The nucleophosmin (NPM1) gene encodes for a phosphoprotein (Döhner et al., 2015; Tarlock and Meshinchi, 2015) which is involved in the biogenesis of ribosome, duplication of centrosome during mitosis, cell proliferation, and apoptosis induction through p53 and p19Arf (Falini et al., 2007). NPM1 mutants have been shown to cause aberrant cytoplasmic localization of NPM1 and NPM1-interacting proteins, as well as impaired function of the nucleolar wild-type NPM1 protein (Döhner et al., 2015; Tarlock and Meshinchi, 2015). NPM1 was found commonly mutated in both adult and childhood AML patients, with a higher incidence reported in adults (24–29%) (Mardis et al., 2009; Grossmann et al., 2012; Cancer Genome Atlas Research, 2013) than in children (3–11%) (Brown et al., 2007; Thol et al., 2011; Liang et al., 2013; Rubio et al., 2016; Shiba et al., 2016). The incidence appeared to be higher in cytogenetically normal AML (CN-AML), with 48% in adults (Döhner et al., 2005) and 15% in childhood AML (Rubio et al., 2016). NPM1 mutations have also been shown to be predictive of a favorable prognosis with a reduced risk of relapse in AML patients (Döhner et al., 2005; Papaemmanuil et al., 2016).

DNA methyltransferase 3 alpha (DNMT3A), isocitrate dehydrogenase 1 and 2 (IDH1 and IDH2) and tet methylcytosine dioxygenase 2 (TET2) are involved in regulating the methylation of the genome (Langemeijer et al., 2011; Shah and Licht, 2011; Lu et al., 2012; Liang et al., 2013; Aslanyan et al., 2014; Ibrahem et al., 2014). DNA methyltransferases (DNMTs) act as a catalyst to convert cytosine to 5-methylcytosine (Shah and Licht, 2011). DNMT3A mutation was first identified in an AML patient by WGS (Ley et al., 2010). This mutation is rare in childhood AML, ranging from 0 to 2% (Ho et al., 2011; Thol et al., 2011; Liang et al., 2013) but the occurrence rate has been reported in 22–26% of adult cases (Ley et al., 2010; Cancer Genome Atlas Research, 2013; Shlush et al., 2014). Patients who harbored DNMT3A mutations showed unfavorable outcomes and higher relapse rates (Marková et al., 2012). An even worse prognosis was observed in CN-AML patients with this mutation and high dose of anthracycline chemotherapy has been recommended for this subgroup of patients (Marková et al., 2012; Patel et al., 2012; Shivarov et al., 2013). A meta-analysis involving 12 studies with a total of 6,377 patients with DNMT3A mutations showed poor prognostic impact on the overall survival, relapse free survival, and event free survival (Tie et al., 2014).

IDH1 mutations that affect the arginine residue at position R132 or R170, and IDH2 at R140 or R172 (Abbas et al., 2010; Radhi et al., 2010; Green et al., 2011; Rakheja et al., 2012; Shih et al., 2012) will impair histone demethylation (Lu et al., 2012). IDH1 mutation was first discovered in 2009 by the WGS in an adult AML genome (Mardis et al., 2009). IDH1 and IDH2 mutations were rarely detected in childhood AML, with a frequency of 0–1 and 1–2% respectively, but a higher incidence was seen in children with CN-AML (Ho et al., 2010, 2011; Andersson et al., 2011; Damm et al., 2011; Liang et al., 2013). In contrast, IDH1 and IDH2 were detected more commonly in adult AML, with a prevalence of 4–9 and 8–19% respectively (Mardis et al., 2009; Ho et al., 2010; Döhner et al., 2015). Remarkably, a higher frequency of IDH1/IDH2 (20%) was detected in adult AML via the WGS and WES approach (Cancer Genome Atlas Research, 2013), suggesting NGS could be a more sensitive approach in detecting subclonal mutations in heterogenous AML blast cells. Notably, patients with IDH1 mutations appeared to have more unfavorable outcomes compared to those with IDH2 mutations (Abbas et al., 2010; Patel et al., 2012) and shorter overall survival compared to patients without IDH mutation (Aref et al., 2015). Moreover, patients with double positive IDH and NPM1 mutations, but FLT3-ITD-negative, showed a shorter relapse free survival (Paschka et al., 2010).

TET2 converts methylcytosine to 5-hydroxymethylcytosine and has a role in regulating myelopoiesis (Shih et al., 2012). Loss of function mutation in TET2 decreased DNA hydroxymethylation (Aslanyan et al., 2014). TET2 was shown to be frequently mutated in adult AML (8–28%) as compared to childhood AML (1–7%) (Langemeijer et al., 2011; Weissmann et al., 2012; Cancer Genome Atlas Research, 2013; Liang et al., 2013; Kao et al., 2014; Kutny et al., 2015). The prognostic effect of TET2 mutation remains controversial. No significant impact on relapse free survival was observed in a large cohort of younger adult AML patients (Gaidzik et al., 2012) and no difference in relapse incidence at 5 years in CN-AML patients with or without TET2 mutations (Ahn et al., 2015), whilst other studies demonstrated that TET2 mutations were correlated with inferior outcomes in favorable-risk CN-AML (Patel et al., 2012) and failure to achieve complete remission in childhood AML (Kutny et al., 2015).

Fms-like Tyrosine Kinase 3 (FLT3) encodes for a receptor tyrosine kinase which is a membrane-bound receptor with an intrinsic tyrosine kinase domain. Internal tandem duplication (ITD) of the juxtamembrane domain and point mutations of the tyrosine kinase domain (TKD) in FLT3 have been shown to result in the constitutive activation of the receptor kinase activity (Nakao et al., 1996; Tarlock and Meshinchi, 2015). In both adult and childhood AML, the incidence of FLT3 ITD (15–37 and 10–21% respectively) was higher compared to FLT3 TKD (10 and 7%) (Cloos et al., 2006; Thol et al., 2011; Damm et al., 2012; Grossmann et al., 2012; Cancer Genome Atlas Research, 2013; Liang et al., 2013; Kao et al., 2014; Rubio et al., 2016; Shiba et al., 2016). FLT3-ITD was associated with poor outcomes and high relapse rate, and the duplication has been shown to persist in both diagnosis and relapse samples with no significant changes in frequency (Kottaridis et al., 2002; Shih et al., 2002; Cloos et al., 2006). Even though allogeneic stem cell transplant or the use of FLT3 inhibitors has been proposed as promising treatment approaches for patients harboring FLT3 mutations, the results have been disappointing due to the high relapse rate (Alvarado et al., 2014). Recently, some studies showed that patients with FLT3-ITD AML should be offered allogeneic stem cell transplant when feasible as the preferred post-remission treatment for better survival outcomes (Ho et al., 2016; Oran et al., 2016). Meanwhile, mutations in FLT3-TKD appeared to be less commonly detected and its prognosis value remains unclear and contradicting (Martelli et al., 2013; Ofran and Rowe, 2013).

Neuroblastoma RAS Viral (v-ras) Oncogene Homolog (NRAS) and Kirsten Rat Sarcoma Viral Oncogene Homolog (KRAS) belong to the RAS GTPase family that encode a membrane-associated guanosine nucleotide phosphate (GTP) binding proteins, and are involved in regulating signal transduction upon binding of ligand to a variety of membrane receptors (Bowen et al., 2005; Berman et al., 2011). The occurrence of NRAS mutations has been reported among adult (9–12%) and childhood AML (12–13%) but KRAS mutations appeared to be more prevalent in adult AML than childhood AML (12 vs. 3–7%) (Mardis et al., 2009; Cancer Genome Atlas Research, 2013; Liang et al., 2013; Kao et al., 2014; Shiba et al., 2016). Larger cohort studies have shown that the clinical outcomes of patients who harbored mutant and wild-type NRAS did not differ, as evidenced by similar complete remission (Bowen et al., 2005; Bacher et al., 2006; Berman et al., 2011) and relapse rates (Bowen et al., 2005; Bacher et al., 2006; Berman et al., 2011).

Proto-Oncogene Receptor Tyrosine Kinase (KIT) is involved in hematopoiesis as well as proliferation and regulation of cell survival. The frequency of KIT mutation in adult AML was reported lower as compared to childhood AML (4 vs. 12–20%) (Cancer Genome Atlas Research, 2013; Liang et al., 2013; Shiba et al., 2016). A higher frequency was reported in core-binding factor acute myeloid leukemia (CBF-AML) with inv(16) and t(8;21), in both children (19–44%) and adults (16–46%) (Jones et al., 2010; Pollard et al., 2010). Based on the meta-analysis conducted by Chen et al. KIT mutations conferred increased relapse risk in adult CBF-AML with t(8;21) (Chen et al., 2016). In contrast, KIT mutations did not show significant impact on relapse risk in childhood CBF-AML patients (Pollard et al., 2010). In vitro studies demonstrated that the use of a tyrosine kinase inhibitor was effective against leukemic cells harboring KIT mutations and served as promising therapeutic approach for AML patients.

CCAAT Enhancer Binding Protein Alpha (CEBPA) is a transcription factor that is involved in regulating the differentiation of neutrophils. CEBPA mutations were mostly located at the N-terminal domain (NTD) and bZip domain (Ho et al., 2009). The frequency of CEBPA mutations reported in childhood AML is generally double that seen in adult AML (2–12 vs. 6–7.5%) (Grossmann et al., 2012; Cancer Genome Atlas Research, 2013; Liang et al., 2013; Rubio et al., 2016; Shiba et al., 2016). Biallelic CEBPA mutations are associated with favorable prognosis in CN-AML whilst monoallelic mutations showed contradicting outcomes in patients (Pastore et al., 2014; Li et al., 2015). Several lines of evidence have shown that the CEBPA mutation served as a good prognosis marker of AML in which patients with CEBPA mutation had significantly longer relapse free overall survival (Renneville et al., 2009a). More strikingly, patients who acquired CEBPA mutation during relapse (absent in diagnostic sample) had favorable outcomes with prolonged survival after relapse, with a 67% 10-year overall survival rate (Tawana et al., 2015).

Runt Related Transcription Factor (RUNX1) encodes for a core-binding factor which binds to the core element of many enhancers and promoters. RUNX1 mutations are more prevalent in AML without complex karyotype, with 5–18% in adult AML (Gaidzik et al., 2011; Grossmann et al., 2012; Cancer Genome Atlas Research, 2013) and 1–6% in childhood AML (Liang et al., 2013; Al-Kzayer et al., 2014). RUNX1 mutation is associated with unfavorable outcome, is predictive of chemotherapy resistance with a refractory rate of 30% and also an increased relapse rate (Gaidzik et al., 2011; Grossmann et al., 2012). Clonal evolution of RUNX1 mutation was reported in the relapsed samples of childhood AML cases (Ismael et al., 2014). Clinical analysis has proven that patients with RUNX1 mutations who received allogeneic HSCT had a 52% 4-year relapse free survival as compared to those treated with conventional intensive post-remission therapy (0% 4-year relapse free survival) (Gaidzik et al., 2011). Hence, allogeneic hematopoietic stem cell transplant (HSCT) is highly recommended for patients harboring RUNX1 mutation as compared to conventional consolidation therapy.

Additional Sex Combs Like 1 Transcriptional Regulator (ASXL1) and Additional Sex Combs Like 2 Transcriptional Regulator (ASXL2) encodes a dual-function chromatin-binding protein, which acts as a transcription activator or repressor (Katoh and Katoh, 2003). ASXL1 mutation is frequently associated with secondary AML evolving from myelodysplastic syndrome (MDS) (Devillier et al., 2012) with a 5–17% occurrence in adult AML, with a higher frequency in cases with intermediate risk cytogenetics (31%) and CN-AML (13%) (Grossmann et al., 2012; Schnittger et al., 2012; Döhner et al., 2015). The incidence of ASXL1 mutation is rare (1%) in childhood AML and increases with age (Liang et al., 2013). However, the prevalence of ASXL2 mutation in both adult (22%) and childhood (26%) AML is almost similar (Micol et al., 2014). In childhood AML, the prognostic implication of ASXL1 and ASXL2 mutations remains inconclusive. Shiba et al. (2016) did not observe the association with inferior outcome, whilst Micol et al. (2014) reported that the mutations were associated with increased relapse risk. Thus, a larger study cohort is needed to confirm the prognostic effect of ASXL1 and ASXL2 in childhood AML. Meanwhile, ASXL1 mutations were predictive of inferior prognosis and contributed significantly to leukemogenesis in male adults with AML with MDS, those older in age and those harboring RUNX1 mutations (Schnittger et al., 2012; Döhner et al., 2015).

BCL6 Corepressor (BCOR) encodes a POZ/zinc finger transcriptional repressor (Huynh et al., 2000), and its loss-of-function mutation has led to the inhibition of proliferation and differentiation of myeloid cells (Cao et al., 2016). The occurrence rate of BCOR mutation in both childhood and adult AML was found to be almost similar, namely 3.4 and 3.8% (Grossmann et al., 2011; Shiba et al., 2016) respectively by using the WES technique. The frequency of BCOR mutations was higher in CN-AML adult patients (17%), and associated with poorer prognosis, with an overall 2-year survival rate of 25.6% (Grossmann et al., 2011). In contrast, the 3-year overall survival was similar for childhood AML patients with or without BCOR mutation (71 vs. 72%) (Shiba et al., 2016).

Lysine (K) Methyltransferase 2A (KMT2A) or formerly known as Mixed-Lineage Leukemia (MLL) is involved in regulating gene expression during early development and hematopoiesis. MLL-PTD was found mutated in 6% of adult AML patients (Döhner et al., 2002; Grossmann et al., 2012) and a lower frequency of MLL-PTD (2%) was reported in childhood AML (Liang et al., 2013). MLL-PTD conferred an inferior prognosis, in which the 3-year event free survival was only 10.5% (Grossmann et al., 2012), and 69% of CN-AML patients with MLL-PTD relapsed after achieving remission (Döhner et al., 2002). Therefore, MLL-PTD is associated with inferior outcomes in AML patients especially those with CN-AML. In addition, inferior event free survival (median 0 vs. 6.8 months) was noted in MLL-PTD AML patients with DNMT3A mutations compared to those without DNMT3A mutations (Kao et al., 2015).

Enhancer of Zeste Homologue 2 (EZH2) is a histone methyltransferase which is involved in transcriptional repression by depositing histone H3 lysine 27 (H3K27) and also plays a pivotal role in hematopoiesis (Yap et al., 2011; Lund et al., 2014). EZH2 controls the balance between cell differentiation and renewal, thus its dysregulation may lead to tumorigenesis (Lund et al., 2014). EZH2 mutations decreased H3K27 trimethylation and increased chromatin relaxation. EZH2 mutations have been detected in ~2% adult AML (Khan et al., 2013) and ~1% childhood AML (Shiba et al., 2016). EZH2 mutation was more frequently detected in acute megakaryoblastic leukemia (AMKL) and AMKL with Down Syndrome (DS-AMKL), 16 and 33% respectively (Yoshida et al., 2013). EZH2 mutations was correlated with poor prognosis and inferior survival (Kawahara et al., 2012; Larsson et al., 2014). 3-Deazaneplanocin (DZNep), an EZH2 inhibitor, may serve as a potential drug for consolidation or maintenance therapy in AML patients to decrease the chance of relapse by eliminating the cancer stem cells which are resistant to conventional chemotherapy such as cytarabine and daunorubicin (Miranda et al., 2009; Horton and Huntly, 2012).

Cohesin is a multiprotein complex which comprises of four primary subunits SMC1A, SMC3, RAD21, and STAG1/STAG2, and is responsible for sister chromatid cohesion, gene expression regulation and DNA repair (Haarhuis et al., 2014; Kim et al., 2016). SMC3 gene mutation was first described by Ding et al. (2012) by comparing the genomic landscape of a primary and matched relapse adult AML using WGS. The SMC3 mutation was found enriched in relapsed clones and may be induced by the damaging effects from cytotoxic chemotherapy. Subsequently, in 2016, Shiba et al. reported that SMC3, RAD21, and STAG2 mutations were detected in 8.3% of 204 childhood AML cases, and the truncating mutations resulted in loss of cohesin function. However, they did not observe an association between the mutations and overall survival rate (Shiba et al., 2016). A larger cohort is therefore needed to provide a better overview of the utility of cohesin as a prognostic biomarker.

Tumor Protein p53 (TP53), is the most well-studied tumor suppressor gene and is involved in cell cycle regulation in response to cellular stress. TP53 is frequently mutated in AML patients with complex karyotypes, with an incidence rate of 69–73% (Haferlach et al., 2008; Grossmann et al., 2012). The incidence rate was reported lower in other cytogenetic subgroups, ranging from 8 to 14% in adult AML and about 1% in childhood AML (Haferlach et al., 2008; Grossmann et al., 2012; Cancer Genome Atlas Research, 2013; Liang et al., 2013; Kao et al., 2014). TP53 mutation was associated with inferior prognosis and higher relapse risk, in which the 3-year overall survival and event free survival was 0% (Grossmann et al., 2012). The observation of a poorer prognosis for TP53 mutation was also independent of age, karyotype and other mutations such as NPM1/FLT3-ITD, CEBPA, RUNX1, WT1, DNMT3A, and IDH2 (Hou et al., 2015). Based on above evidence, TP53 mutation may serve as a promising predictive biomarker of the worst prognosis in AML.

Wilms Tumor 1 (WT1) is a transcription factor involved in urogenital development and was found overexpressed in AML with a role in promoting leukemogenesis (Rampal and Figueroa, 2016). The frequency of WT1 mutation was slightly higher in adult AML compared to childhood AML (6–12.5 vs. 6–8%) (Welch et al., 2012; Cancer Genome Atlas Research, 2013; Liang et al., 2013; Shiba et al., 2016). WT1 mutation was significantly correlated with adverse outcomes in both childhood and adult AML. In adult AML (n = 268, aged 15–50 year), the 4-year overall survival rate was only 22%, and the relapse rate was relatively high (82%) (Renneville et al., 2009b). A similar finding of higher relapse incidence (85.7%) was reported in 470 adult AML patients with WT1 mutations (Hou et al., 2010). The similar pattern was seen in childhood AML with cumulative relapse incidence of 70% in 298 patients (Hollink et al., 2009).

Plant Homeodomain Finger 6 (PHF6) acts as a tumor suppressor gene. PHF6 non-sense and frameshift mutations have been shown to result in loss-of-function alleles (Van Vlierberghe et al., 2011). PHF6 mutations were observed in T-cell acute lymphoblastic leukemias, with an incidence rate of 20% (Van Vlierberghe et al., 2010). In contrast, PHF6 mutations were only detected in 3–8% adult AML (Van Vlierberghe et al., 2011; Ding et al., 2012), and more frequently detected in males than females (Van Vlierberghe et al., 2011). Adverse outcomes had been reported in AML patients with intermediate risk group (Patel et al., 2012). PHF6 mutations were found to be acquired during initiation of leukemogenesis and were also enriched in relapsed samples (Ding et al., 2012), hence could serve as a potential predictive biomarker of AML relapse.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.