversial.

Table 2 shows prognostic relevance of relatively established genetic status in AML. However, therapeutic interventions based on the genetic status should be carefully conducted, because no prospective study confirms the prognostic risk of each mutation. In the future, validation study and meta-analysis should be conducted.

Table 2 Prognostic relevance of mutation status
Full size table

Conclusion

Comprehensive genomic analyses using the next-generation sequencer have revealed a lot of mutations in AML. Although functional and prognostic implications of each mutation are not fully elucidated, novel cooperative effects of mutations are speculated. Particularly, it is important to clarify the interactions of epigenetic modifier mutations with class I and/or class II mutations for understanding the pathogenesis and novel therapeutic targets of AML. To understand the entire picture of molecular pathogenesis in AML, gene rearrangement, gene copy number, DNA methylation and expression profiles are needed to be analyzed with gene mutations. Furthermore, it is also important to validate which mutations should be included for the prognostic stratification system. Since newly identified mutations are recurrently, but infrequently identified in AML, very large-scale prospective studies are required to confirm their prognostic implications. Finally, it is expected that genome research promotes the development of new treatment.