DISCUSSION
The scenario of few recurrent driver mutations in HBs represents a challenge for risk stratification and adjustment of the therapeutic regimen, and for this reason molecular subclassifications, including gene signatures, have been proposed 2,29,30. Expression pattern of 16 genes grouped HBs according to tumor aggressiveness in two groups, called C1 and C2 signatures where β-catenin is upregulated in both signatures with genes being related to hepatocyte markers (GLUL , CYP2E1 , CYP1A1, andAQP9) , cell cycle regulation (E2F5 , BUB1 ,DLG7 ), hepatic stem/progenitor markers (DUSP9 , AFP ,CK19 , and TACSTD1) , and metabolism (APOC4, ALDH2 )29. The differential activation of hepatic stem/progenitor markers (LIN28B, SALL4 , and AFP ), genes related to cancer pathways (TERT, TP53 ), metabolism (NFE2L2 ) and hepatobiliary system (NOCHT1 , HNF ) were also used to stratify HBs in molecular subtypes according to risk30.
Here, we focused on epigenetic markers (expression of DNMTs and TETs as well as global DNA methylation) in addition to genes related to the stages of liver differentiation, in an attempt to perform HB stratification. Hepatocytes are the major cell type of the liver, accounting for ~ 70% of the mass of the adult organ. The anterior portion of the hepatic diverticulum gives rise to the liver from endoderm cells that differentiate in the bi-potential cells known as hepatoblasts 31,32, and liver growth and hepatocyte maturation are processes regulated by genes acting intrinsically in these cells 33. HB is postulated to develop from hepatoblasts, the precursor cells of hepatocytes 2; however, microscopically, HBs are heterogeneous, rarely composed of only one cell type, often exhibiting combinations of epithelial, stromal, mesenchymal and/or undifferentiated cells components11.
Genes associated with DNA methylation exhibited in HBs a pattern of expression directly correlated, which is inverse to the expression pattern of mature hepatocyte markers, highlighting the central role of this specific epigenetic pathway in liver differentiation and as well as in HB stratification. Our data also disclosed a direct correlation ofUHRF1  and AFP  expression in tumors, which was previously described for HCC 34, possibly representing a new finding to be explored in HBs. Additionally, there is a similar expression pattern of HFN4A and FOXA2 , which are nuclear factors that present tissue co-expression and cooperate for hepatic pathway cell commitment35, emphasizing the key role of the hepatic differentiation blockage in HB development.
Besides reinforcing the molecular heterogeneity of these embryonal tumors, we propose a panel of 13 genes for this HB stratification (TET1, TET2, TET3, DNMT1, DNMT3A, UHRF1, ALB, CYP3A4, TDO2 , UGT1A1, AFP , HNF4 A, and FOXA2 ). Remarkably, our data showed that the DNA methylation machinery exerts a key role in the characterization of HBs, directly reflected in diverse DNA methylation content. Six out of the eight genes associated with DNA methylation were determinants for HB stratification, evidencing the importance of the epigenetic machinery for the biology of this embryonal liver cancer.
At the beginning of the embryo development, the epigenetic machinery acts by decreasing the overall level of DNA methylation, allowing the expression of genes associated with pluripotency 36. During the maturation of the organ, there is an increase in the DNA methylation, a process that will culminate in the selective expression of tissue-specific genes, associated with the promotion of cell differentiation. The iPSC and definitive endoderm cell lines clustered separately (Set-4), reinforcing that HBs are composed of cells already compromised with the hepatocyte differentiation. A small group of intermediate-risk tumors exhibited high expression of the epigenetic machinery genes as well as hepatoblast markers (Set-1), a profile suggestive of a precursor stage of hepatocyte differentiation. This HB group present the highest expression of TET1 , TET3 , andUHRF1 among all tumors; in a recent work, we proposed an active demethylation process in HBs, particularly associated with upregulation of TETs and UHRF1 15 and, accordingly, Set-1 tumors exhibited marked global DNA hypomethylation. In the other hand, the Set-3 presented downregulation of the DNA methylation genes and high expression of mature hepatocyte markers, similarly to non-tumoral control livers; they were likely derived from cells compromised with more advanced stages of hepatocyte differentiation, and their global DNA methylation content is also similar to non-tumoral livers, corroborating the hypothesis of origin in late stages of hepatocyte differentiation. Considering that Set-3 comprised HBs from patients who are deceased and/or developed metastases, as well as one sample characterized as HB/HCC features, the major markers of hepatocyte (UGT1A1 , TDO2 , and CYP3A4 ) that are differentially expressed in this group, deserve further investigation as biomarkers of worse prognosis. Finally, the Set-2 contains clinically heterogeneous HBs exhibiting an intermediate expression profile between the Set-1 and Set-3 clusters, with high expression of the DNA methylation genes and downregulation of hepatoblast and mature hepatocyte markers. The Set-2 has a global DNA methylation level intermediary between Set-1 and Set-3, which is in accordance with the hypothesis that these tumor samples have a molecular profile transitional between hepatoblasts and mature hepatocytes.
In a recent analysis involving a large group of 113 samples, HBs presented two epigenetic signatures (Epi-CA and Epi-CB), according to the degree of sample hypomethylation. Moreover, amplification of the oncogenic 14q32 DLK1-DIO3  locus in part of the HBs delineated signatures of moderate or strong expression of genes mapped to this region. Using both findings, a molecular risk stratification of three categories was proposed (MRS-HB) 37. Despite using a smaller cohort of 21 HB samples in our study, we also showed that DNA methylation is a strong biomarker for HB stratification, pinpointing to specific DNA methylation genes as important players.
Precision medicine brings the promise of improvement in diagnosis and matching patients to personalized targeted therapies using genomics, epigenomics, metabolomics and proteomics. One of the major barriers to the implementation of this approach in clinical settings is the intra/inter- tumor heterogeneity, in addition to the rarity in the case of pediatric cancer. In particular, HB is a tumor with a genetic background with low mutational background and few cytogenetic alterations12,30,38–41, in addition to DNA methylation changes13,42.
The present work describes the possibility of molecular stratification of HBs according to markers of hepatocyte differentiation and DNA methylation. Tumor heterogeneity could be overcome using molecular signatures that are linked to clinical data. Here, we evidenced a HB group (Set-3) with a similarity of late stages of hepatocyte differentiation and lower DNA methylation genes that present a tendency to a worse prognosis. Although the other two sets do not have clinical characteristics that are divided evenly between groups, the association of Set-3 with a worse prognosis is in concordance with the literature that have shown metastasis at diagnosis, advanced age of diagnosis27,28,43, mutations at TERT gene30,39 associated with high mortality.
Most drugs based on the mechanisms of anticancer resistance in HB will act on pathways such as p53, tyrosine kinase, cell cycle control, and transcriptional and translational events. A specific gene,CYP3A4, upregulated in Set-3, is related to five drugs used in the treatment of HBs 44, particularly etoposide (a drug used in neoadjuvant chemotherapy), which causes demethylation in liver cells 45,46, and Sorafenib, a kinase inhibitor mainly oxidized by CYP3A4 , with the collaboration ofCYP1A1 and CYP1B1 47. Vincristine, Cyclophosphamide and Isofosfamide are also metabolized by CYP3A448–51. These findings suggest that changes in this gene could lead to resistance to treatment, maybe explaining the worse clinical signs observed in Set-3, with CYP3A4 levels similar to mature hepatocytes.
Together, data here showed the possibility of HB stratification in three groups considering the epigenetic machinery and markers of liver differentiation stages. These groups have levels of DNA methylation according to their stage of cell differentiation, which indicates that the clinical heterogeneity widely described for this tumor, also occurs at the epigenetic level. Therefore, we demonstrate that epigenetics can be an important tool that must be considered in stratification of HB.