Background

Acute myeloid leukemia (AML) is a group of clonal malignant diseases that derive from the hematopoietic stem cells. AML is characterized by a large group of germinal cells, which leads to a loss of normal hematopoietic function [1]. The clinical prognosis of patients with AML is various. The differences in outcomes among AML patients depend on multiple intrinsic factors [2, 3]. With the development of methodologies of massive sequencing, it has been demonstrated that somatic mutations in NPM1, FLT3, CEBPA, IDH1 and IDH2 are connected to prognosis in AML [4]. To be specific, patients with mutated FLT3 have a dismal outcome, while mutations in NPM1 and CEBPA are related with favorable prognosis. The advent of chemotherapy and allogeneic hematopoietic stem cell transplantation (allo-HSCT) has significantly improved AML treatment outcomes [5]. Relapse and refractory of leukemia remain the most disturbing problems in AML patients [6]. Thus, it is urge to explore more reliable and effective prognostic biomarkers to enhance the capacity of prediction and thus improve the outcome of AML by choosing optimal therapeutic approach.

MicroRNAs are short non-coding RNAs, which are implicated in a diverse group of critical cellular mechanisms, such as apoptosis, differentiation, cell cycle progression, and immune responses [7]. Recently, more and more attentions have been focused on the prognostic role of microRNAs in AML. A recent study has shown that the upregulation of miR-181a facilitates better survival of AML patients who are cytogenetically normal [8]. However, AML patients who are cytogenetically normal and upregulated miR-212 and miR-3151 have shorter overall and disease-free survival [9, 10]. However, most microRNA analyses did not differentiate the AML patients treated with chemotherapy and allo-HSCT. Thus, microRNAs may have varied prognostic roles in chemotherapy and allo-HSCT treatment group, respectively.

MiR-25, a member of miR-106b-25 cluster, is located on human chromosome 7q22.1 [11]. Previous studies revealed that miR-25 was involved in many kinds of cancers [12]. It has been identified that miR-25 is a potential biomarker for pediatric AML based on Pipeline of Outlier MicroRNA Analysis (POMA) model [13]. More important, Garzon et al. [14] reported that miR-25 is significantly down-regulated in 122 newly diagnosed AML samples compared with CD34+ normal cells. However, clinical and prognostic role of miR-25 in AML are still unclear. A total of 162 recently diagnosed de novo AML patients were enrolled in this evaluation. The cases were placed into two groups based on the treatment that they received. The present study suggested that miR-25 is a solitary AML prognostic biomarker. Furthermore, our study revealed that allo-HSCT would be more beneficial to patients showing downregulated miR-25.

Patients and methods

Patients

Approximately 162 patients with a diagnosis of de novo AML were included in this study. The data sets used in this investigation were acquired from The Cancer Genome Atlas (TCGA). A single-institution tissue banking strategy endorsed by the human studies committee of Washington University was used in this study. All of the patients provided their written informed consent. AML diagnosis and classification were made according to the French–American–British (FAB) and the World Health Organization (WHO) criteria. The cases were placed into two groups according to the clinical treatment received. Ninety patients accepted chemotherapy, and the rest accepted allo-HSCT.

Gene-expression profiling

The samples from 155 patients both had been obtained mRNA and microRNA expression data. These data were applied to identify the mRNA-expression signature associated with miR-25 expression. The sequencing read count for each miRNA was normalized to Reads per million reads (RPM). The mRNA expression values were logged (base 2) prior to analysis [15]. Spearman correlation was used to correlate the mRNA-expression profile with miR-25 expression. Hierarchical clustering analysis was used to reorder the gene rows. To screen for target genes of miR-25, Targetscan, miRNApath and miRDB website tools were implemented. Gene Ontology enrichment assessment of genes in miR-340 related signature was performed with the Database for Annotation, Visualization, and Integrated Discovery (DAVID).

Statistical analysis

The clinical endpoints of this investigation on treatment outcomes included overall survival (OS) and event-free survival (EFS). OS pertains to the time interval from diagnosis to death or last follow-up of the patient. EFS is described as the time interval from diagnosis to disease progression, relapse, or death attributed to any cause. The patients were assigned to the high or low expression groups based on the median miR-25 expression. Descriptive statistics (median and/or range) were used to summarize patients’ clinical and molecular characteristics. To elucidate the role of miR-25 expression in AML clinical and molecular features, the Pearson Chi-square and Fisher’s exact tests were used to screen for significant differences between two categorical variables. In addition, the Mann–Whitney’s U test was used for continuous variables. For univariable and multivariable analysis, a Cox proportional hazards model was employed to determine the effect of various risk factors on patient OS and EFS. The limited backward elimination procedure was applied to assess hazard ratios (HRs) and P values. Kaplan–Meier analysis was performed to determine the impact of miR-25 expression on OS and EFS. Statistical analysis was conducted with SPSS and GraphPad Prism. Differences among variables were determined to be statistically significant when the P value was < 0.05.

Results

Correlation analysis of miR-25 expression and clinical characteristics

To establish the correlation among miR-25 expression and various clinical profiles, we assigned the patients who underwent chemotherapy and allo-HSCT to one of two groups according to median miR-25 expression levels, respectively. The associations of the clinical features with miR-25 expression levels are summarized in Table 1. In the chemotherapy group, subjects who exhibited upregulated miR-25 had a higher percentage of RUNX1-RUNX1T1 compared to those with downregulated expression (P = 0.026). In addition, high miR-25 expresser involved in more good risk cases of AML (P = 0.002). However, no significant differences were observed in gender, age, WBC count, BM blast, PB blast, FAB subtypes, FLT3-ITD, NPM1, DNMT3A, RUNX1, MLL-PTD, TP53, IDH1 and IDH2 among the high and low miR-25 expression group. In the allo-HSCT group, study participants with upregulated miR-25 exhibited a lower frequency for FLT3-ITD mutations (P = 0.045) compared to those with downregulated miR-25. No significant differences in as far as gender, age, WBC count, BM blast, PB blast, and mutations in the NPM1, RUNX1, DNMT3A, MLL-PTD, IDH1, IDH2, and TP53 genes were observed among the upregulated and downregulated miR-25 groups.

Table 1 Comparison of clinical and molecular characteristics with miR-25 expression in patients with AML
Full size table

Prognostic value of miR-25 profiles in AML patients

We performed Kaplan–Meier analysis and log-rank test to assess the prognostic value of miR-25 profiles in AML patients. The chemotherapy group showed that AML patients with upregulated miR-25 were connected to better EFS (P = 0.019) and OS (P = 0.0086) relative to those with downregulated miR-25 (Fig. 1a, b). However, AML patients who received allo-HSCT did not exhibit any connection among prognosis and miR-25 expression (Fig. 1c, d). These findings revealed that miR-25 may be utilized as a chemotherapy-specific prognostic marker for AML.

Fig. 1
figure 1

Kaplan–Meier survival curves of AML patients stratified based on miR-25 expression. a, b In the chemotherapy group, the high miR-25 expressers had significantly prolonged OS and EFS (n = 90) compared with low miR-25 expressers. c, d There were no significant differences in patients undergoing allo-HSCT between high and low miR-25 groups (n = 72)

Full size image

High level of miR-25 is independently associated with favorable prognosis

To determine whether miR-25 expression could be used as an independent predictor for AML patient survival, we conducted univariate and multivariate Cox analyses. For the chemotherapy group, univariate analysis revealed that the upregulation of miR-25 was connected with longer EFS (HR = 0.598, 95% CI 0.376–0.951, P = 0.030) and OS (HR = 0.556, 95% CI 0.347–0.890, P = 0.015). Furthermore, multivariate cox analysis indicated that miR-25 upregulation was independently connected with longer EFS (HR = 0.561, 95% CI 0.333–0.943, P = 0.029) and OS (HR = 0.502, 95% CI 0.296–0.851, P = 0.011) after adjustment of mutation status for the FLT3-ITD, NPM1, DNMT3A, RUNX1, IDH1, and IDH2 genes and WBC count (Table 2).

Table 2 Univariate and multivariate analyses in patients treated with chemotherapy
Full size table

Univariate analysis of the allo-HSCT group suggested that AML cases harboring FLT3-ITD mutations had shorter EFS (HR = 1.873, 95% CI 1.020–3.437, P = 0.043) and OS (HR = 1.998, 95% CI 1.053–3.788, P = 0.034). Patients with mutations only in the RUNX1 gene exhibited shorter OS (HR = 2.253, 95% CI 1.046–4.849, P = 0.038). Multivariate analysis indicated that FLT3-ITD and RUNX1 remained independent outcome predictors after adjusting for all other prognostic factors (Table 3). However, allo-HSCT patients did not show any significant differences between upregulated and downregulated miR-25 expression.

Table 3 Univariate and multivariate analyses in patients treated with allo-HSCT
Full size table

Allo-HSCT may circumvent poor patient outcomes that are related to downregulated miR-25 expression

To determine whether allo-HSCT therapy could circumvent the severe prognosis that was associated with downregulated miR-25, the whole cohort of 162 cases was split into two groups according to the median miR-25 expression levels. In the downregulated miR-25 group, the AML cases who received allo-HSCT showed significantly longer EFS (HR = 0.515, 95% CI 0.327–0.831, P = 0.0069) and OS (HR = 0.405, 95% CI 0.250–0.639, P = 0.0002) relative to cases who underwent standard chemotherapy alone (Fig. 2a, b). For the upregulated miR-25 group, no obvious differences in EFS (P = 0.969) and OS (P =0.364) were observed among the allo-HSCT and chemotherapy regimens. Thus, the AML patients showing downregulated miR-25 may benefit from treatment with allo-HSCT.

Fig. 2
figure 2

Allo-HSCT treatment circumvents the unfavorable outcomes of AML patients showing downregulated miR-25 expression. a, b A total of 162 cases were placed into two groups according to the median miR-25 expression levels. In the downregulated miR-25 group, the Kaplan–Meier survival curves of AML patients classified based on chemotherapy (n = 52) and allo-HSCT (n = 29) treatment. c, d In the upregulated miR-25 group, the Kaplan–Meier survival curves of AML patients classified based on chemotherapy (n = 38) and allo-HSCT (n = 43) treatment

Full size image

Biological insights into miR-25 profiles in AML

To generate insights into the molecular mechanism of miR-25, we analyzed a gene expression signature that was connected with miR-25 expression among AML cases. An association between the expression of 205 genes and miR-25 was observed. Among these genes, 145 were negatively correlated and 60 were positively correlated with the expression of miR-25 (Fig. 3). MiR-25 expression was inversely correlated with the expression of HOXA and HOXB, as well as the HOX cofactor MEIS1. Notably, these genes are crucial for the leukemogenesis and self-renewal capacities of AML [8, 16, 17]. Furthermore, we discovered that the expression of miR-25 was negatively connected with the levels of the PRDM16, Which involved in AML translocation [18]; CD97, an EGF-TM7 receptor [19]; IRAK1, which activates NF-κB pathways by the interaction with TRAF6 [20]; NFKB2, a pro-inflammatory response gene [21]; MYH9, which predicts unfavorable outcome of AML [22]; HDAC11, a epigenetic regulator. Notably, MYH9 was a in silico predicted target of miR-25.

Fig. 3
figure 3

Heat map of miR-25-associated gene-expression signature in patients with AML. The columns represent patients and the rows represent genes. The columns are ordered from left to right according to increasing expression levels of miR-25. The hierarchical cluster analysis was performed to order rows. The expression levels of various genes are represented by nodes of different colors, ranging from the lowest (green) to the highest (red)

Full size image

Gene Ontology showed that genes involved in cellular metabolic process, system development, immune system process, transcription, hematopoietic or lymphoid organ development, hemopoiesis and myeloid cell differentiation were markedly overrepresented among differentially expressed genes associated with miR-25 expression (Table 4).

Table 4 Gene ontology terms of biological processes in the miR-25 associated expression profile
Full size table

Discussion

AML has been considered to occur as the result of genetic abnormalities, including chromosomal rearrangements, gene deregulations and mutations [23]. The deregulated expression of microRNAs in AML can influence cell proliferation, survival and hematopoietic differentiation [24]. The association of microRNAs with prognosis in heterogeneous patients with AML is still largely unclear. In this evaluation, the upregulated of miR-25 was determined to be an independently favorable prognosticator of AML cases who were administered chemotherapy. Furthermore, allo-HSCT may overcome the poor prognosis of AML cases with low miR-25 expression.

A correlation between aberrant miRNA expression and AML prognosis has been established [25, 26]. However, most of previous microRNA markers is restricted to AML without cytogenetic abnormalities. In our study, univariate and multivariate analyses demonstrated that miR-25 is an independently biomarker for cases administered chemotherapy. High miR-25 expression can predict favorable outcome. The prognostic role of miR-25 is different with previously established prognostic factors in a heterogeneous population of AML. MiR-25, as an independent outcome predictor, may improve the current clinical risk-based classification of patients with AML.

To further understand the biological insight into the molecular mechanism underlying miR-25, we identified genes significantly correlated with miR-25 expression. We discovered that the expression of miR-25 negatively connected with the levels of PRDM16, HOXAs, HOXBs, MEIS1, CD97, IRAK1, NFKB2 and MYH9. HOXA and HOXB gene clusters are the common characters of AML [27, 28]. Of these genes, HOXB4 is positively involved in the renewal of hematopoietic stem cell [29, 30]. A previous study has shown that HOXA9 contributes to the proliferation, apoptosis, and differentiation processes of leukemia [31]. In addition, HOXA9 has been correlated with poor AML prognosis [32]. Prior evaluations have revealed that IRAK1 may be utilized as a therapeutic target for AML, and TRAF6 may be used to activate pathways such as NFKB, MAPK, and AKT [20, 33]. PRDM16, also known as MEL1, is highly homologous to MDS1/EVI1. High expression of PRDM16 can predict the adverse outcome of AML [18]. Moreover, MYH9 has also been predicted as a direct target of miR-25. High expression of MYH9 can induce resistant to chemotherapy and predict poor clinical outcome in AML [22]. Taken together, the miR-25-associated gene-expression profiling analyses provide insights into the leukemogenic role of genes that are either direct or indirect targets of miR-25. Therefore, the miR-25-associated gene-expression signature analysis give novel insights into the oncogenic role of these genes. These miR-25-related genes could contribute to the chemotherapeutic responses of AML patients.

The FMS-like tyrosine kinase 3 (FLT3) gene is pivotal to hematopoietic stem cell proliferation and differentiation [34]. FLT3 mutations take a great account of most frequent genetic aberrations in AML [35]. FLT3-ITD mutation is one of FLT3 mutations, which can keep the tyrosine kinase persistently active, and result in the abnormal proliferation of leukemic cells. Mutations in the FLT-ITD gene have been associated with higher risk for relapse and poor OS and EFS [36]. Consistent with the conclusion, our data suggested that FLT3-ITD mutation is a poor outcome marker in patients undergoing allo-HSCT. These analysis results indicate that allo-HSCT cannot overcome all adverse prognosis of molecular markers. The findings of this study have revealed that allo-HSCT circumvents the poor chemotherapy outcomes that are related to downregulated miR-25 expression. Thus, low miR-25 expression may be employed as a predictor of adverse prognoses among patients who received chemotherapy, as well as identify patients who require strategies in selecting the best treatment regimen, i.e., chemotherapy and/or allo-HCST.

Conclusion

In conclusion, high expression of miR-25 was identified to independently predict favorable survival in a highly heterogeneous population of patients with AML. Our findings may offer more information for the therapeutic strategies and the prediction of patients with AML, which may improve the survival and reduce the relapse of them. More importantly, allo-HSCT circumvents poor chemotherapeutic outcomes in cases with downregulated miR-25. The expression levels of miR-25 may thus be utilized in determining whether chemotherapy or allo-HSCT is the optimal treatment regimen for a specific AML patient.