FGF19 protects hepatocellular carcinoma cells against endoplasmic reticulum stress via activation of FGFR4-GSK3β-Nrf2 signaling

ABSTRACT: The tumor microenvironment induces endoplasmic reticulum (ER) stress in tumor cells, an event that can promote progression, but it is unknown how tumor cells adapt to this stress. In this study, we show that the fibroblast growth factor FGF19, a gene frequently amplified in hepatocellular carcinoma (HCC), facilitates a survival response to ER stress. Levels of FGF19 expression were increased in stressed HCC cells in culture and in a mouse xenograft model. Induction of ER stress required the transcription factor ATF4, which directly bound the FGF19 promoter. In cells where ER stress was induced, FGF19 overexpression promoted HCC cell survival and increased resistance to apoptosis, whereas FGF19 silencing counteracted these effects. Mechanistic investigations implicated glycogen synthase kinase-3β in regulating nuclear accumulation of the stress-regulated transcription factor Nrf2 activated by FGF19. Our findings show how FGF19 provides a cytoprotective role against ER stress by activating a FGFR4-GSK3β-Nrf2 signaling cascade, with implications for targeting this signaling node as a candidate therapeutic regimen for HCC management.

INTRODUCTION
Fibroblast growth factor 19 (FGF19), a member of the hormone-like FGF family, has activity as an ileum-derived postprandial enterokine regulating liver metabolism (1). FGF19 predominantly signals through FGF receptor 4 (FGFR4) in the presence of coreceptor β-Klotho to efficiently activate downstream signaling pathways (2). In humans, increased expression of FGF19 is critical for the development and progression of several cancer types including hepatocellular carcinoma (HCC), breast cancer, prostate cancer, thyroid cancer and cholangiocarcinoma (3-6). FGF19 was recently identified as a driver oncogene in HCC, a cancer known for its high incidence in cancer-related deaths (3,7). FGF19 regulates a variety of hepatocyte cellular functions, such as bile acid metabolism, proliferation, differentiation, and epithelial-mesenchymal transition (EMT) in a FGFR4-dependent manner (8-10). In mice, aberrant signaling through Fgf15, the mouse orthologue to human FGF19, has been implicated in HCC development in a context of chronic liver injury and fibrosis (11). Elevated endoplasmic reticulum (ER) stress is observed in solid tumors as a consequence of microenvironment changes (12). When ER stress occurs, ER functions are altered, and a number of molecular actions collectively identified as the “unfolded protein response” (UPR) are activated to counteract ER-associated damage (13). Several indications suggest that cancer cell adaptation to adverse conditions largely relies on the ability of a cell to perturb ER stress-associated regulatory networks and prevent ER stress-induced cell death (14,15).

It has been reported that either mRNA or secreted protein levels of FGF19 was increased in intestinal epithelial Caco-2 cells following treatment of ER inducers Thapsigargin (TG) or Tunicamycin (TM) (16). However, the molecular mechanisms underlying upregulation of FGF19 under ER stress have not been explored in detail. Nuclear factor E2-related factor 2 (Nrf2), a key mediator involved in anti-oxidative, anti-inflammatory, and mitochondrial protection, can activate an array of genes encoding antioxidant and detoxification enzymes through selective transcriptional binding within antioxidant response elements (ARE) (17). Nrf2 regulatory signals serve protective roles in hepatic inflammation, fibrosis, and regeneration (18). Abundant expression of Nrf2 has been identified in different cancer types including HCC, which may associate with cancer cell proliferation, invasion, and chemoresistance (19,20). The self-protective, antioxidant response of Nrf2 is a complex and highly orchestrated pathophysiological process (18,21), and glycogen synthase kinase 3β (GSK3β) serves as an integration point of a myriad of signaling pathways during the control of Nrf2 activity (22,23). Recent evidence in HCC cells indicates that GSK3β can physically interact with Nrf2 and inhibit Nrf2 activity by nuclear export and subsequent proteasomal degradation in response to late-phase oxidative stress (24). Our previous study has demonstrated that the FGF19/FGFR4 axis inactivates GSK3β by increasing phosphorylation levels of GSK3β in HCC cells (10). Thus, such coordinating events lead us to determine whether a functional relationship between FGF19 and Nrf2 takes place in the tumor microenvironment, specifically within the context of elevated ER stress.

Here we show for the first time that FGF19 is translationally activated by ER stress both in cultured HCC cells and in a mouse-xenograft model. The ER stress-associated activation of FGF19 transcription requires activating transcription factor 4 (ATF4), which increasingly binds to the amino-acid-response element (AARE) within the FGF19 promoter during ER stress. As well, we demonstrate that depletion of FGF19 decreases HCC cell survival and enhances ROS-associated apoptosis upon ER stress. Additionally, FGF19 facilitates the nuclear accumulation of Nrf2 through the FGFR4-GSK3β signaling, which plays a central role in providing FGF19-dependent cytoprotection to HCC cells against ER stress. Thus, by taking into account its cytoprotective role in the tumor microenvironment, FGF19 may represent a promising molecular target in which its inhibition could be used therapeutically against HCC development. HL7702, HepG2, and Hep3B cell lines were obtained from the American Type Culture Collection (ATCC). MHCC97L and MHCC97H cell lines were purchased from the Cell Bank of Chinese Academy of Sciences (CBCAS). All cell lines were maintained in the growth medium according to the manufacturer’s instructions and passage <5 were used in this study. Real-time reverse transcription polymerase chain reaction (RT-PCR), Western blot, transient transfections, lentiviral transduction, and luciferase reporter assays were carried out as described previously (10,25-29). The gene-specific primers used in real-time RT-PCR analysis are listed in Table S1.

All experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) at Augusta University. Six-week-old BALB/c athymic nude mice were inoculated subcutaneously in the double hind flanks with 5×106 HCC cells per 100 µl suspended in dilute Matrigel 1:1 in ice cold phosphate-buffered saline (PBS). Dimethyl sulfoxide (DMSO) or 10 μM TG was intratumorally injected (left for DMSO and right for TG) when the xenografts reached ~150 mm3. The xenografts were harvested at 8 hours post-injection (hpi) and homogenized for real-time RT-PCR analysis. The full-length cDNA of human FGF19, FGFR4, and ATF4 were individually cloned into a pcDNA3.1 (+) expression vector (Life Technologies, Carlsbad, CA). The ARE-luciferase reporter containing three copies of ARE sequences (in head-to-tail orientation) were generated using the pGL3 vector (Promega, Madison, WI,
5 USA). Twenty-one base pair siRNA duplexes targeting human Nrf2 gene (5’-GUAAGAAGCCAGAUGUUAATT-3’), ATF4 gene (5’-GCCUAGGUCUCUUAGAUGATT-3’) and a standard control (Dharmacon siCONTROL nontargeting siRNA) were synthesized by Dharmacon RNA Technologies (Lafayette, CO). FGF19, ATF4 and β-Actin antibodies were purchased from Abcam (Cambridge, MA). Antibodies against Nrf2, p-GSK3β (Ser9), GRP78，c-PARP, Lami B, HO-1 and Bcl-2 were procured from Cell Signaling Technologies (Beverly, MA). GSK3β inhibitors TWS119 and Tideglusib, PI3K inhibitor LY294002, JNK inhibitor SP600125, and MEK inhibitor U0126 were obtained from Selleckchem (Houston, TX). TG and TM, acetaminophen, silymarin, superoxide dismutase (SOD), ethanol, H2O2, DMSO and DAPI were purchased from Sigma-Aldrich (St. Louis, MO). NE-PER® Nuclear and Cytoplasmic Extraction Reagent Kit was purchased from Thermo Scientific (Waltham, MA) and used according to the manufacturer's instructions.

Cell proliferation and viability at the indicated incubation times were determined by CellTiter 96® AQueous One Solution Cell Proliferation Kit and CellTiter-Glo® Luminescent Cell Viability Kit (Promega, Madison, WI), respectively. Apoptosis was determined by the FITC Annexin V/Dead Cell Apoptosis Kit with FITC annexin V and PI (Life Technologies, Carlsbad, CA) and analyzed by flow cytometry using the BD FACSCalibur system (BD Biosciences, Franklin Lakes, NJ). For DAPI staining, cells were stained with 0.5 g/ml DAPI for 1 hour after fixation with 4 % paraformaldehyde. The stained cells were washed with PBS and examined by fluorescence microscopy (Olympus Corporation, Tokyo, Japan). Apoptotic cells were identified by condensation and fragments of nuclei. The apoptotic cells in seven randomly selected fields from triplicate chambers were counted in each experiment under a horizon.
Electrochemical detection of free radical levels released from cells was established in our previous work (10,25,26,30). Briefly, 5×105 HCC cells were seeded in 6-well plates in the presence or absence of various ER stress inducers. Cyclic voltammetry (CV) was used to monitor cellular superoxide (O2•−) generation on CHI760D electrochemical station (ChenHua Instruments, Wuhan, China).

SOD was applied to verify current changes caused by O2•−. Percentages of peak (potential=0.7V; current enhancement) were compared and calculated against the control curve. ChIP assays were performed using a ChIP Kit (Abcam, Cambridge, MA) as we previously described. Briefly, chromatin from cells was crosslinked with 1% formaldehyde for 10 min at room temperature, sheared to an average size of 500bp, and immunoprecipitated with an ATF4 antibody (Abcam). The ChIP-qPCR primers (Forward: 5’ GGAGGCTTGATGCAATCCCGATAA-3’; Reverse: 5’-GGCAGCCTTATATAGTAGCGCCTT-3’) were designed to amplify a proximal promoter region containing the putative ATF4 binding element AARE (5’-CTTGATGCA-3’) at the FGF19 promoter. Each immunoprecipitated DNA sample was quantified using qPCR and all ChIP-qPCR signals were normalized to the input. Statistical analyses were performed with unpaired Student’s t test for two group comparisons and one-way analysis of variance (ANOVA) for multi-group comparisons. Data were presented as mean ± S.E.M of at least three independent experiments. Gene expression data were obtained from GEO database and the Pearson’s correlation coefficient was used to analyze the correlation between the expression levels of two genes. A p-value of 0.05 or less was considered to be significant.

RESULTS
FGF19 is highly expressed in HCC, and the UPR is always activated in HCC, raising the possibility that ER stress-dependent alteration in FGF19 expression is the one of the main mechanisms that underlie liver tumorigenesis. We first tested whether ER stress influenced FGF19 expression in HCC cell lines. HepG2, Hep3B, MHCC97L, and MHCC97H cells were treated with ER stress inducers TG and TM, respectively. Both TG and TM treatment led to a significant increase in FGF19 expression levels in all cell lines examined (Figure 1A and 1B), concomitant with increasing levels of the ER stress marker GRP78 (Figure 1B). The expression of FGF19 in normal human HL7702 hepatocytes was also increased following TG treatment (Figure 1A and 1B). No significant changes of FGFR4 levels were observed in the presence or absence of ER stress (Figure 1B). We next examined the expression status of FGF19 using different triggers of ER stress. Consistent with TG treatment, H2O2, ethanol, and acetaminophen were also able to induce FGF19 upregulation in HepG2 and MHCC97L cells in a dose-dependent manner (Figure 1C). To determine whether elevated FGF19 expression was associated with ER stress in vivo, we developed HepG2 and MHCC97L xenografts in nude mice and then induced ER stress through intratumoral injection of TG. TG treatment led to a significant increase (p=0.007) in FGF19 expression levels in the isolated tumor cells compared with vehicle treatment (Figure 1D). Silymarin, a mix of natural flavonoids, exhibits potent antioxidant function in repressing ER stress (31).

We treated MHCC97L cells with silymarin in the presence of TG or TM for 24 hours. Decreased expression levels of FGF19 were found in our co-treatments, as compared to those with TG or TM single drug treatment (Figure 1E and 1F). These results suggest that blocking cell oxidation abrogates FGF19 upregulation in ER stress. The mouse Fgf15 gene is homologous with the human FGF19 gene and is not expressed in normal mouse liver (11). No Fgf15 expression was detected out in the mouse liver tissues treated with TG, CCl4 or acetaminophen, although Grp78 was markedly upregulated after treatment (Supplementary Figure S1). ATF4 is a critical ER stress-inducible transcription factor (27) implicated as a regulator of ER stress-induced FGF19 upregulation in intestinal epithelial cells (32). Increased ATF4 expression levels were observed in HepG2 and MHCC97L cells following TG exposure (Figure 2A and 2B), suggesting that ATF4 may contribute to an ER stress-induced increase in FGF19 mRNA in HCC cells. To further address this observation, we then knocked down ATF4 using siRNA in HepG2 cells and determined FGF19 expression levels with or without TG treatment. Loss of ATF4 expression attenuated FGF19 upregulation induced by TG (Figure 2C and 2D). In contrast, overexpression of ATF4 upregulated FGF19 expression in MHCC97L cells (Figure 2E). To further determine whether ATF4 transcriptionally regulates FGF19, we performed ChIP assays using a ChIP-grade ATF4 antibody, which demonstrated a specific, direct interaction of the ATF4 protein with the AARE within the FGF19 promoter in MHCC97L cells (Figure 2F). As expected, increased levels of ATF4 at the FGF19 promoter binding site were seen following TG treatment (Figure 2F). Consistently, analysis of microarray datasets from HCC patients revealed a clear positive correlation between ATF4 and FGF19 levels (Supplementary Figure S2). Taken together, these results indicate that ATF4 is upregulated during ER stress, which in turn, transcriptionally activates FGF19 in HCC cells.

We next determined the potential role of FGF19 in ER stress. Forced expression of FGF19 enhanced cell survival and inhibited apoptosis in TG-treated MHCC97L cells (Figure 3A and 3B). Given that excessive reactive free radicals (ROS), such as O2•−, are released to activate stress-response pathways and induce defense mechanisms (33), we developed sensitive electrochemical biosensors to monitor O2•−generation in cancer cells in situ (Supplementary Figure S3). Intracellular O2•− was increased in TG-treated MHCC97L cells (Figure 3C), and this increase was abrogated when FGF19 was overexpressed (Figure 3C), suggesting that FGF19-mediated induction of O2•− levels may contribute to ER stress-induced consequences in HCC cells. SOD alternately catalyzes the dismutation (or partitioning) of O2•− into either O2 or H2O2 (34). In these assays, SOD treatment abrogated current changes trigged by TG (Figure 3C), demonstrating the specificity of the electrochemical biosensors. In contrast, knockdown of FGF19 led to a dramatic decrease in survival rate and an increase in apoptosis in TG-treated MHCC97H cells compared with siRNA no-target control (siNC) treated cells (Figure 3D and 3E). Electrochemical analysis showed that FGF19 depletion enhanced O2•− release following TG treatment (Figure 3F). These findings support a notion that high expression levels of FGF19 help to defend HCC cells against ER stress. A similar tendency was observed in H2O2-treated HCC cells (Supplementary Figure S4), indicating that FGF19 plays a pivotal role in ER stress response.

The FGF19-FGFR4 signaling has been implicated in the development and progression of HCC (2,35). Our previous studies have shown that overexpression of FGF19 in HCC cells enriched secreted FGF19 levels in the supernatant to promote cell growth and migration through activation of FGFR4 (10,25). We have generated FGFR4 knockout MHCC97L cells using a CRISPR-Cas9 system (10). Using these cells, we determined the possible involvement of FGFR4 in ER stress. In the TG treatment, the phenotypes shown in FGF19 knockdown cells (Figure 3), such as decreased cell survival, increased capacity of O2•− release, and enhanced cell apoptosis, were also observed in FGFR4 knockout cells (Figure 4A-4C). To evaluate whether FGF19-FGFR4 axis contributes to the resistance of HCC cells to ER stress, we overexpressed the FGF19 gene in FGFR4 knockout MHCC97L cells before TG treatment. As expected, increased FGF19 expression could not protect HCC cells against ER stress when FGFR4 was depleted (Figure 4D-4F), demonstrating that FGFR4 is a required component during this process. Together, these observations reveal that ER stress can be buffered by mitigating the activation of the FGF19-FGFR4 axis in HCC cells.

Nrf2 is a transcription factor known to regulate expression of a large array of antioxidant and cytoprotective genes to mediate cellular stress responses (36). Since FGF19 impacts the protection of HCC cells against ER stress, we examined the involvement of Nrf2 in the FGF19 regulatory network. FGF19 overexpression resulted in a significant increase in Nrf2 protein levels in HepG2 and MHCC97L cells, while FGF19 knockdown reduced Nrf2 levels in MHCC97H cells (Figure 5A). The anti-apoptosis gene Bcl-2 and anti-oxidant gene HO-1 are downstream targets of Nrf2 (37). In the presence of TG, both genes showed higher expression levels in FGF19 overexpressing cells and lower expression levels in FGF19 knockdown cells as compared to the respective controls (Figure 5A). FGFR4 knockout attenuated Nrf2-dependent expression of Bcl-2 and HO-1 in FGF19 overexpressing MHCC97L cells (Figure 5B), indicating that Nrf2 pathway is heavily dependent on FGF19/FGFR4 activation in HCC cells. Surprisingly, mRNA levels of Nrf2 did not change in HCC cells regardless of FGF19 expression levels (Figure 5A), which prompted us to investigate whether upregulation of FGF19 facilitated nuclear accumulation of Nrf2. Compared with control cells, enriched Nrf2 protein levels were found in the purified nuclear fraction in FGF19 overexpressing MHCC97L cells with concomitant reduction of Nrf2 in cytosol (Figure 5C).

To further explore that increased FGF19 activates Nrf2-mediated activation of ARE-dependent genes, we constructed a luciferase reporter containing a triplicate of ARE sequence motifs (Figure 5D). Co-transfection of full-length FGF19 cDNA with the ARE reporter into MHCC97L cells led to increased luciferase activities (Figure 5D). Similar results were also observed when the full-length FGFR4 cDNA and the ARE luciferase reporter were co-transfected (Figure 5D). To determine the potential clinical relevance of these genes in HCC patients, we assessed GEO datasets and investigated the association between the expression levels of FGF19 and HO-1 or Bcl-2 in HCC clinical samples. This analysis showed a positive linear association either between FGF19 and HO-1 transcript levels as well as between FGF19 and Bcl-2 (Supplementary Figure S2). To elucidate the molecular mechanisms responsible for FGF19-induced Nrf2 accumulation, the status of Keap1 and GSK3β were determined, two critical regulators of Nrf2. As shown in Figure 5A, enhanced phospho-GSK3β levels, but not Keap1 levels, were found in FGF19-overexpressing HCC cells, demonstrating that FGF19 facilitates Nrf2 nuclear accumulation via inactivation of GSK3β by phosphorylation at specific residues. We thus inactivated GSK3β activity using GSK3β inhibitors TW119 and Tideglusib and observed attenuated FGF19-dependent Nrf2 accumulation (Figure 5E). Unlike the GSK3β inhibitors, however, treatment with PI3K, ERK and JNK inhibitors did not affect luciferase activities of ARE reporter in MHCC97L cells where FGF19 was overexpressed (Figure 5F). Taken together, these results demonstrate that GSK3β contributes to Nrf2 nuclear accumulation triggered by ER stress-induced FGF19-FGFR4 activation in HCC cells.

Persistent accumulation of Nrf2 in the nucleus has been implicated in enhancing oncogenesis (38). Since high expression of FGF19 promotes nuclear translocation of Nrf2, we depleted Nrf2 in FGF19 overexpressing MHCC97L cells to examine the functional role of Nrf2 in FGF19-dependent cell survival. Knockdown of Nrf2 using siRNA remarkably attenuated the proliferation rate induced by FGF19 overexpression (Supplementary Figure S5). Most importantly, in the treatment of TG, Nrf2 knockdown blunted the increase in FGF19-dependent cell survival (Figure 6A) and disrupted the ability to remove O2•− (Figure 6B). The same phenotypes were also observed in the presence of H2O2 (Supplementary Figure S6). Furthermore, overexpression of FGF19 enhanced resistance to apoptotic cell death triggered by TG, but this was blunted when Nrf2 was depleted in these cells (Figure 6C and 6D). Loss of Nrf2 expression also attenuated the FGF19-dependent induction of HO-1 and Bcl-2 in TG-treated MHCC97L cells (Figure 6E). Nrf2 is positioned at the interface of oxidative and ER stress pathways (38). To determine whether FGF19 promotes Nrf2 to translocate into nucleus during ER stress, HepG2 or MHCC97H cells were treated with TG in the presence or absence of FGF19 expression to determine the accumulation status of Nrf2 within the nucleus. Based on our results, TG-induced FGF19 upregulation was associated with increased nuclear accumulation of Nrf2, whereas Nrf2 levels in the nucleus were reduced to almost beyond detection when FGF19 was depleted (Figure 6F). These observations indicate that FGF19 plays a cytoprotection role through activation of Nrf2 in HCC cells when under ER stress.

DISCUSSION
ER stress has been implicated in the pathogenesis of viral hepatitis, insulin resistance, hepatosteatosis, nonalcoholic steatohepatitis, and other liver diseases that increase risk of HCC (39,40). ER stress is also present in HCC when the tumor microenvironment experiences nutrient deprivation, hypoxia, and imbalance between production and removal of ROS (41,42). However, the mechanisms underlying cellular adaptation to ER stress during development and progression of HCC are still unclear. Our study showed that ER stress promoted ATF4-mediated transcriptional activation of FGF19 in HCC cells and consequently enhanced the prosurvival and anti-apoptotic response of HCC cells during ER stress (Figure 7). These findings demonstrate that FGF19 is involved in tumorigenesis and adaptation to extreme environments, which may represent a legitimate target for cancer therapeutics. In this work, we revealed a novel functional role of FGF19 in promoting HCC cell survival under acute ER stress. FGF19 secreted from HCC cells functions as endocrine or paracrine factors to activate the FGFR4 receptor residing on the surface of HCC cells to increase GSK3β phosphorylation and, in turn, promote Nrf2 translocation to the nucleus (Figure 7). Nuclear accumulation of Nrf2 transcriptionally activates target genes such as HO-1 and Bcl-2 in HCC cells, leading to a significant decrease in ROS-associated apoptosis induced by ER stress (Figure 7). Thus, it appears that HCC cells create a dependence on the FGF19 stress response pathway to maintain a permissive growth environment.

Therefore, inhibition of the FGF19-FGFR4 axis may prevent tumor development of apoptosis is a prominent mode of cytotoxicity for many chemotherapeutic drugs. Apoptosis occurs by distinct pathways that involve the cell surface, mitochondria, or ER. Certain drugs, such as sorafenib, the first-line systemic therapy for advanced HCC, can induce ER stress-mediated apoptosis independent of the MEK-ERK pathway (43,44). Our group has recently demonstrated that hyperactivation of FGF19-FGFR4 axis is one of the main mechanisms of sorafenib resistance in HCC cells, and blocking it can overcome the resistance to sorafenib (25). The present study implies that FGF19-FGFR4-GSK3β-Nrf2 signaling cascades may be a main pathway involved in sorafenib-induced ER stress. Either FGF19 or FGFR4 represents a promising target in developing novel cancer therapies against FGF19-FGFR4-dependent cancers. FGF19 is an endocrine hormone of the FGF family that regulates bile acid, carbohydrate, lipid, and energy metabolism (10,45). Although a neutralizing antibody specific against FGF19 (1A6) has been developed to block the binding of FGF19 to FGFR4 and neutralize the tumorigenic oncogenic activity of FGF19 in HCC cells, it has an unexpected effect in preventing the beneficial functions of FGF19 in bile acid homeostasis (4,45).

To achieve successful therapy, M70, an engineered and nontumorigenic FGF19 variant carrying 3 amino acid substitutions (A30S, G31S, and H33L) and a 5 amino acid deletion, has been developed (45,46). Unlike 1A6, M70 can effectively block the tumorigenic effects associated with wild-type FGF19, without compromising its role in bile acid homeostasis (45,46). Nevertheless, the other encompassing for disrupting FGF19-FGFR4 interaction is by specifically targeting FGFR4 activation. BLU9931, the first highly selective anti-FGFR4 Roblitinib inhibitor, exhibits strong ability in suppressing FGFR4 activity and decreasing tumor size in HCC xenograft models that maintain an intact FGFR4 signaling pathway (47).

In summary, we present evidence that ER stress-induced FGF19 upregulation and FGFR4 activation facilitates HCC cell survival and anti-apoptosis via the FGFR4-GSK3β-Nrf2 signaling cascade. Thereby, our findings make FGF19 an interesting, emerging molecular target for potential therapeutic intervention for patients with HCC.