Introduction
Genome-wide studies in patients with myeloid malignancies have provided major insights into the pathogenesis of these diseases,
1,2 and especially in acute myeloid leukemia (AML), an increasing panel of genetic markers has been identified that constitute a base for risk stratification and therapeutic decision making.
1,3–6
Recently, the tet oncogene family member 2 (
TET2) gene was identified to be mutated in a variety of myeloid disorders.
7 Subsequent sequencing analysis revealed
TET2 mutations (
TET2mut) in 7% to 23% in de novo AML
8–11 and 14% to 55% in other myeloid malignancies.
7,8,12–15 In some of these instances,
TET2mut were associated with uniparental disomies that involve
TET2 on 4q24.
7,16,17 Thus far, a leukemogenic role of ten-eleven translocation (
TET) family gene members (
TET1,
TET2, and
TET3) was only known for
TET1, which is involved as a translocation partner in
MLL-rearranged AML and rarely in acute lymphoblastic leukemias.
18
The
TET family members have two highly conserved regions, an
N-terminal cysteine-rich domain followed by a 2-oxoglutarate (2OG) -Fe(II) oxygenase characteristic double-stranded b-helix.
16,19 Recently, TET proteins were also found to be homologues of 2OG-Fe(II) oxygenases catalyzing the conversion of thymine to β-
d-glucosyl hydroxymethyl uracil in trypanosomes
20; in humans, 2OG oxygenases have been shown to be involved in various biologic functions including histone demethylation and DNA repair.
21
TET1 is an enzyme involved in the conversion of 5- methylcytosine (5mC) to 5-hydroxymethylcytosine (5hmC) in DNA, which is a process thought to play an important role in DNA demethylation and, thus, epigenetic regulation.
20 On the basis of sequence homology, the TET2 protein is expected to also play a role in chromatin remodeling,
16 and it was demonstrated that TET2 also converts 5mC into 5hmC.
22 Furthermore, it was shown that
TET2mut samples display uniformly low levels of 5hmC compared with normal controls, supporting a functional relevance of
TET2mut in leukemogenesis.
Recently, Figueroa et al
10 found that activating mutations of
IDH1/
2 are mutually exclusive with mutations of the 2OG-dependent
TET2. Because
IDHmut interfere with the production of 2OG (α-ketoglutarate) by the aberrant production of 2-hydroxyglutarate, it was hypothesized that the 2OG-dependent catalytic function of
TET2 might be hampered by
IDHmut, which would explain the mutual exclusiveness of these mutations. In line,
IDHmut were associated with similar epigenetic defects as
TET2mut, and
IDHmut impaired the catalytic function of TET2.
10 Thus, there is evidence that
TET2mut and
IDHmut may lead to a biologically redundant hypermethylation phenotype. In accordance with current multistep pathogenesis models of leukemogenesis,
TET2mut co-occur with other mutations (eg,
JAK2V617F in myeloproliferative neoplasia or
NPM1 in AML), but until recently, it was still questionable whether
TET2mut represent an early pathogenic event.
16,19 Data from two independent conditional mouse models demonstrated that Tet2 haploinsufficiency leads to increased stem-cell self-renewal and myeloproliferation, which, thereby, suggested that monoallelic
TET2mut are an early event in leukemogenesis because they can contribute to myeloid transformation.
23,24
With regard to the prognostic impact of
TET2mut, initial studies in small cohorts with AML revealed inconclusive results.
8,9 A recent study of a large cohort with AML reported an adverse prognostic impact in the molecular favorable-risk cytogenetically normal (CN) -AML group,
11 whereas there was no impact of
TET2mut in the intermediate-risk-I group as defined by the current European LeukemiaNet (ELN) criteria.
3
In this study, we aimed to further explore the frequency and clinical impact of TET2mut in a large cohort of genetically and clinically well-characterized younger adult patients with AML. In addition, by using gene-expression profiling (GEP) in 333 patients, we sought to determine whether TET2mut might be associated with a strong gene-expression signature that would further delineate TET2mut AML as a biologic subset of AML.
Results
Frequency and Types of TET2mut
We found 60 TET2mut samples in 783 patients (7.6%), with six patients who exhibited two mutations (Data Supplement). Mutations were distributed all over the gene and most commonly affected exon 3 (n = 22), exon 11 (n = 26, which encodes the 2OG-binding domain), and exon 10 (n = 6) but also occurred in other exons (ie, exons 4 [n = 3], 6 [n = 3], 7 [n = 1], 8 [n = 1], and 9 [n = 4], but not exon 5; Appendix Fig A1, online only). All mutations but one were heterozygous. Frameshift mutations, which resulted from insertions or deletions (n = 16; 24%), and nonsense mutations (n = 13; 20%), which were predicted to result in protein truncation, accounted for approximately one-half of mutations; the remaining mutations were single nucleotide substitutions that led to missense mutations (n = 37; 56%). Analysis of germline materials in 13 patients (derived from buccal swaps or remission BM and/or PB) showed that all missense changes were acquired and not present in germline DNA.
Association of TET2mut With Clinical Characteristics and Genetic Alterations
Patients with
TET2mut were in trend older and had in trend higher WBC counts (
P = .08 and
P = .11, respectively;
Table 1). There was no significant difference with respect to platelet counts, BM or PB blast counts, and type of AML between
TET2mut and
TET2 wild type (
TET2wt).
TET2mut were found in all major cytogenetic subsets, and there was also no association with other AML-associated molecular markers (
NPM1mut [
P = .35],
FLT3-ITD [
P = .26], and
FLT3-TKD mutations [
P = .35],
CEBPAmut [
P = .24], and
RUNX1mut [
P = .99]), with the exception of
IDHmut that were almost mutually exclusive with
TET2mut (
P < .001;
Table 1).
Within the subgroup of CN-AML, patients with
TET2mut were older, and they had higher WBC counts (
P = .03 and
P = .03, respectively), and mutations were again inversely correlated with
IDHmut (
P = .005;
Table 2).
Response to Induction Therapy
In the entire cohort, no significant differences between patients with
TET2wt and
TET2mut with respect to rates of CR (71% [513 of 723 patients] and 73% [44 of 60 patients], respectively;
P = .77), refractory disease (18% [129 of 723 patients] and 15% [nine of 60 patients];
P = .72), and early or hypoplastic death (11% [81 of 723 patients] and 12% [seven of 60 patients];
P = .83;
Table 3) were found. In multivariable analysis, age, WBC, cytogenetic risk group,
FLT3-ITD, and
NPM1mut were significantly associated with CR achievement;
TET2mut had no impact (Data Supplement). Within the subgroup of patients with CN-AML, there was a trend toward higher CR rates in patients with
TET2mut (
P = .12;
Table 3; Data Supplement).
We also evaluated the clinical impact of
TET2mut according to the ELN classification that groups CN-AML into a molecular favorable group (
CEBPAmut and/or
NPM1mut without
FLT3-ITD) and unfavorable group (intermediate-I, all remaining patients with CN-AML).
3 There was no significant difference with regard to CR rates in the ELN favorable-risk group (77.8% [seven of nine patients] in
TET2mut v 84.7% [72 of 85 patients] in
TET2wt;
P = .63), whereas in the intermediate-I risk group, we observed a significantly higher CR rate for patients with
TET2mut (90.0% [18 of 20 patients] in
TET2mut v 66.2% [143 of 216 patients] in
TET2wt;
P = .04;
Table 3). In multivariable analysis of the intermediate-I risk group,
TET2mut was a significant factor that predicted CR achievement (
P = .03, odds ratio [OR], 5.59; 95% CI, 1.20 to 26.10; Data Supplement).
Survival Analysis
The median follow-up time for survival was 6.5 years (95% CI, 6.3 to 6.72 years); the estimated 4-year relapse-free survival (RFS) and overall survival (OS) of the entire cohort were 43% (95% CI, 0.40% to 0.48%) and 43% (95% CI, 0.40% to 0.47%), respectively. Univariable survival analysis on the end points event-free survival (EFS), cumulative incidence of relapse, RFS, and OS showed no significant differences between patients with
TET2wt and
TET2mut (
P = .67,
P = .32,
P = .49, and
P = .45, respectively;
Figs 1A and
1B).
Subset analyses in patients with CN-AML (n = 330) showed no significant impact of
TET2mut on EFS, cumulative incidence of relapse, RFS, and OS, (
P = .94,
P = .33,
P = .22, and
P = .36, respectively;
Figs 1C and
1D). Similarly, ELN subgroup analyses did not show an impact of
TET2mut on EFS, RFS, or OS, neither in the favorable-risk (
P = .65,
P = .81, and
P = .64, respectively) nor in the intermediate-I group (
P = .81,
P = .10, and
P = .38, respectively;
Figs 1E and
1F).
In multivariable analysis, TET2mut had no impact on RFS, EFS, and OS in both the entire cohort (RFS: OR 0.96; P = .84; EFS: OR, 0.93; P = .65; OS: OR, 0.94; P = .72) and subgroup of patients with CN-AML (RFS: OR, 1.06 [P = .83]; EFS: OR, 0.83 [P = .41]; OS: OR, 0.93 [P = .77]; Data Supplement). The only variables that consistently appeared in all models for survival end points were age, WBC, type of AML, cytogenetic risk, NPM1mut, and FLT3-ITD (Data Supplement). Similarly, multivariable analyses within the ELN subgroups also revealed no significant impact of TET2mut on outcome (Data Supplement).
Explorative Outcome Analyses According to TET2 and IDH Mutational Status
The observation that
TET2mut are mutually exclusive with
IDHmut,10,11 and recent data that suggested that these two aberrations might have a common mechanism of action
10 prompted us to perform clinical correlations for a cohort that comprised both the
TET2mut and
IDHmut groups. Compared with patients with
TET2wt/IDHwt, patients with
TET2mut and/or
IDHmut were associated with older age (
P < .001), higher BM blast counts (
P = .02), and the cytogenetic intermediate-risk group (
P < .001; Data Supplement). Univariable analysis revealed an inferior outcome as reflected by in trend shorter RFS and shorter OS for patients with
TET2mut and/or
IDHmut (
P = .12 and
P = .03, respectively;
Figs 2A and
2B); in multivariable analysis, however,
TET2mut/IDHmut was not independent of other known prognostic markers (Data Supplement).
Evaluation of the combined
TET2mut/IDHmut group within the CN-AML subgroup also revealed a significant association with older age (
P = .02) and inferior OS (
P = .03;
Fig 2), but again did not confer independent prognostic information (Data Supplement).
TET2mut-Associated Gene-Expression Pattern
To evaluate the possible impact of TET2mut on disease biology via epigenetic deregulation, we compared gene-expression profiles from 31 patients with TET2mut with 302 patients with TET2wt. Class-comparison analysis revealed a distinct TET2mut-associated gene-expression profile that comprised 124 genes (P < .005), which was also shared by a group of TET2wt samples (Fig A1; Data Supplement). Of note, a significant proportion of these TET2mut-like instances comprised IDHmut leukemia samples. Significantly deregulated candidate genes included CD56 with higher expression levels and JAK2 and MEF2C with lower expression levels in TET2mut AML. Pathway-comparison analysis for Biocarta pathways revealed several pathways to be significantly enriched in this signature, such as the monocyte and its surface molecules pathway (including cell adhesion molecules [PECAM1] and integrins [ITGAL, ITGB5, ITGA6]) and candidates belonging to the nuclear factor of activated T cells pathway (including MEF2C, EDN1, MAPK1, and RAF1; Data Supplement).
Discussion
TET2mut have been found in various myeloid neoplasms, and their clinical impact has been investigated by several groups.
7–9,12–15 However, there are only a few studies that evaluated the frequency and clinical role of
TET2mut within large cohorts of patients with AML derived from prospective clinical trials, and thus the clinical relevance still remains uncertain.
In the study by Abdel-Wahab et al,
8 TET2mut were found in 12% of patients with AML and were associated with decreased OS. In contrast, the French ALFA (Acute Leukemia French Association) cooperative group reported
TET2mut in 20% of 147 patients with AML and found no significant correlation with outcome in 111 patients who had achieved a CR (both in the entire cohort and in 54 patients with CN-AML).
9 In a recent study by the CALGB (Cancer and Leukemia Group B) of 427 adult patients with CN-AML,
TET2mut were found in 23% of patients. Mutations correlated with older age and higher WBC, and they predicted for inferior survival in the ELN molecular favorable-risk group of patients.
11 Although we also observed a trend for an association of
TET2mut with older age and higher WBC in our AML cohort, we did not find a prognostic impact of
TET2mut on clinical outcome in the entire patient cohort or subgroups of CN-AML or ELN molecular favorable-risk AML.
These discrepancies might have been due to differences in study populations because the CALGB study included only de novo AML patients, and older patients (> 60 years of age) were also enrolled. The higher
TET2mut incidence of 23% might reflect the age distribution that ranged from 18 to 83 years, with a median age of 66 years for patients with
TET2mut.
11 Similarly, the ALFA study reported a much higher
TET2mut incidence in 36 patients with AML who did not achieve a CR (27% compared with 12% in 111 patients with a CR), a cohort enriched for elderly patients.
9 In contrast, our study was restricted to younger patients with AML (age range, 18 to 60 years) with a median age of 47.5 years and included de novo and secondary/therapy-related AML. In agreement with the incidence of
TET2mut of 7.6% in our cohort, a recent ECOG study reported somatic
TET2mut in 7.3% of 385 patients with AML age 60 years or younger.
10
Furthermore, on the basis of the age difference, the proportion of patients who received intensive therapies is likely to be different among the previously mentioned studies, which might also partly account for the differences in outcome. Although all patients in the CALGB study received intensive cytarabine/daunorubicin-based first-line therapy, only younger patients were assigned to a more intensive consolidation therapy. In our study, all patients with cytogenetic intermediate risk received intensive double-induction therapy and consolidation with repetitive cycles of high-dose cytarabine or autologous transplantation, and allogeneic transplantation was received by patients with a matched-related donor.
25
In accordance with both the CALGB and ECOG studies that reported a significantly lower frequency of
IDHmut in
TET2mut,
10,11 we also made the observation of almost mutual exclusiveness of
TET2mut and
IDHmut. This result points to a common pathomechanism, and this hypothesis is further strengthened by the fact that the altered enzymatic function of mutated
IDH converts 2OG into 2-hydroxyglutarate,
33 and TET2 function is 2OG-dependent. Thus,
IDHmut-associated 2OG level changes might influence the catalytic TET2 function converting 5mC into 5hmC, which is also impaired in patients with
TET2mut who display uniformly low levels of 5hmC.
22 In agreement, epigenetic deregulation reflected by low 5hmC levels was also observed in a fraction of patients with
TET2wt,
22 and patients with
IDHmut were shown to present with low 5hmC levels via the inhibition of TET2 function,
10 which explains a common hypermethylation phenotype of
IDHmut and
TET2mut AML. However, our combined analysis of patients with
TET2mut and
IDHmut did not reveal an impact on clinical outcome independent of other known markers.
These findings that
TET2mut and
IDHmut acted as epigenetic regulators, which may be biologically redundant, were further strengthened by our GEP data. The
TET2mut-associated expression pattern was not restricted to patients with
TET2mut but was also shared by a large number of patients with
TET2wt, including patients with
IDHmut and
TET2mut-like
IDHwt. Similarly, Metzeler et al
11 and Figueroa et al
10 found no strong gene-expression patterns for
TET2mut or
IDHmut AML, respectively,
10,11 but it was shown that
TET2mut AML displayed an hypermethylation signature overlapping with that of patients with
IDHmut.
10 In addition, other yet unknown factors might be involved in epigenetic deregulation such as recently discovered mutations in
EZH2,
34 ASXL1,
35 and
DNMT3A.
36
Nevertheless, the gene-expression pattern seems to reflect in part the biologic impact of mutated
TET2, as it is consistent with the recently observed impaired myeloid differentiation after shRNA-mediated
TET2 knockdown, which resulted in an expansion of monocyte/macrophage lineages.
22 In addition, the inverse correlation of
TET2mut and
JAK2 expression warrants further investigation as a result of a potential interplay of these genes in disease initiation and progression,
37 and the deregulation of the NFAT pathway members might play a pathogenic role in
TET2mut AML.
38
With regard to the pathogenic relevance of
TET2mut, recent conditional mouse models proposed that
TET2mut occurs in a stem/progenitor cells, which creates a predisposition to the development of myeloid malignancy.
23,24 On the basis of these studies,
TET2mut might represent an early event that, in cooperation with secondary mutations, drives the phenotype of the disease. However, in AML, there are several findings that argue against this hypothesis:
TET2mut were spread over all cytogenetic subgroups and co-occurred with aberrations such as inv (16) or mutations of
NPM1 that are usually found in de novo AML but not as secondary events. Similarly, in our study,
TET2mut was not enriched in secondary or treatment-associated AML. Clonality studies as well as longitudinal studies that evaluated the potential involvement of
TET2mut in clonal evolution have to be performed to further elucidate the role of
TET2mut in the development of AML.
In conclusion, in our study of younger adult patients with AML, TET2mut were only found in approximately 8% of patients and were not associated with a clinical phenotype or outcome. The impact of TET2mut will need to be revaluated in the light of additional gene mutations that influenced epigenetic regulation in AML.