Acute myeloid leukemia (AML) is rising in the population with dismal outcome. It is urgent to identify the high-risk patients under the standard chemotherapy at the time of the initial diagnosis. Currently, the well-established biomarkers include clinical and genetic features. For instance, chromosomal abnormalities have been proved as an effective risk stratification tool. However, approximately 50% of AML patients have normal karyotypes. They are currently defined as CN-AML patients and can be further stratified by mutated genes [
21,
22]. However, these biomarkers often interact or cluster with each other. For example,
NPM1 mutations coexist with mutations in
FLT3-ITD and/or
DNMT3A and/or
IDH1/2. Consequently, the predictive value of
NPM1 mutations would be confounded by these factors. Thus, identifying independent predictors is becoming more and more important in clinical practices. circRNAs are believed to be novel biomarkers in several solid tumors. But very little is known on whether it can be used as an independent predictor for AML patients.
In this study, we enrolled 30 CN-AML patients with poor survival and selected 30 cases with long-term survival by matching the well-established factors such as age, WBC, and mutations in FLT3-ITD, CEBPA, NPM1, DNMT3A, IDH1, and IDH2 genes. By the propensity score analysis, we excluded the confounders and identified 308 circular RNAs. We further focused on a GO term of carbon-oxygen lyase activity (GO:0016835) and identified hsa_circ_0075451 as a potential survival predictor. In order to confirm the independent prognostic value, we estimated the sample size and enrolled enough patients to perform stratification analysis and multivariate analysis in an independent cohort of CN-AML patients. Based on these analyses, we proved hsa_circ_0075451 to be a reliable and independent predictor in the training and validated cohort.
In addition, via the analysis of the different expression of targeted genes combining with sequences interaction analyzed in silico, we constructed the circRNA-miRNA-mRNA regulatory network (Fig.
4). hsa_circ_0075451 might sponge 10 miRNAs to upregulate 84 genes, which are involved in multiply pathways that may lead to leukemia progression. A wealth of evidence has suggested that 10 miRNAs (miR-512-5p, miR-330-5p, miR-326, miR-338-3p, miR-515-5p, miR-873, miR-766-3p, miR-940, miR-661, and miR-492) can function as tumor suppressors in different types of cancer such as head and neck squamous cell carcinoma, melanoma, gastric cancer, breast cancer, colon cancer, ovarian cancer, human non-small cell lung cancer, and cervical cancer [
23‐
30]. Furthermore, some miRNAs have the ability to inhibit some oncogenes of
EXPH5,
DAPK1,
ITGA2,
PLXNA2,
SLC4A7,
DSG2,
PPP1R9A,
KIAA1549, and
TANC1 in the RAS signaling pathway [
31];
PTPRD,
MMP16,
AFF3,
IFNLR1, and
PRDM16 in the nuclear factor kappa B signaling pathway [
32];
CD72 and
CARD11 in the BCR signaling pathway [
33];
TERT,
EIF3A,
PRKCZ,
ITGA2,
CDC37,
TNR, and
PRLR in the PI3K-Akt signaling pathway; and
SMO and
KIF3A in the hedgehog signaling pathway [
34]. Others might inhibit genes of
FNBP1L,
ZC4H2,
MMP16,
PPP1R9A,
AFF3,
MAGI3,
KIAA1217,
ZFHX3,
TANC1,
PRKCZ, and
PDZD2 which were identified as a prostate cancer dependency-regulating RNA splicing [
35];
FNBP1L,
SLC4A7,
DSG2,
MAGI3,
KIAA1217,
TANC1,
PRKCZ,
NCAM1, and
OLFM4 involving in cell-to-cell interactions that facilitate cell adhesion and inversion [
36]; and
LCN2 (
NGAL) and
OLFM4 (
GW112) selectively express in colon cancer and serve as selective targets [
37]. Notably, some genes are involved in epigenetic and metabolic changes. For example,
BCAT1 enhances branched-chain amino acid production and promotes myeloid leukemia progression [
38]. While
AASS controls the first two steps in the lysine degradation pathway, lysine modification takes part in the transcript regulation reported in the pan-cancer analysis [
39].
PRDM16 acts as a transcription coregulator that controls the development of brown adipocytes and also involves in the GO:0046974 term of histone methyltransferase activity (H3-K9 specific). Recently,
PRDM16 was shown to play an important role in the pathogenesis of MDS and AML [
40]. Finally, we identified a regulatory axis of hsa_circ_0075451 -| miR-330-5p/miR-326-|
PRDM16 in AML cells by the dual luciferase report assay. Taken together, these target genes specifically for
PRMD16 coexist in multiple functions of transduction signaling, transcription regulation, epigenetic modification, and metabolic change. Thus, we further measured the cellular metabolites to understand the metabolic features of high expressers. As a result, we found highly expressed hsa_circ_0075451 was associated with higher levels of cellular energy sources such as fructose, methionine, glutamine, arginine, serine, and isoleucine and lower levels of dodecanoic acid. These results suggest that blasts with upregulated hsa_circ_0075451 expression can affect fatty acid synthesis using other metabolites like BCAA as substrates. Therefore, this may explain the reason that hsa_circ_0075451 causes the poor survival may be via sponging these miRNAs and in turn leading to overexpression of oncogenes. Recent study identified fructose as a crucial factor in blasts’ survival and drug resistance [
41]. In our study, cellular fructose was increased in patients with highly expressed circRNA, implying another underlying mechanism of overexpressed fructose in AML.
There are also some limitations to this study. Firstly, the training group might not be large enough to identify all of the aberrant circRNAs. Secondly, patients in the training group were selected from the prospective clinical trial, while the validated group was enrolled from the single center based on the retrospective study. Thirdly, the selected patients for mRNA profiling and metabolic analyses were not from the same cases. Therefore, upstream changes of mRNAs coding enzymes might not translate exactly into the downstream changes of metabolites. In spite of some limitations, this is one of the few studies to identify independent and reliable circRNAs with prognostic significance by the discovery and validation study design.