Background

Acute myeloid leukemia (AML) is a clonal disease characterized by amplification of immature myeloid progenitors with differentiation arrest in the bone marrow (BM), finally resulting in hematopoietic failure [1]. AML is a clinically, cytogenetically and molecularly heterogeneous disease with variable clinical outcomes [1]. The features of morphology, immunology, cytogenetics and molecular biology (MICM) are the basis for AML diagnosis [1]. Cytogenetic abnormalities also provide the most important prognostic information of AML [2]. Molecular biological alterations, such as gene mutations and aberrant gene expression, also play important roles in leukemogenesis and predict treatment response and patient survival [2]. Therefore, the identification of biological markers to develop a better prognostic, diagnostic and therapeutic risk stratification for AML is of great importance.

The inhibitor of differentiation/DNA binding (ID) genes (ID1/ID2/ID3/ID4) encode ID proteins that are transcriptional regulators controlling the timing of cell fate determination and differentiation in stem and progenitor cells during physiological development [3, 4]. It was suggested that ID proteins could have key roles in cancer development [3, 4]. At the same time, dysregulated ID gene expression was linked to tumor growth, invasiveness, metastasis, angiogenesis and patient survival [3, 4]. ID1 and ID2 overexpression has been shown to correlate with enhanced malignant potential in various types of cancers including AML [3, 4]. Although increased ID1 expression was observed in AML patients, the prognostic value of ID1 overexpression remains controversial [5,6,7]. In addition, the prognostic effect of ID2 overexpression in AML was reported in our previous study [8]. In contrast, ID4 functioned as a tumor suppressor presenting a paradigm shift in the context of ID1 and ID2 during the process of tumorigenesis and leukemogenesis [3]. ID4 hypermethylation was an independent factor that affected clinical outcome and predicted leukemic transformation in patients with myelodysplastic syndrome (MDS) [9]. However, the function of ID3 and its expression pattern in AML are not completely understood. Herein, we systematically explored the expression and clinical implications of ID genes expression in AML.

Materials and methods

Patients from public datasets and our hospital

The identification cohort included 173 AML patients with ID gene (ID1/ID2/ID3/ID4) expression data (RNA-Seq V2 data) from public The Cancer Genome Atlas (TCGA) datasets [10]. All AML patients received standard chemotherapy as induction therapy. Following induction therapy, 100 patients underwent chemotherapy only, whereas the remaining 73 patients underwent auto/allo-hematopoietic stem cell transplantation as consolidation treatment. The ID gene expression in AML compared with controls was analyzed by GEPIA [11].

The validation cohort contained 107 AML patients treated at the Affiliated People’s Hospital of Jiangsu University. Patients with antecedent hematological diseases or therapy-related AML were excluded. The clinical characteristics of the AML patients are shown in Supplementary Table S1. BM samples were collected from AML patients once they were diagnosed. A total of 32 healthy donors served as normal controls. The age of the AML patients (median 57, range 18–87) showed no significant differences from that of the controls (median 52, range 20–66) (P > 0.05). This current study was approved by the Ethics Committee of the Affiliated People’s Hospital of Jiangsu University, and all the individuals provided written informed consent.

RNA isolation and reverse transcription

BM mononuclear cells (BMMNCs) were separated through gradient centrifugation using Lymphocyte Separation Medium (Solarbio, Beijing, China), and then used for total RNA extraction by TRIzol reagent (Invitrogen, Carlsbad, CA). Reverse transcription was performed to synthesize cDNA as reported [12,13,14].

RT–qPCR

The detection of ID3 and ABL1 (housekeeping gene) mRNA was determined by real-time quantitative PCR (RT–qPCR) using AceQ qPCR SYBR Green Master Mix (Vazyme, Piscataway, NJ). The primers applied for ID3 expression detection were 5’-ACTCAGCTTAGCCAGGTGGA-3’ (forward) and 5’-AAGCTCCTTTTGTCGTTGGA-3’ (reverse), whereas those for ABL1 expression detection were 5’-TCCTCCAGCTGTTATCTGGAAGA-3’ (forward) and 5’-TCCAACGAGCGGCTTCAC-3’ (reverse). The relative ID3 mRNA level was measured according to the 2−∆∆Ct method [12,13,14].

Bioinformatics analysis

All procedures regarding the bioinformatics analysis were carried out as described in our previous studies [15, 16]. To obtain the differentially expressed genes/miRNAs (DEGs/DEmiRs), analysis of the RNA sequencing data was conducted based on the raw read counts with the R/Bioconductor package “edgeR”. All statistical analyses were controlled for the false discovery rate (FDR) by the Benjamini–Hochberg procedure.

Statistical analysis

Statistical analysis was carried out based on the SPSS 20.0 and GraphPad 5.0 software. Comparisons of continuous and categorical variables were conducted using the Mann–Whitney U test/Kruskal–Wallis test and Pearson’s χ2 test/Fisher’s exact test, respectively. Kaplan–Meier analysis (log-rank test) and Cox regression (proportional hazards model, backward method) were used to analyze the effect of ID1/ID2/ID3/ID4 expression on survival including disease-free survival (DFS) and overall survival (OS). The ability of ID3 expression to discriminate in AML patients from controls was evaluated by the receiver operating characteristic (ROC) curve and area under the ROC curve (AUC). A two-sided P value less than 0.05 was considered statistically significant in all analyses.

Results

Identification of reduced ID3 expression among ID genes correlated with prognosis in AML from public TCGA datasets

We first searched GEPIA to determine the expression of ID genes (ID1/ID2/ID3/ID4) in AML. As presented in Fig. 1, the expression of ID1 and ID3 was markedly downregulated (both P < 0.001), and the expression of ID2 was greatly upregulated in AML (P < 0.001), whereas ID4 showed no dramatic difference in expression (P > 0.05).

Fig. 1 
figure 1

Expression of ID genes in AML. a ID1 expression in AML; b ID2 expression in AML; c ID3 expression in AML; d ID4 expression in AML. *: P < 0.001. The expression of ID genes expression in AML compared with controls is analyzed in GEPIA (http://gepia.cancer-pku.cn/)

Full size image

Next, to investigate the prognostic significance of the ID genes in AML, we evaluated the impact of ID gene expression on OS and DFS times by Kaplan–Meier analysis. When analyzing the prognostic value, the AML patients were divided into two groups by the median level of ID gene expression. In all the AML patients, only high ID2 expression was markedly associated with a shorter OS time (P = 0.023), whereas the other ID members did not affect either OS or DFS time (P > 0.05) (Fig. 2). In cytogenetically normal AML (CN-AML) patients, lower ID3 and ID4 expression was nearly or markedly correlated with shorter OS (P = 0.027 and 0.034, respectively) and DFS (P = 0.037 and 0.056, respectively), whereas the other ID members did not affect either OS or DFS times (P > 0.05) (Fig. 2).

Fig. 2
figure 2

Prognostic value of ID genes in AML. The prognostic effect of ID genes (ID1/ID2/ID3/ID4) on overall survival and disease free survival were analyzed by Kaplan–Meier method

Full size image

Finally, we further investigated the impact of ID gene expression on OS and DFS times in AML by Cox regression analysis. In all the AML patients, the expression of ID1, ID2 and ID3 independently affected the OS time (P = 0.016, 0.039 and 0.028, respectively) (Table 1), whereas the expression of ID1 and ID3 independently affected the DFS time (P = 0.043 and 0.022, respectively) (Supplementary Table S2). Among CN-AML patients, only ID3 expression independently affected both the OS and DFS times (P = 0.030 and 0.041, respectively) (Table 1 and Supplementary Table S2).

Table 1 Cox regression univariate and multivariate analysis of variables for overall survival in AML patients
Full size table

Taken together, these results suggest that ID3 expression may be most valuable prognostic biomarker among the ID genes in AML, especially CN-AML, and it was selected for further analysis.

Clinical significance of ID3 expression and its correlation with gene mutations in AML

To further analyze the clinical relevance of ID3 expression in AML, the AML patients from TCGA dataset were divided into two groups by the median ID3 expression level. Comparisons of clinicopathological features, including age, sex, white blood cell (WBC) count, peripheral blood (PB)/BM blasts, French–American–British (FAB) classifications, cytogenetics and gene mutations, between the two groups (low and high ID3 expression) in both the total AML and the CN-AML cohort are shown in Table 2. In all AML patients, low ID3 expression was greatly correlated with higher WBC counts and PB/BM blasts (P < 0.001, = 0.001 and = 0.002, respectively). Moreover, low ID3 expression was markedly correlated with normal karyotype and intermediate cytogenetic risk (P = 0.004 and 0.014, respectively). Based on the results, we further compared ID3 expression between AML patients with different cytogenetic risks, and confirmed the significant differences (P = 0.036, Fig. 3a). In addition, low ID3 expression was markedly associated with FLT3 and NPM1 mutations as well as wild-type TP53 (P = 0.018, 0.011 and 0.028, respectively). Similarly, we further determined ID3 expression between AML patients with and without these gene mutations. Expectedly, marked differences were observed in subgroups divided by FLT3 and NPM1 status (P = 0.005 and 0.003, respectively, Fig. 3b-c), whereas a trend was observed in subgroups divided by TP53 and CEBPA status (P = 0.063 and 0.088, respectively, Fig. 3d-e). In CN-AML, the above significant differences were not observed (Table 2).

Fig. 3
figure 3

The associations of ID3 expression with cytogenetic risks/genetic abnormalities in AML. a ID3 expression among different cytogenetic risks of AML. b ID3 expression in AML patients with and without FLT3 mutations. c ID3 expression in AML patients with and without NPM1 mutations. d ID3 expression in AML patients with and without CEBPA mutations. e ID3 expression in AML patients with and without TP53 mutations

Full size image
Table 2 Correlation of ID3 expression with clinic-pathologic characteristics in AML patients
Full size table

The independent prognostic value of ID3 expression in AML

Because a marked correlation was found between ID3 expression and common prognostic factors such as WBC, cytogenetics and gene mutations, we performed multivariate analysis by Cox regression to confirm the independent prognostic impact of ID3 expression in AML after adjusting for the prognosis-related factors. Multivariate Cox regression analysis indicated that ID3 expression was an independent risk factor affecting OS (P = 0.022, Table 3) and DFS (P = 0.043 and Supplementary Table S3) in CN-AML patients.

Table 3 Cox regression multivariate analysis of variables for overall survival in AML patients
Full size table

Molecular signatures correlated with ID3 expression in AML

To investigate the biological network caused by aberrant ID3 expression in AML, we first analyzed the transcriptomes of the two groups of patients (low and high ID3 expression) from the TCGA dataset. Based on the conditions of |log2 FC|>1.5, FDR < 0.05 and P < 0.05, a total of 839 DEGs (706 downregulated and 133 upregulated) between the low and high ID3 expression groups were identified (Fig. 4a-b and Supplementary Table S4). The top 100 downregulated DEGs, such as SLIT3 and ID4, are reported to have antitumor activities in AML [9, 17, 18]. Moreover, Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses [19,20,21] revealed that these DEGs are involved in multiple biological processes and the PI3K/AKT signaling pathway (Fig. 4c-d).

Fig. 4
figure 4

Biological network of aberrant ID3 expression in AML. a Expression heatmap of differentially expressed mRNAs/lncRNAs between low and high ID3 expression groups in AML (|log2 FC|>1.5, FDR < 0.05 and P < 0.05). b Volcano plot of differentially expressed mRNAs/lncRNAs between low and high ID3 expression groups in AML. c Gene Ontology (GO) analysis of differentially expressed mRNAs/lncRNAs. d Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis of differentially expressed mRNAs/lncRNAs. e Expression heatmap of differentially expressed miRNAs between low and high ID3 expression groups in AML (FDR < 0.05 and P < 0.05)

Full size image

We next revealed 72 DEmiRs (38 downregulated and 34 upregulated) between the low and high ID3 expression groups according to the conditions of FDR < 0.05 and P < 0.05) (Fig. 4e and Supplementary Table S4). The top 10 downregulated DEmiRs, including miR-139, miR-195, miR-203, miR-497 and miR-144, are reported to have antitumor effects in AML [22,23,24,25,26]. The top 10 upregulated DEmiRs, such as miR-196a, are reported to have protumor effects in AML [27]. Moreover, upregulated DEmiRs (potentially negatively associated with ID3 expression), such as miR-326, have been reported as potential microRNAs that directly target ID3 [28].

Validation of ID3 expression and its discriminative capacity in AML patients from our hospital

Given the results above, we further validated the expression of ID3 in the BMMNC samples of 107 newly diagnosed AML patients and 32 healthy donors as normal controls from our hospital. The expression of ID3 was extremely decreased in AML patients compared with normal controls (P = 0.001, Fig. 5a). Moreover, ROC analysis indicated that ID3 expression may serve as a prospective biomarker for discriminating AML patients from controls, with an AUC of 0.701 (95% CI: 0.598–0.805) (P = 0.001, Fig. 5b). These results confirmed the low expression pattern of ID3 in AML and revealed that ID3 expression might serve as a latent biomarker that is helpful for the diagnosis of AML.

Fig. 5
figure 5

Validation of ID3 expression and its discriminative capacity in AML. a The relative expression of ID3 in AML. b ROC curve analysis of ID3 expression in distinguishing AML controls

Full size image

Discussion

Dysregulation of ID gene expression has been revealed in various human cancers including AML, and was also associated with clinical outcome. Recently, Lu et al. using bioinformatics methods revealed that increased expression of ID1 and ID2 was correlated with poorer and better survival times, respectively, whereas ID3 and ID4 expression was not correlated with survival in lung adenocarcinoma patients [29]. Similarly, abnormal expression of ID genes may affect the occurrence and prognosis of lung cancer, and may be associated with cell metabolism and transcriptional regulation by using bioinformatics analysis [30]. These same results were further identified in breast cancer [31]. In the current study, by the bioinformatics analysis, we found that the expression of ID1 and ID3 was downregulated in AML, whereas the expression of ID2 was upregulated. Moreover, only abnormal ID3 expression may serve as an independent prognostic biomarker in AML and ID1/ID2 expression may independently affect clinical outcome in total AML. Previously, a few studies have reported the prognostic significance of ID gene expression in AML. Tang et al. revealed that high ID1 expression was correlated with adverse prognosis in AML [5]. However, a later study demonstrated that overexpression of ID1 was not an independent prognostic biomarker in young CN-AML patients [6]. Interestingly, our previous study indicated that overexpression of ID1 was correlated with higher karyotypic risk classification and served as an independent risk factor in young (age < 60 years) non-M3 patients [7]. Meanwhile, overexpression of ID2 was a frequent event in patients with AML and predicted poor chemotherapy response and clinical outcome [8]. Conversely, promoter hypermethylation-mediated ID4 repression was linked to disease progression in MDS and poor prognosis in AML. Altogether, these different results may be attributed to the differences in ethnicity and in AML subtype distribution. Accordingly, further studies are needed to validate the expression and clinical implications of the ID genes in AML.

In the present study, we mainly focused on ID3 expression in AML based on the bioinformatics identification and experimental validation. For the first time, we revealed that ID3 expression could serve as a prognostic predictor in AML. Notably, it is very interesting that ID3 could independently affect OS but not DFS by multivariate Cox regression analysis. We deduced that the role of aberrant ID3 expression in AML survival was not directly mediated by influencing leukemia development but could affect multiple factors that lead to all-cause death in AML. Previously, only May et al. revealed that ID2 and ID3 protein expression mirrored granulopoietic maturation and discriminated between acute leukemia subtypes [32]. However, numerous studies have investigated the expression and prognostic value of ID3 in human solid tumors. Xu et al. demonstrated that ID3 played a tumor suppressive role in papillary thyroid cancer and impeded metastasis by inhibiting E47-mediated epithelial to mesenchymal transition (EMT) [33]. Huang et al. indicated that ID3 could enhance the stemness of intrahepatic cholangiocarcinoma by gaining the transcriptional activity of β-catenin and could act as a potential biomarker in predicting response to adjuvant chemotherapeutics [34]. Moreover, ID3 overexpression was correlated with medulloblastoma seeding and is a poor prognostic factor in medulloblastoma patients [35]. Sharma et al. revealed that ID1 and ID3 overexpression alleviated all three cyclin-dependent kinase inhibitors (CDKN2B, -1 A, and − 1B) resulting in a more aggressive prostate cancer phenotype [36]. Expression of ID1 and ID3 was increased in human invasive lobular carcinoma compared with invasive ductal carcinoma, associated with poor prognosis uniquely in patients with invasive lobular carcinoma and correlated with the upregulation of angiogenesis and matrisome-related genes [37]. In addition to the above results, several studies have also reported the value of the combination of ID3 expression with other members in cancer prognosis. For instance, Antonângelo et al. showed that ID1, ID2 and ID3 coexpression was associated with prognosis in stage I/II lung adenocarcinoma patients treated with surgery and adjuvant chemotherapy [38]. Additionally, ID1 and ID3 coexpression was correlated with a poor clinical outcome in patients with locally advanced non-small cell lung cancer treated with definitive chemoradiotherapy [39]. The combined expression of VPREB3 and ID3 was used to develop a new helpful tool for the routine diagnosis of mature aggressive B-cell lymphomas [40]. All these results suggested the prognostic value of ID3 expression in diverse human cancers.

The functional role of ID3 has also been widely investigated, and was reported to be associated with diverse biological processes such as angiogenesis, apoptosis, cell cycle regulation/proliferation, cell migration/invasion, epithelial-to-mesenchymal transition, stem cell renewal and signaling [3]. Although we did not validate the direct role of ID3 in AML in this study, we identified the association of ID3 with PI3K/AKT signaling by bioinformatics methods. Moreover, the association of low ID3 expression with FLT3 mutation was also observed in AML patients. Similarly, Chen et al. demonstrated that miR-212-5p was involved in the progression of non-small cell lung cancer through the activation of PI3K/Akt signaling pathway by targeting ID3 [41]. Zhang et al. indicated that Per2 downregulated ID3 expression via the PTEN/AKT/Smad5 axis to inhibit glioma cell proliferation [42]. Moreover, ID3 was reported to play a significant role in reversing cisplatin resistance in human lung adenocarcinoma cells by regulating the PI3K/Akt pathway [43]. Accordingly, further functional studies are needed to confirm the direct role of ID3 in AML biology.

The regulatory mechanism of ID3 expression was preliminarily studied. Xu et al. demonstrated that hypermethylation of the CpG island at the promoter region of ID3 was the main contributor to the repression of this gene [33]. In addition, several studies also revealed the regulatory potential of miRNAs. Zhao et al. found that miR-326 could bind to ID3, which accelerated the development of medulloblastoma [28]. Moreover, high ID3 expression by silencing miR-212-5p expression suppressed the activity of the PI3K/Akt signaling pathway and consequently promoted apoptosis and inhibited proliferation in lung cancer cells [41]. Herein, we also observed the association of ID3 with several miRNAs such as miR-1259, miR-508, miR-9, miR-944, let-7b, miR-141, and miR-223. However, only miR-326 was confirmed by previous studies [28]. Accordingly, further studies are needed to confirm the direct association of ID3 with these miRNAs.

Conclusion

In summary, the bioinformatics analysis and experimental verification demonstrate that low ID3 expression independently affects OS and DFS in patients with CN-AML, which might be seen as a potential prognostic indicator in CN-AML.