Purpose:

Hepatocellular carcinoma (HCC) is characterized by high intertumor heterogeneity of genetic drivers. Two multitarget tyrosine kinase inhibitors (TKI), lenvatinib and sorafenib, are used as standard-of-care chemotherapeutics in patients with advanced HCC, but a stratification strategy has not been established because of a lack of efficacious biomarkers. Therefore, we sought biomarkers that indicate lenvatinib-susceptible HCC.

Experimental Design:

We performed genetic screening of HCC driver genes involved in TKI susceptibility using a novel HCC mouse model in which tumor diversity of genetic drivers was recapitulated. A biomarker candidate was evaluated in human HCC cell lines. Secreted proteins from HCC cells were then screened using mass spectrometry. Serum and tumor levels of the biomarker candidates were analyzed for their association and prediction of overall survival in patients with HCC.

Results:

We found that lenvatinib selectively eliminated FGF19-expressing tumors, whereas sorafenib eliminated MET- and NRAS-expressing tumors. FGF19 levels and lenvatinib susceptibility were correlated in HCC cell lines, and FGF19 inhibition eliminated lenvatinib susceptibility. Lenvatinib-resistant HCC cell lines, generated by long-term exposure to lenvatinib, showed FGF19 downregulation but were resensitized to lenvatinib by FGF19 reexpression. Thus, FGF19 is a tumor biomarker of lenvatinib-susceptible HCC. Proteome and secretome analyses identified ST6GAL1 as a tumor-derived secreted protein positively regulated by FGF19 in HCC cells. Serum ST6GAL1 levels were positively correlated with tumor FGF19 expression in patients with surgically resected HCC. Among patients with serum ST6GAL1-high HCC who underwent TKI therapy, lenvatinib therapy showed significantly better survival than sorafenib.

Conclusions:

Serum ST6GAL may be a novel biomarker that identifies lenvatinib-susceptible FGF19-driven HCC.

Translational Relevance

Treatment options for patients with advanced-stage hepatocellular carcinoma (HCC) are limited. Two multitarget tyrosine kinase inhibitors, lenvatinib and sorafenib, are used as systemic standard-of-care chemotherapeutics, but a stratification strategy has not been established because of a lack of predictive biomarkers for the efficacy of these agents. In this study, we established a novel HCC mouse model that recapitulates the diversity of genetic drivers of tumors that are observed in human patients with HCC. Using this model, we found that HCC-expressing FGF19 is susceptible to lenvatinib and that ST6GAL1, which is a tumor-derived secreted protein downstream of FGF19, may be useful as a serum biomarker for identification of FGF19-driven lenvatinib susceptibility in patients with HCC. Our findings not only clarify an HCC genetic subtype that is susceptible to lenvatinib, but they also provide clinical evidence of the possible utility of a novel serum biomarker for optimal drug selection.

Hepatocellular carcinoma (HCC) is the fourth leading cause of cancer-related deaths worldwide (1). Patients with early-stage HCC can undergo curative treatments such as local ablation and surgical resection. However, HCC recurs in more than half of these patients within 2 years and eventually progresses into the incurable advanced stage (2). In addition, a significant proportion of patients with HCC are initially diagnosed at an advanced stage, partly due to the poor implementation rates of surveillance programs (3). Treatment options for patients with advanced stage HCC are very limited. Sorafenib, a tyrosine kinase inhibitor (TKI), was the only effective systemic first-line chemotherapy for HCC for more than a decade (4). Recently, a phase III randomized trial (REFLECT) comparing the efficacy of lenvatinib, another TKI, with that of sorafenib in patients with unresectable HCC was conducted and revealed the noninferiority of lenvatinib compared with sorafenib with regard to overall survival (OS) in these patients. On the basis of this trial, lenvatinib has been used as another first-line systemic chemotherapy option for advanced HCC since 2018. Most recently, atezolizumab and bevacizumab combination therapy demonstrated superiority over sorafenib monotherapy with regard to OS and progression-free survival in a phase III randomized trial (IMbrave 150 trial; ref. 5) and will become the new first-line therapy for advanced HCC. However, lenvatinib and sorafenib will still be important alternative first-line or second-line therapies. In terms of their molecular mechanism, lenvatinib and sorafenib block VEGFRs and platelet-derived growth factor receptors (PDGFR). In addition, lenvatinib uniquely targets FGFRs, whereas sorafenib targets Raf kinase (4, 6). Although the two TKIs exhibit different mechanisms of action, neither a clinical recommendation nor criteria for optimal TKI selection exist due to a lack of biomarkers for prediction of the efficacy of these agents.

Recent large-scale cancer genome sequencing projects have significantly advanced our understanding of cancer genetic drivers (7). Regarding HCC, comprehensive sequencing of over 500 patients has revealed that recurrent somatic mutations and copy-number alterations are present in a variety of cancer genes at minor frequencies, which reflects the extreme intertumor heterogeneity of genetic drivers (7, 8). In contrast to classic cytotoxic chemotherapeutic drugs, TKIs target specific molecules and cancer signaling pathways. Therefore, the diversity of genetic drivers may have a great impact on the antitumor effects of TKIs, and precise assessment of this diversity may help to identify an HCC genetic subtype that is susceptible to each available TKI, leading to the identification of an efficacious biomarker. Drug assessment in vivo is also important, because TKIs not only kill tumor cells but also block angiogenesis and influence a variety of cell types in the tumor microenvironment. However, such research has been hindered by a lack of appropriate HCC animal models for evaluation of the impacts of a variety of cancer genes on drug susceptibility. Furthermore, to be clinically applicable, biomarkers must not only be tumor based but must also be able to be noninvasively detected in the serum of patients with HCC susceptible to each TKI therapy, because advanced HCC cases are often diagnosed through dynamic CT or MRI (9, 10) and treated without tumor biopsy.

In this study, to simultaneously assess the individual effects of a variety of genetic drivers on TKI susceptibility in HCC in vivo, we created a novel HCC mouse model recapitulating the diversity of tumor genetic drivers via transposon-based intrahepatic delivery of a pooled oncogene cDNA library. A genetic screen of this model revealed that tumors expressing FGF19 were susceptible to lenvatinib in vivo. Subsequently, we showed that FGF19-expressing human hepatoma cells were susceptible to lenvatinib and acquired resistance to this agent by FGF19 downregulation in vitro. Next, to discover biomarkers of FGF19-driven HCC, we comprehensively evaluated tumor-secreting proteins by cellular proteomics and secretome analyses. We identified a new regulatory link between FGF19 and the secreted protein ST6GAL1 in HCC cells and showed that a high serum ST6GAL1 level was a reliable diagnostic marker of FGF19-driven highly malignant HCC. Finally, we obtained clinical evidence that ST6GAL1 may be a useful serum biomarker for selection of patients with HCC for whom lenvatinib will yield better survival benefits than sorafenib.

Generation of a pooled oncogene cDNA library

We aimed to generate an oncogene library covering as many signaling pathways involved in HCC pathogenesis (8, 11, 12) as possible. We first selected eight known oncogenes that are either recurrently mutated or copy number–amplified in a recent large-scale HCC genome sequencing project from The Cancer Genome Atlas (TCGA) Research Network (7); these genes included CTNNB1 (Wnt/β-catenin pathway), NFE2L2 (Oxidative stress pathway), NRAS (Ras/Raf/ERK pathway), FGF19 (RTK pathway), MET (RTK pathway), CCND1 (Cell cycle pathway), MCL1 (Apoptosis pathway), and MYC (Telomere maintenance pathway). We also selected two known oncogenes, AKT (PI3K pathway) and YAP (Hippo pathway), the dysregulation of which is known to be involved in HCC pathogenesis (13–15). cDNA sequences for these 10 oncogenes and GFP were individually cloned into the piggyBac (PB) transposon vector (pPB-CAG-EBNXN; provided by Junji Takeda) along with corresponding unique barcodes and luciferase cDNA. We generated a pooled oncogene cDNA library by mixing all these plasmids in equal amounts. The detailed cloning methods and barcode sequences are provided in the Supplementary Materials and Methods and Supplementary Table S1, respectively.

Animal experiments

C57BL/6J male mice were purchased from Charles River Laboratories Japan and treated with humane care under approval from the Animal Care and Use Committee of Osaka University Medical School (Suita, Japan). To deliver oncogene cDNA into hepatocytes, we used the hydrodynamic tail vein injection (HTVi) method, which has been reported to result in transgene expression in 5%–40% of hepatocytes one day after injection (16). We also used PB transposon vectors containing oncogene cDNA in combination with transposase plasmids, which allowed the cDNA in the plasmid to become integrated into the genome of hepatocytes; this in turn led to its constitutive expression. Forty-four micrograms of pooled cDNA library and 8.8 μg of transposon plasmid were dissolved in 0.1 mL/g body weight of saline and injected for 5 seconds into male C57BL/6J mice at 8 weeks of age via the tail vein. Two days after injection, the luciferase signal was examined with an IVIS system and d-luciferin (Summit Pharmaceuticals International Corporation) to confirm intrahepatic delivery of the pooled library. Two weeks after injection, liver tumor formation was identified with an IVIS Lumina (Summit Pharmaceuticals International Corporation) and by CT scan (microCT2, Rigaku Corporation) with contrast agent (Iopamiron, Bayer). Mice with liver luciferase activity were randomly assigned to a lenvatinib treatment group, a sorafenib treatment group, and corresponding vehicle groups. They were treated by oral gavage with 30 mg/kg lenvatinib (Eisai), 60 mg/kg sorafenib (Bayer) or vehicle once a day. The dose of these drugs was determined on the basis of previous reports (17, 18). Their condition was monitored every day, and the animals were euthanized when they reached humane endpoints defined according to the Animal Care and Use Committee guidelines of Osaka University Medical School (Suita, Japan) and the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines (19). Liver tumors that developed in mice were isolated, and the long diameter of tumors over 5 mm were used for the analysis. The tumors were examined by histologic analysis, molecular analysis, and barcode sequencing analysis. The animals were maintained with water and standard mouse chow under specific pathogen-free conditions.

Sequence analysis

Total DNA was extracted from the mouse liver tumors with a DNeasy Kit (Qiagen). The barcode sequences integrated into the genomic DNA were amplified by PCR using Illumina-adapted indexed primers and Platinum PCR SuperMix High Fidelity (Thermo Fisher Scientific). The PCR products were purified with a PCR purification kit (Qiagen) and AMPure beads (Beckman Coulter). Each PCR amplicon was quantified using a Qubit Fluorometer (Thermo Fisher Scientific) and an Agilent 2100 Bioanalyzer (Agilent), and the products were pooled. Then, 75-bp single-end sequencing was performed on a flow cell of a HiSeq 2500 platform (Illumina). The sequencing reads were converted into FASTQ format and demultiplexed using bcl2fastq2 v2.17.1.14 (Illumina). Reads with unique barcode sequences were assigned to different genes and counted with Cutadapt 1.9.2 (cutadapt, RRID:SCR_011841). The relative cDNA abundance of each oncogene in each tumor was calculated by dividing the read counts of each unique barcode by the total read counts in a single tumor. Further information about the primers and PCR protocols is given in Supplementary Tables S2 and S3.

Human serum and histologic analyses

Serum samples were obtained from 76 patients with HCC who underwent curative hepatectomy at Osaka University (Suita, Japan) and from 96 patients with advanced HCC treated with TKI therapy at Osaka University (Suita, Japan) and other institutions participating in the Osaka Liver Forum. HCC tissues were obtained from 62 of the 76 patients with surgically resected HCC. All patients provided written informed consent, and the study design was consistent with the principles of the Declaration of Helsinki. The protocol of the study using patient serum and resected tissues was approved by the Institutional Review Board (IRB) Committees at Osaka University Hospital (Suita, Japan, IRB No. 19438). Serum FGF19 and ST6GAL1 levels were examined by ELISA using human FGF19 (DF1900, R&D Systems) and human ST6GAL1 (ab243669, Abcam) ELISA kits, respectively, following the manufacturers' protocols. Tumor tissues were fixed in 10% neutral-buffered formaldehyde solution, embedded in paraffin, and cut into thin slices. The liver sections were stained with hematoxylin and eosin (H&E) and a primary FGF19 antibody (Atlas Antibodies, catalog no. HPA036082, RRID:AB_10669665, 1:300, Sigma-Aldrich). Heat-mediated antigen retrieval was performed using citrate retrieval buffer (Agilent) and a Decloaking Chamber NxGen (Biocare Medical). The FGF19 staining intensity and positive area in liver tumor sections were quantified using Hybrid Cell Count software (Keyence). Tumors with an average FGF19-stained area of more than 10% in four high-power fields were categorized into the FGF19-high group. The rest of the tumors were categorized into the FGF19-low group.

Statistical analysis

Statistical analysis was performed with Student t tests or Mann–Whitney U tests to assess differences between unpaired groups with parametric or nonparametric distributions, respectively. One-way ANOVA followed by the Tukey–Kramer post hoc test or the Kruskal–Wallis test was performed for parametric or nonparametric multiple comparisons, respectively. χ2 tests or Fisher exact tests were used to analyze categorical data. Correlations were assessed using the Pearson product-moment correlation coefficient. In the resected cohort, the end point of relapse-free survival (RFS) was defined as the time from the day of hepatectomy until the first objective observation of disease progression or death from any cause. In the TKI cohort, the end point of OS was defined as the time from the day of treatment initiation until death from any cause. The Kaplan–Meier method and log-rank test were used to analyze differences in OS and RFS. Univariate and multivariate Cox proportional hazards regression models were used to analyze factors associated with improved RFS and OS. We used Cox hazards regression models in the TKI cohort to study the interaction of treatment with pretreatment status. To assess the diagnostic performance of serum biomarkers, ROC curve analysis was performed, and the AUC was used to evaluate the predictive power. The cut-off value was determined with Youden J statistic (20). Otherwise, the statistical analyses used are indicated in the figure legends. A P value <0.05 was considered to indicate statistical significance. Prism ver.8.4.2 for Windows (GraphPad Prism, RRID:SCR_002798) and JMP 13 (SAS Institute Inc. RRID:SCR_014242) were used for the analyses.

All other information regarding the materials and methods is provided in the Supplementary Data. The plasmid information and the real-time PCR primer information are shown in Supplementary Table S4 and S5, respectively.

FGF19-driven tumors are susceptible to lenvatinib in vivo

To rapidly screen the impacts of individual HCC genetic drivers on TKI susceptibility in vivo, we first sought to establish an HCC mouse model mimicking the diversity of tumor genetic drivers. On the basis of the findings of a previous large-scale human HCC genome sequencing project and other studies (7, 13–15), we selected 10 HCC oncogenes, AKT, CCND1, CTNNB1, FGF19, MCL1, MET, MYC, NFE2L2, NRAS, and YAP, which are involved in the major signaling pathways that are dysregulated in human HCC (8, 11, 12), and aimed to randomly and constitutively express these oncogenes in mouse livers. We cloned the individual cDNA sequences into PB transposon vectors along with unique barcodes and luciferase (Fig. 1A) and confirmed that the constructs were highly expressed in vitro when individually transfected (Supplementary Fig. S1A). We then generated a pooled cDNA library by mixing all these plasmids together and delivering them into the livers of C57BL/6 mice together with the PB transposase-expressing plasmid by hydrodynamic injection via the tail vein. This induced random genomic integration of these oncogene cDNA sequences and their corresponding barcodes in hepatocytes. Within 14 days after injection, the mice showed strong intrahepatic luciferase signals and had developed multiple macroscopic liver tumors (Fig. 1BD). These tumors showed trabecular patterns with polygonal tumor cells on histological analysis (Fig. 1E) and produced high levels of Afp and Gpc3 (Fig. 1F), indicating that they were HCC tumors. Importantly, we assessed the relative abundance of each oncogene in each tumor in this model by sequencing the molecular barcodes within the tumor genome. Sequencing analysis revealed that the genetic drivers of these tumors were highly heterogeneous (Supplementary Fig. S1B).

Figure 1.

FGF19-driven tumors are susceptible to lenvatinib in vivo. A, Schematic of the vector construct. The barcode sequence, cytomegalovirus (CMV) promoter, oncogene cDNA, and luciferase cDNA were tandemly cloned into the piggyBac (PB) vector within inverted terminal repeat (ITR) regions on both sides. B–F, C57BL/6J mice received hydrodynamic tail vein injection (HTVi) of a PB-based pooled oncogene cDNA library or vehicle. Intrahepatic luciferase activity and liver images were examined by IVIS and micro-computed tomography (micro CT) 14 days after injection. Representative images of IVIS (B) and micro CT scans (C, top) and macro images of liver tumors (C, bottom) are shown. The number of liver tumors per mouse was detected by micro CT at 14 days after injection (n = 5–11; *, P < 0.05; D). H&E staining of a tumor lesion in the liver of a genetically heterogeneous liver tumor model mouse (the scale bar indicates 100 μm) (E). qPCR analysis of Afp and Gpc3 mRNA levels in tumor (T) and nontumor (NT) tissues (n = 3–16 each; *, P < 0.05; AU, arbitrary unit; F). G–I, At 14 days after injection of the pooled oncogene cDNA library, mice were treated with either vehicle or lenvatinib (30 mg/kg/day) until humane endpoints were reached. The OS of liver tumor–bearing mice since initiation of treatment with vehicle or lenvatinib (n = 14–18 each) was determined (G). H&E staining of liver tumors was performed after treatment with vehicle or lenvatinib (the scale bar indicates 100 μm; H). Upon treatment with vehicle or lenvatinib, genomic DNA (gDNA) was extracted from liver tumors, and barcode sequences integrated into the gDNA were quantified by next-generation sequencing. The relative cDNA abundance of each oncogene in a single tumor was calculated by dividing the read count of each unique barcode in the single tumor by the total read counts in that tumor (n = 26–51 each; *, P < 0.05; I).

Figure 1.

FGF19-driven tumors are susceptible to lenvatinib in vivo. A, Schematic of the vector construct. The barcode sequence, cytomegalovirus (CMV) promoter, oncogene cDNA, and luciferase cDNA were tandemly cloned into the piggyBac (PB) vector within inverted terminal repeat (ITR) regions on both sides. B–F, C57BL/6J mice received hydrodynamic tail vein injection (HTVi) of a PB-based pooled oncogene cDNA library or vehicle. Intrahepatic luciferase activity and liver images were examined by IVIS and micro-computed tomography (micro CT) 14 days after injection. Representative images of IVIS (B) and micro CT scans (C, top) and macro images of liver tumors (C, bottom) are shown. The number of liver tumors per mouse was detected by micro CT at 14 days after injection (n = 5–11; *, P < 0.05; D). H&E staining of a tumor lesion in the liver of a genetically heterogeneous liver tumor model mouse (the scale bar indicates 100 μm) (E). qPCR analysis of Afp and Gpc3 mRNA levels in tumor (T) and nontumor (NT) tissues (n = 3–16 each; *, P < 0.05; AU, arbitrary unit; F). G–I, At 14 days after injection of the pooled oncogene cDNA library, mice were treated with either vehicle or lenvatinib (30 mg/kg/day) until humane endpoints were reached. The OS of liver tumor–bearing mice since initiation of treatment with vehicle or lenvatinib (n = 14–18 each) was determined (G). H&E staining of liver tumors was performed after treatment with vehicle or lenvatinib (the scale bar indicates 100 μm; H). Upon treatment with vehicle or lenvatinib, genomic DNA (gDNA) was extracted from liver tumors, and barcode sequences integrated into the gDNA were quantified by next-generation sequencing. The relative cDNA abundance of each oncogene in a single tumor was calculated by dividing the read count of each unique barcode in the single tumor by the total read counts in that tumor (n = 26–51 each; *, P < 0.05; I).

Close modal

Next, to search for HCC oncogenes involved in lenvatinib susceptibility, we treated tumor-bearing mice with lenvatinib or vehicle until humane endpoints were reached and compared oncogene abundance levels. The lenvatinib-treated group showed a significantly longer OS than the vehicle-treated group (Fig. 1G). Histologically, lenvatinib-treated tumors became necrotic (Fig. 1H). Sequencing analysis of tumor molecular barcodes showed that FGF19 cDNA was significantly less abundant in lenvatinib-treated tumors than in vehicle-treated tumors (Fig. 1I), whereas the other oncogenes exhibited no differences in cDNA abundance between groups, suggesting that lenvatinib selectively depleted FGF19-expressing tumors. We also used this model to search for genes involved in sorafenib susceptibility and found that the cDNA abundance of MET and NRAS was significantly reduced in sorafenib-treated tumors (Supplementary Fig. S1C). Because sorafenib is a well-known Raf kinase inhibitor (21), this model is excellent for screening of oncogenes expressed in drug-sensitive tumors. Collectively, our in vivo screening results suggest that FGF19-driven HCC may be susceptible to lenvatinib.

FGF19-expressing human HCC cells are susceptible to lenvatinib and acquire resistance upon FGF19 downregulation in vitro

We further studied the involvement of tumor FGF19 signaling in lenvatinib efficacy in vitro. Among nine human HCC cell lines, the HuH7 and Hep3B cell lines showed higher FGF19 mRNA and protein levels (Fig. 2A and B) and markedly lower half-maximal inhibitory concentration (IC50) values for lenvatinib than the others (Fig. 2C). Knockdown of FGF19 in both cell lines eliminated lenvatinib susceptibility (Fig. 2D). Next, lenvatinib-resistant HCC cell lines were established via exposure of Huh7 and Hep3B cells to 5 μmol/L lenvatinib for 3 months. We confirmed that the resistant lines (Huh7-LenR and Hep3B-LenR) became refractory to a high dose of lenvatinib (Fig. 2E) and found that FGF19 levels were significantly lower in these cells than in parental cells (Fig. 2F). We then lentivirally transduced a doxycycline-inducible FGF19 expression plasmid into the Hep3B and Hep3B-LenR cell lines and selected the transduced cells with puromycin. Doxycycline administration upregulated FGF19 expression in both cell lines (Fig. 2G) and restored the lenvatinib susceptibility of Hep3B-LenR cells (Fig. 2H). These data support the hypothesis that FGF19-expressing HCC cells are susceptible to lenvatinib.

Figure 2.

FGF19-expressing human HCC cells are susceptible to lenvatinib and acquire resistance upon FGF19 downregulation in vitro. A, qPCR analysis of FGF19 mRNA levels in nine hepatoma cell lines (n = 3 each; AU, arbitrary units). B, Western blot (WB) analysis of FGF19 and ACTB protein levels in nine hepatoma cell lines. C, The IC50 values of lenvatinib in nine hepatoma cell lines were examined by WST-8 assay (n = 4 each). D, Two days after transfection with negative control (NC) or FGF19 siRNA, Hep3B and HuH7 cells were treated with lenvatinib for 72 hours, and cell viability was assessed by WST-8 assay (n = 6 each; *, P < 0.05 vs. NC siRNA with DMSO; AU, arbitrary units). E, Hep3B and HuH7 cells were treated with 5 μmol/L lenvatinib or DMSO for 3 months, and lenvatinib-resistant cells were generated. The parental and resistant cells were treated with lenvatinib for 72 hours, and cell viability was assessed by WST-8 assay (n = 6 each; *, P < 0.05 vs. 0 μmol/L; AU, arbitrary units). F, qPCR analysis of FGF19 mRNA levels in parental cells and lenvatinib-resistant Hep3B and HuH7 cells (n = 3 each; *, P < 0.05; AU, arbitrary units). G, Parental and lenvatinib-resistant Hep3B cells transduced with a doxycycline-inducible FGF19-expressing vector were treated with doxycycline at the indicated concentrations for 24 hours. The protein levels of FGF19 and ACTB were assessed by WB. H, Parental and lenvatinib-resistant Hep3B cells transduced with a doxycycline-inducible FGF19-expressing vector were treated with 100 μmol/L doxycycline and lenvatinib at the indicated concentrations for 72 hours. Cell viability was examined by WST-8 assay (n = 6 each; *, P < 0.05 vs. 0 μmol/L; AU, arbitrary units).

Figure 2.

FGF19-expressing human HCC cells are susceptible to lenvatinib and acquire resistance upon FGF19 downregulation in vitro. A, qPCR analysis of FGF19 mRNA levels in nine hepatoma cell lines (n = 3 each; AU, arbitrary units). B, Western blot (WB) analysis of FGF19 and ACTB protein levels in nine hepatoma cell lines. C, The IC50 values of lenvatinib in nine hepatoma cell lines were examined by WST-8 assay (n = 4 each). D, Two days after transfection with negative control (NC) or FGF19 siRNA, Hep3B and HuH7 cells were treated with lenvatinib for 72 hours, and cell viability was assessed by WST-8 assay (n = 6 each; *, P < 0.05 vs. NC siRNA with DMSO; AU, arbitrary units). E, Hep3B and HuH7 cells were treated with 5 μmol/L lenvatinib or DMSO for 3 months, and lenvatinib-resistant cells were generated. The parental and resistant cells were treated with lenvatinib for 72 hours, and cell viability was assessed by WST-8 assay (n = 6 each; *, P < 0.05 vs. 0 μmol/L; AU, arbitrary units). F, qPCR analysis of FGF19 mRNA levels in parental cells and lenvatinib-resistant Hep3B and HuH7 cells (n = 3 each; *, P < 0.05; AU, arbitrary units). G, Parental and lenvatinib-resistant Hep3B cells transduced with a doxycycline-inducible FGF19-expressing vector were treated with doxycycline at the indicated concentrations for 24 hours. The protein levels of FGF19 and ACTB were assessed by WB. H, Parental and lenvatinib-resistant Hep3B cells transduced with a doxycycline-inducible FGF19-expressing vector were treated with 100 μmol/L doxycycline and lenvatinib at the indicated concentrations for 72 hours. Cell viability was examined by WST-8 assay (n = 6 each; *, P < 0.05 vs. 0 μmol/L; AU, arbitrary units).

Close modal

ST6GAL1 is a tumor-derived secreted protein positively regulated by FGF19 in HCC

Next, we examined tumor-derived secreted proteins to identify serum biomarkers for FGF19-driven HCC. To this end, we performed stable isotope labeling with amino acids in cell culture (SILAC)-based cellular proteomic and secretomic analyses on FGF19-expressing Hep3B and Huh7 HCC cells with or without FGF19 knockdown (Fig. 3A; Supplementary Tables S6–S9). We identified 6 proteins (CTGF, CCL20, GDF15, ST6GAL1, ORM1, and SERPINI1) whose expression was decreased in both the lysates and supernatants of the two cell lines when FGF19 was silenced with siRNA (Fig. 3B). We then assessed the expression levels of each of the six genes upon siRNA-mediated inhibition of FGF19 and validated that CTGF, GDF15, and ST6GAL1 were significantly downregulated by FGF19 knockdown with two siRNAs (Fig. 3C). We next assessed the expression levels of these genes in lenvatinib-resistant cell lines (Hep3B-LenR and HuH7-LenR) with markedly reduced FGF19 expression (Fig. 2F). The expression levels of GDF15 and ST6GAL1 were significantly lower in both lenvatinib-resistant cell lines than in their parental cell lines (Fig. 3D). In addition, we examined their expression levels in liver tumors generated from a pooled library of oncogenic cDNA sequences. Consistent with the results of barcode sequencing (Fig. 1I), the mRNA expression levels of FGF19 in tumors were significantly reduced by lenvatinib treatment (Fig. 3E). Importantly, GDF15 and ST6GAL1 levels were also significantly decreased by lenvatinib treatment (Fig. 3F). We further assessed the correlations between the gene expression of FGF19 and the two secreted proteins in HCC cell lines and human HCC tissues. FGF19 levels were significantly correlated with ST6GAL1 levels but not with GDF15 levels in the nine human hepatoma cell lines (Fig. 3G) or HCC tissues from TCGA cohort (Fig. 3H). Collectively, the results indicated that the secreted protein ST6GAL1 was the protein most positively correlated with FGF19 in HCC. We thus expect that ST6GAL1 could be a surrogate serum marker of FGF19-driven HCC.

Figure 3.

ST6GAL1 is a tumor-derived secreted protein positively regulated by FGF19 in HCC. A and B, Hep3B cells and HuH7 cells were stably labeled with amino acids containing isotopes in cell culture for 10 passages. Then, the cells were transfected with negative control (NC) or FGF19 siRNA for 3 days, and the cell lysates and supernatants were analyzed with proteomics. A diagram depicting the stable isotope labeling with amino acids in cell culture (SILAC)-based cellular proteomic and secretomic strategies is shown (A). Proteins showing greater than 0.5-fold downregulation in FGF19 siRNA-transfected cells compared to NC siRNA-transfected cells were selected and compared between Hep3B and HuH7 cells. Sixty proteins in supernatants and 37 proteins in cell lysates were shared between Hep3B and HuH7 cells, and six proteins identified in both supernatants and cell lysates are listed (B). C, Hep3B and HuH7 cells were transfected with either NC or FGF19 siRNA. The mRNA levels of CTGF, CCL20, GDF15, ST6GAL1, ORM1, and SERPINI1 were analyzed by qPCR 72 hours after transfection (n = 4 each; *, P < 0.05 vs. NC siRNA; AU, arbitrary units). D, mRNA levels of FGF19, CTGF, GDF15, and ST6GAL1 in lenvatinib-resistant or parental Hep3B and HuH7 cells (n = 4 each; *, P < 0.05; AU, arbitrary units). mRNA levels of FGF19 (E), Gdf15, and St6gal1 (F) in liver tumors obtained from a genetically heterogeneous liver tumor mouse model upon treatment with vehicle or lenvatinib (n = 12–15 each; *, P < 0.05; AU, for arbitrary units). G, mRNA levels of ST6GAL1 and GDF15 in 9 hepatoma cell lines (n = 3 each; left) and their correlations with FGF19 mRNA levels (right; n.s., not significant; AU, arbitrary units, respectively). H, Correlations between FGF19 and ST6GAL1 mRNA levels (top) and between FGF19 and GDF15 mRNA levels (bottom) in 373 patients with HCC registered in TCGA database portal.

Figure 3.

ST6GAL1 is a tumor-derived secreted protein positively regulated by FGF19 in HCC. A and B, Hep3B cells and HuH7 cells were stably labeled with amino acids containing isotopes in cell culture for 10 passages. Then, the cells were transfected with negative control (NC) or FGF19 siRNA for 3 days, and the cell lysates and supernatants were analyzed with proteomics. A diagram depicting the stable isotope labeling with amino acids in cell culture (SILAC)-based cellular proteomic and secretomic strategies is shown (A). Proteins showing greater than 0.5-fold downregulation in FGF19 siRNA-transfected cells compared to NC siRNA-transfected cells were selected and compared between Hep3B and HuH7 cells. Sixty proteins in supernatants and 37 proteins in cell lysates were shared between Hep3B and HuH7 cells, and six proteins identified in both supernatants and cell lysates are listed (B). C, Hep3B and HuH7 cells were transfected with either NC or FGF19 siRNA. The mRNA levels of CTGF, CCL20, GDF15, ST6GAL1, ORM1, and SERPINI1 were analyzed by qPCR 72 hours after transfection (n = 4 each; *, P < 0.05 vs. NC siRNA; AU, arbitrary units). D, mRNA levels of FGF19, CTGF, GDF15, and ST6GAL1 in lenvatinib-resistant or parental Hep3B and HuH7 cells (n = 4 each; *, P < 0.05; AU, arbitrary units). mRNA levels of FGF19 (E), Gdf15, and St6gal1 (F) in liver tumors obtained from a genetically heterogeneous liver tumor mouse model upon treatment with vehicle or lenvatinib (n = 12–15 each; *, P < 0.05; AU, for arbitrary units). G, mRNA levels of ST6GAL1 and GDF15 in 9 hepatoma cell lines (n = 3 each; left) and their correlations with FGF19 mRNA levels (right; n.s., not significant; AU, arbitrary units, respectively). H, Correlations between FGF19 and ST6GAL1 mRNA levels (top) and between FGF19 and GDF15 mRNA levels (bottom) in 373 patients with HCC registered in TCGA database portal.

Close modal

FGF19 may positively regulate ST6GAL1 via STAT3 phosphorylation

We investigated the molecular mechanism by which FGF19 transcriptionally regulated ST6GAL1. We found that FGF19 overexpression significantly increased ST6GAL1 expression in Hep3B-LenR cells and enhanced STAT3 phosphorylation (Supplementary Fig. S2A and S2B). In contrast, FGF19 knockdown decreased ST6GAL1 expression in Hep3B cells (Fig. 3C) and inhibited STAT3 phosphorylation (Supplementary Fig. S2B). Furthermore, suppression of STAT3 activation by the inhibitor Stattic or siRNA significantly decreased ST6GAL1 expression but did not affect the FGF19 levels in Hep3B cells (Supplementary Fig. S2C and S2D). Taken together, these results suggest that FGF19 may positively regulate ST6GAL1 via STAT3 phosphorylation.

Serum ST6GAL1 is a biomarker of FGF19-driven highly malignant HCC

We then investigated the association between serum ST6GAL1 levels and tumor FGF19 expression in 62 patients with HCC who underwent curative surgical resection retrospectively. The patient background information is shown in Supplementary Table S10. IHC analysis revealed that approximately a quarter of the patients with HCC exhibited high levels of FGF19 expression in tumor cells (Fig. 4A). Patients with FGF19-high HCC showed significantly more advanced disease and shorter RFS after curative resection (Fig. 4B and C), suggesting that FGF19-high HCC is highly malignant, in agreement with previous reports (22). Serum ST6GAL1 levels were significantly higher in patients with FGF19-high HCC than in those with FGF19-low HCC (Fig. 4D, left) and were significantly positively correlated with FGF19-positive areas in tumor sites (Fig. 4D, right). In addition, patients with FGF19-high HCC were identified with high sensitivity and specificity using the cutoff serum ST6GAL1 value of 19.1 ng/mL determined by the Youden index (Fig. 4E). Moreover, serum ST6GAL1 levels were positively correlated with tumor size and disease stage (Fig. 4F), and patients with high serum ST6GAL1 levels had significantly shorter RFS than those with low levels (Fig. 4G). In sharp contrast, serum FGF19 levels were not different between patients with FGF19-high HCC and patients with FGF19-low HCC and could not be used to discriminate the groups (Supplementary Fig. S3A and S3B). Furthermore, serum FGF19 levels did not affect disease stage or RFS in patients with HCC (Supplementary Fig. S3C and S3D). Taken together, our data indicate that ST6GAL1 could be a reliable serum biomarker for identification of patients with FGF19-expressing HCC, which exhibits high malignancy potential.

Figure 4.

Serum ST6GAL1 is a biomarker of FGF19-driven highly malignant HCC. A–E, Sixty-two surgically resected HCC tissue sections were stained for FGF19 and divided into two groups based on FGF19 expression: the FGF19-high group (n = 15) and the FGF19-low group (n = 47). Representative images of FGF19 staining in FGF19-high and FGF19-low tumors are shown (the scale bar indicates 100 μm; A). The tumor grades of patients with FGF19-high and FGF19-low HCC as assessed by the tumor–node–metastasis (TNM) staging system (B), the RFS of patients with FGF19-high and -low HCC (C), the serum ST6GAL1 levels of patients with FGF19-high and FGF19-low HCC (D, left), as examined by ELISA (*, P < 0.05), and the correlation between the FGF19-positive area and the serum ST6GAL1 levels (D, right) are shown. ROC curves for diagnosis of patients with FGF19-high HCC by serum ST6GAL1 levels were constructed (AUC, area under the curve). The cut-off value in each ROC curve was determined with Youden J statistic (gray dot; E). F and G, Serum ST6GAL1 levels were examined by ELISA in 79 patients with HCC who underwent surgical resection. The correlations between serum ST6GAL1 levels and tumor size (F, left) and between serum ST6GAL1 levels and tumor grade according to the TNM staging system (F, right) were investigated (*, P < 0.05). The RFS times of patients with HCC with high and low serum ST6GAL1 levels are shown (G).

Figure 4.

Serum ST6GAL1 is a biomarker of FGF19-driven highly malignant HCC. A–E, Sixty-two surgically resected HCC tissue sections were stained for FGF19 and divided into two groups based on FGF19 expression: the FGF19-high group (n = 15) and the FGF19-low group (n = 47). Representative images of FGF19 staining in FGF19-high and FGF19-low tumors are shown (the scale bar indicates 100 μm; A). The tumor grades of patients with FGF19-high and FGF19-low HCC as assessed by the tumor–node–metastasis (TNM) staging system (B), the RFS of patients with FGF19-high and -low HCC (C), the serum ST6GAL1 levels of patients with FGF19-high and FGF19-low HCC (D, left), as examined by ELISA (*, P < 0.05), and the correlation between the FGF19-positive area and the serum ST6GAL1 levels (D, right) are shown. ROC curves for diagnosis of patients with FGF19-high HCC by serum ST6GAL1 levels were constructed (AUC, area under the curve). The cut-off value in each ROC curve was determined with Youden J statistic (gray dot; E). F and G, Serum ST6GAL1 levels were examined by ELISA in 79 patients with HCC who underwent surgical resection. The correlations between serum ST6GAL1 levels and tumor size (F, left) and between serum ST6GAL1 levels and tumor grade according to the TNM staging system (F, right) were investigated (*, P < 0.05). The RFS times of patients with HCC with high and low serum ST6GAL1 levels are shown (G).

Close modal

Serum ST6GAL1 is a biomarker of lenvatinib-susceptible HCC

On the basis of our experimental evidence showing the high lenvatinib susceptibility of FGF19-driven HCC, we hypothesized that baseline serum ST6GAL1 levels would also be useful biomarkers for prediction of lenvatinib susceptibility in patients with HCC. To test this hypothesis, we examined the pretreatment serum ST6GAL1 levels of 96 patients with advanced HCC who later underwent either lenvatinib or sorafenib treatment and analyzed their association with prognosis. The patient background information is shown in Supplementary Table S11. There were no differences in serum ST6GAL1 levels or OS between lenvatinib-treated and sorafenib-treated patients (Fig. 5A and B). We then divided the patients into ST6GAL1-high and ST6GAL1-low groups based on the cut-off serum ST6GAL1 level used to identify FGF19-high HCC (Fig. 4E). Among the 62 patients with HCC with low serum ST6GAL1 levels, OS was not different between lenvatinib-treated patients and sorafenib-treated patients (Fig. 5C, left). On the other hand, among the 34 patients with HCC with high serum ST6GAL1 levels, lenvatinib-treated patients showed significantly longer OS than sorafenib-treated patients (Fig. 5C right; Supplementary Fig. S4). We also found that the serum ST6GAL1 levels were significantly decreased upon lenvatinib treatment in patients with high pretreatment serum ST6GAL1 levels (Supplementary Fig. S5). In addition to the Child-Pugh score, Barcelona Clinic Liver Cancer (BCLC) stage, and serum α-fetoprotein (AFP) level, lenvatinib treatment was identified as one of the predictive factors of better OS in the ST6GAL1-high group by univariate Cox proportional hazard analysis. Finally, a multivariate Cox proportional hazard analysis identified lenvatinib treatment and low serum AFP level as independent factors that contribute to better OS in ST6GAL1-high tumors (Table 1). Taken together, our clinical observations suggest that ST6GAL1 may be useful for selecting patients with HCC for whom lenvatinib therapy will provide a better survival benefit than sorafenib therapy.

Figure 5.

Serum ST6GAL1 is a biomarker of lenvatinib-susceptible HCC. A and B, In 96 patients with advanced HCC who were treated with either lenvatinib or sorafenib, baseline serum ST6GAL1 levels were examined by ELISA (A), and OS was assessed by the Kaplan–Meier method (B; n = 31 for sorafenib and 65 for lenvatinib; n.s., not significant). C, Ninety-six patients with advanced HCC were divided into ST6GAL1-high and ST6GAL1-low groups based on the cut-off value of serum ST6GAL1 levels used to identify FGF19-driven HCC. The OS times of 62 patients treated with either lenvatinib or sorafenib among patients with low serum ST6GAL1 levels (n = 19 for sorafenib and 43 for lenvatinib; C, left) and the OS times of 34 patients treated with either lenvatinib or sorafenib among patients with high serum ST6GAL1 levels (n = 12 for sorafenib and 22 for lenvatinib; C, right) are shown.

Figure 5.

Serum ST6GAL1 is a biomarker of lenvatinib-susceptible HCC. A and B, In 96 patients with advanced HCC who were treated with either lenvatinib or sorafenib, baseline serum ST6GAL1 levels were examined by ELISA (A), and OS was assessed by the Kaplan–Meier method (B; n = 31 for sorafenib and 65 for lenvatinib; n.s., not significant). C, Ninety-six patients with advanced HCC were divided into ST6GAL1-high and ST6GAL1-low groups based on the cut-off value of serum ST6GAL1 levels used to identify FGF19-driven HCC. The OS times of 62 patients treated with either lenvatinib or sorafenib among patients with low serum ST6GAL1 levels (n = 19 for sorafenib and 43 for lenvatinib; C, left) and the OS times of 34 patients treated with either lenvatinib or sorafenib among patients with high serum ST6GAL1 levels (n = 12 for sorafenib and 22 for lenvatinib; C, right) are shown.

Close modal
Table 1.

Univariate and multivariate analysis of factors associated with the OS in patients with high serum ST6GAL1 levels.

CategoryUnivariate analysis HR (95% CI)PMultivariate analysis HR (95% CI)P
Age (years) ≧70/<70 1.27 (0.38–5.72) 0.72   
Sex Male/female 0.86 (0.23–5.64) 0.85   
Etiology Others/HCV 1.06 (0.33–3.40) 0.92   
Child-Pugh B/A 7.13 (2.21–27.06) 0.001 3.56 (0.77–16.14) 0.10 
Maximum tumor size ≧50 mm/<50 mm 1.49 (0.33–5.04) 0.57   
Tumor n ≧7/<7 2.71 (0.87–9.32) 0.084   
BCLC stage C/B 4.06 (1.24–18.19) 0.019 2.89 (0.74–14.72) 0.12 
AFP (ng/mL) ≧37.4/<37.4 3.86 (1.13–17.7) 0.031 6.43 (1.29–46.24) 0.021 
Treatment Sorafenib/lenvatinib 3.27 (1.02–12.3) 0.046 4.66 (1.22–21.86) 0.023 
CategoryUnivariate analysis HR (95% CI)PMultivariate analysis HR (95% CI)P
Age (years) ≧70/<70 1.27 (0.38–5.72) 0.72   
Sex Male/female 0.86 (0.23–5.64) 0.85   
Etiology Others/HCV 1.06 (0.33–3.40) 0.92   
Child-Pugh B/A 7.13 (2.21–27.06) 0.001 3.56 (0.77–16.14) 0.10 
Maximum tumor size ≧50 mm/<50 mm 1.49 (0.33–5.04) 0.57   
Tumor n ≧7/<7 2.71 (0.87–9.32) 0.084   
BCLC stage C/B 4.06 (1.24–18.19) 0.019 2.89 (0.74–14.72) 0.12 
AFP (ng/mL) ≧37.4/<37.4 3.86 (1.13–17.7) 0.031 6.43 (1.29–46.24) 0.021 
Treatment Sorafenib/lenvatinib 3.27 (1.02–12.3) 0.046 4.66 (1.22–21.86) 0.023 

In this study, to screen genetic drivers involved in the susceptibility of HCC to lenvatinib in vivo, we generated a novel liver tumor mouse model that reflects the tumor genetic complexity observed in human HCC. With this model, we found that FGF19-driven HCC is more susceptible to lenvatinib therapy than HCC driven by other oncogenes in vivo. We also found that MET or NRAS-driven HCC is more susceptible to sorafenib. This finding may reflect the kinase inhibition profile of sorafenib in that it can block Raf signaling downstream of MET and NRAS. On the contrary, several studies have reported that the HGF/Met signaling pathways were involved in sorafenib resistance (23–26). Although the exact reason for this discrepancy is unclear, one reason for this difference might be that most of these published studies used sorafenib-resistant HCC cell lines generated in vitroby long-term exposure of the cells to sorafenib and focused on susceptibility in vitro. In contrast, our assay system considered the tumor microenvironment and thus assessed the susceptibility of tumor cells to the drug in an in vivo context.

While a variety of HCC mouse models have been established, including chemical- or diet-induced models, genetically engineered models, and cell-line or patient-derived xenograft (PDX) models (27, 28), our model has several advantages for preclinical in vivoscreening of drug susceptibility. First, our model is established through a transposon-based system for random and traceable intrahepatic genomic integration of multiple transgenes in a pooled format. When used together with deep sequencing-based deconvolution of the integrated transgenes, this system enables simultaneous screening of drug susceptibility related to multiple genetic drivers in vivo. In this study, we pooled 10 HCC oncogenes, but the system could be easily manipulated to assess any gene sets of interest. In addition, multiple liver tumors develop within 2 weeks in our model; thus, drug susceptibility screening can be completed within 2 months. These advantages eliminate the need for time- and cost-consuming processes for generation of multiple-knockout/transgenic mice or preparation of hundreds of mice for one-by-one in vivogenetic modification. Furthermore, our model does not require an immunocompromised host, and the tumors endogenously develop under conditions of physiologic immunocompetence; thus, it can be used to study the effects of any drug on immune cells, vascular endothelial cells, and stromal cells in the tumor microenvironment. In contrast, a caveat of our model is that our system may require further modification or validation for the evaluation of tumor suppressor genes (TSG), because an short hairpin RNA or CRISPR vector instead of a cDNA expression vector has to be used to target TSGs, which may cause potential bias due to differences in the plasmid backbone. In addition, the significance of TERT, which is one of the most important HCC driver genes, is generally difficult to evaluate in a mouse model because telomerase is expressed in adult mice, and telomeres in mice are much longer than those in humans (29). Nevertheless, given our findings, our new model may be a valuable tool for preclinical in vivo drug assessment, especially in the context of genetic diversity.

FGF19 is an experimentally proven oncogenic driver of HCC (30–32). A previous TCGA study and other studies have shown that gene amplification frequently occurs at the FGF19 gene locus in patients with HCC (7, 31). FGF19 mainly performs oncogenic functions in tumor cells via paracrine and autocrine signals through FGFR4 binding (33, 34); thus, FGF19-driven HCC could be amenable to lenvatinib therapy, which has been reported by several groups (18, 35). Considering that FGF19 is a secreted protein, our cell- and animal-based experimental results prompted us to hypothesize that serum FGF19 levels could be used to identify patients with FGF19-driven HCC and thus could be biomarkers for prediction of lenvatinib efficacy. Unexpectedly, however, serum FGF19 levels were not associated with intratumoral FGF19 expression in patients with HCC and thus could not be used to detect patients with FGF19-driven HCC. Similarly, a recent report has shown that pretreatment serum FGF19 levels cannot be used to predict the efficacy of lenvatinib treatment in patients with HCC (36). FGF19 is a bile acid-mediated hormone that physiologically regulates bile acid synthesis, cholesterol metabolism, and insulin sensitivity (37). While FGFR4, a receptor of FGF19, is highly expressed in the liver (38), FGF19 is expressed mainly in the ileum in healthy people (39). Serum FGF19 levels are upregulated in patients with chronic hepatitis and are positively associated with cholestasis and fibrosis (40, 41). An IHC study has shown that intrahepatic bile duct cells and endothelial cells in patients with hepatitis express FGF19, which could be responsible for the elevated serum FGF19 levels in these patients (41). Considering that serum FGF19 levels were associated with neither disease grade nor postoperative tumor-free survival in our patients with HCC, serum FGF19 levels might better reflect background FGF19 protein secretion by the liver and/or intestine than secretion by tumors themselves in patients with HCC.

In this study, with the help of comprehensive proteomics, we found that the tumor-derived secreted protein ST6GAL1 was positively regulated by FGF19 in HCC cells. Subsequently, we confirmed the validity of serum ST6GAL1 levels as reliable biomarkers for identification of patients with FGF19-driven HCC. Moreover, patients with HCC with high serum ST6GAL1 levels exhibited more advanced disease and shorter RFS, suggesting that HCC with an active FGF19-ST6GAL1 axis has increased malignancy potential. Aberrant glycosylation is a universal feature of cancer cells, and abnormal sialylation is especially widely observed in a variety of cancers (42). ST6GAL1 is a sialyltransferase enzyme that catalyzes the transfer of sialic acid from cytidine monophosphate-sialic acid to galactose-containing substrates and can modify glycoproteins and glycolipids (43). ST6GAL1 is overexpressed in various cancer types, including pancreatic, breast, prostate, and ovarian cancer, and plays a significant role in tumor progression (42). Its expression is also positively correlated with the metastatic ability of murine HCC (44), and its overexpression promotes proliferation and migration of human HCC cells (45). Given this information and our clinical findings, the potential oncogenic role of FGF19-ST6GAL1 signaling in patients with HCC might be worth evaluating.

In our cohort of 96 patients with advanced HCC who underwent TKI therapy, OS was not different between lenvatinib-treated and sorafenib-treated patients. However, in a subgroup of patients with HCC with high serum ST6GAL1 levels, OS was significantly longer for lenvatinib-treated patients than for sorafenib-treated patients. In this subgroup, lenvatinib treatment was one of the predictive factors of improved OS, suggesting the potential usefulness of ST6GAL1 as a serum biomarker for optimal TKI selection. Because serum ST6GAL1 levels are associated with FGF19 expression in HCC, this finding suggests that patients with HCC with high serum ST6GAL1 levels may have FGF19-driven tumors and may thus be more susceptible to lenvatinib than those with low serum ST6GAL1 levels. This result might also be related to a recent finding that FGF19/FGFR4 signaling contributes to sorafenib resistance in HCC (46). Meanwhile, our study has the following limitations: the correlation between tumor ST6GAL1/FGF19 mRNA levels and serum ST6GAL1 levels was not evaluated in our cohorts, and our analysis was retrospective. A further large-scale prospective study is needed to validate the usefulness of our potential biomarker in real-world clinical practice.

In conclusion, we established a novel liver tumor mouse model that can be used for simultaneous and rapid assessment of drug susceptibility related to genetic drivers and showed that FGF19-driven aggressive HCC is susceptible to lenvatinib treatment. We also found that ST6GAL1 is a tumor-derived secreted protein downstream of FGF19 that may be useful as a serum biomarker for identification of FGF19-driven patient with HCC for whom lenvatinib treatment will be beneficial.

Y. Myojin is listed as a co-inventor on a pending patent regarding a new tumor biomarker owned by Osaka University. T. Kodama reports grants from Japan Agency for Medical Research and Development and Japan Society for the Promotion of Science during the conduct of this study; personal fees from Eisai Co., Eli Lilly Co., Bayer, and Chugai Pharmaceutical Co. outside the submitted work; and is listed as a co-inventor on a pending patent regarding a new tumor biomarker. Y Hayashi reports personal fees from Chugai Pharmaceutical, Novartis Pharma K.K., and Merck Biopharma outside the submitted work. H. Hikita reports personal fees from Chugai Pharmaceutical Co. outside the submitted work. T. Takehara reports grants from Japan Agency for Medical Research and Development and Japan Society for the Promotion of Science during the conduct of the study; personal fees from Eisai Co. and Chugai Pharmaceutical Co. outside the submitted work; and is listed as a co-inventor on a pending patent regarding a new tumor biomarker. No disclosures were reported by the other authors.

Y. Myojin: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing-original draft. T. Kodama:Conceptualization, funding acquisition, methodology, writing-original draft. K. Maesaka: Resources, formal analysis. D. Motooka: Formal analysis. Y. Sato:Data curation, formal analysis. S. Tanaka: Resources. Y. Abe: Data curation, formal analysis. K. Ohkawa: Resources. E. Mita: Resources. Y. Hayashi: Methodology. H. Hikita: Methodology. R. Sakamori: Methodology. T. Tatsumi: Methodology. A. Taguchi: Data curation, formal analysis, funding acquisition, writing-review and editing. H. Eguchi: Resources. T. Takehara: Conceptualization, supervision, funding acquisition, project administration, writing-review and editing.

This work was supported by the Japan Agency for Medical Research and Development (AMED) under grant numbers JP20fk0210074 (to T. Kodama), JP19cm0106437 (to T. Kodama and T. Takehara), and 20fk0210082h0001 (to Y. Abe and A. Taguchi) and by a Grant-in-Aid for Scientific Research (to T. Kodama) from the Ministry of Education, Culture, Sports, Science, and Technology, Japan under grant numbers 17K09422 and 20H03661.

The authors thank the following doctors for providing the serum obtained from patients with advanced HCC: H. Hagiwara from Kansai Rosai Hospital, Y. Inui from Hyogo Prefectural Nishinomiya Hospital, T. Yakushijin from Osaka General Medical Center, S. Tamura from Minoh City Hospital, M. Inada from Toyonaka Municipal Hospital, H. Fukui from Yao Municipal Hospital, and H. Ogawa from Nishinomiya Municipal Central Hospital. They thank Kentaro Taki for acquiring the mass spectrometry data. They thank Toru Okamoto for providing the pFTRE-pGK-puro, pCAG-c-Myc-2A-Luc, pCAG-Nras-2A-Luc, pCAG-GFP-2A-Luc, and pFTRE-pGK-Puro plasmids; Xin Chen for providing the pT3-Myr-AKT-HA, pT3-N90beta-catenin, pT3-EF1a-c-MET, and pT3-EF1a-YAPS127A plasmids; William Hahn for providing the pBABE puro cyclinD1 HA plasmids; Roger Davis for providing the pCDNA3.1-hMcl-1 plasmid; Randall Moon for providing NC16 pCDNA3.1 FLAG NRF2; Junji Takeda for providing the pPB-CAG-EBNXN and pCMV-hyPBase plasmids; Didier Trono for providing the psPAX2 plasmid; and Bob Weinberg for providing the pCMV-VSV-G plasmid. They thank Eisai for providing the lenvatinib.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisementin accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Bray
F
,
Ferlay
J
,
Soerjomataram
I
,
Siegel
RL
,
Torre
LA
,
Jemal
A
. 
Global cancer statistics 2018: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries
.
CA Cancer J Clin
2018
;
68
:
394
424
.
2.
Tabrizian
P
,
Jibara
G
,
Shrager
B
,
Schwartz
M
,
Roayaie
S
. 
Recurrence of hepatocellular cancer after resection: patterns, treatments, and prognosis
.
Ann Surg
2015
;
261
:
947
55
.
3.
Kew
MC
. 
Hepatocellular carcinoma in developing countries: prevention, diagnosis and treatment
.
World J Hepatol
2012
;
4
:
99
104
.
4.
Llovet
JM
,
Ricci
S
,
Mazzaferro
V
,
Hilgard
P
,
Gane
E
,
Blanc
JF
, et al
Sorafenib in advanced hepatocellular carcinoma
.
N Engl J Med
2008
;
359
:
378
90
.
5.
Finn
RS
,
Qin
S
,
Ikeda
M
,
Galle
PR
,
Ducreux
M
,
Kim
TY
, et al
Atezolizumab plus bevacizumab in unresectable hepatocellular carcinoma
.
N Engl J Med
2020
;
382
:
1894
905
.
6.
Kudo
M
,
Finn
RS
,
Qin
S
,
Han
KH
,
Ikeda
K
,
Piscaglia
F
, et al
Lenvatinib versus sorafenib in first-line treatment of patients with unresectable hepatocellular carcinoma: a randomised phase 3 non-inferiority trial
.
Lancet
2018
;
391
:
1163
73
.
7.
Cancer Genome Atlas Research Network
. 
Electronic address wbe, cancer genome atlas research N. comprehensive and integrative genomic characterization of hepatocellular carcinoma
.
Cell
2017
;
169
:
1327
41.e23
.
8.
Schulze
K
,
Imbeaud
S
,
Letouzé
E
,
Alexandrov
LB
,
Calderaro
J
,
Rebouissou
S
, et al
Exome sequencing of hepatocellular carcinomas identifies new mutational signatures and potential therapeutic targets
.
Nat Genet
2015
;
47
:
505
11
.
9.
Kokudo
N
,
Takemura
N
,
Hasegawa
K
,
Takayama
T
,
Kubo
S
,
Shimada
M
, et al
Clinical practice guidelines for hepatocellular carcinoma: The Japan Society of Hepatology 2017 (4th JSH-HCC guidelines) 2019 update
.
Hepatol Res
2019
;
49
:
1109
13
.
10.
Kanwal
F
,
Singal
AG
. 
Surveillance for hepatocellular carcinoma: current best practice and future direction
.
Gastroenterology
2019
;
157
:
54
64
.
11.
Llovet
JM
,
Montal
R
,
Sia
D
,
Finn
RS
. 
Molecular therapies and precision medicine for hepatocellular carcinoma
.
Nat Rev Clin Oncol
2018
;
15
:
599
616
.
12.
Sieghart
W
,
Losert
D
,
Strommer
S
,
Cejka
D
,
Schmid
K
,
Rasoul-Rockenschaub
S
, et al
Mcl-1 overexpression in hepatocellular carcinoma: a potential target for antisense therapy
.
J Hepatol
2006
;
44
:
151
7
.
13.
Kodama
T
,
Yi
J
,
Newberg
JY
,
Tien
JC
,
Wu
H
,
Finegold
MJ
, et al
Molecular profiling of nonalcoholic fatty liver disease-associated hepatocellular carcinoma using SB transposon mutagenesis
.
Proc Natl Acad Sci U S A
2018
;
115
:
E10417
E26
.
14.
Zhang
S
,
Zhou
D
. 
Role of the transcriptional coactivators YAP/TAZ in liver cancer
.
Curr Opin Cell Biol
2019
;
61
:
64
71
.
15.
Akula
SM
,
Abrams
SL
,
Steelman
LS
,
Emma
MR
,
Augello
G
,
Cusimano
A
, et al
RAS/RAF/MEK/ERK, PI3K/PTEN/AKT/mTORC1 and TP53 pathways and regulatory miRs as therapeutic targets in hepatocellular carcinoma
.
Expert Opin Ther Targets
2019
;
23
:
915
29
.
16.
Bell
JB
,
Podetz-Pedersen
KM
,
Aronovich
EL
,
Belur
LR
,
McIvor
RS
,
Hackett
PB
. 
Preferential delivery of the sleeping beauty transposon system to livers of mice by hydrodynamic injection
.
Nat Protoc
2007
;
2
:
3153
65
.
17.
Wilhelm
SM
,
Carter
C
,
Tang
L
,
Wilkie
D
,
McNabola
A
,
Rong
H
, et al
BAY 43–9006 exhibits broad spectrum oral antitumor activity and targets the RAF/MEK/ERK pathway and receptor tyrosine kinases involved in tumor progression and angiogenesis
.
Cancer Res
2004
;
64
:
7099
109
.
18.
Matsuki
M
,
Hoshi
T
,
Yamamoto
Y
,
Ikemori-Kawada
M
,
Minoshima
Y
,
Funahashi
Y
, et al
Lenvatinib inhibits angiogenesis and tumor fibroblast growth factor signaling pathways in human hepatocellular carcinoma models
.
Cancer Med
2018
;
7
:
2641
53
.
19.
Kilkenny
C
,
Browne
WJ
,
Cuthill
IC
,
Emerson
M
,
Altman
DG
. 
Improving bioscience research reporting: the ARRIVE guidelines for reporting animal research
.
PLoS Biol
2010
;
8
:
e1000412
.
20.
Youden
WJ
. 
Index for rating diagnostic tests
.
Cancer
1950
;
3
:
32
5
.
21.
Adnane
L
,
Trail
PA
,
Taylor
I
,
Wilhelm
SM
. 
Sorafenib (BAY 43–9006, Nexavar), a dual-action inhibitor that targets RAF/MEK/ERK pathway in tumor cells and tyrosine kinases VEGFR/PDGFR in tumor vasculature
.
Methods Enzymol
2006
;
407
:
597
612
.
22.
Hyeon
J
,
Ahn
S
,
Lee
JJ
,
Song
DH
,
Park
CK
. 
Expression of fibroblast growth factor 19 is associated with recurrence and poor prognosis of hepatocellular carcinoma
.
Dig Dis Sci
2013
;
58
:
1916
22
.
23.
Firtina Karagonlar
Z
,
Koc
D
,
Iscan
E
,
Erdal
E
,
Atabey
N
. 
Elevated hepatocyte growth factor expression as an autocrine c-Met activation mechanism in acquired resistance to sorafenib in hepatocellular carcinoma cells
.
Cancer Sci
2016
;
107
:
407
16
.
24.
Chen
S
,
Xia
X
. 
Long noncoding RNA NEAT1 suppresses sorafenib sensitivity of hepatocellular carcinoma cells via regulating miR-335-c-Met
.
J Cell Physiol
2019
;
234
:
14999
5009
.
25.
Xiang
QF
,
Zhan
MX
,
Li
Y
,
Liang
H
,
Hu
C
,
Huang
YM
, et al
Activation of MET promotes resistance to sorafenib in hepatocellular carcinoma cells via the AKT/ERK1/2-EGR1 pathway
.
Artif Cells Nanomed Biotechnol
2019
;
47
:
83
9
.
26.
Qu
Z
,
Wu
J
,
Luo
D
,
Jiang
C
,
Ding
Y
. 
Exosomes derived from HCC cells induce sorafenib resistance in hepatocellular carcinoma both in vivoand in vitro
.
J Exp Clin Cancer Res
2016
;
35
:
159
.
27.
Bakiri
L
,
Wagner
EF
. 
Mouse models for liver cancer
.
Mol Oncol
2013
;
7
:
206
23
.
28.
Nazzal
M
,
Sur
S
,
Steele
R
,
Khatun
M
,
Patra
T
,
Phillips
N
, et al
Establishment of a patient-derived xenograft tumor from hepatitis C-associated liver cancer and evaluation of imatinib treatment efficacy
.
Hepatology
2020
;
72
:
379
88
.
29.
Cheng
Zhao Y
,
Zhang
F
,
Zhang
J
,
Wang
S
,
Zhu
J
. 
Engineering a humanized telomerase reverse transcriptase gene in mouse embryonic stem cells
.
Sci Rep
2019
;
9
:
9683
.
30.
Nicholes
K
,
Guillet
S
,
Tomlinson
E
,
Hillan
K
,
Wright
B
,
Frantz
GD
, et al
A mouse model of hepatocellular carcinoma: ectopic expression of fibroblast growth factor 19 in skeletal muscle of transgenic mice
.
Am J Pathol
2002
;
160
:
2295
307
.
31.
Sawey
ET
,
Chanrion
M
,
Cai
C
,
Wu
G
,
Zhang
J
,
Zender
L
, et al
Identification of a therapeutic strategy targeting amplified FGF19 in liver cancer by Oncogenomic screening
.
Cancer Cell
2011
;
19
:
347
58
.
32.
Zhou
M
,
Luo
J
,
Chen
M
,
Yang
H
,
Learned
RM
,
DePaoli
AM
, et al
Mouse species-specific control of hepatocarcinogenesis and metabolism by FGF19/FGF15
.
J Hepatol
2017
;
66
:
1182
92
.
33.
Yang
X
,
Liaw
L
,
Prudovsky
I
,
Brooks
PC
,
Vary
C
,
Oxburgh
L
, et al
Fibroblast growth factor signaling in the vasculature
.
Curr Atheroscler Rep
2015
;
17
:
509
.
34.
French
DM
,
Lin
BC
,
Wang
M
,
Adams
C
,
Shek
T
,
Hotzel
K
, et al
Targeting FGFR4 inhibits hepatocellular carcinoma in preclinical mouse models
.
PLoS One
2012
;
7
:
e36713
.
35.
Hoshi
T
,
Watanabe Miyano
S
,
Watanabe
H
,
Sonobe
RMK
,
Seki
Y
,
Ohta
E
, et al
Lenvatinib induces death of human hepatocellular carcinoma cells harboring an activated FGF signaling pathway through inhibition of FGFR-MAPK cascades
.
Biochem Biophys Res Commun
2019
;
513
:
1
7
.
36.
Chuma
M
,
Uojima
H
,
Numata
K
,
Hidaka
H
,
Toyoda
H
,
Hiraoka
A
, et al
Early changes in circulating FGF19 and ang-2 levels as possible predictive biomarkers of clinical response to lenvatinib therapy in hepatocellular carcinoma
.
Cancers
2020
;
12
:
293
.
37.
Sandhu
DS
,
Baichoo
E
,
Roberts
LR
. 
Fibroblast growth factor signaling in liver carcinogenesis
.
Hepatology
2014
;
59
:
1166
73
.
38.
Repana
D
,
Ross
P
. 
Targeting FGF19/FGFR4 pathway: a novel therapeutic strategy for hepatocellular carcinoma
.
Diseases
2015
;
3
:
294
305
.
39.
Potthoff
MJ
,
Kliewer
SA
,
Mangelsdorf
DJ
. 
Endocrine fibroblast growth factors 15/19 and 21: from feast to famine
.
Genes Dev
2012
;
26
:
312
24
.
40.
Wang
X
,
Chen
L
,
Wang
H
,
Cai
W
,
Xie
Q
. 
Modulation of bile acid profile by gut microbiota in chronic hepatitis B
.
J Cell Mol Med
2020
;
24
:
2573
81
.
41.
Brandl
K
,
Hartmann
P
,
Jih
LJ
,
Pizzo
DP
,
Argemi
J
,
Ventura-Cots
M
, et al
Dysregulation of serum bile acids and FGF19 in alcoholic hepatitis
.
J Hepatol
2018
;
69
:
396
405
.
42.
Garnham
R
,
Scott
E
,
Livermore
KE
,
Munkley
J
. 
ST6GAL1: a key player in cancer
.
Oncol Lett
2019
;
18
:
983
9
.
43.
Chakraborty
A
,
Dorsett
KA
,
Trummell
HQ
,
Yang
ES
,
Oliver
PG
,
Bonner
JA
, et al
ST6Gal-I sialyltransferase promotes chemoresistance in pancreatic ductal adenocarcinoma by abrogating gemcitabine-mediated DNA damage
.
J Biol Chem
2018
;
293
:
984
94
.
44.
Han
Y
,
Liu
Y
,
Fu
X
,
Zhang
Q
,
Huang
H
,
Zhang
C
, et al
miR-9 inhibits the metastatic ability of hepatocellular carcinoma via targeting beta galactoside alpha-2,6-sialyltransferase 1
.
J Physiol Biochem
2018
;
74
:
491
501
.
45.
Wang
L
,
Li
S
,
Yu
X
,
Han
Y
,
Wu
Y
,
Wang
S
, et al
α2,6-Sialylation promotes immune escape in hepatocarcinoma cells by regulating T cell functions and CD147/MMP signaling
.
J Physiol Biochem
2019
;
75
:
199
207
.
46.
Gao
L
,
Wang
X
,
Tang
Y
,
Huang
S
,
Hu
CA
,
Teng
Y
. 
FGF19/FGFR4 signaling contributes to the resistance of hepatocellular carcinoma to sorafenib
.
J Exp Clin Cancer Res
2017
;
36
:
8
.