The impact of FGF19/FGFR4 signaling inhibition in antitumor activity of multi-kinase inhibitors in hepatocellular carcinoma

FGF19/FGFR4 autocrine signaling is one of the main targets for multi-kinase inhibitors (MKIs). However, the molecular mechanisms underlying FGF19/FGFR4 signaling in the antitumor effects to MKIs in hepatocellular carcinoma (HCC) remain unclear. In this study, the impact of FGFR4/ERK signaling inhibition on HCC following MKI treatment was analyzed in vitro and in vivo assays. Serum FGF19 in HCC patients treated using MKIs, such as sorafenib (n = 173) and lenvatinib (n = 40), was measured by enzyme-linked immunosorbent assay. Lenvatinib strongly inhibited the phosphorylation of FRS2 and ERK, the downstream signaling molecules of FGFR4, compared with sorafenib and regorafenib. Additional use of a selective FGFR4 inhibitor with sorafenib further suppressed FGFR4/ERK signaling and synergistically inhibited HCC cell growth in culture and xenograft subcutaneous tumors. Although serum FGF19high (n = 68) patients treated using sorafenib exhibited a significantly shorter progression-free survival and overall survival than FGF19low (n = 105) patients, there were no significant differences between FGF19high (n = 21) and FGF19low (n = 19) patients treated using lenvatinib. In conclusion, robust inhibition of FGF19/FGFR4 is of importance for the exertion of antitumor effects of MKIs. Serum FGF19 levels may function as a predictive marker for drug response and survival in HCC patients treated using sorafenib.

www.nature.com/scientificreports/ but not in HCC. Multi-kinase inhibitors (MKIs), such as sorafenib, regorafenib, and lenvatinib, which inhibit multiple tyrosine kinase type receptors, are generally used to treat advanced HCC, although the survival benefits are limited [8][9][10] . Since the pharmacodynamic properties of the MKIs are quite different (Supplementary  Table S1) 11,12 , understanding the molecular mechanisms of MKI activity is essential for both drug selection and prognosis prediction. Fibroblast growth factor 19 (FGF19), a secreted protein from the small intestine, has unique specificity for FGF receptor 4 (FGFR4) 13 . FGF19/FGFR4 signaling regulates many biological processes such as cellular proliferation, differentiation, angiogenesis, and the metabolism of bile acid, glucose and lipids [14][15][16][17] . Of note, autocrine secretion of FGF19 has been reported in a variety of cancers, including HCC 18 . Although FGF19 transgenic mice develop HCC, crossing with FGFR4 knockout mice rescues the phenotype 19,20 . Together, these findings suggest the importance of FGF19/FGFR4 signal regulation in HCC. Although FGFR4 is one of the main targets of MKIs, the role of FGF19/FGFR4 signaling inhibition remains to be elucidated.
In this study, we examined the contribution of FGF19/FGFR4 signaling to the antitumor effects in culture and in an in vivo transplant model. To evaluate the expression of FGF19 and FGFR4, clinicopathological analyses using HCC surgical specimens were conducted. We also investigated whether serum FGF19 levels can function as a predictive marker for treatment response of MKIs using enzyme-linked immunosorbent assay (ELISA).

Basal expression of FGF19/FGFR4 in HCC cells.
To investigate the role of FGF19/FGFR4 signaling, we first evaluated the expression of FGF19 and FGFR4 in human HCC cell lines, namely Huh7, JHH7, HepG2 and PLC/PRF/5 cells. Of note, FGF19 mRNA expression varied among these cells and was consistent with FGF19 production in the culture supernatant (Fig. 1A,B). FGF19 mRNA expression in Huh7 and JHH7 cells and the FGF19 concentration in supernatants of these cells were significantly higher than those in HepG2 and PLC/ PRF/5 cells. Subsequent Western blotting demonstrated that FGFR4 was expressed in all cell lines examined (Fig. 1C). Taken together, FGF19 production was concomitant with FGFR4 expression in Huh7 and JHH7 cells (Fig. 1D).
Loss-of-function assays of FGFR4 in HCC cells. We next conducted FGFR4-knockdown assay using short hairpin RNA (shRNA) for FGFR4. shRNA targeting luciferase (Luc) was used as a control. Two different shRNAs, sh-FGFR4-1 and sh-FGFR4-2, markedly repressed FGFR4 protein expression (Fig. 1E). Although both shRNAs markedly repressed the growth of all cell lines examined, this repression was more evident in FGF19 + / FGFR4 + cell lines (Huh7 and JHH7 cells, Fig. 1F). We then examined the effects of recombinant FGF19 treatment in control and FGFR4-knockdown HCC cells. FGF19 altered proliferation in dose-dependent manner in control cells ( Fig. 2A). As expected, FGF19-induced cell growth was lost in FGFR4-knockdown HCC cells. These results suggested that FGF19/FGFR4 signaling is closely associated with the cell growth activity of HCC cells.
Next, the expression of the adapter protein of FGFR4 and downstream target of FGFR4 were examined in Huh7 and JHH7 cells treated with recombinant FGF19 and/or BLU-9931, a selective FGFR4 inhibitor. Western blotting analyses using phospho-FRS2α (pFRS2α) and phospho-ERK (pERK) antibodies revealed that phosphorylation of both proteins was prominently suppressed by a selective FGFR4 inhibitor in not only parental cells but also recombinant FGF19-treated cells (Fig. 2B).

Combined use of a selective FGFR4 inhibitor and MKIs.
To examine the effects of MKIs on pFRS2α and pERK levels, Western blotting was performed in HCC cells treated with MKIs at varying concentrations ( Fig. 3A). Decreases in pFRS2α and pERK levels were observed in HCC cells treated with MKIs in a dosedependent manner. Although 10 μM sorafenib and regorafenib treatment resulted in FGFR4 signaling inhibition, only 1 μM lenvatinib was sufficient. Cotreatment of a selective FGFR4 inhibitor (1 μM) and sorafenib or regorafenib successfully inhibited pFRS2α and pERK expression (Fig. 3B). Next, we examined the effects of the combined use of a selective FGFR4 inhibitor and MKIs (sorafenib or regorafenib) on cell growth activity. Additional treatment of a selective FGFR4 inhibitor enhanced cell growth inhibition followed by sorafenib or regorafenib treatment (Fig. 3C). Of note, synergistic effects of a selective FGFR4 inhibitor and sorafenib or regorafenib on the proliferation of HCC cells were confirmed by calculating the combination index (CI < 1).

MKI and/or selective FGFR4 inhibitor treatment in a xenograft transplantation model.
To examine the anti-tumor effects of MKIs and a selective FGFR4 inhibitor in vivo, we conducted xenograft transplantation using Non-obese diabetic/severe combined immunodeficiency (NOD/SCID) mice. Sorafenib (10 mg/Kg) and/or a selective FGFR4 inhibitor (30 mg/Kg) was administered daily after the subcutaneous tumors reached 10 mm in diameter. Although tumor growth was obviously inhibited by the sorafenib and selective FGFR4 inhibitor treatment, it was further suppressed by the co-treatment of sorafenib and a selective FGFR4 inhibitor (Fig. 4A). Although the tumor volume in each group was not significantly different, there was a stronger tendency for tumor regression in the combination group than in the single agent group and in control group.
Next, subcutaneous tumors were subjected to immunohistochemical analyses. Of note, phosphorylation of ERK was inhibited by the administration of a FGFR4 inhibitor. Ki-67 and caspase3 immunostaining demonstrated that the co-treatment of sorafenib and selective FGFR4 inhibitor efficiently inhibited cell proliferation and induced apoptosis (Fig. 4B).
Focusing on the co-expression of FGF19/FGFR4, we compared the patient characteristics between FGF19 + FGFR4 + HCC (n = 33) and FGF19 − FGFR4 − HCC (n = 39) (Fig. 5C). Clinicopathological examination demonstrated that FGF19 + FGFR4 + HCC was significantly associated with a high serum level of AFP (p = 0.001) and low differentiation pathology (p = 0.003). In addition, they were subjected to the prognostic analyses by the Kaplan-Meier method. The presence of vascular invasion or tumor satellite nodules are well known factors implicated in recurrence 22 . In fact, the recurrence-free survival (RFS) in patients with vascular invasion or tumor   www.nature.com/scientificreports/ no statistically significant difference between FGF19 + FGFR4 + HCC and FGF19 − FGFR4 − HCC groups regarding vascular invasion or tumor satellite nodules (p = 0.134). Although there were no significant differences, the RFS in patients with FGF19 + FGFR4 + HCC was slighter shorter than that in patients with FGF19 − FGFR4 − HCC (p = 0.086, Fig. 5D).
Potential of serum FGF19 as a marker for MKI treatment effects. Lastly, the prognosis of HCC patients treated using sorafenib or lenvatinib was assessed by the Kaplan-Meier method based on serum FGF19 levels before MKI treatment. The cut-off value of serum FGF19 (200 pg/mL) was used to divide HCC cases into 2 groups (FGF19 low and FGF19 high groups) according to our previous report 23 . There was no clinical feature difference between serum FGF19 high and FGF19 low patients treated using sorafenib and lenvatinib (Supplementary  Tables S2 and S3). In sorafenib-administered cases, FGF19 low patients exhibited a significantly longer progression-free survival (PFS) and overall survival (OS) than FGF19 high patients (Fig. 6A, B). The median PFS in the www.nature.com/scientificreports/ FGF19 low and FGF19 high groups was 139 days and 86 days, respectively. Similarly, the median OS was 494 days and 353 days, respectively. In contrast, the PFS and OS of FGF19 high patients were modestly longer than those of FGF19 low patients in lenvatinib-administered cases, but there were no significant differences between two groups. The median PFS in the FGF19 low and FGF19 high groups was 90 days and 195 days, respectively. Similarly, the median OS was 464 days and 733 days, respectively (Fig. 6C, D).

Discussion
Although driver gene mutations, such as in the TERT promoter, beta-catenin, and Tp53, are often observed in HCC, the frequency of mutations in other genes is not high 24,25 . As druggable gain-of-function mutations, such as in the epidermal growth factor receptor (EGFR) gene, are seldom detected, no therapeutic agents targeting a single gene mutation have been clinically applied 26 . Considering the genetic characteristics of HCC mentioned above, MKIs, inhibitors of multiple tyrosine kinase receptors, may be reasonable for HCC treatment. FGFR4 is one of the important targets for MKIs. FGF19/FGFR4 signaling functions as an essential cell growth signal in many solid tumors 27,28 . Together, inhibition of FGF19/FGFR4 signaling may play a role in the anti-tumor effects of MKIs against HCC. In this study, we investigated the impact of the regulation of FGF19/FGFR4 signaling on treatment using MKIs, namely sorafenib, regorafenib and lenvatinib.
To gain insight into the role of FGFR4, we conducted the loss-of-function assay of FGFR4 in culture. FGFR4knockdown markedly suppressed cell growth in Huh7 and JHH7 cells (FGF19 + FGFR4 + HCC cells), but not in HepG2 and PLC/PRF/5 cells (FGF19 − FGFR4 + cells). In addition, recombinant FGF19 treatment increased cell proliferation in Huh7, JHH7, HepG2 and PLC/PRF/5 cells (FGF19 + FGFR4 + HCC cells and FGF19 − FGFR4 + cells). These results suggested that FGF19/FGFR4 signaling activity directly affects the proliferation of HCC cells. We next examined the effects of MKIs on FGFR4-mediated ERK signaling by Western blotting. Of note, lenvatinib www.nature.com/scientificreports/ markedly reduced pFRS2 and pERK levels compared with sorafenib and regorafenib. Furthermore, co-treatment of sorafenib or regorafenib and a selective FGFR4 inhibitor, BLU-9931 caused synergistic cell growth inhibition accompanied by reduced pFRS2 and pERK. Taken together, growth inhibition following lenvatinib treatment may be due to suppression of the FGFR4-mediated RAS/RAF/MAPK pathway. Our analyses revealed that the combined use of sorafenib or regorafenib with a selective FGFR4 inhibitor suppressed FGFR4-mediated ERK signaling, thereby producing additional anti-tumor effects. A selective FGFR4 inhibitor, BLU-554, was previously applied in clinical trials with favorable outcomes in some HCC patients 29 . Pan-FGFR inhibitors have also been recently developed and applied in clinical studies on different cancer types [30][31][32] . As FGFR1 activation also function in sorafenib resistance 33 , co-treatment of sorafenib or regorafenib and pan-FGFR inhibitors may be useful.
FGF19 gene amplification was reported in 10-20% of HCC cases. However, in the present analyses utilizing TCGA data, FGF19 amplification was not correlated with its RNA expression. Few genes with an increased copy number have an increased protein level 34 . Our clinicopathological examination revealed that FGF19 and FGFR4 were concomitantly expressed in approximately one-third of the HCC tissue samples. FGF19 + FGFR4 + HCC is associated with biological malignancy, represented by low differentiation pathology and unfavorable prognosis. Together, these results suggest that FGF19/FGFR4 signaling is activated in some HCC cases and this may be closely related to progression.
Biomarkers are utilized in many aspects of medical treatment for cancer patients such as early diagnosis, recurrence diagnosis and prediction of prognosis 35 . Among them, predictive biomarkers for treatment outcome are important in HCC patients. Ideally, they should be functional, and evaluated before drug selection and administration 36 . As FGF19 and FGFR4 exhibited similar expression pattern in HCC tissues, we considered the possibility that measurement of autocrine secretion of FGF19 in HCC patients sera reflects the dependency upon FGF19/FGFR4 signaling. As expected, high serum FGF19 levels were associated with an unfavorable OS and PFS in patients treated using sorafenib, but not in patients treated using lenvatinib. Taking into consideration the marked inhibitory effects of lenvatinib on FGF19/FGFR4 signaling in culture compared with sorafenib, these findings are reasonable.
HCC is one of the typical hypervascular tumors, and angiogenesis plays an essential role in its development and progression 37 . MKIs inhibit both cell growth and tumor angiogenesis through the suppression of angiogenic receptor tyrosine kinases such as FGFR, vascular endothelial growth factor (VEGF) receptor (VEGFR) and platelet-derived growth factor receptor 38 . Lenvatinib has been reported to have excellent anti-tumor and anti-angiogenic effects in VEGF-overexpressing HCC xenograft models 39 . Further analyses are necessary to clarify whether these angiogenic markers or their combined use with FGF19 can also function as the markers for lenvatinib treatment response.
Our study demonstrated that robust inhibition of FGF19/FGFR4 is of importance for the exertion of antitumor effects by MKIs. In addition, serum FGF19 levels may act as a predictive marker for drug response and survival in HCC patients treated using MKIs.

Data collection and analysis from TCGA-Liver Hepatocellular Carcinoma (LIHC). RNA sequenc-
ing dataset (ID: TCGA.LIHC.sampleMap/HiSeqV2) and whole genome microarray dataset (ID: TCGA.LIHC. sampleMap/Gistic2_CopyNumber_Gistic2_all_thresholded.by_genes) were downloaded using the UCSC Xena Browser (https ://xenab rowse r.net/). In those datasets, the two classes of phenotypes were named "primary tumor" and "solid tissue normal", only those samples in the clinical category of "primary tumor" were for this study. RNA sequencing dataset shows the gene-level transcription estimates, as in log2(x + 1) transformed RSEM normalized count. Whole genome microarray dataset shows the gene-level copy number estimates values, as in − 2, − 1, 0, 1, 2 using GISTIC2 method. We defined − 2, − 1 and 0 as amplification negative and 1 and 2 as amplification positive. Subsequently, Both FGF19 expression and its amplification in HCC were subjected to secondary analyses.
Statistical analysis. Data are presented as the mean with standard deviation (SD), or the median with minimum to maximum and interquartile range (IQR). Statistical differences in the quantitative values between groups were determined by either Student's t test or the Mann-Whitney U test. Chi-square test was used for categorical values. The relationship between FGF19 and FGFR4 expression based on immunohistochemical staining was analyzed by the Kendall rank correlation coefficient. The log-rank test was used to analyze survival data. The level of significance was set as p < 0.05. All statistical analyses were performed using the SPSS statistical software version 24 (IBM, Chicago, IL, USA). www.nature.com/scientificreports/