Acute myeloid leukemia (AML) is the most common malignancy of acute leukemia in humans, comprising approximately 80 percent of the cases (Short et al., 2018). Compared with adults, the incidence rate in children is relatively rare, but with disproportionate mortality (Elgarten and Aplenc, 2020). As recently reported, the pediatric acute myeloid leukemia (pAML) accounts for 25% of pediatric leukemia with a low overall survival ratio of 70% (Taga et al., 2016). Due to the decades of study, the treatments for pAML has been greatly improved by advances in the hematopoietic stem cell transplantation (HSCT) method, chemotherapy, supportive care, and optimal risk stratification (Tarlock and Meshinchi, 2015; Paulraj et al., 2019). Based on the various pAML subtypes, the clinical pAML procedures are conducted separately. For instance, children with de novo AML are administrated with a standard treatment, which includes four or five cycles of myelosuppressive chemotherapy with cytarabine and anthracyclines followed by a HSCT for a subset of patients. At the same time, children with acute promyelocytic leukemia (APL) are recommended with an all-trans retinoic acid (ATRA)–combined regimen. In contrast, myeloid leukemia children with Down syndrome (ML-DS) are generally converged with a less intensive regimen (Ribeiro, 2014).

The worldwide incidence of pAML is heterogeneous because of the variable prevalence of the risk factors. As reported, host factors such as age, race, and germline predisposition impact outcomes contribute to a large amount of pAML (Radhi et al., 2010; Shimada, 2017). In addition to the host factors, the elevated level of peripheral white blood cells (WBCs) has been also suggested as an unfavorable aspect for pAML (Woods et al., 1996). For adult AML, growing evidence based on retrospective assessments supports the association between the disease and the abnormal proliferation and differentiation of a clonal population of myeloid stem cells (De Kouchkovsky and Abdul-Hay, 2016). Except for the chromosomal rearrangements, the particular genetic mutations play a key role in the formation of AML, causing more than 97% of cases (Patel et al., 2012). Compared with adult AML, the pathological mechanism underlying pAML is still poorly understood and the molecular landscape behind pAML remains quite distinct (Bolouri et al., 2018). The treatments for pAML have evolved with years of work. If detected at an early stage, the pAML could be administrated with various choices of procedures. Unfortunately, for the majority of patients, pAML is diagnosed at a later stage so that the long-term prognosis remains unsatisfactory.

MicroRNAs (miRNAs) are emerging as a promising candidate for the molecular mechanism behind pAML, which attracts attention of many researchers. Previously, a study by Obulkasim et al. (2017)suggested that the biological subgroups of pAML were reflected by a common miRNA expression pattern, while the separate subtypes of pAML have distinct miRNA expression patterns. The miRNAs belong to a class of small endogenous RNAs that regulate gene expression post-transcriptionally and play a role in gene silencing and translation inhibition by binding to target genes. The miRNAs represent a highly conserved class of tissue-specific genes that have been found in all eukaryotic cells preserved across species since their discovery in 1993 (Vishnoi and Rani, 2017; Lu and Rothenberg, 2018). Generally speaking, they are short RNA molecules with 19–25 nucleotides in size. A single miRNA can target hundreds of mRNAs and influence the expression of many genes. The appropriate maintenance of miRNA expression is required for a balanced physiological environment.

Since the underlying molecular mechanism of pAML remains unclear, the disorder has brought great difficulty to clinical treatment. To address these issues, in this study, we used a combination of bioinformatics analysis and external clinical specimen experiments to establish differentially expressed gene profiling. Meanwhile, the potential signaling axis for pAML formation had also been explored. All of these promising outcomes enriched the precise early diagnosis of the disease, which provided tremendous help for the pAML study.

Even with years of hard work for the breakthroughs in pAML, the prevention, early screening, diagnosis, treatments, and overall outcomes are still inferior, highlighting the need for improved targeted therapies. In this study, we comprehensively compared differentially expressed mRNAs as well as miRNAs between pAML and control groups based on two independent databases. In this integrated study, the 10 mRNAs and three miRNAs were hypothesized as key regulators for pAML formation. Previously, more and more attention had been paid to the interaction between the signal transducer and activator of transcription 3 (STAT3) and AML in adults. Increases in cytokine ligands, such as IL-6, trigger intracellular tyrosine phosphorylation of STAT3, which is seen in up to 50% of AML cases and signifies a worse prognosis (Schuringa et al., 2000). In addition to STAT3, NPM1 and CEBPA confer to notable class II mutations while DNA-methylation–related genes DNMT3A, TET2, and IDH-1 and IDH-2 contribute to the class II mutation group for AML, which are found in about 27% and more than 40% of AML cases, respectively (Takahashi, 2011; Klein et al., 2018). However, the majority of these mutations were not significantly expressed in the pAML group specimens from our analysis, which might reflect a tremendous difference between pediatrics and adults for AML initiation as well as development. The DNA methylation events in somatic mutations are suggested to be highly prevalent in adults. Inversely, the structural alterations in methyltransferase genes are more prevalent in young children, but rarer or even absent in adults. These may explain the differences for the AML between children and adults considering differentially expressed gene profiles.

Previously, the connection between miRNA and leukemogenesis was built on the cell proliferation as well as cell adhesion regulation by particular miRNA families (Mott et al., 2007; Volinia et al., 2010). These are in dire need of better understanding of the functions of miRNA in pAML since most of the work focuses on the adult AML (Danen-van Oorschot et al., 2012). A study by Zhang et al. (2009) demonstrated the miR-expression differences between FAB-M1, FAB-M2, and FAB-M3 groups in pAML. Another study suggested the specific expression pattern of miR-99a, miR-125b, and let-7c in pAML and the upregulations of these factors stimulated leukemogenesis by switching the balance between TGF-ß and Wnt signaling pathways (Bonifant et al., 2018). In a large cohort, Obulkasim and his colleagues analyzed 665 miRNAs on 165 pediatric AML samples, which provided 14 key clusters for the differentially expressed miRNAs (Lu and Rothenberg, 2018). In this study, they claimed pAML samples with MLL rearrangements were classified with 89% accuracy using 37 miRNA expression signatures; 14 miRNAs were highly expressed and 23 were lowly expressed compared to the samples without MLL rearrangements. In consistent with our findings here, hsa-miR-133, hsa-miR-181, and hsa-miR-195 were strikingly repressed in the pAML samples. Moreover, they supposed that hsa-miR-181 and hsa-miR-195 were functional as tumor suppressors while hsa-miR-133 was only used to classify MLL-rearranged samples in pAML.

Except for miRNA analysis, we also hypothesized the key mRNAs for pAML, including ZC3H15, BCLAF1, PPIG, DNTTIP2, SRSF11, KTN1, UBE3A, PRPF40A, TMED5, and GNL2. ZC3H15 represents zinc finger CCCH-type containing 15, which is an immediate early erythropoietin response gene. Using the co-immunoprecipitation method, it can be found that ZC3H15 is functional via the signaling adapter protein tumor necrosis factor receptor–associated factor 2 (TRAF-2) with the NF-κB pathway for AML formation (Capalbo et al., 2013). Here, in this study, we also observed the key roles of ZC3H15 for pAML. Meanwhile, the NF-κB pathway was suggested as the primary signaling pathway by KEGG analysis (as shown in Figure 2). Taken these together, ZC3H15 may be a central factor for the pAML process through the NF-κB signaling pathway, raising an interesting direction for future pathological mechanism studies underlying pAML. BCLAF1 was originally initiated as a key of apoptosis and repressor of transcription, which is associated with antiapoptotic members of the Bcl2 family for various developmental processes such as T-cell activation and so on (Sarras et al., 2010). SRSF11 was previously demonstrated as a novel TERC-binding protein, which localizes to nuclear speckles and associates with active telomerase enzyme for cell cycle manipulation (Lee et al., 2015). Based on its function as a nuclear speckle–targeting factor essential for telomerase association with telomeres, SRSF11 has been deeply explored in cancer research. UBE3A is a dual-function protein, which consists of ubiquitin ligase as well as transcriptional co-activator function and has been shown to play a fundamental role in the modulation of synaptic function and plasticity (Vatsa and Jana, 2018). Except for these, the association between selected hub genes from this study and pAML has not been fully explored, which calls for a great point for future studies. Since ZC3H15 and BCLAF1 were suggested as two primary hub genes for pAML. We further investigated the expression level of them between control and pAML groups. Initially, we analyzed the expression in the cell line and observed the distinct expression pattern for both of them (data not shown). We preferred to utilize peripheral blood samples since pAML refers to a blood disease. Using clinical specimens, both ZC3H15 and BCLAF1 displayed a remarkable increased expression level in the pAML compared with the control group (as shown in Table 1), which confirmed the functions of these genes in clinical verification. It is worth noting that all the top hub genes showed increased expressions in the pAML group, which could be explained by the downregulations of top three miRNAs (as shown in Figures 1, 3). All of these deserve further investigation.

To conclude, in the light of unclear pathological mechanisms as well as missing potential biomarkers for pAML, we systematically generated a differential expression profile of mRNAs and miRNAs and provided several potential miRNA-dependent signaling axes for pAML patients. All the work here offered novel insights for the future pAML research.