Transcriptional Profiling of Pure Fibrolamellar Hepatocellular Carcinoma Reveals an Endocrine Signature Gabriel G. Malouf,1* Sylvie Job,2* Valerie Paradis,3 Monique Fabre,4 Laurence Brugie`res,5 Pierre Saintigny,6 Laure Vescovo,2 Jacques Belghiti,7 Sophie Branchereau,8 Sandrine Faivre,1 Aurelien de Reynie`s,2** and Eric Raymond1** Fibrolamellar hepatocellular carcinoma (FLC) is a rare subtype of liver cancer occurring mostly in children and young adults. We have shown that FLC comprises two separate entities: pure (p-FLC) and mixed-FLC (m-FLC), differing in clinical presentation and course. We show that p-FLCs have a distinct gene expression signature different from that of m-FLCs, which have a signature similar to that of classical hepatocellular carcinomas. We found p-FLC profiles to be unique among 263 profiles related to diverse tumoral and nontumoral liver samples. We identified two distinct molecular subgroups of p-FLCs with different outcomes. Pathway analysis of p-FLCs revealed ERBB2 overexpression and an up-regulation of glycolysis, possibly leading to compensatory mitochondrial hyperplasia and oncocytic differentiation. Four of the sixteen genes most significantly overexpressed in p-FLCs were neuroendocrine genes: prohormone convertase 1 (PCSK1); neurotensin; delta/notch-like EGF repeat containing; and calcitonin. PCSK1 overexpression was validated by immunohistochemistry, yielding specific, diffuse staining of the protein throughout the cytoplasm, possibly corresponding to a functional form of this convertase. Conclusion: p-FLCs have a unique transcriptomic signature characterized by the strong expression of specific neuroendocrine genes, suggesting that these tumors may have a cellular origin different from that of HCC. Our data have implications for the use of genomic profiling for diagnosis and selection of targeted therapies in patients with pFLC. (HEPATOLOGY 2014;59:2228-2237)
F
ibrolamellar hepatocellular carcinoma (FLC) has been described as a rare variant of hepatocellular carcinoma (HCC).1-5 FLC is typically diagnosed as a symptomatic bulky tumor mass in the liver of a young adult with no history of cirrhosis or hepatic dysfunction.2,6 The pathological features of FLC were
described as nests and cords of oncocytic cells with prominent nuclei and eosinophilic granular cytoplasms surrounded by a fibrous lamellar stroma.3 We showed that tumors with all these pathological characteristics, known as pure FLC (p-FLC), have clinical presentations and behaviors different from those of mixed FLC
Abbreviations: ADE, adenoma; AJCC, American Joint Committee on Cancer; ANOVA, analysis of variance; CALCA, calcitonin; CIR, cirrhosis; DNER, delta/ notch-like EGF repeat containing; DYS, dysplasia; ECM, extracellular matrix ; EMT, epithelial to mesenchymal transition; FLC, fibrolamellar hepatocellular carcinoma; GO, Gene Ontology; HB, hepatoblastoma; HCC, hepatocellular carcinoma; IHC, immunohistochemical; KEGG, Kyoto Encyclopedia of Genes and Genomes; LDH, lactate dehydrogenase; m-FLC, mixed FLC; mRNA, messenger RNA; nc-HCC, noncirrhotic liver; NL, normal liver; NTS, neurotensin; PC, principal component; PCA, principal component analysis; PCSK1, prohormone convertase 1; p-FLC, pure FLC; qRT-PCR, quantitative reverse transcription polymerase chain reaction; TGF-b, transforming growth factor beta. From the 1Department of Medical Oncology, Beaujon University Hospital and INSERM U728, Clichy, France; 2Programme Cartes d’Identit e des Tumeurs, Ligue Nationale Contre le Cancer, Paris, France; 3Department of Pathology, Beaujon University Hospital, Clichy, France; 4Department of Pathology, Institut Gustave Roussy, Villejuif, France; 5Department of Pediatrics, Institut Gustave Roussy, Villejuif, France; 6Department of Thoracic Oncology, University of Texas, MD Anderson Cancer Center, Houston, TX; 7Department of Hepato-biliary Surgery, Beaujon University Hospital, Clichy, France; and 8Department of Pediatric Surgery, Bic^e tre University Hospital, University Paris XII, Paris, France. Received May 29, 2013; accepted January 14, 2014. This work was supported by Cartes d’Identit e des Tumeurs (CIT) program of the Ligue Nationale Contre le Cancer, Foundation Nelia et Amadeo Barletta (FNAB) and the Association pour la Recherche et l’Enseignement en Canc e rologie (AAREC). *Cofirst authors (contributed equally to this work). **Cosenior authors (contributed equally to this work). 2228
HEPATOLOGY, Vol. 59, No. 6, 2014
(m-FLC).6 In particular, p-FLCs typically occur in young patients (35 20 ng/mL Yes No ND Vascular invasion Yes No ND Satelite nodules Yes No ND Adjacent normal Available Biological replicate Yes No AJCC stage I-II III-IV ND
nc-HCC (n 5 7)
m-FLC (n 5 5)
p-FLC (n 5 14)
Fisher’s Exact P Value
6 1 0
4 1 0
3 10 1
0.02
7 0
5 0
3 11
0.0001
3 4 0
3 2 0
1 13 0
0.036
2 5 0
2 3 0
3 10 1
NS
6 1 0
1 4 0
4 9 1
0.05
5
2
3
NS
0 7
0 5
3 11
NS
5 2 0
2 3 0
9 4 1
NS
Abbreviation: ND, not defined; AFP, alpha-fetoprotein; NS, not significant.
p-FLCs overexpressed several neuroendocrine genes, including prohormone convertase 1 (PSCK1).17-19
Patients and Methods Patient Characteristics and Tumor Selection. A total of 39 liver samples were included in this study: 29 primary tumors, obtained from complete surgical resections, snap-frozen immediately after surgery and stored in liquid nitrogen, and 10 tumor-adjacent normal liver (NL) samples obtained from 10 of these patients. The 29 tumor samples included 17 p-FLCs (corresponding to 14 patients), 5 m-FLCs, and 7 HCCs arising in noncirrhotic liver (nc-HCC). Clinical and pathological characteristics of the corresponding patients are summarized in Table 1.
Address reprint requests to: Eric Raymond, M.D., Ph.D., Department of Medical Oncology and INSERM-U728, H^ opital Beaujon, APHP–Paris 7 Diderot, 100 Boulevard du Gen e ral Leclerc, 92110 Clichy, France. E-mail:
[email protected]; fax: 133-1-40 87 54 87. C 2014 by the American Association for the Study of Liver Diseases. Copyright V View this article online at wileyonlinelibrary.com. DOI 10.1002/hep.27018 Potential conflict of interest: Nothing to report.
2230
MALOUF, JOB, ET AL.
Gene Expression Profiles. HG-U133-plus-2.0 Affymetrix array data were obtained for our 39 samples (E-MTAB-1503). Gene expression of PCSK1, delta/notch-like EGF repeat containing (DNER), calcitonin (CALCA), and NTS was measured in duplicate by quantitative reverse transcription polymerase chain reaction (qRT-PCR) reactions for all of these 39 samples (Supporting Table 1). Four public transcriptomic data sets (GSE6764, GSE14520, E-TABM-36, and E-MEXP-1851) corresponding to 220 liver tissues were collected20-24; they included 141 HCCs, 25 hepatoblastomas (HBs), three hepatocellular adenomas (ADEs), 17 liver dysplasia (DYS) lesions, 13 cirrhosis (CIR) lesions, and 21 nontumoral livers. Finally, two p-FLC samples and two tumor adjacent normal liver samples were provided by Kannangai et al.22 Transcriptomic data for 96 lung cancer cell lines were also collected from ArrayExpress (E-MTAB-37). Gene Expression Analyses. Affymetrix expression profiles for all 263 liver samples were normalized intra-25 and interseries.26 Consensus clustering analysis was performed as described elsewhere.27 We used moderated t tests28 to identify differentially expressed genes. We used four criteria to identify specific markers of p-FLC: moderated t test (P < 1e-3), foldchange (>8), area under the curve (>0.90), and low basal levels (log2 intensity 0.985 within duplicates; Supporting Fig. 1B). mFLC and nc-HCC samples clustered with other cancerous liver samples (Fig. 1). The restriction of PCA analysis to our series confirmed the specific gene expression profile of p-FLCs (Supporting Fig. 3). Three of the fourteen cases were found to have gene expression profiles similar to those of m-FLCs and nc-HCCs. These three cases corresponded to the oldest patients, with a median age of 58.3 versus 20.6 years for the other p-FLC patients (analysis of variance [ANOVA]: P < 5e-4). These three outliers were excluded from the signaling pathway analysis. Pathway Deregulation in p-FLCs. We carried out pathway analysis, comparing the expression profiles of p-FLCs with those of other HCCs (m-FLCs and ncHCCs) and, separately, with those of NL, to identify pathways differentially regulated in p-FLCs. Pathways displaying significant deregulation in p-FLCs included the glycolysis pathway, the insulin pathway, the hypoxia pathway, and the Erbb pathway (Table 2). The glycolysis pathway was found to be up-regulated in p-FLC (Fig. 2A), because at least one regulator of each step of glycolysis was overexpressed, with the exception of the step regulated by phosphoglycerate mutase 2. Key glycolytic genes, such as phosphofructokinase, muscle (PFKM), phosphofructokinase, platelet (PFKP), lactate dehydrogenase (LDH)B and hexokinase 2, were found to be overexpressed in p-FLCs (Fig. 2A). Some of these genes are known not only to promote glycolysis, but also to down-regulate the Krebs cycle and mitochondrial respiration. This is the case for LDHB (ANOVA, P 5 1.95e-12), LDHA (ANOVA, P 5 1.55e-03), and pyruvate dehydrogenase kinase 1 (ANOVA, P 5 3.05e-02).29-31 These results are consistent with previous reports on oncocytic tumors revealing that functional and structural impairments of respiratory complex may force cells to rely on glycolysis for energy production, as the result of a dysfunction of oxidative phosphorylation.32 In addition to up-regulation of
HEPATOLOGY, Vol. 59, No. 6, 2014
MALOUF, JOB, ET AL.
2231
Fig. 1. PCA of the 263 gene expression profiles related to 39 sample profiles from our series and 224 sample profiles extracted from five external data sets (first and second components). Annotation is based on histological samples annotation.
glycolysis, we observed the concomitant deregulation of hypoxia and insulin-signaling pathways (Supporting Fig. 4), although we were unable to draw firm conclusions about whether these pathways were activated or inhibited. Collectively, these data suggest that a Warburg effect occurs in p-FLCs. Another finding of interest is the significance of the Erbb pathway, driven by overexpression of the ERBB2
Table 2. List of Pathways Differentially Altered Between p-FLC and m-FLC, nc-HCC, and NL
Pathway Names
Glycolysis GO:0006096: glycolysis KEGG 00010: glycolysis/gluconeogenesis Insulin KEGG 04910: insulin-signaling pathway GO:0050796: regulation of insulin secretion Hypoxia GO:0001666: response to hypoxia Erbb KEGG 04012: ErbB-signaling pathway Central nervous system GO:0048678: response to axon injury
p-FLC vs. m-FLC/ nc-HCC Tukey’s P Value
p-FLC vs. NL Tukey’s P Value
1.67E-07 6.55E-07
6.55E-10 2.94E-11
4.64E-07 1.07E-03
2.38E-03 4.55E-02
6.39E-04
5.91E-04
2.58E-05
9.45E-05
4.35E-04
4.25E-03
oncogene (ANOVA, P 5 1.83e-08; Fig. 2B). Activation of ERBB2 may occur through two different mechanisms: (1) overexpression or (2) dimerization with another ERBB receptor through ligand-mediated stimulation.33 Herein, the expression level of two ERBB2 ligands (EREG and AREG) was concomitantly higher in p-FLCs (Fig. 2B). ERBB2 staining by IHC yielded a strong diffuse cytoplasmic staining associated with partial membranous staining in the majority of p-FLCs cases (Fig. 2C). As expected, expression profiles of the three p-FLC outliers were similar to expression profiles of mFLC and nc-HCC samples (Supporting Fig. 5). p-FLCs Overexpress Neuroendocrine Genes. Finally, we investigated genes strongly expressed in pFLC, as compared to other types of HCC (m-FLC and nc-HCC) and NL. We identified 16 genes, (Supporting Table 3) of which four are neuroendocrine genes: PCSK1, DNER, NTS, and CALCA. We confirmed the overexpression of these genes by RT-PCR (Fig. 3A). Whole exome sequencing of eight pFLCs did not reveal any mutation in these four genes (data not shown), consistent with the COSMIC database showing a mutation rate less than 0.35% for any of these genes in tumor samples. PCSK1 has been reported to be overexpressed specifically in various subtypes of neuroendocrine tumors.17,18 We investigated the expression
Fig. 2. (A) Heat map of the expression level of genes involved in glycolysis. The successive steps related to glycolysis are indicated by successive black and gray boxes. For genes differentially up-regulated in p-FLC, the ANOVA P values are indicated next to the gene. (B) Box plots of the gene expression levels of ERBB2, AREG, and EREG in p-FLC, nc-HCC/m-FLC, and NL. (C) Diffuse cytoplasmic and partial membranous immunostaining of ERBB2 in p-FLC, NL, conventional HCC (HCC) and m-FLC.
HEPATOLOGY, Vol. 59, No. 6, 2014
MALOUF, JOB, ET AL.
Table 3. List of Pathways Differentially Altered Between the Two Pure Fibrolamellarp-FLC Subgroups (pFLC-A and pFLC-B) Pathway Name
Collagen and ECM GO:0005584: collagen type I GO:0031012 : ECM TGF-b GO:0007179: TGF-b receptor-signaling pathway KEGG 04350: TGF-b-signaling pathway Vascular regulation GO:0048407: platelet-derived growth factor binding GO:0001944 : vasculature development Cellular adhesion GO:0007155: cell adhesion GO:0031589: cell-substrate adhesion EMT GO:0010717: regulation of EMT
Tukey’s P Value
