Genetic alterations of m6A regulators predict poorer survival in acute myeloid leukemia

1Methylation of N6 adenosine (m6A) is the most abundant form of messenger RNA (mRNA) modification in eukaryotes. It is known to play crucial roles in the regulation of gene expression, protein translation, and splicing in normal biology. m6A regulatory enzymes consist of “writers” METTL3 and METTL14, “readers” YTHDF1 and YTHDF2, and “erasers” FTO and ALKBH5. m6A perturbation mediated via knockdown or knockout of these enzymes can cause cell death, decreased cell proliferation, impaired self-renewal capacity, and developmental defects. For example, ablation of METTL3 perturbs embryonic stem cell differentiation. Depletion of FTO and ALKBH5 leads to obesity and impairment of spermatogenesis, respectively. Silencing of m6A methyltransferase can result in modulation of the TP53 signaling pathway of relevance to tumorigenesis. More recently, overexpression of FTO has been shown to promote leukemogenesis. It is therefore surprising that genetic alterations affecting m6A regulatory genes have not been explored in human cancers, including leukemia. Hence, there is a compelling reason to determine whether mutations, deletions, and amplifications of m6A regulatory genes are enriched in leukemia subtypes. Clinicopathological associations including patient survival have not previously been reported.

2Here, we curate mutations, including point mutations, deep deletions, and amplifications of the best characterized m6A regulatory genes, METTL3, METTL14, YTHDF1, YTHDF2, FTO, and ALKBH5. Deep deletions are possibly homozygous deletions as measured using the Genomic Identification of Significant Targets in Cancer algorithm (GISTIC). Four distinct types of hematological malignancies were sequenced by the Cancer Genome Atlas Research (TCGA) Network: acute myeloid leukemia (AML), multiple myeloma (MM), acute lymphoblastic leukemia (ALL), and chronic lymphocytic leukemia (CLL), and genetic data has been made available via cBioPortal. Mutations of m6A regulatory genes were found in 2.6% (5/191) of AML, 2.4% (5/205) of MM, 1.0% (1/106) of ALL, and 0% (0/666) of CLL (Additional file 1: Figure S1a). For AML, we further identified variation in gene copy number in 10.5% (20/191) of patients (Additional file 2: Table S1). There was a comparable frequency of copy number loss measured as shallow deletion (possibly heterozygous deletion) using GISTIC (n = 19) and copy number gain (n = 13) of m6A regulatory genes (Additional file 1: Figure S1b). Among these, copy number loss of ALKBH5 is the most frequent in this AML cohort (12/191, 6.3%). Notably, 4.7% (9/191) of AML patients had concomitant copy number gain or loss of more than one m6A regulatory gene (Additional file 2: Table S1). In four of these nine cases, a copy number gain of an m6A writer was detected concomitantly with a shallow/deep deletion of an m6A eraser (Additional file 2: Table S1), indicating a potential synergistic alteration of m6A regulatory enzymes that may lead to increased levels of RNA m6A modification. Shallow deletions of METTL14, FTO, and ALLBH5 were significantly associated with reduced mRNA expression of these genes (Additional file 3: Figure S2). Copy number gain of METTL14 was significantly associated with an increase in its expression (Additional file 3: Figure S2). Thus, shallow deletion and copy number gain may result in the reduced and increased expression of m6A regulatory genes, respectively.

3We determined whether mutations and copy number variations (CNVs) of m6A regulatory genes are associated with clinicopathological and molecular features of AML. Mutations and/or CNVs of METTL3, METTL14, YTHDF1, YTHDF2, FTO, and ALKBH5 as a group were significantly associated with poorer cytogenetic risk in AML (P < 0.0001, Table 1). Additionally, we observed a marked increased in TP53 mutations (P < 0.0001, Table 1) but a significant lack of NPM1 and FLT3 mutations (P < 0.005, Table 1) in AML patients harboring genetic alterations of m6A regulatory genes. These clinicopathological and molecular features were also associated with CNVs of m6A regulatory genes alone (Table 1). However, they were not associated with mutations of m6A regulatory genes alone (Table 1), which may be due to the small number of cases with mutations (n = 5).

4We further determined whether shallow/deep deletion of ALKBH5 is associated with the clinicopathological and molecular features. Consistent with our findings in m6A regulatory genes overall, shallow/deep deletion of ALKBH5 was significantly associated with poorer cytogenetic risk and the presence of TP53 mutation in this AML cohort (P < 0.0001, Additional file 4: Table S2). NPM1 and FLT3 mutations were absent in AML patients with shallow/deep deletion of ALKBH5 (Additional file 4: Table S2).

5We performed Kaplan-Meier analysis to investigate the impact of genetic alterations in m6A regulatory genes on overall (OS) and event-free survival (EFS) in patients with AML. As a group, patients with a mutation of any of the genes encoding m6A regulatory enzymes had a worse OS (P = 0.007) and EFS (P < 0.0001, Fig. 1a). Inferior OS and EFS were also evident in patients who had mutations and/or CNVs of these genes (Fig. 1b) and in those with shallow/deep deletion of ALKBH5 (Fig. 1c).

6Of all clinicopathological and molecular features considered for this de novo AML cohort, older age (>60 years), white blood cell count > median (15,200 per mm3), unfavorable cytogenetic risk, and DNMT3A and TP53 mutations were significantly associated with inferior OS and/or EFS in univariate analyses (Additional file 5: Figure S3 and Additional file 6: Figure S4). We therefore examined the impact of m6A regulatory gene mutations and/or CNVs on the outcome of AML patients with poor risk genotypes. Alterations of m6A regulatory genes as a group were associated with inferior OS and EFS in patients regardless of age (Additional file 7: Figure S5). These genetic alterations did not confer a worse OS or EFS in patients with unfavorable cytogenetic risk, white blood cell count > median, or DNMT3A mutations (Additional file 8: Figure S6).

7We further determined the survival of AML patients based on whether they exhibited combined TP53 mutations and genetic alterations of m6A regulatory genes. Almost all patients with mutated TP53 (93.6%, Table 1) had ≥1 genetic alteration(s) of m6A regulatory gene(s). This group of patients had worse OS and EFS than patients who did not have any of these genetic alterations (Additional file 9: Figure S7a). There is a non-significant trend in patients with wild-type TP53 in combination with genetic alterations of m6A regulatory genes to exhibit inferior EFS compared to patients without genetic alterations of these genes (Additional file 9: Figure S7a).

8Because mutations, deletions, amplifications, and/or CNVs of m6A regulatory genes were relatively confined to patients with wild-type FLT3 and NPM1 (95.6%, Table 1), we determined whether these genetic alterations impact OS and EFS stratified by FLT3 or NPM1 mutation status. Inferior OS and EFS were observed in patients with wild-type FLT3 who had ≥1 genetic alteration(s) of m6A regulatory gene(s) (P < 0.0001, Additional file 9: Figure S7b). Notably, these patients also had worse OS (P < 0.041) and EFS (P < 0.042) compared to patients who had mutant FLT3 but no genetic alteration of m6A regulatory genes (Additional file 9: Figure S7b). Genetic alterations of m6A regulatory genes as a group were also significantly associated with a worse OS and EFS in patients with wild-type NPM1 (P < 0.0001, Additional file 9: Figure S7c). Integration of molecular analyses of m6A regulatory genes may be useful to determine a poorer outcome in AML patients who have neither been classified as “poor risk” due to the presence of FLT3 mutations nor better outcome conferred by NPM1 mutations, particularly within a group of TP53 wild-type patients.

9In a multivariate Cox proportional hazard model that includes variables associated with poorer survival, genetic alterations of m6A regulatory genes as a group were not an independent prognostic factor for OS (Fig. 1d). However, genetic alterations of m6A regulatory genes did independently predict poorer OS (hazard ratio = 2.073; 95% CI, 1.13–3.80; P = 0.018) when TP53 mutation was excluded from the model (Fig. 1d). Similar results were observed in multivariate analyses to predict EFS (Fig. 1d). Our results support a strong association between genetic alterations of m6A regulatory genes and TP53 mutation. The fact that one is confounding the other in predicting patients’ outcome suggests that both may be complementary in the pathogenesis and/or maintenance of AML.

10Identification of novel biomarkers and molecular targets to guide the development of anti-leukemic therapies remains a major challenge. Particularly for AML, the molecular markers to define subtypes and prognosis are under continuous refinement. Given that m6A modification to RNA has broad physiological functions, its impairment may be associated with the development and progression of diverse cancers, including leukemia. The current WHO classification highlights epigenetic modifiers as being mutated early during the clonal evolution of AML. Novel genetic subgroups now include mutation in genes that encode splicing regulators, TP53, and other epigenetic modifiers.

11Our present study is the first to determine the clinicopathological associations and impact of genetic alterations affecting m6A regulatory genes in AML. We found a striking association between genetic alterations of these genes as a group and TP53 mutations (Table 1). Importantly, genetic alterations of m6A regulatory genes are associated with inferior outcome in AML patients, although this may be confounded by the adverse impact of TP53 mutations on survival (Additional files 6: Figure S4 and 9: Figure S7). It has been established that loss of the m6A methyltransferase, METTL3, resulted in alternative splicing and gene expression changes of >20 genes involved in the TP53 signaling pathway including MDM2, MDM4, and P21 in a human liver cancer cell line. It is plausible that genetic alterations of m6A modifiers, TP53, and/or its regulator/downstream targets contribute in complementary pathways to the pathogenesis and/or maintenance of AML. Further studies in larger AML cohorts would assist in confirming our findings and spur future research into the functional role of m6A RNA modification in AML and its link to tumorigenesis pathways, especially TP53 signaling.