High Expression of TET1 Predicts Poor Survival in Cytogenetically Normal Acute Myeloid Leukemia From Two Cohorts.

Ten-Eleven-Translocation 1 (TET1) plays a role in the DNA methylation process and gene activation. Recent reports suggest TET1 acts as an oncogene in leukemia development. However, the clinical relevance and biological insight of TET1 expression in cytogenetically normal acute myeloid leukemia (CN-AML) is unknown. In this study, quantification of TET1 transcript by real-time quantitative PCR in bone marrow blasts was performed in 360 CN-AML patients. As a result, high TET1 expression was more common in M0/M1 morphology and genes of NPM1 mutations, and underrepresented in CEBPA double allele mutations in our AML patients. In addition, we found overexpression of TET1 was associated with an inferior overall survival and event free survival in the two independent cohorts. Notably, mRNA and miRNA integrative analyses showed aberrant expression of several hub oncogenes appear to be regulated by some miRNAs like miR-127-5p, miR-494, miR-21 and miR-616 in high TET1 expressers. In conclusion, the TET1 gene expression might serve as a reliable predictor for patients survival in AML.

Introduction

Acute myeloid leukemia (AML) is a group of molecular and clinical heterogeneity of hematological malignances with adverse clinical outcome. The poor outcome may partly be explained by the lack of molecular markers to risk stratification for personalized therapy. Although chromosomal abnormalities have been validated as an effective tool for the risk stratification, chromosomal lesions are identified in approximately 50% of all AML patients. In contrast, about 50% of all AML cases are cytogenetically normal (CN) (Rowe, 2016). In order to refine classification for CN-AML patients, molecular diagnosis such as NPM1, FLT3-ITD, and CEBPA mutational analysis is crucial (Becker et al., 2010). Additionally, abnormal expression of some oncogenes like MN1, ERG, EVI1, the brain and acute leukemia cytoplasmic and isocitrate dehydrogenase 1 have been described previously in studies from other and our groups as prognostic biomarkers (Damiani et al., 2013, Ma et al., 2015, Haferlach et al., 2012).

Recently, with the extensive research of leukemogenesis and the development of new techniques, more and more novel biomarkers relevant to epigenetic alterations have emerged as candidates for disease detection, diagnosis and prognosis. For example, 5-hydroxymethylcytosine (5hmC), also called the “sixth DNA base”, is catalyzed by the Ten-Eleven-Translocation (TET) family proteins, which are involved in the DNA demethylation process. Several studies demonstrated the expression of the TET1 protein and 5hmC is markedly reduced in a wide range of solid tumors like colon cancer (Neri et al., 2015) and gastric cancer (Frycz et al., 2014), suggesting that TET1 functions as a tumor suppressor in these types of cancers. With respect to hematologic disease like leukemia, studies demonstrated an up-regulation of TET1 was observed in MLL-rearranged leukemia and plays an indispensable oncogenic role in the development of MLL-rearranged leukemia in vitro and in vivo (Huang et al., 2016, Ittel et al., 2013, Huang et al., 2013a). Further studies revealed that several essential downstream direct target genes of MLL fusions, such as HOXA9, MEIS1, and PBX3 have been shown to be critical for the development, maintenance and leukemia stem cells (LSCs) self-renewal of MLL-rearranged leukemia are also direct target genes of TET1 (Huang et al., 2016, Wong et al., 2007, Zhu et al., 2016). Taken together, these results highlight TET1 functions as a potent oncogenic role in MLL-arranged leukemia. However, it is not yet known about the clinical relevance and the upstream regulation mechanism of TET1 expression in non-MLL-arranged leukemia, especially CN-AML. In this study, we found high TET1 expression is associated with poor survival in the two independent cohort of AML patients, and provide the putative miRNA and mRNA regulation mechanism to decipher the biologic insights in high TET1 expressers with CN-AML.

Methods

Patients

AML patients with age > 14 at diagnosis were included in this study. We collected the clinical data of 360 AML patients from medical records between January 2010 and July 2016. WHO classification, conventional cytogenetic banding assay, and molecular analyses were performed centrally in Zhejiang Institute of Hematology (ZIH), China as previously described in AML diagnosis (Wang et al., 2013). Patients were treated with intensive induction chemotherapy as previous reported (Jin et al., 2013). After the complete remission induction, younger patients were treated with a high-dose of cytarabine-based chemotherapy, whereas older patients were treated in an individualized manner decided by the physicians, as described previously (Wang et al., 2013). In the consolidation therapy, twenty-nine patients in this cohort received allogeneic transplantation. In addition, the published TCGA data was used as an external validation (https://tcga-data.nci.nih.gov/tcga/) (Cancer Genome Atlas Research N et al., 2013). All of the patients provided written informed consent to participate in the study. This study was approved by the Research Ethics Committee of the First Affiliated Hospital, College of Medicine, Zhejiang University.

Cytogenetic and Gene Mutation Analysis

Chromosomal abnormalities and gene mutations of NPM1, FLT3-ITD, CEBPA, DNMT3A, IDH1 and IDH2 were analyzed by as previously described (Chen et al., 2014). The researchers conducted the above cytogenetics analyses blinded to both the levels of TET1 expression and clinical outcome.

Quantitative Reverse Transcriptase-PCR

The researchers conducted PCR experiments without the knowledge of clinical outcome. The methods for RNA extraction, complementary DNA synthesis and quantitative PCR were reported previously (Yu et al., 2017). The commonly used housekeeping (HK) genes like β-actin, α-tubulin, GUSB and GAPDH were measured. In order to select the suitable HK genes in our cohort, we performed GeNorm analysis using R package “NormqPCR” (Perkins et al., 2012). As a result, the M values of β-actin, α-tubulin, GUSB and GAPDH were 0.11, 0.08, 0.08 and 0.07, respectively. Thus, GAPDH was selected as the most stable HK gene owing to the lowest M-value. PCR reactions were performed in a total volume of 25 μl containing of 1 μl of 100 ng/μl sample cDNA, 12.5 μl of 2 × PCR Mix, 1 μl of 0.5 μM of each primer, and 10.5 μl of ddH2O. mRNA levels were normalized to GAPDH housekeeping gene. The following primers were used for quantitative PCR:

TET1 5′- TTCGTCACTGCCAACCTTAG-3′ (sense) and 5′-ATGCCTCTTTCACTGGGTG -3′ (antisense); TET2 5′- CCCTTCTCCGATGCTTTCTG-3′ (sense) and 5′-TGGGTTATGCTTGAGGTGTTC-3′ (antisense); TET3 5′- AAGACACCTCGCAAGTTCC-3′ (sense) and 5′-GTTGGTCACCTGGTTCTGATAG-3′ (antisense); GAPDH (control), 5′-ATGGGGAAGGTGAAGGTCG-3′ (sense) and 5′-GGGTCATTGATGGCAACAATATC-3′ (antisense).

MicroRNA Experiments

For the miRNA measurement, total RNA was obtained using a miRNeasy Mini Kit (Cat # 217004,QIAGEN, GmbH, Germany) following the manufacturer's instructions and subsequently checked for a RNA integrity number (RIN) to inspect RNA integration by an Agilent Bioanalyzer 2100 (Agilent technologies, Santa Clara, CA, US). The detailed methods for the miRNA experiments were described previously (Yu et al., 2017). We deposited the raw and processed data at the National Center for Biotechnology Information Gene Expression Omnibus (repository number GSE103431).

Gene Expression Arrays

Affymetrix Human Genome U133 Plus 2.0 Array GeneChip microarrays were used to assess the total RNA samples. The clinical annotation and microarray data of 12 and 10 samples were published previously (Shen et al., 2011, Wang et al., 2013). GSE27187 raw data were downloaded from the GEO database (www.ncbi.nlm.nih.gov/geo). The raw data of 22 samples were pre-processed using functional normalization in the affy R package (Gautier et al., 2004). The Cancer Genome Atlas (TCGA) mRNA gene expression profiling data were generated by Affymetrix Human Genome U133Plus2.0 GeneChips. Gene expression of TCGA data was used for extended analyses. The relative expression of TET1 in the TCGA cohort was calculated by the ratio of TET1 to GAPDH transcript levels.

Definition of Clinical End Points and Statistical Analysis

Descriptive statistics including frequency counts, median and range were used to describe patient characteristics. The main objective of this study was to explore the prognostic value of TET1 expression in AML patients. Overall survival (OS) was defined as time from date of diagnosis until death due to any cause or the last follow-up, and event-free survival (EFS) was defined as time from date of diagnosis until removal from study due to non-complete remission, relapse, or death. The adjusted variables like age, WBC, ELN favorable genotype, genes of DNMT3A, IDH1 and IDH2 mutations, BMT and treatment protocols are regarded as the well-established predictors for AML patients. Based on the rule of sample size calculation: 10 events per variable (Collins et al., 2015), we should enroll 300 patients with 90 deaths (the 3-year survival rate of Chinese patients is 30% (Wang et al., 2013)) to assess 9 variables in the Cox regression model. Because EFS estimated multiple clinical events including death, disease relapse, and treatment response, EFS was selected to estimate the cutoff value of TET1 expression. To explore the patients with the best EFS, we subdivided the development cohort into four quartiles (Q1: < 25%, Q2: 25%~50%, Q3: 50%~75%, Q4:>75%) according to the expression value of TET1 (Fig. S1). The survival curves between Q1 and Q2 were comparable, while those in Q3 and Q4 had similarly poor EFS compared with those in Q1. Thus, we dichotomized these patients into high and low TET1 groups based on the median value (Fig. 1). Comparison of survival curves were based on the Gehan-Wilcoxon and log-rank test. The proportional-hazards assumption was checked for each variable before fitting Cox models. A limma package (Ritchie et al., 2015) in R software was used to test for the difference of microRNA and mRNA signatures between high and low TET1 expressers. Hierarchical clustering based on expression levels of these mRNAs was performed and visualized by heatmap. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were created with ClueGO v2.3.4 (Bindea et al., 2013) and visualizing by Cytoscape software (Shannon et al., 2003). The toppgene analysis was done on the platform (https://toppgene.cchmc.org/prioritization.jsp). Interaction of miRNA and mRNA integrative analyses in silico were using the mirtar platform (http://mirtar.mbc.nctu.edu.tw/human/index.php). All statistical analyses were conducted with R statistic packages, version 3.4.1 (www.r-project.org). The two-sided level of significance was set at p-value < 0.05.

Fig. 1

Survival curves of CN-AML patients. Kaplan-Meier estimates of EFS (A) and OS (B) by high and low TET1 expression for our patients, and EFS (C) and OS (D) for patients from the TCGA cohort, respectively.

Results

Characteristics of CN-AML Patients with High TET1 Expression

TET1 expressions were measured in BM samples from 360 CN-AML patients at diagnosis. The median value of TET1 transcript level was 0.73 expression value of TET1/GAPDH with the interquartile range from 0.39 to 1.37. We classified patients into high and low expression based on the median value. The median and interquantile range of TET1 expressions were 1.38(0.97,2.21) and 0.39(0.26,0.53) for high and low expressers, respectively. Clinical characteristics of patients with high TET1 expression are summarized in Table 1. High TET1 expressers were more frequent in AML FAB subtype M0 (15.6% vs. 6.1%) and M1 (12.2% vs. 6.1%) morphology (P = 0.010). High expressers had the lower prevalence of CEBPA double allele mutations (9% vs. 20%, P = 0.005) than low expressers. Patients with high TET1 expression had a tendency for higher frequency of NPM1 mutations (32.5% vs. 23.3%, P = 0.071). Additionally, TET1 expressions were positively correlated to TET2 and TET3 expressions. However, there was no statistically significant correlation between TET1 expression and other variables including age, sex, white blood cell counts (WBC), hemoglobin levels, platelet counts, percentage of bone marrow blasts, genes of FLT3-ITD, DNMT3A, IDH1 and IDH2 mutations, and treatment protocols (Table 1).

Table 1

Characteristics of CN-AML patients by high and low TET1 expression.

Variables	Low expression	High expression	P value
Number(%)	180(50%)	180(50%)
Male, n(%)	107(59.4)	103(57.2)	0.748
Age, median(IQR), years	53.00(42.00,65.00)	55.00(39.00,63.00)	0.812
WBC, median(IQR), × 109/L1	12.10(2.75,49.12)	11.64(2.40,63.97)	0.825
HB, median(IQR), g/L2	87.00(69.75,103.25)	84.00(65.65,103.00)	0.292
PLT, median(IQR), × 109/L3	49.50(24.75,92.50)	46.00(26.00,88.25)	0.886
BM blast, median(IQR),%4	64.50(39.12,78.00)	67.50(41.88,83.12)	0.211
FAB classification, n(%)5			0.010
M0	11(6.1)	28(15.6)
M1	11(6.1)	22(12.2)
M2	94(52.2)	83(46.1)
M4	12(6.7)	7(3.9)
M5	48(26.7)	37(20.6)
M6	4(2.2)	3(1.7)
Genes mutations, n(%)
FLT3-ITD	37(20.8)	36(20.2)	1.000
NPM1	41(23.3)	55(32.5)	0.071
CEBPADM6	34(20.0)	15(9.0)	0.005
DNMT3A	26(15.7)	19(11.9)	0.340
IDH1	36(22.6)	26(16.0)	0.157
IDH2	22(13.9)	26(16.7)	0.533
ELN favorable group, n(%)	55(30.9)	46(25.7)	0.330
Treatment, n(%)7			0.060
DA	49(27.2)	37(20.6)
HAA	21(11.7)	36(20.0)
IA	110(61.1)	107(59.4)
BMT, n(%)	11(6.1)	18(10.0)	0.245
TET1, median(IQR)	0.39(0.26,0.53)	1.38(0.97,2.21)	< 0.001
TET2, median(IQR)	2.17(1.68,2.93)	3.01(2.19,4.21)	< 0.001
TET3, median(IQR)	0.57(0.41,0.81)	1.03(0.71,1.70)	< 0.001

Abbreviations:1WBC, white blood cell; 2HB, hemoglobin; 3PLT, platelet counts; 4BM, bone marrow; 5FAB, French–American–British classification systems; 6DM: Double-allele. 7The protocols used for induction therapy in different groups including HAA, homoharringtonine-based treatment (homoharringtonine 2 mg/m2/day for 3 days, cytarabine 75 mg/m2 twice daily for 7 days, aclarubicin 12 mg/m2 daily for 7 days) regiment; DA, daunorubicin 45 mg/m2 daily for 3 days and cytarabine 100 mg/m2 daily for 7 days; IA, idarubicin 6–8 mg/m2 daily for 7 days and aclarubicin 20 mg/m2 daily for 5 days. IQR, interquantile. BMT, bone marrow transplantation. ELN (European leukemia Net) favorable genotype represents NPM1 mutant and FLT3-ITD negative or double allele CEBPA mutations.

Overexpression of TET1 Associated with Poor Clinical Outcome in CN-AML Patients

In our patients, three years overall survival (OS) rate and event free survival (EFS) rate were 34% and 31%, respectively. High TET1 expressers (n = 180) had more adverse OS, EFS, and complete remission rate compared to low expressers (n = 180) (Table S1, Fig. 1A–B). However, there was no significant difference between high and low TET2 or TET3 expressers for OS and EFS (Figs. S2–3). Moreover, in the subgroup analyses we found TET1 expressions can further stratify CN-AML patients with favorable ELN genotype into different subgroups with respect to OS and EFS (Fig. S4). In order to exclude the potential confounders factors, we conducted multivariate analyses. As shown in Table 2, TET1 expression was still an independent prognostic factor after adjusting for age, WBC, ELN risk groups, and genes of DNMT3A, IDH1 and IDH2 mutations regardless of CR [OR (95%CI), 0.354(0.204, 0.602); P < 0.001], OS [HR (95%CI), 1.614(1.154, 2.256); P = 0.005] and EFS [HR (95%CI), 1.607(1.169, 2.21); P = 0.004]. Moreover, we also conducted landmark analysis by including patients whose survival was > 30 days in order to ignore the cause of induction death by intense chemotherapy. As a result, high expression of TET1 was still independently associated with poor OS [HR (95%CI), 1.652(1.139, 2.396), P = 0.010], EFS [HR (95%CI), 1.629(1.148, 2.31), P = 0.010] and CR[OR (95%CI), 0.336(0.184, 0.598); P < 0.001] in the multivariate survival analyses (Table S2).

Table 2

Multivariable analyses of clinical outcome for patients with CN-AML.

Variables	Complete remission		Overall survival		Event free survival
	P value	OR (95%CI)	P value	HR (95%CI)	P value	HR (95%CI)
TET1 expression (High vs.Low)	< 0.001	0.354(0.204, 0.602)	0.005	1.614(1.154, 2.256)	0.004	1.607(1.169, 2.210)
Age	0.023	0.519(0.294, 0.912)	< 0.001	2.132(1.506, 3.016)	< 0.001	1.996(1.436, 2.776)
WBC	0.062	0.600(0.349, 1.020)	< 0.001	1.976(1.401, 2.787)	0.001	1.724(1.242, 2.394)
ENL favorable group	0.009	2.300(1.249, 4.355)	< 0.001	0.401(0.260, 0.620)	< 0.001	0.466(0.311, 0.696)
DNMT3A	0.267	0.660(0.313, 1.372)	0.001	2.087(1.362, 3.198)	< 0.001	1.941(1.297, 2.904)
IDH1	0.035	0.460(0.221, 0.943)	0.042	1.596(1.016, 2.506)	0.019	1.667(1.088, 2.554)
IDH2	0.173	0.611(0.299, 1.239)	0.658	0.901(0.569, 1.428)	0.881	1.034(0.669, 1.599)
Treatment protocols
HAA vs. DA	0.024	2.731(1.154, 6.643)	0.06	0.601(0.354, 1.022)	0.084	0.645(0.392, 1.060)
IA vs. DA	0.001	3.098(1.652, 5.929)	0.001	0.52(0.354, 0.766)	< 0.001	0.565(0.389, 0.820)
BMT	0.251	1.909(0.669, 6.349)	0.014	0.368(0.166, 0.817)	0.097	0.566(0.289, 1.108)

Abbreviations: WBC, white blood cell; The treatment protocols used for induction therapy in different groups including HAA, homoharringtonine-based treatment (homoharringtonine 2 mg/m2/day for 3 days, cytarabine 75 mg/m2 twice daily for 7 days, aclarubicin 12 mg/m2 daily for 7 days) regiment; DA, daunorubicin 45 mg/m2 daily for 3 days and cytarabine 100 mg/m2 daily for 7 days; IA, idarubicin 6–8 mg/m2 daily for 7 days and aclarubicin 20 mg/m2 daily for 5 days. BMT, bone marrow transplantation. ELN (European leukemia Net) favorable genotype represents NPM1 mutant and FLT3-ITD negative or double allele CEBPA mutations.

Validation of the Prognostic Impact of TET1 Expression in an Independent Cohort of CN-AML Patients

In this study, we firstly selected 91 CN-AML patients with normal karyotype based on cytogenetic classification from the TCGA cohort to validate our results. We defined the cutoff value using the same method based on the median value of TET1 expression in the TCGA cohort. Consequently, poor survivals were still observed in patients with high TET1 expressers (Fig. 1C–D). However, no significant differences of OS and EFS were observed in the TCGA cohort with respect to TET2 or TET3 expressions (Fig. S5). Secondly, in the 197 total AML patients, TET1 expressions (continuous variable, log2 transformation) were associated with poor OS [HR(95%CI): 1.13(0.99,1.28), P = 0.075] and EFS[HR(95%CI): 1.15(1.02,1.29), P = 0.024], respectively. Thus, the expression of TET1 was indeed a poor predictor in different populations.

mRNA Expression Profiling Associated with High TET1 Expressers

In order to further obtain biological insights of the aberrant TET1 expressions on leukemia development, we conducted the metaanalysis in gene expression profiles from 91 CN-AML from the TCGA cohort (Cancer Genome Atlas Research N et al., 2013) and 22 CN-AML patients in our previously published researches (Shen et al., 2011, Wang et al., 2013). As a result, a total of 1127 genes including 863 upregulated genes and 264 downregulated genes were identified in high expressers compared to low expressers using the false discovery rate (FDR) < 0.05 as threshold values (Fig. S6 and Table S3). In the KEGG pathways enrichment analysis, we found TET1 high expressers involved in the upregulated pathways such as cell cycle, purine metabolism, pyrimidine metabolism, RNA transport and ribosome biogenesis pathways (Fig. S7), and the downregulated pathways such as influenza A, endocytosis, MAPK singling pathway, proteoglycans in cancer and tuberculosis (Fig. S8). Notably, we found several upregulated hub genes involving at least 2 different pathways such as RPP40, RPP38, POP1, POLD2, POLR1C, POLR3D, POLR2D, POLR1E, POLR2F and ZNRD1 (Fig. S7) and downregulated hub genes of HSPA6, TLR4, MAPK13, MAP2K1, FGFR1, CD14 and TLR2 in high TET1 expressers (Fig. S8). We also found 4 genes like FAM69B, NPR3, CD34, SLC37A1 in RNA-sequencing group 6 of the NEJM AML study (Cancer Genome Atlas Research N et al., 2013) were upregulated in high TET1 expressers. Additionally, 230 TET1 co-expressed genes in the TCGA cohort (Jamalpour et al., 2017) can be found as part of differently expressed genes in high TET1 expressers (Table S3). Moreover, these differently expressed genes were enriched in 16 molecular functions: RNA and DNA binding, methyltransferase activity, DNA binding transcription factor activity, lipopolysaccharide receptor activity, etc. (Table S4).

Changes of miRNA Expression in High TET1 Expression

In order to exclude the effect of variables on the microRNA expression, we applied the propensity score analysis to match each CN-AML patient with high and low TET1 expression based on variables are like age, WBC, and genes of FLT3-ITD, NPM1, CEBPA, DNMT3A, IDH1 and IDH2 mutations in our cohort (Table S5). Thus, we selected 12 paired samples with high and low TET1 expression to assess the differences of microRNA (miR) expression. As a result, we identified 199 differently expressed miRs including 111 downregulated and 88 upregulated in high TET1 expressers (P-value < 0.05, Fig. 2A, Table S6). Among these miRs, up-regulation of miR-127 and miR-494, down-regulation of miR-21 and miR-616 were seen significant changes in high TET1 expressers from an independent cohort of TCGA patients with CN-AML (Fig. 2B–E).

Fig. 2

Heatmap plot illustrating the microRNAs expression between high and low TET1 expression(A). Validation of microRNAs expression (B-E) in patients with low vs high TET1 expression in TCGA cohort.

Integrative Analysis of mRNA and miRNA Interaction

In order to further understand the regulated mechanism, we conducted the integrative analysis of mRNA and miRNA interaction between high and low expressers. Specifically, thirty-five genes upregulated in high TET1 expressers were predicted to be targeted by miR-616 and miR-21 (Fig. 3). By contrast, we found 22 genes downregulated in high TET1 expressers were predicted to be targeted by high expression of miR-127-5p and miR-494, respectively.

Fig. 3

MiRNAs and mRNAs interactive effects in the high TET1 expressers. Red color labels represents overexpression of miRNAs and mRNAs and blue color represents downregulated expression in the high TET1 expressers. The targeting relationships between miRNAs and mRNAs are shown by solid lines.

After combining the differentially expressed miRNAs and pathways together, we found miR-494 and miR-127-5p were upregulated in high TET1 expressers (Table S5 and Fig. S8), and genes of RAB11FIP5, RAB11FIP1 in the endocytosis pathway and ITGAX in the tuberculosis pathway were downregulated (Table S3). In silico analysis, miR-127-5p can regulate genes of RAB11FIP1 in the endocytosis pathway and ITGAX in the tuberculosis pathway, and miR-494 can regulate RAB11FIP5 in endocytosis pathway (Fig. S8), implicating some novel miRNA-mRNA regulated pathways occurring in high TET1 expressers.

Discussion

In the MLL-rearranged leukemia, TET1 acts as an oncogene. However, the clinical relevance and prognostic significance of TET1 expression in non-MLL-rearranged leukemia, particularly in CN-AML are still unknown. In this study, we found CN-AML patients with high TET1 expression obtained higher frequency of NPM1 mutations and FAB M0/1 morphology. These results supported the hypothesis that increasing TET1 expression might block hematopoietic stem cell differentiation. Taken together, the clinical characteristics of high TET1 expressers implied TET1 expression level might be associated with a poor outcome. As expected, we found that high TET1 expressers harbored poor overall survival in the two independent cohorts of CN-AML patients. This result is also supported by the study that TET1 acts as an oncogene in leukemia development (Huang et al., 2013a).

In order to further understand the underlying biological insights in high TET1 CN-AML, we analyzed genes expression profiles in two independent cohorts of CN-AML patients. First, we found 1127 significantly expressed genes in high TET1 expressers. These interesting genes were involved in 10 pathways together with several hub genes, which can be used as potential drug targets in the future. Specifically, in high TET1 expressers the downregulated hub gene like RAB11FIP1 has been described as tumor suppressors, and the downregulated pathways involving in innate immune reaction like endocytosis, tuberculosis, influenza A and MAPK signaling pathways might lead to leukemia development (Boulay et al., 2016). At the same time, high TET1 expressers exhibit high activity of DNA and protein turnover, presenting with upregulation of purine and pyrimidine metabolism, RNA transport and ribosome biogenesis. These results might help us to explain the poor survival of the high TET1 expression in CN-AML.

In order to further understand the biological links, we analyzed miRNAs profiles relevant to TET1 expression. As a result, we found four miRNAs, miR-127-5p, miR-494, miR-616 and miR-21, with robust changes in high TET1 expressers from the different cohorts of patients. The function of miR-494 and miR-127 is reported as an oncogene in the different context of tissues. For example, several authors have described them as oncomiRs in lung cancer (Shi et al., 2017), colorectal cancers (Sun et al., 2014), lymphoma and leukemia (Shi et al., 2017, Diakos et al., 2010). Here, we found miR-127-5p and miR-494 were positively correlated with the high TET1 expression, supporting their function as oncogenes in CN-AML. Moreover, their oncogenic function can be deciphered by their putative targeting genes, which have been descripted as tumor suppressors like ITGAX (Ma et al., 2015) and SLC46A2 (Kim et al., 2015). Notably, we also found downregulated SGSH expression was predicted to be targeted by miR-127 and miR-494 in high TET1 expressers (Fig. 3). SGSH is one of enzymes in the lysosomal degradation, which often correlates with poor prognosis and increased recurrence in many cancers (Piao and Amaravadi, 2016). Recently, targeting lysosomes is reported as a novel therapeutic strategy to eradicate imatinib-resistant chronic myelogenous leukemia cells (Piao and Amaravadi, 2016, Puissant et al., 2010). These results implied miR-127 and miR-494 might subvert lysosomal proteases via downregulated expression of SGSH leading to leukemia development. Thus, the poor prognosis of high TET1 expressers maybe explained by the aberrant expressions of miR-127 and miR-494 and their targeting genes. With respect to the function of miR-21, it has been reported miR-21 is required for maintaining the imatinib-resistant phenotype of CML stem cells (Wang et al., 2015), and upregulation of miR-21 is a poor prognostic marker in acute lymphoblastic leukemia (ALL) (Labib et al., 2017). These reports supported miR-21 as an oncogene in ALL and CML leukemia. However, downregulated expression of miR-21 in AML was also reported (Diaz-Beya et al., 2013) and seemed to act as a tumor suppressor. Here, we found miR-21 was significantly down-regulated in high TET1 expressers, supporting the function as a tumor suppressor. The exact function of miR-616 is not known in hematological disease, but it is thought to play a tumor suppressor role in CN-AML. In the current study, we found miR-616 was downregulated in high TET1 expressers and negatively correlated to the expression of several targeting genes which have been demonstrated as a tumor promoter like CHD6V (Douet-Guilbert et al., 2015), TCF4 (Ishiguro et al., 2016), VANGL1 (Cetin et al., 2015), or associated with poor clinical outcome such as WARS2 (Huang et al., 2013b) and NOP16 (Zhang et al., 2014). With regards to the differentially expressed miRNAs and their targeting KEGG pathways, we found miR-494 and miR-127-5p was respectively involved in downregulation of the tuberculosis and endocytosis pathways by targeting genes of RAB11FIP5, ITGAX and RAB11FIP1 in high TET1 expressers (Fig. S8). Thus, targeting miR-494 expression and miR-127-5p may be a novel treatment strategy in CN-AML with high TET1 expressers. These results also need more experimental clarification.

In conclusion, we uncovered high TET1 expression could predict unfavorable survival in CN-AML patients. It is worthy to note that the prognostic value of the biomarker is dependent on the different treatment protocols. For example, the poor predictor of DNMT3A mutations might disappear when AML patients received high dose of daunorubicin induction therapy (Sehgal et al., 2015). Therefore, an analogous result might occur if patients with high TET1 expression receive the strikingly intensive treatment. Additionally, we found four miRNAs and a group of targeting genes coexpressed with TET1. These biological links can help us answer the questions that high TET1 expression contributes to poor clinical outcome. At the same time, these hub genes and the miRNA interaction uncover several important regulatory networks, but luciferase reporter assays are required to further study in the future. Thus, this study indicated routine measurement of TET1 expression at diagnosis can add to risk classification and therapy decision making for CN-AML patients.