Advertisement

Hepatoblastomas with carcinoma features represent a biological spectrum of aggressive neoplasms in children and young adults

Open AccessPublished:May 13, 2022DOI:https://doi.org/10.1016/j.jhep.2022.04.035

      Highlights

      • HBCs display a combination of HB and HCC histological and molecular features.
      • HBCs are genetically unstable hepatocellular neoplasms with higher mutation and CNA burdens than HBs.
      • HBCs are associated with poor outcomes and patients may benefit from specialized aggressive therapeutic algorithms.

      Background & Aims

      Hepatoblastoma (HB) and hepatocellular carcinoma (HCC) are the predominant liver cancers in children, though their respective treatment options and associated outcomes differ dramatically. Risk stratification using a combination of clinical, histological, and molecular parameters can improve treatment selection, but it is particularly challenging for tumors with mixed histological features, including those in the recently created hepatocellular neoplasm not otherwise specified (HCN NOS) provisional category. We aimed to perform the first molecular characterization of clinically annotated cases of HCN NOS.

      Methods

      We tested whether these histological features are associated with genetic alterations, cancer gene dysregulation, and outcomes. Namely, we compared the molecular features of HCN NOS, including copy number alterations, mutations, and gene expression profiles, with those in other pediatric hepatocellular neoplasms, including HBs and HCCs, as well as HBs demonstrating focal atypia or pleomorphism (HB FPAs), and HBs diagnosed in older children (>8).

      Results

      Molecular profiles of HCN NOS and HB FPAs revealed common underlying biological features that were previously observed in HCCs. Consequently, we designated these tumor types collectively as HBs with HCC features (HBCs). These tumors were associated with high mutation rates (∼3 somatic mutations/Mb) and were enriched with mutations and alterations in key cancer genes and pathways. In addition, recurrent large-scale chromosomal gains, including gains of chromosomal arms 2q (80%), 6p (70%), and 20p (70%), were observed. Overall, HBCs were associated with poor clinical outcomes.

      Conclusions

      Our study indicates that histological features seen in HBCs are associated with combined molecular features of HB and HCC, that HBCs are associated with poor outcomes irrespective of patient age, and that transplanted patients are more likely to have good outcomes than those treated with chemotherapy and surgery alone. These findings highlight the importance of molecular testing and early therapeutic intervention for aggressive childhood hepatocellular neoplasms.

      Lay summary

      We molecularly characterized a class of histologically aggressive childhood liver cancers and showed that these tumors are clinically aggressive and that their observed histological features are associated with underlying recurrent molecular features. We proposed a diagnostic algorithm to identify these cancers using a combination of histological and molecular features, and our analysis suggested that these cancers may benefit from specialized treatment strategies that may differ from treatment guidelines for other childhood liver cancers.

      Graphical abstract

      Keywords

      Introduction

      Hepatoblastoma (HB) and hepatocellular carcinoma (HCC) are the most common primary liver malignancies in children, adolescents, and young adults. HB is usually diagnosed in young children and has shown an unequivocal rise in incidence, representing the pediatric cancer with the highest average annual percent increase in incidence over the past 3 decades.
      • Hubbard A.K.
      • Spector L.G.
      • Fortuna G.
      • Marcotte E.L.
      • Poynter J.N.
      Trends in international incidence of pediatric cancers in children under 5 years of age: 1988–2012.
      HCCs are more commonly seen in adolescents and young adults. When diagnosed in children, HCCs are often associated with underlying genetic, metabolic, and inflammatory liver conditions.
      • Hubbard A.K.
      • Spector L.G.
      • Fortuna G.
      • Marcotte E.L.
      • Poynter J.N.
      Trends in international incidence of pediatric cancers in children under 5 years of age: 1988–2012.
      Clinical presentations, treatment options, and outcomes differ dramatically between HB and HCC, with 5-year overall survival rates near 70% for patients with HB and 30% for those with HCC.
      • Czauderna P.
      • Lopez-Terrada D.
      • Hiyama E.
      • Häberle B.
      • Malogolowkin M.H.
      • Meyers R.L.
      Hepatoblastoma state of the art: pathology, genetics, risk stratification, and chemotherapy.
      HBs are embryonal neoplasms that histologically recapitulate liver developmental stages.
      • Ranganathan S.
      • Lopez-Terrada D.
      • Alaggio R.
      Hepatoblastoma and pediatric hepatocellular carcinoma: an update.
      They may be either exclusively epithelial or mixed epithelial and mesenchymal, and are characterized by aberrant Wnt pathway activation that is most often associated with genetic alterations in CTNNB1.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      HBs often respond to chemotherapy, and the combination of chemotherapy—most commonly cisplatin and doxorubicin—and surgery is an effective treatment for a majority of patients. However, chemoresistance or incomplete response to chemotherapy often leads to unfavorable clinical outcomes.
      • Ueda Y.
      • Hiyama E.
      • Kamimatsuse A.
      • Kamei N.
      • Ogura K.
      • Sueda T.
      Wnt signaling and telomerase activation of hepatoblastoma: correlation with chemosensitivity and surgical resectability.
      Comparatively, pediatric HCCs are a histologically and biologically diverse group of neoplasms
      • Haines K.
      • Sarabia S.F.
      • Alvarez K.R.
      • Tomlinson G.
      • Vasudevan S.A.
      • Heczey A.A.
      • et al.
      Characterization of pediatric hepatocellular carcinoma reveals genomic heterogeneity and diverse signaling pathway activation.
      that are often clinically aggressive, chemotherapy-resistant, and only curable by complete surgical resection.
      • Czauderna P.
      • Lopez-Terrada D.
      • Hiyama E.
      • Häberle B.
      • Malogolowkin M.H.
      • Meyers R.L.
      Hepatoblastoma state of the art: pathology, genetics, risk stratification, and chemotherapy.
      Hepatocellular malignant neoplasms in children have been traditionally classified as either HB or HCC based on their histology, however, some tumors exhibit mixed or overlapping features. These include tumors showing focal or diffuse macrotrabecular pattern, and tumors that display significant cytologic atypia, with coarse chromatin, nucleoli, pseudo inclusions, atypical mitoses, or even marked nuclear pleomorphism and anaplasia. A second group of tumors, sometimes classified as mixed HB and HCC, demonstrate distinct areas that show characteristic HB histological features admixed with other areas with HCC features. Guidelines for the diagnosis and treatment of these tumors are still evolving. Tumors with mixed or overlapping HB and HCC features were first documented over a decade ago and designated as transitional-cell liver tumors. These clinically aggressive tumors were identified in older children, had aggressive histological features, and were hypothesized to arise from cells that were transforming from HB to HCC.
      • Prokurat A.
      • Kluge P.
      • Kościesza A.
      • Perek D.
      • Kappeler A.
      • Zimmermann A.
      Transitional liver cell tumors (TLCT) in older children and adolescents: a novel group of aggressive hepatic tumors expressing beta-catenin.
      ,
      • Zimmermann A.
      The emerging family of hepatoblastoma tumours: from ontogenesis to oncogenesis.
      In 2014, international pediatric liver cancer experts and representatives of the leading pediatric cancer oncology consortia that oversee liver cancer therapeutic trials drafted the first International Consensus Classification of Pediatric Liver Neoplasms. The consortia formally recognized the existence of tumors that demonstrate either combined or overlapping histological HB and HCC features, and it created the provisional hepatocellular neoplasm not otherwise specified (HCN NOS) diagnostic category in an effort to identify and characterize these neoplasms.
      • López-Terrada D.
      • Alaggio R.
      • de Dávila M.T.
      • Czauderna P.
      • Hiyama E.
      • Katzenstein H.
      • et al.
      Towards an international pediatric liver tumor consensus classification: proceedings of the Los Angeles COG liver tumors symposium.
      Despite this, only 5 HCN NOSs have been molecularly characterized to date.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      ,
      • Eichenmuller M.
      • Trippel F.
      • Kreuder M.
      • Beck A.
      • Schwarzmayr T.
      • Haberle B.
      • et al.
      The genomic landscape of hepatoblastoma and their progenies with HCC-like features.
      ,
      • Carrillo-Reixach J.
      • Torrens L.
      • Simon-Coma M.
      • Royo L.
      • Domingo-Sàbat M.
      • Abril-Fornaguera J.
      • et al.
      Epigenetic footprint enables molecular risk stratification of hepatoblastoma with clinical implications.
      Results from these studies suggested that HCN NOSs carry characteristic molecular alterations, including TERT promoter mutations, and that the prognosis of HCN NOS patients is particularly poor.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      We note that current protocols for diagnosing and classifying pediatric hepatocellular neoplasms are exclusively based on histopathology and may be insufficient to accurately differentiate between HBs, HCN NOSs, and HCCs because of limited tumor sampling, the often heterogeneous nature of these tumors, and their overlapping histological features.
      • Prokurat A.
      • Kluge P.
      • Kościesza A.
      • Perek D.
      • Kappeler A.
      • Zimmermann A.
      Transitional liver cell tumors (TLCT) in older children and adolescents: a novel group of aggressive hepatic tumors expressing beta-catenin.
      ,
      • Eichenmuller M.
      • Trippel F.
      • Kreuder M.
      • Beck A.
      • Schwarzmayr T.
      • Haberle B.
      • et al.
      The genomic landscape of hepatoblastoma and their progenies with HCC-like features.
      Herein, we analyzed the molecular features and outcomes of the largest cohort of histologically confirmed HCN NOSs to date. In addition, our search for tumors that display cytologic atypia, nuclear pleomorphism and anaplasia, and HCC-like macrotrabecular pattern, identified tumors that demonstrated these features focally but did not meet the HCN NOS diagnostic criteria. We descriptively designated these tumors as HBs that demonstrate focal macrotrabecular pattern, pleomorphism or anaplasia (HB FPAs), and their molecular profiles and outcomes to those of HCN NOS, classical HBs,
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      and pediatric HCCs.
      • Haines K.
      • Sarabia S.F.
      • Alvarez K.R.
      • Tomlinson G.
      • Vasudevan S.A.
      • Heczey A.A.
      • et al.
      Characterization of pediatric hepatocellular carcinoma reveals genomic heterogeneity and diverse signaling pathway activation.

      Patients and methods

      Tumor histopathology and patient selection

      With institutional review board approval, we searched the Texas Children’s Hospital pathology database to identify HB, HCN NOS, and HCC cases that were diagnosed or seen in consultation at the Texas Children’s Hospital between 2004 and 2016. Cases were selected based on age, diagnoses, observations in surgical pathology reports, histological reviews, and material available for molecular profiling. Histological reviews of glass slides from all cases—including diagnostic biopsies, resection, transplant, and metastases samples—were performed by 3 pathologists (MJF, DLT, and TLP) who confirmed the features in the original reports and selected representative areas of tumor most suitable for molecular testing. Representative tumor areas of either both components when possible (biphasic HCN NOS cases) or with the most atypia or pleomorphism (equivocal HCN NOS and HB FPAs) were selected for profiling studies. Selected clinical information was collected for these patients, as allowed by the IRB protocol; see Table S1. From 170 cases, we identified a total of 25 HCN NOS cases and 10 HB FPAs that were diagnosed in patients younger than 8 years. We also identified 5 unambiguous classical HB cases of patients that were older than 8 years at diagnosis, as high-risk controls to enable molecular profile comparisons. Note that analyses of the molecular profiles of these tumors also incorporated other controls using previously reviewed samples, including HBs
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      and HCCs.
      • Haines K.
      • Sarabia S.F.
      • Alvarez K.R.
      • Tomlinson G.
      • Vasudevan S.A.
      • Heczey A.A.
      • et al.
      Characterization of pediatric hepatocellular carcinoma reveals genomic heterogeneity and diverse signaling pathway activation.
      We identified HCN NOS as tumors with concurrent histological features of HB and HCC and classified them as either biphasic HCN NOS or equivocal HCN NOS (Fig. 1). We note that this histological distinction is not a part of the HCN NOS Consensus Classification. However, the distinction was made for this study to facilitate our molecular investigation, and all HCN NOS cases were classified as either biphasic HCN NOS or equivocal HCN NOS based on histology. Biphasic HCN NOS (n = 13) had discernable tumor areas with distinct HB or HCC features, while equivocal HCN NOS (n = 12) were characterized by cytological features and growth patterns that were intermediate between HB and HCC, and, hence, also difficult to classify. Equivocal HCN NOS cases were histologically more homogeneous overall, with intermediate HB and HCC features, including higher nucleus-to-cytoplasm ratios, occasional pleomorphism, increased mitotic count, and, sometimes, macrotrabecular pattern (Fig. 1). To further illustrate the histological features of HCN NOS cases, we included selected histological images and immunostains of biphasic HCN NOS (Figs. S1-5) and equivocal HCN NOS (Fig. S6) as supplementary figures. We note that some HCN NOS cases had been previously diagnosed descriptively as mixed tumors or as transitional liver cell tumors, a designation that has been superseded by the international consensus classification HCN NOS. Patient ages ranged from 11 months to 23 years, and both the youngest and oldest patients were diagnosed with biphasic HCN NOS (Fig. 2). The 23-year-old patient was the only young-adult HCN NOS identified in our database and was included in the study to illustrate the wide age range of patients with HCN NOS at diagnosis.
      Figure thumbnail gr1
      Fig. 1Representative histopathology of the pediatric hepatocellular neoplasms in the study.
      (A) Representative pediatric HCCs (4x). (B) Two equivocal HCN NOS cases exhibiting homogeneous cell patterns with higher nuclear-to-cytoplasmic ratios, slight pleomorphism, increased mitotic count (10x). Equivocal HCN NOS cases may display a macro-trabecular growth pattern (not illustrated). (C) Biphasic HCN NOS exhibiting distinct histological patterns in separate nodules. Areas of typical HBs adjacent to HBs with carcinomatous features can be distinctly identified (4x). (D) HB FPAs showing histological features of HBs but containing foci or cell clusters with atypical cytologic features including increased pleomorphism, vesicular chromatin with prominent nucleoli, and higher nuclear-to-cytoplasmic ratio (10x). (E) Epithelial fetal HBs without atypical or carcinomatous features (10x). Tumor illustrations use lighter colors to refer to HB-like cells and darker colors to refer to HCC-like cells. HB, hepatoblastoma; HB FAPs, HBs demonstrating focal atypia or pleomorphism; HCC, hepatocellular carcinoma; HCN NOS, hepatocellular neoplasm not otherwise specified. (This figure appears in color on the web.)
      Figure thumbnail gr2
      Fig. 2Pathology of 40 HBs, including 12 equivocal HCN NOS cases, 13 biphasic HCN NOS cases, 10 HB FPAs in patients aged <8, and 5 classical HBs in patients aged >8.
      (A) Patient outcomes included (A) alive with no evidence for recurrence for at least 2 years, (DOD) dead of disease, (R) relapsed, and (NA) not available. Treatment strategies included (C) chemotherapy and (T) transplantation. Available data for some of these patients included pretext scores, the presence of metastatic lesions, age at diagnosis, risk classification by CHIC-HS, Cairo et al., and Sumazin et al., as well as outcomes; note that the high-risk HB class C2 by Cairo et al. was marked as high-risk for readability. (B) Excluding patients without available outcomes data, the presence of metastases correlated with patient outcomes and had a hazard ratio of 1.9, p = 0.02 (FET). Transplanted patients had significantly better outcomes with a hazard ratio of 3.3, p = 6E-4 (FET). When combined, HCN NOSs and HB FPAs had worse outcomes than our older patient cohort, with a hazard ratio of 1.5. (C) Our patients differed by age, most patients with equivocal NOS were older than those with biphasic NOS. (D) TERT promoter mutations were mainly observed in tumors that were diagnosed in older patients with no significant association with histological or clinical subtypes; p = 6E-8, t test. HB, hepatoblastoma; HB FAPs, HBs demonstrating focal atypia or pleomorphism; HCN NOS, hepatocellular neoplasm not otherwise specified. (This figure appears in color on the web.)
      Cases that were not classified as HCN NOS but whose diagnostic reports mentioned atypical histological features such as focal anaplasia, significantly increased pleomorphism, or prominent macrotrabecular pattern, were classified as HB FPAs. HB FPAs focally displayed cells with prominent nucleoli and higher nucleus-to-cytoplasm ratio, that were occasionally arranged in a macrotrabecular pattern. Macrotrabecular pattern identification required the presence of 5 or more cell-thick trabeculae, as previously described.
      • López-Terrada D.
      • Alaggio R.
      • de Dávila M.T.
      • Czauderna P.
      • Hiyama E.
      • Katzenstein H.
      • et al.
      Towards an international pediatric liver tumor consensus classification: proceedings of the Los Angeles COG liver tumors symposium.
      While these features are most often seen in post-chemotherapy specimens, they can be identified in pre-treatment diagnostic samples
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      and their underlying biology and prognostic significance remain unknown. Our selection criteria required the identification of these features at either diagnosis or resection. HB FPAs often contain pleomorphic components with large (>2-3x) tumor cells that may have irregular nuclear contours, large hyperchromatic nuclei, intranuclear inclusions, prominent nucleoli, and may present atypical mitoses. We note that 7/10 of the HB FPAs in our cohort displayed a macrotrabecular pattern. Representative histological features of HB FPAs are illustrated in Fig. S7. To investigate the outcomes of HB FPAs independently of other commonly recognized risk factors, we selected patients with HB FPAs that were younger than 8 years at diagnosis (n = 10). Current clinical trial protocols stratify patients that are older than 8 years as high risk
      • Cairo S.
      • Armengol C.
      • Maibach R.
      • Häberle B.
      • Becker K.
      • Carrillo-Reixach J.
      • et al.
      A combined clinical and biological risk classification improves prediction of outcome in hepatoblastoma patients.
      • Haeberle B.
      • Rangaswami A.
      • Krailo M.
      • Czauderna P.
      • Hiyama E.
      • Maibach R.
      • et al.
      The importance of age as prognostic factor for the outcome of patients with hepatoblastoma: analysis from the Children's Hepatic tumors International Collaboration (CHIC) database.
      • Meyers R.L.
      • Maibach R.
      • Hiyama E.
      • Häberle B.
      • Krailo M.
      • Rangaswami A.
      • et al.
      Risk-stratified staging in paediatric hepatoblastoma: a unified analysis from the Children's Hepatic tumors International Collaboration.
      and this age-selection criterion allowed us to distinguish the biology and outcomes of patients with HB FPAs from older patients with high-risk HBs that did not display histological features of either HCN NOS or HB FPAs.
      Our study also included 5 HBs that were diagnosed in older children (greater than 8 years) and showed unambiguous HB histology with either typical epithelial or epithelial and mesenchymal components and without histological features of HCN NOS or HB FPAs. These rare cases were chosen as clinical high-risk HB controls to compare them to HBs from younger children, HCN NOS, HB FPAs, and HCCs. Note that 4 of these 5 older HB cases were transplanted and that 4 of these 5 cases developed metastatic disease (Fig. 2A). Finally, samples from HBs reported by Sumazin et al.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      and HCCs from Haines et al.
      • Haines K.
      • Sarabia S.F.
      • Alvarez K.R.
      • Tomlinson G.
      • Vasudevan S.A.
      • Heczey A.A.
      • et al.
      Characterization of pediatric hepatocellular carcinoma reveals genomic heterogeneity and diverse signaling pathway activation.
      were selected for profiling as additional controls, based on tissue availability.

      Clinical data collection, risk stratification, and patient outcome evaluation

      We profiled 40 HCN NOS, HB FPAs, and HBs that were diagnosed in older patients. Treatment data were available for 37/40 of these patients. These patients were treated as follows, 16 patients with HCN NOS were treated with chemotherapy followed by surgical resection, and 7 were transplanted; 7 of the patients with HB FPAs were treated with chemotherapy and resection, and 2 were transplanted; 1 older patient with HB was treated with chemotherapy and resection, and 4 were transplanted (Fig. 2). PRETEXT data were available for 34/40 patients, of whom 12 were PRETEXT 4, 18 were PRETEXT 3, 2 were PRETEXT 2, and 2 were PRETEXT 1. Most patients (n = 30, 88%) were either PRETEXT 3 or 4 at diagnosis. We note that PRETEXT 3 and PRETEXT 4 HBs do not have adjoining liver sections that are free of disease, and while sometimes feasible, complete resection of these cases is challenging and depends on their response to preoperative chemotherapy.
      • Towbin A.J.
      • Meyers R.L.
      • Woodley H.
      • Miyazaki O.
      • Weldon C.B.
      • Morland B.
      • et al.
      2017 PRETEXT: radiologic staging system for primary hepatic malignancies of childhood revised for the Paediatric Hepatic International Tumour Trial (PHITT).
      ,
      • Roebuck D.J.
      • Aronson D.
      • Clapuyt P.
      • Czauderna P.
      • de Ville de Goyet J.
      • Gauthier F.
      • et al.
      2005 PRETEXT: a revised staging system for primary malignant liver tumours of childhood developed by the SIOPEL group.
      Patient outcomes data were used to evaluate risk stratification and compare treatment outcomes. When evaluating patient outcomes, we defined good outcomes as event-free survival for at least 2 years, and poor patient outcomes as relapse or death from disease during a median follow-up of 5.5 years. We compared risk stratification predictions by 3 stratification models following published protocols. CHIC-HS risk stratification was based on patient age at diagnosis, multiple variables that affect tumor resectability including PRETEXT staging and VPEFR annotation factors, serum AFP (alpha-fetoprotein) concentration, and the presence of metastases.
      • Meyers R.L.
      • Maibach R.
      • Hiyama E.
      • Häberle B.
      • Krailo M.
      • Rangaswami A.
      • et al.
      Risk-stratified staging in paediatric hepatoblastoma: a unified analysis from the Children's Hepatic tumors International Collaboration.
      Risk stratification by Cairo et al. was based on clustering Z-score normalized expression profiles of a 16-gene signature that includes both upregulated and downregulated genes in HB.
      • Cairo S.
      • Armengol C.
      • De Reynies A.
      • Wei Y.
      • Thomas E.
      • Renard C.A.
      • et al.
      Hepatic stem-like phenotype and interplay of Wnt/beta-catenin and Myc signaling in aggressive childhood liver cancer.
      Gene expression was Z-score normalized using the mean and standard deviations observed in 8 non-cancer liver samples for each gene, and samples were clustered with and without normalized expression from Cairo et al. at a perfect agreement. Risk stratification by Sumazin et al. was based on a combination of RNA and DNA markers, including (1) AFP, SALL4, DKK1, HMGA2, and LIN28B expression estimates, (2) inferred activity of NRF2 and YAP1 based on their target’s expression profiles, (3) NFE2L2 and TERT promoter mutations, and (4) SALL4 amplification.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.

      Genetic profiling

      Genome-wide copy numbers were estimated using DNA isolated from formalin-fixed paraffin-embedded (FFPE) tumor tissue (39 cases) using the Affymetrix OncoScan FFPE Assay Kit and OncoScan Console 1.3, and data were analyzed by Affymetrix CHAS 3.1 and OncoScan Nexus Copy Number 7.5 after BWA alignment to GRCh37/hg19. There was not enough material for OncoScan profiling of one of our older patient HB samples. Affymetrix OncoScan detects copy number alterations (CNAs) at a 50 kb resolution and was extensively validated in our laboratory for clinical use. Of the 39 cases, 37 were successfully interrogated by targeted sequencing—the Texas Children’s Hospital Pediatric Solid Tumor (PST) panel—to detect the presence or absence of mutations in 967 kb of human DNA that includes coding, splicing, and regulatory regions that comprise 2,247 exons in 124 genes as well as the promoter region of TERT (Table S2). Barcoded DNA libraries were constructed using KAPA HyperPlus Kit and capture hybridization-based target enrichment using a custom-designed Roche Nimblegen SeqCap probe set, which was followed by sequencing on the Illumina® MiSeq® System and sequence alignment using NextGENe v2.4.1.2 and BWA 0.7.
      Variant calling and annotation were performed by NextGENe and Platypus to produce merged outputs of annotated variants, followed by a literature review to determine the significance and classification of each variant.
      • Li M.M.
      • Datto M.
      • Duncavage E.J.
      • Kulkarni S.
      • Lindeman N.I.
      • Roy S.
      • et al.
      Standards and guidelines for the interpretation and reporting of sequence variants in cancer: a joint consensus recommendation of the Association for Molecular Pathology, American Society of Clinical Oncology, and College of American Pathologists.
      To verify capture assay results, CTNNB1 exon 3/4, NFE2L2 exon 2, and TERT promoter mutation analyses were performed as previously described.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      For cases without viable RNA for CTNNB1 reverse-transcription PCR mutation analysis, the same primer set was utilized with genomic DNA for long PCR mutation analysis.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      Predicted somatic mutations and CNAs are given in Tables S2 and S3, respectively. In addition, CNA predictions based on OncoScan analyses and comparative genomic hybridization/single nucleotide polymorphism array analyses of 9 HCCs
      • Haines K.
      • Sarabia S.F.
      • Alvarez K.R.
      • Tomlinson G.
      • Vasudevan S.A.
      • Heczey A.A.
      • et al.
      Characterization of pediatric hepatocellular carcinoma reveals genomic heterogeneity and diverse signaling pathway activation.
      and 45 HBs,
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      respectively, were used as controls for the OncoScan-profiled HCN NOS, HB FPA, and older HB samples. Note that these control HB and HCC samples were not selected for risk, age, or patient outcomes. Note that in heterogeneous tumors, when possible, tumor areas with the most aggressive histological features, as determined by our team of pathologists, were selected for profiling. While one region per tumor was selected for most tumors, we profiled both the HB-like and HCC-like regions of 5 biphasic HCN NOS samples by OncoScan to compare genetic stability estimates, as well as chromosomal and segmental abnormalities across the regions.

      RNA-expression profiling

      We profiled the gene expression of 37 FFPE HB samples using the NanoString nCounter PanCancer Pathways Panel, which includes 800 probes against cancer genes and 30 HB-specific genes (Table S3). The HB-specific genes were selected based on differential expression and risk predictive potential observed in previously reported HB expression profiling efforts and to represent key pathways of hepatocarcinogenesis. Multiplexed measurements of gene expression through digital readouts of the relative abundance of mRNA transcripts were performed as follows: hybridization of RNA to fluorescent reporter probes and capture probes, purification of the target/probe complexes using nCounter Prep Plates and nCounter Cartridges on the nCounter Prep Station, and analysis using the nCounter Digital Analyzer. Gene expression profiles across samples were normalized to equate positive control probes, and expression estimates were Z-score normalized for each probe independently.
      In addition to our samples, we profiled RNAs of 7 low-risk HBs and 4 HCCs using the same nCounter panel. These HBs and HCCs were compiled and previously profiled using alternative technologies by Sumazin et al.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      and Haines et al.,
      • Haines K.
      • Sarabia S.F.
      • Alvarez K.R.
      • Tomlinson G.
      • Vasudevan S.A.
      • Heczey A.A.
      • et al.
      Characterization of pediatric hepatocellular carcinoma reveals genomic heterogeneity and diverse signaling pathway activation.
      respectively. To profile RNAs in single HB cells, we dissociated tumor cells from fresh samples taken from a patient-derived xenograft model of a primary HB
      • Woodfield S.E.
      • Shi Y.
      • Patel R.H.
      • Jin J.
      • Major A.
      • Sarabia S.F.
      • et al.
      A novel cell line based orthotopic xenograft mouse model that recapitulates human hepatoblastoma.
      using an enzyme mix in a gentleMACS C Tube (Miltenyi Biotec). Cells were spun, strained, and washed in PBS (Life Technologies) and BSA 0.04% (VWR), pooled and loaded into the Chromium Next GEM Single Cell 5′ Library and Gel Bead Kit version 1.1 workflow (10X Genomics). Single-cell RNA and cell hashing library preparation was performed according to the manufacturer’s instructions. Libraries were sequenced by novoGene (Davis, California, USA) using the NovaSeq 6000 System and an S4 2 ×150 kit (Illumina). As noted for genetic profiling, tumor areas with the most aggressive histological features, as determined by our team of pathologists, were selected for profiling.

      Results

      Patient outcomes

      The majority (59%, Fig. 2B) of patients with HCN NOS and HB FPAs had poor outcomes. However, all patients with PRETEXT stages 1 and 2 cancers (n = 4), which are expected to be fully resected, had good outcomes. Consequently, outcome-predictive features were most applicable for classifying patients at PRETEXT stages 3 and 4. Not surprisingly, liver transplantation
      • Meyers R.L.
      • Tiao G.
      • De Goyet J.D.V.
      • Superina R.
      • Aronson D.C.
      Hepatoblastoma state of the art: pre-treatment extent of disease, surgical resection guidelines and the role of liver transplantation.
      ,
      • Junco P.T.
      • Cano E.M.
      • Dore M.
      • Gomez J.J.
      • Galán A.S.
      • Vilanova-Sánchez A.
      • et al.
      Prognostic factors for liver transplantation in unresectable hepatoblastoma.
      was significantly predictive of improved patient outcomes in our single-institution cohort, with a hazard ratio of 3.3 (95% CI 2.6 to 4.4) for good outcomes (event-free survival for at least 2 years), assuming a normal sampling distribution. In total, 7/9 (78%) of our transplanted patients with HCN NOS and HB FPAs had good outcomes. By comparison, the presence of metastases (n = 17), which is also known to correlate with patient outcomes,
      • Trobaugh-Lotrario A.D.
      • Meyers R.L.
      • O’Neill A.F.
      • Feusner J.H.
      Unresectable hepatoblastoma: current perspectives.
      had a hazard ratio of 1.9 (95% CI 1.4–2.5) for poor outcomes (relapse or death from disease during a median follow-up of 5.5 years). It was significantly predictive of outcomes in our patient cohort, with p = 0.02 (Fig. 2B). Interestingly, metastases detection did not reduce the predictive ability of transplantation, and transplanted patients had improved outcomes after conditioning on metastatic disease (p <1E-3 by FET using integrated using Fisher’s Method across conditions). In summary, considering HCN NOS, HB FPAs, and older patients with HB, we found near-significant and significant associations between outcomes and metastatic disease or transplantation at p = 0.02 and p = 6E-4 by FET, respectively. Excluding the older HB cohort resulted in significance estimates of p = 0.02 and p = 0.01 by FET, respectively. In addition, and in part because of the risk and age biases in our selection criteria, age at diagnosis was not predictive of patient outcomes. We note that risk evaluations play a role in patient selection for transplantation and that transplantation may not be available for some of the sickest patients, but we cannot account for this confounding factor in our tests.
      Outcomes for older patients with HB were expected to be poor
      • Meyers R.L.
      • Maibach R.
      • Hiyama E.
      • Häberle B.
      • Krailo M.
      • Rangaswami A.
      • et al.
      Risk-stratified staging in paediatric hepatoblastoma: a unified analysis from the Children's Hepatic tumors International Collaboration.
      and patients with PRETEXT stages 3 and 4 HBs that are older than 8 years of age at the time of diagnosis are uniformly considered high risk and are treated aggressively, including early liver transplantation for unresectable tumors. We evaluated the correlation between age and outcomes in our age-unbalanced cohort where patients with equivocal HCN NOS were older and those with HB FPAs were younger, on average (Fig. 2C). Our set of older HBs included 3 patients with good and 2 with poor outcomes, respectively. All 3 patients with good outcomes were transplanted and 1 of the transplanted patients died (Fig. 2A,B). In contrast, only 40% of patients with HCN NOS and HB FPAs had good outcomes. Notably, 75% of older patients with HB, and 78% of transplanted patients with HCN NOS or HB FPAs had good outcomes. Finally, our observations confirmed previous reports based on smaller patient cohorts, which suggested that TERT promoter mutations have predictive prognostic value.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      We established that these mutations are predominantly observed in cancers of older patients (Fig. 2D). Note that risk classification models, including CHIC-HS
      • Meyers R.L.
      • Maibach R.
      • Hiyama E.
      • Häberle B.
      • Krailo M.
      • Rangaswami A.
      • et al.
      Risk-stratified staging in paediatric hepatoblastoma: a unified analysis from the Children's Hepatic tumors International Collaboration.
      and models based on molecular features by Cairo et al.
      • Cairo S.
      • Armengol C.
      • De Reynies A.
      • Wei Y.
      • Thomas E.
      • Renard C.A.
      • et al.
      Hepatic stem-like phenotype and interplay of Wnt/beta-catenin and Myc signaling in aggressive childhood liver cancer.
      and Sumazin et al.,
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      suggested that nearly all patients in our cohort were high risk (Fig. 2A), with the Cairo model suggesting that all tested patients are high risk, and the Sumazin model agreeing for all but 2 patients. Both these patients, 1 with a PRETEXT-2 biphasic HCN NOS and the other with a PRETEXT-4 HB FPA, were resected but not transplanted and had good outcomes.

      Mutation rates

      Tumor biopsies from our patient cohort—37 of whom were profiled by PST—suggested a much higher mutation rate than that previously observed for HBs.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      Our profiles revealed an average of over 3 coding mutations per biopsy, or a rate of 3.1 somatic mutations per Mb of sequenced DNA—a 16-fold higher mutation rate relative to previous observations.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      We note that this mutation rate was estimated based on a targeted region that may be enriched for alterations and should be verified by a genome-wide approach. However, this small coding region (<1 Mb) included as many alterations in our tumors as the expected total number of coding alterations in HBs.
      • Cairo S.
      • Armengol C.
      • Maibach R.
      • Häberle B.
      • Becker K.
      • Carrillo-Reixach J.
      • et al.
      A combined clinical and biological risk classification improves prediction of outcome in hepatoblastoma patients.
      Moreover, while tumors of older patients with HB are known to have higher somatic coding-mutation rates,
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      most patients under 8-years of age in our cohort (11/19, 58%) had 3 or more coding mutations in PST-targeted regions and even patients aged 4-years and under had more than 3 mutations in these regions, on average; see Fig. 3A and Table S2.
      Figure thumbnail gr3
      Fig. 3Unlike low-risk HBs, which harbor few genetic alterations outside of CTNNB1, our cohort revealed recurrent mutations, CNAs, and dysregulation of multiple cancer genes and pathways.
      (A) Consistent with previous reports, all tumors had alterations in Wnt-signaling pathway genes, including CTNNB1 and APC. In addition, we observed recurrent mutations and CNAs—defined as appearing in multiple patients—in genes and regions that are commonly altered in HCCs as well as genes that are associated with pluripotency signaling and common cancer pathways including PI3K-AKT and mTOR signaling. HBs were sorted by patient age at diagnosis; no significant associations between mutation rates or types with histological or clinical subtypes were observed. Mutated genes in each pathway are given in . (B) Unsupervised clustering of the expression profiles of non-cancer liver samples, HCCs, low-risk HBs, HCN NOS cases, and HB FPAs revealed clusters that are enriched with dysregulated genes from at least 4 cancer pathways. These included, from left to right, clusters composed of (1) all profiled HCCs and some older HB FPAs, and HCN NOS cases; (2) all profiled low-risk HBs and some HB FPAs and HCN NOS cases; (3) some older HB FPAs, and HCN NOS cases; and (4) non-tumor samples. Clusters 1-3 tumors had marked Wnt-signaling pathway activation even in tumors with no identified activating mutations in Wnt-signaling pathway genes. Cluster 2 tumors had lower GI and were seen in younger patients. Cluster 3 tumors had intermediate pathway activation compared to Cluster 1 tumors that showed high activation of NF-kB signaling and Cluster 2 tumors that showed high PIK-AKT signaling. Consequently, we labeled Cluster 1 HCC-like, Cluster 2 HB-like, Cluster-3 Intermediate (INTMD), and Cluster 4 Normal. Note that not all gene clusters were enriched for KEGG-pathway gene dysregulation. CNAs, copy number alterations; GI, genetic instability; HB, hepatoblastoma; HB FAPs, HBs demonstrating focal atypia or pleomorphism; HCC, hepatocellular carcinoma; HCN NOS, hepatocellular neoplasm not otherwise specified; INTMD, Intermediate; LOH, loss of heterozygosity; WT, wild-type. (This figure appears in color on the web.)
      Profiled tumors were enriched with mutations and alterations in key cancer genes and pathways (Tables S4 and S5). All patients had alterations in Wnt-signaling pathway genes, and most patients had mutations in genes reported in HCC (84%) even after excluding the commonly altered CTNNB1. Most patients (57%) also had alterations in pathways that are associated with stem-cell pluripotency, PI3K-AKT, or mTOR signaling; see Fig. 3A. The most common mutations were observed in the TERT promoter (51%), FGFR4 (13%), KMT2C (10%), KEAP1 (8%), RPS6KA3 (8%), CDK12 (8%), NOTCH1 (8%), BRCA2 (8%), ARID1A (5%), ARID1B (5%), EP300 (5%), MAPK1 (5%) and PIK3CA (5%); see Table S2 for a complete list. Gains at the MDM4 (51%), SALL4 (24%), FGF19 (14%), and CCND1 (8%) loci, as well as losses at the CDKN2A (11%) locus were also observed in multiple patients (Fig. 3A). Confirming observations by Hirsch et al.,
      • Hirsch T.Z.
      • Pilet J.
      • Morcrette G.
      • Roehrig A.
      • Monteiro B.J.
      • Molina L.
      • et al.
      Integrated genomic analysis identifies driver genes and cisplatin-resistant progenitor phenotype in pediatric liver cancer.
      we identified recurrent CNAs and loss of heterozygosity at chromosome 11p in 38% of our samples. A complete map of CNAs, including significant focal deletions and amplifications, is given in Fig. 4 and Table S3. We note that TERT promoter, KMT2C, and FGFR mutations are hallmarks of aggressive cancers,
      • Zehir A.
      • Benayed R.
      • Shah R.H.
      • Syed A.
      • Middha S.
      • Kim H.R.
      • et al.
      Mutational landscape of metastatic cancer revealed from prospective clinical sequencing of 10,000 patients.
      while KEAP1 mutations and SALL4, CCND1, and CDKN2A CNAs have been previously associated with poor outcomes in patients with HB.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      ,
      • Eichenmuller M.
      • Trippel F.
      • Kreuder M.
      • Beck A.
      • Schwarzmayr T.
      • Haberle B.
      • et al.
      The genomic landscape of hepatoblastoma and their progenies with HCC-like features.
      ,
      • Iolascon A.
      • Giordani L.
      • Moretti A.
      • Basso G.
      • Borriello A.
      • Della Ragione F.
      Analysis of CDKN2A, CDKN2B, CDKN2C, and cyclin Ds gene status in hepatoblastoma.
      Figure thumbnail gr4
      Fig. 4The copy number landscape of HCCs, older HBs, equivocal and biphasic HCN NOS cases, HB FPAs, and HBs with classical histologies included common and differentiating patterns of whole-arm and focal gains and losses.
      These included whole-arm gains of chromosome 1q that were common to all HCCs and high-risk HB subtypes but not found in classical HBs; gains of chromosome 2q that were typical of high-risk HBs; recurrent gains of chromosome 6p in HCN NOS cases and some older HBs; and gains of chromosome 20p that were typical of HCN NOS and HB FPAs. Chromosome-arm losses are common in HCCs and are present in some older patients with HB, but, interestingly, chromosome 4q losses were observed in HCCs, HB FPAs diagnosed in younger children, and one of the HBs diagnosed in an older child. Copy number profiles of classical HBs were retrieved from Sumazin et al. (2017) and included no significantly recurrent CNAs. CNAs, copy number alterations; HB, hepatoblastoma; HB FAPs, HBs demonstrating focal atypia or pleomorphism; HCC, hepatocellular carcinoma; HCN NOS, hepatocellular neoplasm not otherwise specified. (This figure appears in color on the web.)

      Cancer gene and pathway dysregulation

      We profiled gene expression in 37/40 (93%) of our HBs using the NanoString nCounter assay, including 10 equivocal HCN NOS, 12 biphasic HCN NOS, 10 HB FPAs, and 5 HBs of older patients with classical HB histologies. Unsupervised clustering of these profiles—together with control profiles of 7 low-risk HBs,
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      7 non-cancer pediatric liver samples, and 4 pediatric non-fibrolamellar HCCs with upregulated Wnt-signaling pathway genes and no detected viral infection
      • Haines K.
      • Sarabia S.F.
      • Alvarez K.R.
      • Tomlinson G.
      • Vasudevan S.A.
      • Heczey A.A.
      • et al.
      Characterization of pediatric hepatocellular carcinoma reveals genomic heterogeneity and diverse signaling pathway activation.
      —revealed marked differences between the profiles of cancer and non-cancer samples and similarities between the expression profiles of HCN NOS, HB FPAs, low-risk HBs, and HCCs. Namely, as shown in Fig. 3B, all non-cancer profiles clustered together, as did all low-risk HB profiles, and all HCC profiles. However, while profiles of some HCN NOS, HB FPA, and older HB cases clustered with the HCC profiles, others clustered with the low-risk HB profiles, and a third group formed an independent cluster. Interestingly, HB FPAs did not cluster with HCCs, while older HBs did not cluster with low-risk HBs. On the other hand, 50% of the equivocal HCN NOS profiles, nearly all of which were collected from older patients (83% were older than 8 years at diagnosis), clustered with low-risk HBs. Only one biphasic HCN NOS (MOLR310), who was younger than 2 years at diagnosis, clustered with low-risk HBs. We note that, for all biphasic HCN NOS cases, profiled samples were selected to include cells from tumor regions with HCC-like features. However, to compare abnormalities across the regions we profiled both the HB-like and HCC-like regions of 5 biphasic HCN NOS cases by OncoScan. The result, depicted in Figs. S1-5 suggested that, on average, HCC-like regions are marked by greater genomic instability in terms of both chromosome-arm gains and losses; the number of expected chromosome-arm events in HCC-like regions was over 4 times greater than in HB-like regions.
      Identified clusters were enriched with dysregulated genes from multiple cancer pathways. Namely, all tumor profiles showed upregulation of Wnt-signaling pathway genes. The cluster containing low-risk HBs included cancers with upregulated PI3K-AKT signaling pathway genes (p <2E-3, gene set-enrichment analysis [GSEA]), and the cluster containing HCCs included cancers with upregulated NF-kB signaling pathway genes (p <6E-3, GSEA). Interestingly the third cluster, which is composed of HCN NOS, HB FPAs, and older HB profiles that did not cluster with low-risk HBs or HCCs, showed intermediate expression levels of genes in these pathways; see Fig. 5A. Consequently, we named the 4 main clusters in Fig. 3B HCC-like, HB-like, Intermediate, and Normal. We note that the intermediate category was highly heterogeneous and included 3 biphasic HCN NOS cases, 2 equivocal HCN NOS cases, 4 HB FPAs, and 2 older HBs. Expression profiles of HB-like samples and Intermediate samples showed significant upregulation of Cell cycle, DNA replication, Homologous recombination, Fanconi Anemia, and PI3K-AKT pathway genes with p <1E-2 and p <5E-2 by GSEA, for HB-like samples and Intermediate samples, respectively. Expression profiles of HCC-like samples and Intermediate samples showed significant upregulation of NK-κB signaling, Focal adhesion, Rheumatoid arthritis, and MAPK signaling pathway genes with p <8E-3 and p <5E-2 by GSEA for HCC-like samples and Intermediate samples, respectively (Fig. 5A). Age at diagnosis of patients represented in this category varied widely, including 3 and 4 patients that were diagnosed before the age of 4 and after the age of 12, respectively.
      Figure thumbnail gr5
      Fig. 5Genotypes and RNA-expression profiles of our tumors suggested that HCN NOS cases and HB FPAs fall within a biological spectrum between low-risk HBs and HCCs and that their genetic instability may be associated with observed pathway dysregulation.
      (A) Our INTMD expression cluster showed intermediate cancer-pathway activation compared to our HCC-like and HB-like clusters. (B) It also demonstrated an intermediate level of genomic instability—as calculated by the genome-wide CNA burden in each tumor—with HCC tumors having the highest genomic instability, on average. (C) High-risk tumors were enriched for chr1q amplification, which significantly coincided with higher MDM4 expression. (D) MDM4 expression profiles were inversely correlated with p53 activation in our dataset (Nanostring) and in a second RNA-expression (Affymetrix) data set
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      (Pearson correlation). (E) Single-cell profiling of a high-risk HB showed a significant correlation between MDM4 expression profiles and p53 activation in thousands of profiled cells. HB, hepatoblastoma; HB FAPs, HBs demonstrating focal atypia or pleomorphism; HCC, hepatocellular carcinoma; HCN NOS, hepatocellular neoplasm not otherwise specified; INTMD, Intermediate.

      CNAs and genetic instability

      To benchmark CNA detection in our set, we compared our profiles to profiles of 9 pediatric HCCs
      • Haines K.
      • Sarabia S.F.
      • Alvarez K.R.
      • Tomlinson G.
      • Vasudevan S.A.
      • Heczey A.A.
      • et al.
      Characterization of pediatric hepatocellular carcinoma reveals genomic heterogeneity and diverse signaling pathway activation.
      —including 4 HCCs with upregulated Wnt-signaling pathway genes—and 45 HBs,
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      see Fig. 4. Our results confirmed that HBs have fewer recurrent CNAs and that HCCs have recurrent chromosome-1q gains (90%) and losses in multiple chromosomes, including 1p (70%), 4 (30%), 11 (40%), 15 (40%), 16 (40%), and 18 (30%). Profiles of HBs in older children demonstrated more frequent chromosome-arm losses than other HBs in the cohort, but all older HBs had chromosome-1q gains. The majority (90%) of HCN NOS and HB FPAs had chromosome-1q gains, but unlike older HBs and HCCs, we observed widespread recurrent large-scale gains in these samples, including gains of chromosomal arms 2q (80%), 6p (70%), and 20p (70%), which were also observed in Wnt pathway-activated HCCs.
      • Haines K.
      • Sarabia S.F.
      • Alvarez K.R.
      • Tomlinson G.
      • Vasudevan S.A.
      • Heczey A.A.
      • et al.
      Characterization of pediatric hepatocellular carcinoma reveals genomic heterogeneity and diverse signaling pathway activation.
      Interestingly, chromosome 4q losses were observed in HCCs, HB FPAs, and some older HBs, which suggest that these losses are not associated with age. We used CNAs to evaluate genetic instability at genome scales, scoring each tumor based on the number of nucleotides that were identified as gained or lost by OncoScan. Our analysis suggested that tumor samples identified as HB-like (Fig. 3B) based on gene expression were the most genetically stable, while tumors that were identified as HCC-like were the most genetically unstable (Fig. 5B).
      Chromosome-1q gains were the most frequent genetic event in our high-risk cancer cohort and were detected in nearly 90% of HCN NOS and HB FPAs. Genes with expression profiles that were significantly correlated with their CNA profiles (p <0.01) included THEM4, EFNA1, H3F3A, IL6R, and MDM4 (Fig. 5C). Gains at the MDM4 locus have been previously associated with poor outcomes in HBs
      • Arai Y.
      • Honda S.
      • Haruta M.
      • Kasai F.
      • Fujiwara Y.
      • Ohshima J.
      • et al.
      Genome-wide analysis of allelic imbalances reveals 4q deletions as a poor prognostic factor and MDM4 amplification at 1q32. 1 in hepatoblastoma.
      and may inhibit p53 activity.
      • Woodfield S.E.
      • Shi Y.
      • Patel R.H.
      • Chen Z.
      • Shah A.P.
      • Whitlock R.S.
      • et al.
      MDM4 inhibition: a novel therapeutic strategy to reactivate p53 in hepatoblastoma.
      Indeed, analysis of normalized expression of a panel of p53 target genes that have been studied in HCCs
      • Ally Adrian
      • Balasundaram Miruna
      • Carlsen Rebecca
      • Chuah Eric
      • Clarke Amanda
      • Dhalla Noreen
      • et al.
      TCGA
      Comprehensive and integrative genomic characterization of hepatocellular carcinoma.
      and included in our NanoString platform—CAPN2, CDKN1A, DUSP5, EPHA2, FAS, GADD45A, NFKBIA, and STAT3—suggested that MDM4 expression in both our and the Sumazin et al. datasets were significantly (p <0.01) inversely correlated with p53 activity (see Fig. 5D). In addition, single-cell RNA sequencing of a MOLR442-derived xenograft showed that MDM4 intratumoral RNA-expression variability is inversely correlated with p53 activity (Fig. 5E). We note that fluorescence in situ hybridization verified recurrent chromosome 4q34.3-35.2 losses in HCCs (56%), older HBs (75%), and HB FPAs (41%), and recurrent chromosome 6q12.1-25.3 gains in HCN NOS cases (68%).

      Discussion

      Accurate risk stratification is key to improving treatment selection and, consequently, outcomes of children diagnosed with hepatocellular neoplasms.
      • Schmid I.
      • von Schweinitz D.
      Pediatric hepatocellular carcinoma: challenges and solutions.
      Traditionally, risk stratification has been based on clinical, imaging, and histological parameters,
      • Towbin A.J.
      • Meyers R.L.
      • Woodley H.
      • Miyazaki O.
      • Weldon C.B.
      • Morland B.
      • et al.
      2017 PRETEXT: radiologic staging system for primary hepatic malignancies of childhood revised for the Paediatric Hepatic International Tumour Trial (PHITT).
      ,
      • Meyers R.L.
      • Tiao G.
      • De Goyet J.D.V.
      • Superina R.
      • Aronson D.C.
      Hepatoblastoma state of the art: pre-treatment extent of disease, surgical resection guidelines and the role of liver transplantation.
      ,
      • Trobaugh-Lotrario A.D.
      • Meyers R.L.
      • O’Neill A.F.
      • Feusner J.H.
      Unresectable hepatoblastoma: current perspectives.
      but recent studies, including efforts led by single institutions
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      ,
      • Cairo S.
      • Armengol C.
      • Maibach R.
      • Häberle B.
      • Becker K.
      • Carrillo-Reixach J.
      • et al.
      A combined clinical and biological risk classification improves prediction of outcome in hepatoblastoma patients.
      ,
      • Cairo S.
      • Armengol C.
      • De Reynies A.
      • Wei Y.
      • Thomas E.
      • Renard C.A.
      • et al.
      Hepatic stem-like phenotype and interplay of Wnt/beta-catenin and Myc signaling in aggressive childhood liver cancer.
      as well as national and international consortia,
      • López-Terrada D.
      • Alaggio R.
      • de Dávila M.T.
      • Czauderna P.
      • Hiyama E.
      • Katzenstein H.
      • et al.
      Towards an international pediatric liver tumor consensus classification: proceedings of the Los Angeles COG liver tumors symposium.
      ,
      • Meyers R.L.
      • Maibach R.
      • Hiyama E.
      • Häberle B.
      • Krailo M.
      • Rangaswami A.
      • et al.
      Risk-stratified staging in paediatric hepatoblastoma: a unified analysis from the Children's Hepatic tumors International Collaboration.
      ,
      • Towbin A.J.
      • Meyers R.L.
      • Woodley H.
      • Miyazaki O.
      • Weldon C.B.
      • Morland B.
      • et al.
      2017 PRETEXT: radiologic staging system for primary hepatic malignancies of childhood revised for the Paediatric Hepatic International Tumour Trial (PHITT).
      have proposed improvements by incorporating molecular biomarkers of prognostic significance. These biomarkers, including gene expression profiles,
      • Cairo S.
      • Armengol C.
      • De Reynies A.
      • Wei Y.
      • Thomas E.
      • Renard C.A.
      • et al.
      Hepatic stem-like phenotype and interplay of Wnt/beta-catenin and Myc signaling in aggressive childhood liver cancer.
      and genetic
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      ,
      • Cairo S.
      • Armengol C.
      • Maibach R.
      • Häberle B.
      • Becker K.
      • Carrillo-Reixach J.
      • et al.
      A combined clinical and biological risk classification improves prediction of outcome in hepatoblastoma patients.
      and epigenetic tumor features,
      • Carrillo-Reixach J.
      • Torrens L.
      • Simon-Coma M.
      • Royo L.
      • Domingo-Sàbat M.
      • Abril-Fornaguera J.
      • et al.
      Epigenetic footprint enables molecular risk stratification of hepatoblastoma with clinical implications.
      have shown promise in retrospective studies and are now being prospectively evaluated on patients enrolled in AHEP1531, a Children's Oncology Group pediatric liver cancer therapeutic clinical trial, and the international PHITT trial, which assign intervention based on CHIC risk stratification.
      • Lim I.I.P.
      • Bondoc A.J.
      • Geller J.I.
      • Tiao G.M.
      Hepatoblastoma—the evolution of biology, surgery, and transplantation.
      While the clinical risk stratification of patients with HB has greatly improved over the past 20 years, current risk stratification algorithms include no clinically validated molecular biomarkers. In addition, the histological classification of a subset of pediatric hepatocellular tumors with either intermediate or mixed histological features of HB and HCC remains challenging. To improve risk stratification and to standardize the diagnoses of these tumors, the International Pediatric Liver Tumor Consensus Classification System created the provisional category HCN NOS. HCN NOS cases are characterized by mixed HB and HCC histological features, including the presence of macrotrabecular pattern and solid growth, acinar formations, multinuclear giant tumor cells, and high mitotic rates. Tumors with these features were first reported in descriptive and limited studies
      • Prokurat A.
      • Kluge P.
      • Kościesza A.
      • Perek D.
      • Kappeler A.
      • Zimmermann A.
      Transitional liver cell tumors (TLCT) in older children and adolescents: a novel group of aggressive hepatic tumors expressing beta-catenin.
      ,
      • Zimmermann A.
      The emerging family of hepatoblastoma tumours: from ontogenesis to oncogenesis.
      that did not propose diagnostic criteria but suggested that these tumors originate from transitional liver cells, happen in older children, and may be associated with clinically aggressive behavior.
      • Meyers R.L.
      • Maibach R.
      • Hiyama E.
      • Häberle B.
      • Krailo M.
      • Rangaswami A.
      • et al.
      Risk-stratified staging in paediatric hepatoblastoma: a unified analysis from the Children's Hepatic tumors International Collaboration.
      To date, only a handful of tumors with these features have ever been molecularly profiled
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      ,
      • Eichenmuller M.
      • Trippel F.
      • Kreuder M.
      • Beck A.
      • Schwarzmayr T.
      • Haberle B.
      • et al.
      The genomic landscape of hepatoblastoma and their progenies with HCC-like features.
      ,
      • Carrillo-Reixach J.
      • Torrens L.
      • Simon-Coma M.
      • Royo L.
      • Domingo-Sàbat M.
      • Abril-Fornaguera J.
      • et al.
      Epigenetic footprint enables molecular risk stratification of hepatoblastoma with clinical implications.
      and both their molecular and clinical features remain uncharacterized.
      To address open questions about HCN NOS biology and outcomes, we performed the first systematic molecular and clinical study of 25 HCN NOS cases using samples banked at the Texas Children’s Hospital. Our results confirmed that HCN NOS shares molecular features with HB and HCC, suggested that HCN NOS may be diagnosed in young children, and demonstrated that patients with HCN NOS have poor outcomes (57%) irrespective of age at diagnosis. We classified HCN NOS cases into biphasic and equivocal histological subtypes because of their distinct histological presentations and hypothesized that these subtypes may also be biologically distinct. However, our analysis revealed biological and clinical similarities. Both histological subtypes were associated with poor outcomes: while only 20% of patients with HB are expected to have poor outcomes, 66% and 45% of patients with biphasic and equivocal HCN NOS had poor outcomes, respectively. CNA profiles of tumor samples from both subtypes demonstrated recurrent large-scale chromosomal gains and their RNA-expression profiles clustered together. Our results suggested that, at least at the resolution of our assays, the clinical and biological profiles of biphasic and equivocal HCN NOS cases were similar.
      Interestingly, a systematic histological review identified a subset of HBs that did not fulfill the HCN NOS diagnostic criteria but demonstrated focal pleomorphism, anaplasia, or other features typically seen in HCC, either at diagnosis or after chemotherapy. We designated these tumors as HBs that demonstrated histological HCC features focally (HB FPAs) and included them in our study to compare their characteristics to those of other HBs, HCN NOS, and HCC. Molecular profiling of HB FPAs and HCN NOS suggested that these tumors share common molecular features with both HBs and HCCs, including recurrent genetic alterations that have been observed in Wnt pathway-activated HCCs, and similar patterns of cancer gene dysregulation. In patients younger than 8 years at diagnosis, HB FPAs were not associated with better outcomes than HCN NOS, with 67% of patients having poor outcomes. In total, the majority (59%) of patients with HCN NOS and HB FPAs in our dataset had poor outcomes, and the majority (54%) with good outcomes received liver transplantations. Because of their similar outcomes and histological and biological features, we designated these aggressive hepatocellular neoplasms, including HCN NOS and HB FPAs, as HBs with HCC features (HBCs).
      Our analyses suggested that HBCs can be histologically recognized and confirmed by molecular testing as outlined in Fig. 6. This proposed diagnostic algorithm uses histology to classify pediatric hepatocellular neoplasms into HBs, HCCs, and candidate HBCs. Histologically heterogenous regions of candidate HBCs are then molecularly tested using select genetic biomarkers to confirm HBC diagnoses. We note that the absence of our proposed genetic biomarkers does not rule out HBCs because of the limitations of our study, which may have not detected the entire spectrum of genetic and epigenetic abnormalities. Interestingly, our study suggested a broad age range for patients with HBCs, as well as age-related genetic alterations and cancer gene dysregulation. Namely, we observed significant correlations between older age at diagnosis and TERT promoter mutations, and we observed age-related trends in the clustering of RNA-expression profiles for some HB, HCC, and HBC subtypes. These results indicate that TERT promoter mutations, which have previously been identified as prognostic biomarkers, may be passenger mutations that are indirectly associated with the correlation between age and patient outcomes. However, HBC mutation rates, recurrent CNAs, and cancer gene dysregulation were not significantly associated with age at diagnosis. Indeed, while expression profiles of some tumor samples from older patients clustered together, our HCC-like expression cluster (Fig. 3B) included patients with HCN NOS as young as 2 years and as old as 23 years at diagnosis.
      Figure thumbnail gr6
      Fig. 6Pediatric hepatocellular malignant neoplasms are categorized by histological review and assigned to specific HB subtypes, including HB FPA, equivocal and biphasic HCN NOS, and HCC.
      Histological HB patterns include fetal, embryonal, epithelial, mesenchymal, small-cell undifferentiated, and cholangioblastic. Fetal HBs are well-differentiated, uniform in size with round nuclei and have minimal mitotic activity. Embryonal HBs show higher nuclear to cytoplasm ratio, angulated nuclei, and may form primitive tubules. Small-cell undifferentiated HB components consist of cells with minimal pale amphophilic cytoplasms and round to oval nuclei with fine chromatin. Cholangioblastic HBs contain a characteristic bile duct component, usually at the periphery of epithelial islands. Fibrolamellar and non-fibrolamellar HCCs are characterized by larger cells that display greater nuclear pleomorphism with prominent nucleoli, pseudo inclusions and abnormal mitoses, growing either in trabeculae, nests, and solid sheets. HCCs may develop in the presence of underlying liver disease. HB FPAs demonstrate focal atypia, anaplasia, and may show macrotrabecular or solid growth pattern. HCN NOS cases include cells with features that are commonly associated with both HBs and HCCs (equivocal), or distinct HB-like and HCC-like regions (biphasic HCN NOS cases). While the HB and HCC subtypes mentioned can be diagnosed by histology alone, we identified genetic features that are characteristic of HB FPAs and HCN NOS cases, and the detection of these features was sufficient to confirm HBC diagnoses. The most common—this is not an exhaustive list—of these were mutations in the TERT promoter (51%), FGFR4 (13%), KMT2C (10%), KEAP1 (8%), RPS6KA3 (8%), CDK12 (8%), NOTCH1 (8%), BRCA2 (8%), ARID1A (5%), ARID1B (5%), EP300 (5%), MAPK1 (5%) and PIK3CA (5%). Gains at MDM4 (51%), SALL4 (24%), FGF19 (14%), and CCND1 (8%), as well as losses at CDKN2A (11%) and LOH at chromosome 11p (38%). HB, hepatoblastoma; HB FAPs, HBs demonstrating focal atypia or pleomorphism; HCC, hepatocellular carcinoma; HCN NOS, hepatocellular neoplasm not otherwise specified; LOH, loss of heterozygosity. (This figure appears in color on the web.)
      We retrospectively compared outcomes for patients with HBCs to those of patients with other HBs, including patients with HB who were classified as high risk because of their age at diagnosis (older than 8 years).
      • Meyers R.L.
      • Maibach R.
      • Hiyama E.
      • Häberle B.
      • Krailo M.
      • Rangaswami A.
      • et al.
      Risk-stratified staging in paediatric hepatoblastoma: a unified analysis from the Children's Hepatic tumors International Collaboration.
      In total, 80% of HBs are expected to have good outcomes,
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      but only 60% of the older high-risk HBs in our dataset had good outcomes. Similarly, only 41% of all HBCs and 30% of unresectable (PRETEXT 3 and 4) HBCs in our dataset had good outcomes, compared to 78% of transplanted HBCs. Our analysis suggests that transplanted patients with HBC had significantly better outcomes than those treated by chemotherapy and surgery alone, and that patients with HBCs may benefit from aggressive and early interventions, including transplant referral when complete resection is not feasible. We believe that this hypothesis merits further testing at a larger scale and in prospective trials that would account for surgical staging, clinical, histological, and molecular tumor features. Most importantly, prospective trials that aim to estimate the benefit of liver transplantation for patients with HBCs must have uniform treatment protocols and account for patient suitability and the availability of candidate organs for transplantation. These confounding factors could not be accounted for in our retrospective study.
      To investigate the molecular features of HBCs, we compared their genetic profiles and RNA-expression profiles to those of classical HBs, other high-risk HBs, and HCCs. Pediatric tumors are known to be more genetically stable than adult tumors, and HBs are known to be more stable than most pediatric tumors.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      ,
      • Haines K.
      • Sarabia S.F.
      • Alvarez K.R.
      • Tomlinson G.
      • Vasudevan S.A.
      • Heczey A.A.
      • et al.
      Characterization of pediatric hepatocellular carcinoma reveals genomic heterogeneity and diverse signaling pathway activation.
      ,
      • Gröbner S.N.
      • Worst B.C.
      • Weischenfeldt J.
      • Buchhalter I.
      • Kleinheinz K.
      • Rudneva V.A.
      • et al.
      The landscape of genomic alterations across childhood cancers.
      Indeed, other than recurring mutations in CTNNB1 and a few other Wnt pathways genes, HBs have been reported to have fewer than 3-coding mutations per tumor
      • Ranganathan S.
      • Lopez-Terrada D.
      • Alaggio R.
      Hepatoblastoma and pediatric hepatocellular carcinoma: an update.
      ,
      • Cairo S.
      • Armengol C.
      • Maibach R.
      • Häberle B.
      • Becker K.
      • Carrillo-Reixach J.
      • et al.
      A combined clinical and biological risk classification improves prediction of outcome in hepatoblastoma patients.
      or 0.2 somatic mutations per Mb of sequenced DNA.
      • Sumazin P.
      • Chen Y.
      • Treviño L.R.
      • Sarabia S.F.
      • Hampton O.A.
      • Patel K.
      • et al.
      Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
      Our analyses suggested that HBCs are more genetically unstable than other HBs, that they have up to 3-coding mutations per Mb of sequenced DNA, and that they carry recurrent genetic alterations that target cancer genes that are commonly mutated in HCCs, as well as commonly altered cancer pathways including PI3K-AKT and mTOR signaling (see Table 1). Interestingly, we also identified mutations in pluripotency signaling and developmental genes, including FGFR3, FGFR4, HRAS, NOTCH1, and EP300. While these genes are less commonly altered in HCCs, mutations in these genes have been identified in other adult and pediatric cancers. Moreover, RNA-expression analyses suggested that cancer-pathway dysregulation may be associated with genetic instability. Namely, cancers with HB-like expression profiles showed greater genetic stability, and cancers with HCC-like expression profiles were more unstable. More specific observations included the correlation between chromosome-1q gains, which were the most common alterations in HBCs, and MDM4 expression and p53-pathway deactivation.
      Table 1Summary of biological features that are common to HBs, HBCs, and HCCs.
      HBsHBCsHCC
      Mutation rates0.2 somatic mutations per Mb3 somatic mutations per Mb3-5 somatic mutations per Mb
      Recurrently altered cancer genesCTNNB1, APC, NFE2L2, ARID1A, KMT2D, RPS6KA3CTNNB1, ARID1A, ARID1B, BRCA2, CCND1, CDKN2A, EP300, FGF19, FGFR4, KEAP1, KMT2C, KMT2D, MAPK1, MDM4, NFE2L2, NOTCH1, PIK3CA, RPS6KA3, SALL4, TERT promoterCTNNB1, KEAP1, ARID1A, ARID2, ATM, AXIN, CCND1, FAK, FGF19, IGF2, KMT2D, MET, MYC, NFE2L2, PD1, RAS, RB1, RPS6KA3, TERTpromoter, TP53, VEGFA
      Pathways enriched for mutationsWnt-signaling pathwayStem-cell pluripotency, Wnt, PIK, AKT/mTOR, and RAS/MAPK signaling pathways; HCC pathwayWnt, AKT/MTOR, and RAS/MAPK signaling; Telomere maintenance; Immune checkpoint pathway
      Recurrent CNAs1q, 2, 8, 20, 22 gains, 4q, 1p losses, unbalanced translocations involving 1q12-211q, 2q, 6p, 20 gains; 1p, 4q, 11, 15, 16 segmental and whole-arm losses1q gains; 1p, 4, 11, 15, 16, and widespread losses and whole-arm losses; whole-arm gains genome-wide in Wnt HCCs
      HB mutations, other than in CTNNB1 and APC, have been documented predominantly in high-risk cases.
      CNAs, copy number alterations; HBs, hepatoblastomas; HBCs, HBs with HCC features; HCC, hepatocellular carcinoma.
      In summary, our retrospective single-institution study suggested that HCN NOS and HB FPAs are aggressive hepatocellular tumors that share histological and molecular features with both HB and HCC. We collectively referred to these cancers as HBCs and proposed integrating histological and molecular testing to accurately diagnose HBCs. We note that not all high-risk HBs are HBCs. Indeed, multiple risk stratification models have been proposed for HB, and some high-risk HBs do not display HBC histology or biology. Our analysis suggested that Wnt-driven hepatocellular tumors in children represent a biological and clinical spectrum, and that HBCs are intermediate tumors that are genetically and phenotypically distinct from classical HBs and pediatric HCCs. Further characterization will be necessary to better define the molecular profile of HCN NOS and to differentiate this group from high-risk HB and Wnt-driven HCC in children and young adults. Our analyses also suggested that these cancers can afflict younger patients and that their biological hallmarks are recurrent genetic mutations and whole chromosome-arm gains. In our cohort, liver transplantation was key to improving HBC outcomes, but conclusive evidence for risk stratification and recommendations for their treatment will only be achieved by identifying and characterizing this important group of tumors in upcoming international clinical trials. Such trials should consider integrating both histological features and molecular biomarkers to better diagnose, risk stratify, and select therapies for children with high-risk hepatocellular neoplasms, including HBCs.

      Abbreviations

      CNA, copy number alterations; FFPE, formalin-fixed paraffin-embedded; GSEA, gene set-enrichment analysis; HB, hepatoblastoma; HBCs, HBs with HCC features; HB FAPs, HBs demonstrating focal atypia or pleomorphism; HCC, hepatocellular carcinoma; HCN NOS, hepatocellular neoplasm not otherwise specified; PST, Texas Children’s Hospital Pediatric Solid Tumor Panel; TCH, Texas Children’s Hospital.

      Financial support

      CPRIT award RP180674, the European Union’s Horizon 2020 research and innovation programme under grant agreement 826121, the Schindler Foundation, and NCI award R21CA223140.

      Authors' contributions

      PS, TLP, SFS, MJF, and DHL conceived the project. EFH, HK, AO, RM, GT, JG, SR, AAR, MJF, DHL obtained and processed patient specimens. EFH, KRA, CRP, AP, KS, SEW, JHG, SAV, AH, AR, KEF, RA, KRP, MJP, DHL performed experiments. PS, TLP, SFS, HRK, MU, MJNP, JLE, KEF, KRP, and DHLT analyzed data. PS, TLP, SFS, RA, KRP, MJF, and DHL wrote the manuscript.

      Data availability statement

      Raw data, including DNA and single-cell RNA sequencing, NanoString expression estimates, and OncoScan CNA predictions will be made freely available from the European Nucleotide Archive and Gene Expression Omnibus, under study accessions PRJEB46350, and GSE208217.

      Conflict of interest

      The authors have no conflict of interests to report.
      Please refer to the accompanying ICMJE disclosure forms for further details.

      Acknowledgements

      We thank the BCM GARP core for their assistance in performing molecular expression profiling. We thank our anonymous reviewers for critique that helped improve the quality of the manuscript.

      Supplementary data

      References

        • Hubbard A.K.
        • Spector L.G.
        • Fortuna G.
        • Marcotte E.L.
        • Poynter J.N.
        Trends in international incidence of pediatric cancers in children under 5 years of age: 1988–2012.
        JNCI Cancer Spectr. 2019; 3: pkz007
        • Czauderna P.
        • Lopez-Terrada D.
        • Hiyama E.
        • Häberle B.
        • Malogolowkin M.H.
        • Meyers R.L.
        Hepatoblastoma state of the art: pathology, genetics, risk stratification, and chemotherapy.
        Curr Opin Pediatr. 2014; 26: 19-28
        • Ranganathan S.
        • Lopez-Terrada D.
        • Alaggio R.
        Hepatoblastoma and pediatric hepatocellular carcinoma: an update.
        Pediatr Dev Pathol. 2020; 23: 79-95
        • Sumazin P.
        • Chen Y.
        • Treviño L.R.
        • Sarabia S.F.
        • Hampton O.A.
        • Patel K.
        • et al.
        Genomic analysis of hepatoblastoma identifies distinct molecular and prognostic subgroups.
        Hepatology. 2017; 65: 104-121
        • Ueda Y.
        • Hiyama E.
        • Kamimatsuse A.
        • Kamei N.
        • Ogura K.
        • Sueda T.
        Wnt signaling and telomerase activation of hepatoblastoma: correlation with chemosensitivity and surgical resectability.
        J Pediatr Surg. 2011; 46: 2221-2227
        • Haines K.
        • Sarabia S.F.
        • Alvarez K.R.
        • Tomlinson G.
        • Vasudevan S.A.
        • Heczey A.A.
        • et al.
        Characterization of pediatric hepatocellular carcinoma reveals genomic heterogeneity and diverse signaling pathway activation.
        Pediatr Blood Cancer. 2019; 66e27745
        • Prokurat A.
        • Kluge P.
        • Kościesza A.
        • Perek D.
        • Kappeler A.
        • Zimmermann A.
        Transitional liver cell tumors (TLCT) in older children and adolescents: a novel group of aggressive hepatic tumors expressing beta-catenin.
        Med Pediatr Oncol. 2002; 39: 510-518
        • Zimmermann A.
        The emerging family of hepatoblastoma tumours: from ontogenesis to oncogenesis.
        Eur J Cancer. 2005; 41: 1503-1514
        • López-Terrada D.
        • Alaggio R.
        • de Dávila M.T.
        • Czauderna P.
        • Hiyama E.
        • Katzenstein H.
        • et al.
        Towards an international pediatric liver tumor consensus classification: proceedings of the Los Angeles COG liver tumors symposium.
        Mod Pathol. 2014; 27: 472-491
        • Eichenmuller M.
        • Trippel F.
        • Kreuder M.
        • Beck A.
        • Schwarzmayr T.
        • Haberle B.
        • et al.
        The genomic landscape of hepatoblastoma and their progenies with HCC-like features.
        J Hepatol. 2014; 61: 1312-1320
        • Carrillo-Reixach J.
        • Torrens L.
        • Simon-Coma M.
        • Royo L.
        • Domingo-Sàbat M.
        • Abril-Fornaguera J.
        • et al.
        Epigenetic footprint enables molecular risk stratification of hepatoblastoma with clinical implications.
        J Hepatol. 2020; 73: 328-341
        • Cairo S.
        • Armengol C.
        • Maibach R.
        • Häberle B.
        • Becker K.
        • Carrillo-Reixach J.
        • et al.
        A combined clinical and biological risk classification improves prediction of outcome in hepatoblastoma patients.
        Eur J Cancer. 2020; 141: 30-39
        • Haeberle B.
        • Rangaswami A.
        • Krailo M.
        • Czauderna P.
        • Hiyama E.
        • Maibach R.
        • et al.
        The importance of age as prognostic factor for the outcome of patients with hepatoblastoma: analysis from the Children's Hepatic tumors International Collaboration (CHIC) database.
        Pediatr Blood Cancer. 2020; e28350
        • Meyers R.L.
        • Maibach R.
        • Hiyama E.
        • Häberle B.
        • Krailo M.
        • Rangaswami A.
        • et al.
        Risk-stratified staging in paediatric hepatoblastoma: a unified analysis from the Children's Hepatic tumors International Collaboration.
        Lancet Oncol. 2017; 18: 122-131
        • Towbin A.J.
        • Meyers R.L.
        • Woodley H.
        • Miyazaki O.
        • Weldon C.B.
        • Morland B.
        • et al.
        2017 PRETEXT: radiologic staging system for primary hepatic malignancies of childhood revised for the Paediatric Hepatic International Tumour Trial (PHITT).
        Pediatr Radiol. 2018; : 1-19
        • Roebuck D.J.
        • Aronson D.
        • Clapuyt P.
        • Czauderna P.
        • de Ville de Goyet J.
        • Gauthier F.
        • et al.
        2005 PRETEXT: a revised staging system for primary malignant liver tumours of childhood developed by the SIOPEL group.
        Pediatr Radiol. 2007; 37: 123-132
        • Cairo S.
        • Armengol C.
        • De Reynies A.
        • Wei Y.
        • Thomas E.
        • Renard C.A.
        • et al.
        Hepatic stem-like phenotype and interplay of Wnt/beta-catenin and Myc signaling in aggressive childhood liver cancer.
        Cancer Cell. 2008; 14: 471-484
        • Li M.M.
        • Datto M.
        • Duncavage E.J.
        • Kulkarni S.
        • Lindeman N.I.
        • Roy S.
        • et al.
        Standards and guidelines for the interpretation and reporting of sequence variants in cancer: a joint consensus recommendation of the Association for Molecular Pathology, American Society of Clinical Oncology, and College of American Pathologists.
        J Mol Diagn. 2017; 19: 4-23
        • Woodfield S.E.
        • Shi Y.
        • Patel R.H.
        • Jin J.
        • Major A.
        • Sarabia S.F.
        • et al.
        A novel cell line based orthotopic xenograft mouse model that recapitulates human hepatoblastoma.
        Scientific Rep. 2017; 717751
        • Meyers R.L.
        • Tiao G.
        • De Goyet J.D.V.
        • Superina R.
        • Aronson D.C.
        Hepatoblastoma state of the art: pre-treatment extent of disease, surgical resection guidelines and the role of liver transplantation.
        Curr Opin Pediatr. 2014; 26: 29-36
        • Junco P.T.
        • Cano E.M.
        • Dore M.
        • Gomez J.J.
        • Galán A.S.
        • Vilanova-Sánchez A.
        • et al.
        Prognostic factors for liver transplantation in unresectable hepatoblastoma.
        Eur J Pediatr Surg. 2019; 29: 28-32
        • Trobaugh-Lotrario A.D.
        • Meyers R.L.
        • O’Neill A.F.
        • Feusner J.H.
        Unresectable hepatoblastoma: current perspectives.
        Hepatic Med Evid Res. 2017; 9: 1
        • Hirsch T.Z.
        • Pilet J.
        • Morcrette G.
        • Roehrig A.
        • Monteiro B.J.
        • Molina L.
        • et al.
        Integrated genomic analysis identifies driver genes and cisplatin-resistant progenitor phenotype in pediatric liver cancer.
        Cancer Discov. 2021; 10: 2524-2543
        • Zehir A.
        • Benayed R.
        • Shah R.H.
        • Syed A.
        • Middha S.
        • Kim H.R.
        • et al.
        Mutational landscape of metastatic cancer revealed from prospective clinical sequencing of 10,000 patients.
        Nat Med. 2017; 23: 703-713
        • Iolascon A.
        • Giordani L.
        • Moretti A.
        • Basso G.
        • Borriello A.
        • Della Ragione F.
        Analysis of CDKN2A, CDKN2B, CDKN2C, and cyclin Ds gene status in hepatoblastoma.
        Hepatology. 1998; 27: 989-995
        • Arai Y.
        • Honda S.
        • Haruta M.
        • Kasai F.
        • Fujiwara Y.
        • Ohshima J.
        • et al.
        Genome-wide analysis of allelic imbalances reveals 4q deletions as a poor prognostic factor and MDM4 amplification at 1q32. 1 in hepatoblastoma.
        Genes Chromosom Cancer. 2010; 49: 596-609
        • Woodfield S.E.
        • Shi Y.
        • Patel R.H.
        • Chen Z.
        • Shah A.P.
        • Whitlock R.S.
        • et al.
        MDM4 inhibition: a novel therapeutic strategy to reactivate p53 in hepatoblastoma.
        Sci. Rep. 2021; 11: 1-17
        • Ally Adrian
        • Balasundaram Miruna
        • Carlsen Rebecca
        • Chuah Eric
        • Clarke Amanda
        • Dhalla Noreen
        • et al.
        • TCGA
        Comprehensive and integrative genomic characterization of hepatocellular carcinoma.
        Cell. 2017; 169: 1327-1341. e1323
        • Schmid I.
        • von Schweinitz D.
        Pediatric hepatocellular carcinoma: challenges and solutions.
        J Hepatocellular Carcinoma. 2017; 4: 15-21
        • Lim I.I.P.
        • Bondoc A.J.
        • Geller J.I.
        • Tiao G.M.
        Hepatoblastoma—the evolution of biology, surgery, and transplantation.
        Children. 2019; 6: 1
        • Gröbner S.N.
        • Worst B.C.
        • Weischenfeldt J.
        • Buchhalter I.
        • Kleinheinz K.
        • Rudneva V.A.
        • et al.
        The landscape of genomic alterations across childhood cancers.
        Nature. 2018; 555: 321-327