Hepatocellular carcinoma (HCC), one of the leading causes of cancer-related death, develops from premalignant lesions in chronically damaged livers. Although it is well established that FGF19 acts through the receptor complex FGFR4-β-Klotho (KLB) to regulate bile acid metabolism, FGF19 is also implicated in the development of HCC. In humans, FGF19 is amplified in HCC and its expression is induced in the liver under cholestatic and cirrhotic conditions. In mice, ectopic overexpression of FGF19 drives HCC development in a process that requires FGFR4. In this study, we describe an engineered FGF19 (M70) that fully retains bile acid regulatory activity but does not promote HCC formation, demonstrating that regulating bile acid metabolism is distinct and separable from tumor-promoting activity. Mechanistically, we show that FGF19 stimulates tumor progression by activating the STAT3 pathway, an activity eliminated by M70. Furthermore, M70 inhibits FGF19-dependent tumor growth in a rodent model. Our results suggest that selectively targeting the FGF19–FGFR4 pathway may offer a tractable approach to improve the treatment of chronic liver disease and cancer. Cancer Res; 74(12); 3306–16. ©2014 AACR.
FGF19 (also called FGF15 in rodents) is an endocrine hormone of the FGF family that regulates bile acid, carbohydrate, lipid, and energy metabolism (1). FGF19 selectively binds to FGFR4, which can be further enhanced by coreceptor KLB (2–5). The interaction between FGF19 and FGFR4 is crucial in repressing hepatic expression of cholesterol 7α-hydroxylase (CYP7A1), the first and rate-limiting enzyme in the conversion of cholesterol into bile acids (6–9). Interestingly, FGF19–FGFR4 signaling is also implicated in hepatocellular tumorigenesis. In humans, FGF19 is coamplified with cyclin D1 (CCND1) at approximately 15% frequency in hepatocellular carcinoma (HCC) on 11q13.3 (10). Clonal growth and tumorigenicity of HCC cells harboring the 11q13.3 amplicon can be inhibited by RNAi-mediated knockdown of FGF19, as well as an anti-FGF19 antibody (10). Transgenic mice with ectopic expression of FGF19 in skeletal muscle develop HCC at the age of 10- to 12-month-old (11). This tumorigenic activity is thought to be mediated by the liver-enriched FGFR4, because inactivation of FGFR4 via gene knockout or by a neutralizing antibody reduces tumor burden in FGF19 transgenic mice (12, 13). Therefore, targeting the FGF19–FGFR4 pathway has the potential for treating HCC. However, early efforts have encountered setbacks. Severe toxicity was observed in non-human primates by an anti–FGF19-neutralizing antibody due to on-target inhibition of endogenous FGF19, leading to dysregulation of bile acid metabolism (14). Hence, alternative approaches are needed to overcome these barriers to developing successful therapy.
In an effort to eliminate the tumorigenic activity of FGF19 without compromising its beneficial role in bile acid homeostasis, we established an in vivo liver tumorigenicity model in mice to evaluate FGF19-induced hepatocarcinogenicity. Using an adeno-associated virus (AAV)–mediated gene delivery approach (15), we introduced FGF19 transgene in mice and evaluated a panel of FGF19 variants in vivo to identify tumor-free variants. In this article, we show that one such variant, M70, fully retains the biologic activity of FGF19 in regulating bile acid homeostasis while devoid of tumorigenicity. Notably, we found that M70 binds to and activates FGFR4, which is assumed to mediate FGF19-associated tumorigenicity (13, 16). Mechanistically, we show that FGF19 stimulates tumor progression by activating the STAT3 pathway, an activity completely eliminated by M70. Furthermore, M70 inhibits FGF19-dependent tumor growth in a rodent model.
Materials and Methods
Human FGF19 (NM_005117), human FGFR4 (NM_022963), mouse FGFR4 (NM_008011), human KLB (NM_175737), and mouse KLB (NM_031180) cDNAs were purchased from Genecopoeia. Mutations were introduced in the FGF19 constructs using the QuickChange Site-Directed Mutagenesis Kit (Stratagene).
AAV293 cells (Agilent Technologies) were cultured in DMEM (Mediatech) supplemented with 10% FBS and 1× antibiotic–antimycotic solution (Mediatech). The cells were transfected with three plasmids [AAV transgene, pHelper (Agilent Technologies), and AAV2/9] for viral production. Viral particles were purified using a discontinued iodixanal (Sigma-Aldrich) gradient and resuspended in PBS with 10% glycerol and stored at −80°C. Viral titer or genome copy number was determined by qPCR.
All animal studies were approved by the Institutional Animal Care and Use Committee at NGM. Mice were housed in a pathogen-free animal facility at 22°C under controlled 12-hour light/12-hour dark cycle. All mice were kept on standard chow diet (Harlan Laboratories; Teklad 2918) and autoclaved water ad libitum. Male mice were used unless otherwise specified. C57BL/6J, FVB/NJ, BDF, ob/ob, and db/db mice were purchased from The Jackson Laboratory. Heterozygous rasH2 transgenic mice were obtained from Taconic. All animals received a single 200 μL intravenous injection of 3 × 1011 genome copies of AAV via tail vein on day 1. Animals were euthanized and livers were collected 24 or 52 weeks after dosing with AAV. Liver weight and liver tumor nodule numbers were recorded upon necropsy.
Histologic and immunohistochemical analysis
Formalin-fixed paraffin-embedded tissue sections were stained with hematoxylin and eosin (H&E) for histologic assessment at Nova Pathology. For immunohistology, anti-PCNA (Dako), anti-Ki67 (Dako), anti-glutamine synthetase (Thermo Fisher Scientific), or anti–β-catenin antibodies (Cell Signaling Techology) were used. Biotinylated secondary antibody, ABC-HRP reagent and DAB colorimetric peroxidase substrate (Vector Laboratories) were used for detection.
Measurement of serum FGF19 protein
Whole blood was collected from mouse tail snips into capillary tubes (Becton Dickenson). Levels of human FGF19 and variants were measured in serum using an ELISA assay (Biovendor). The assay recognizes both FGF19 and M70 in an indistinguishable manner.
Serum levels of liver enzymes (ALT, AST, and ALKP), triglycerides, total cholesterol, HDL-C and LDL-C were measured using enzymatic reactions on COBAS Integra 400 clinical analyzer (Roche Diagnostics). Concentrations of total bile acids in serum were determined using a 3α-hydroxysteroid dehydrogenase method (Diazyme).
Gene expression analysis
Total RNA was extracted from tissues or cells using the RNeasy Kit (Qiagen). qRT-PCR analysis was performed using QuantiTect multiplex qRT-PCR master mix (Qiagen) and premade primers and probes (Life Technologies). Reactions were performed in triplicates on an Applied Biosystems 7900HT Sequence Detection System. Relative mRNA levels were calculated by the comparative threshold cycle method using GAPDH as the internal standard.
Expression of recombinant proteins
FGF19 and M70 were produced in E. coli and purified as described with modification (17). Methionine was added to the mature peptides to facilitate expression in bacteria. Proteins were then purified to homogeneity via successive rounds of ion-exchange and hydrophobic interaction chromatography. Protein sequence was confirmed via LC/MS and monodispersity via SEC-HPLC (TOSOH TSKgelG3000).
Surface plasmon resonance assays
Surface plasmon resonance (SPR) experiments were performed on a Biacore T200 instrument at 25°C. For direct binding between FGFR4 and ligands, anti-Fc antibody was immobilized on a CM5 sensor chip using standard amine-coupling procedure and mouse FGFR4(ECD)-Fc (R&D Systems) was captured subsequently. KD values were determined using the Biacore T200 evaluation software V.1 using a 1:1 binding model.
Solid-phase binding assay
Ninety-six–well ELISA plates (Thermo Fisher Scientific) were coated with goat-anti-Fc, blocked with 3% BSA, and then incubated with 1 μg/mL mouse FGFR4(ECD)-Fc (R&D Systems) in PBS containing 3% BSA. The plates were incubated with various concentrations of FGF19 or M70 [in the presence of 20 μg/mL heparin (Sigma) and 1 μg/mL mouse KLB (R&D Systems)] in PBS/3% BSA. The bound proteins were detected using biotinylated FGF19-specific polyclonal antibody (R&D Systems) followed by streptavidin–horseradish peroxidase and the TMB peroxidase colorigenic substrate (KPL).
Rat L6 myoblasts (ATCC) were cultured in DMEM supplemented with 10% FBS at 37°C under 5% CO2. Cells were transiently transfected with expression vectors encoding mouse KLB, mouse FGFR4, GAL4-Elk1 transcriptional activator (pFA2-Elk1; Stratagene), and firefly luciferase reporter–driven GAL4-binding sites (pFR-luc; Stratagene) using Fugene 6 transfection reagent (Roche Applied Science).
CYP7A1 expression in primary hepatocytes and in mice
Primary hepatocytes from mouse or rat livers (Life Technologies) were plated on collagen I-coated 96-well plates (Becton Dickinson) and incubated overnight in Williams' E media supplemented with 100 nmol/L dexamethasone and 0.25 mg/mL Matrigel. Cells were treated with recombinant FGF19 or M70 proteins for 24 hours. CYP7A1 expression in cell lysates was determined by qRT-PCR analysis. For assessing CYP7A1 regulation in vivo, 12-week-old db/db mice were injected intraperitoneally (i.p.) with FGF19 or M70 proteins. Mice were euthanized 4 hours after dosing and livers were harvested for qRT-PCR analysis.
In vivo signaling analysis
The db/db mice (9–11-week-old; The Jackson Laboratory) were given intraperitoneal injections (1 mg/kg) of FGF19 or M70 recombinant proteins. Livers were collected after injection and snap-frozen in liquid nitrogen. Signaling proteins were detected in liver lysates with antibodies to pSTAT3 (Cell Signaling Technology #9145), STAT3 (Cell Signaling Technology #8768), or antibody cocktail I (Cell Signaling Technology #5301).
All results are expressed as the mean ± SEM. One-way ANOVA followed by the Dunnett post-test was used to compare data from multiple groups (GraphPad Prism). When indicated, unpaired the Student t test was used to compare two treatment groups. A P value of 0.05 or smaller was considered statistically significant.
An AAV-mediated transgene system for evaluation of hepatocellular tumorigenesis in vivo
AAV-mediated gene delivery provides a means to achieve continuous transgene expression without inflammatory responses that are commonly associated with other viral vectors (18). Sustained expression of up to 1 year has been observed with the AAV gene delivery method when introduced into adult mice (19). The first AAV vector was recently approved as a treatment for a genetic disorder in humans (20).
In the previously reported FGF19 transgenic models, FGF19 was ectopically expressed in the skeletal muscle, a nonphysiologic site of FGF19 expression (8, 11). Under pathologic conditions, such as cirrhosis or cholestasis, FGF19 expression is induced in the liver (21–24). As an alternative to the conventional approach of generating transgenic mice, we introduced FGF19 via AAV in 6 to 12-week-old mice (Fig. 1A). The primary tissue of transgene expression using this method is liver (25).
Multiple mouse strains were evaluated for latency and robustness of FGF19-mediated liver tumor formation (Table 1). A control AAV virus encoding GFP (AAV–GFP), did not promote liver tumor formation in the mouse strains tested (Table 1). In general, mice injected with AAV–GFP exhibited similar phenotype as saline-injected animals (data not shown). For simplicity, only results from AAV–GFP-injected animals were shown as controls in following studies. Interestingly, the tumor latency varied depending upon the mouse genetic background. In particular, among several mouse strains tested, db/db mice exhibited the shortest latency and high tumor penetrance, with the appearance of multiple, large, raised tumor nodules protruding from the liver surface 24 weeks following AAV–FGF19 delivery (Fig. 1B). This finding is consistent with the observation that mutations in the leptin receptor are frequently found in cirrhotic livers and are linked to HCC in human (26, 27). The db/db mice, which carry a genetic defect in the leptin receptor (28), provide a clinically relevant genetic context for evaluating candidate HCC-promoting genes.
|.||FGF19 .||Control .|
|Mouse strain||24 Weeks||52 Weeks||24 Weeks||52 Weeks|
|.||FGF19 .||Control .|
|Mouse strain||24 Weeks||52 Weeks||24 Weeks||52 Weeks|
NOTE: Various strains of mice (6–12-week-old) were injected with 3 × 1011 genome copies of AAV vectors encoding FGF19 or a control gene (GFP). Tumor incidence was determined at 24 or 52 weeks after AAV administration and expressed as number of animals with liver tumor over the total number of animals in the group.
Abbreviation: n.d., not determined.
AAV-mediated delivery of wild-type FGF19 transgene into db/db mice resulted in high circulating levels of FGF19, reaching approximately 1 μg/mL 1 week after a single tail vein injection and persisting throughout the 24-week study period (Fig. 1C). Twenty-four weeks after gene delivery, the mice were euthanized and subjected to necropsy. Visible tumor nodules on the entire surface of the liver were counted (Fig. 1D). The maximum diameter of the liver tumor nodules was recorded (Fig. 1D). Occasionally, a few liver tumor nodules were observed in db/db mice injected with control virus or saline, probably reflecting an increased background level of hepatic tumorigenesis in this genetic model (Fig. 1D; data not shown).
Microscopic examination classified the AAV–FGF19-induced in situ liver tumors as solid HCC, which resembled those reported in FGF19 transgenic animals (Fig. 1E). Cellular proliferative status, examined by immunohistochemical staining for Ki-67 and PCNA, indicated that the tumors were highly proliferative. Similar to the tumors observed in FGF19 transgenic mice, liver tumors in AAV–FGF19 mice were glutamine synthetase–positive, suggestive of a pericentral origin (Fig. 1E; ref. 11). Liver tumors from AAV–FGF19 mice also showed increased nuclear staining for β-catenin (Fig. 1E). Taken together, these data suggest that AAV-mediated transgene expression in mice provides a robust system to evaluate FGF19-induced hepatocarcinogenesis in vivo.
M70 is an engineered, nontumorigenic FGF19
Using the AAV-mediated gene delivery method, we evaluated a panel of FGF19 variants in db/db mice for their tumorigenicity. A FGF19 variant carrying 3 amino acid substitutions (A30S, G31S, and H33L) and a 5-amino acid deletion, referred as M70, was selected for further studies (Fig. 2A).
Whereas ectopic expression of FGF19 promoted significant liver tumor formation in db/db mice (15.6 ± 2.8 tumor nodules/liver), livers from mice with high systemic exposure to M70 for 24 weeks were completely free of hepatic tumor nodules (Fig. 2B). FGF19-expressing mice exhibited a significant increase in liver weight (Fig. 2C), which closely correlates with liver tumor burden. In contrast, mice expressing M70 did not show any increase in liver weight (Fig. 2C). Similar results were obtained when the liver-to-body weight ratio was calculated (Fig. 2D and Fig. 2E). Average serum concentration of M70 was 2 to 3 μg/mL in these mice, about 10,000-fold higher than circulating FGF19 levels in human (Fig. 2F). Histologic analysis of livers from M70-expressing mice showed no evidence of neoplastic lesions associated with FGF19 overexpression in mice, including hepatocellular dysplasia, hepatocellular adenomas, or HCC (Fig. 2G). Moreover, as indicated by Ki-67 staining, overexpression of M70 did not promote hepatocellular proliferation observed in mice expressing FGF19 (Fig. 2G). Furthermore, although liver tumor lesions in FGF19-expressing mice became prominently stained for glutamine synthetase (a marker for pericentral hepatocytes), no increased expression of glutamine synthease was observed in the liver of M70-expressing mice (Fig. 2G). Finally, no liver toxicity, as reflected by serum levels of liver enzymes, was observed following 24 weeks of exposure to M70 (Fig. 2H). Additional serum parameters (e.g., glucose, triglycerides, and cholesterol) are included in Supplementary Table S1. Taken together, these results demonstrate that M70 lacks the ability to promote hepatocellular tumorigenesis in db/db mice.
We also evaluated the liver tumorigenic potential of M70 in a rasH2 transgenic mouse model. CB6F1-RasH2 mice hemizygous for a human H-RAS transgene have been extensively used as an accelerated alternative to the conventional 2-year carcinogenicity assessment in rodents (29). Sensitive to genotoxic and nongenotoxic carcinogens, rasH2 mice develop both spontaneous and induced neoplasms earlier than wild-type mice. Because activation of RAS signaling pathway is frequently observed in human HCC (30), this strain provides a relevant genetic background for studying hepatocarcinogenicity.
During the course of a 52-week study, rasH2 mice expressing FGF19 or M70 had a significant reduction of body weight gain compared with control mice (Fig. 3A). After 52 weeks of continuous exposure, clear differences in liver morphology were observed in mice expressing FGF19 compared with those expressing M70. Gross morphologic changes, including the appearance of multiple tumor nodules, were observed in the livers of mice expressing FGF19 (Fig. 3B). In contrast, the livers from mice expressing M70 showed normal gross morphology and were completely free of tumor nodules (Fig. 3B). A low level of spontaneous liver tumor formation was observed in control rasH2 mice (Fig. 3B). Moreover, M70-expressing animals showed a dramatic decrease in liver weight compared with mice expressing FGF19 (Fig. 3C). M70 also normalized the ratio of liver to body weight in rasH2 mice (Fig. 3D). The serum levels of FGF19 and M70 in these mice are comparable, 155 ± 28 and 209 ± 22 ng/mL, respectively (Fig. 3E).
H&E-stained liver sections from these mice were evaluated for the presence of neoplastic lesions (Fig. 3F). In addition, anti-glutamine synthetase staining was carried out as a marker of FGF19-induced liver tumor (Fig. 3F). rasH2 mice expressing FGF19 displayed a variety of cellular abnormalities, including hepatocellular adenoma and HCCs. Remarkably, none of the livers from mice expressing M70 exhibited histologic evidence of neoplastic lesions (Fig. 3F). Consistent with results of histologic analysis, increased hepatic expression of Ki-67 and AFP, an embryonic hepatic protein often induced in HCC (31), was observed in FGF19-expressing rasH2 mice, but not in mice expressing M70 (Fig. 3G).
In summary, unlike FGF19, prolonged exposure to high levels of M70 did not promote liver tumor formation, in either db/db or rasH2 mice.
M70 binds and activates FGFR4 in vitro and in vivo
To elucidate the molecular mechanism that underlies the inability of M70 to induce liver tumors, we assessed the interaction of M70 to the known receptor complex for FGF19. SPR analysis was used to measure direct binding of either M70 or FGF19 to FGFR4. As shown in Fig. 4A and B, the affinity of M70 for FGFR4 was comparable with that of FGF19 (dissociation constant KD = 134 ± 47 and 167 ± 5 nmol/L, respectively). Using a solid-phase assay, M70 interacted with the FGFR4–KLB receptor complex (Fig. 4C), with the presence of KLB dramatically increased ligand-receptor affinity. The dissociation constant of M70 binding to the FGFR4–KLB receptor complex, indicated a high-affinity interaction that was virtually identical to that of FGF19 (KD = 2.14 and 2.49 nmol/L; respectively).
We also evaluated the ability of M70 to activate its receptors in a cell-based assay using rat L6 cells transfected with an FGF-responsive GAL-Elk1 luciferase reporter (9, 32). In this assay, effective binding of a ligand to FGFR results in the activation of the endogenous ERK kinase pathway, leading to subsequent activation of a chimeric transcriptional activator comprising of an Elk1 activation domain and a GAL4 DNA–binding domain. L6 cells lack functional FGFR or KLB and are only responsive to FGF19 when cotransfected with cognate receptors (data not shown). M70 activated intracellular signaling pathways in L6 cells coexpressing FGFR4 and KLB as effectively as FGF19 (EC50 = 38 and 52 pmol/L for M70 and FGF19, respectively; Fig. 4D). In contrast, signaling in cells transfected with FGFR4 alone was much less responsive to either ligand, showing a >500-fold reduction in potency upon addition of either FGF19 or M70 (Fig. 4D). These results suggest that the formation of a ternary complex between FGFR4–KLB coreceptors and the cognate ligands is important for potent activation of intracellular signaling.
In addition, we analyzed FGFR4 pathway activation in Hep3B, a human HCC cell line that expresses KLB and, among the FGFR isoforms, predominantly FGFR4. Recombinant M70 protein induced phosphorylation and activation of ERK with a similar potency and efficacy as wild-type FGF19 (EC50 = 0.38 and 0.37 nmol/L for M70 and FGF19, respectively; Fig. 4E).
Finally, we assessed a physiologically relevant activity of M70. FGF19 have been implicated in the regulation of hepatic bile acid metabolism in humans and in rodents (7, 8). FGF19 potently represses hepatic expression of CYP7A1, in a process that requires FGFR4 (8, 9). We evaluated the ability of M70 to regulate CYP7A1 in primary hepatocytes. Upon addition to the culture media, M70 effectively repressed CYP7A1 expression in primary hepatocytes derived from mouse or rat liver, showing an activity comparable with that of wild-type FGF19 (Fig. 4F).
To evaluate the acute effects of M70 administration on hepatic expression of CYP7A1 in vivo, mice were injected i.p. with recombinant M70 or FGF19 protein at doses ranging from 0.001 to 10 mg/kg (Fig. 4G). A single i.p. injection of M70 potently suppressed the expression of CYP7A1 mRNA with an ED50 value of 1.29 μg/kg (Fig. 4G). Consistent with the repression of bile acid synthesis, serum levels of total bile acids were lower in mice with long-term exposure to M70 (Fig. 4H). These data demonstrate that systemic administration of M70 can potently and rapidly trigger FGFR4-mediated response in vivo.
In summary, M70 and wild-type FGF19 exhibit a comparable profile of biologic activity that leads to the activation of ERK and suppression of CYP7A1.
M70 exhibits differential signaling pathway activation compared with FGF19
M70 binds to the FGFR4 receptor complex and activates the intracellular signaling pathway leading to CYP7A1 repression, but does not promote liver tumor formation in either db/db or rasH2 mouse models. The identification and characterization of M70 allows us to define two distinct and separable biologic processes regulated by the FGF19–FGFR4 pathway, bile acid homeostasis, and tumorigenesis.
To elucidate the molecular basis for this lack of tumorigenic potential, we analyzed the activation of key signaling proteins involved in tumorigenesis, including ERK, PI3K/AKT, STATs, and WNT/β-catenin pathways. M70 and FGF19 proteins (1 mg/kg) were injected i.p. into db/db mice. Livers were collected and phosphorylation of signaling proteins was measured by immunoblotting. Consistent with the ability of both molecules to signal in cultured primary hepatocytes, FGF19 and M70 stimulated ERK phosphorylation to a similar extent in liver tissues in vivo. In line with previous reports on the role of FGF19 in modulating hepatic protein synthesis (33), both wild-type FGF19 and M70 induced phosphorylation of ribosomal S6 protein (Fig. 5). These data are consistent with results described in the previous sections that M70 retains activity on the FGFR4–KLB receptor complex. Neither M70 or FGF19 had any effect on hepatic levels of phosphorylated AKT, nor did they significantly activate GSK3β and β-catenin at any of the time points tested (Fig. 5).
Remarkably, FGF19 induced STAT3 phosphorylation 2 hours after dosing (Fig. 5A). This effect lasted to 4 hours after dosing (data not shown). In contrast, M70 completely lacked the ability to induce STAT3 phosphorylation (Fig. 5A). The pSTAT3 activation by FGF19 is likely due to noncell autonomous mechanisms on the liver, because no induction of pSTAT3 was observed 15 minutes after protein injection or in primary mouse hepatocyte culture (data not shown). Consistent with these observations of differential STAT3 phosphorylation and activation, STAT3 target genes, including Survivin, Cyclin D1, and Bcl-XL, were dramatically upregulated in db/db and rasH2 livers expressing FGF19, but not M70 (Fig. 5B and C). Because STAT3 is an oncogene frequently activated in HCC (34), its activation by FGF19 provides a plausible mechanism for FGF19-induced hepatocarcinogenicity. Conversely, the inability of M70 to activate the STAT3 pathway could contribute to its lack of tumorigenicity in vivo.
Thus, M70 only activates a subset of signaling pathways downstream of its receptors. This property, a hallmark of selective modulator or “biased agonist” (35), suggests that M70 may act as a selective modulator of FGFR4 to regulate metabolism without causing tumorigenicity.
M70 inhibits FGF19-mediated tumor formation
Our observations suggest that M70 behaves as a selective modulator (or “biased ligand”) to activate the metabolic signaling of, but not the tumorigenic pathway from, FGFR4. Next, we examined whether the biased agonism of M70 could inhibit FGF19-dependent tumor formation.
The db/db mice were injected with AAV–FGF19, with or without 10-fold molar excess of AAV-M70. Mice were necropsied 24 weeks after transgene expression and the livers were excised for analysis. Although ectopic expression of FGF19 in db/db mice promoted the formation of tumor nodules on the hepatic surface, livers from mice expressing both FGF19 and M70 were completely free of tumor nodules (Fig. 6A). Liver weights of mice coexpressing FGF19 and M70 were significantly lower relative to mice expressing FGF19 only (Fig. 6B). The ratios of liver to body weight in M70 and FGF19 cotreated mice were not significantly different from those of control mice (Fig. 6C). The serum levels of FGF19 were 94 ± 12 ng/mL when dosed alone, and the combined serum level of FGF19 and M70 was 453 ± 169 ng/mL (Fig. 6D). Histologic analysis of the livers confirmed that, unlike FGF19-expressing mice, mice coexpressing M70 and FGF19 did not exhibit any histologic evidence of liver tumors (Fig. 6E). These data clearly demonstrate that M70 can effectively block tumor formation induced by wild-type FGF19.
These results are consistent with the notion that M70 acts as a biased ligand that is capable of antagonizing wild-type FGF19 in tumorigenic signaling, and demonstrates the potential of using a selective modulator, such as M70, to suppress FGF19-dependent tumor growth.
FGF19 is an endocrine hormone of the FGF family that regulates bile acid, carbohydrate, lipid, and energy metabolism (33, 36, 37). Among FGFRs, FGF19 binds selectively to FGFR4 (2). A direct link between FGF19 and FGFR4 in hepatocellular oncogenesis is established in a mouse study in which FGF19-mediated liver tumorigenesis was abrogated in FGFR4 KO mice (13). Although predominantly expressed in the ileum, FGF19 is induced in liver under cholestatic conditions caused by either destruction or obstruction of bile duct, cirrhosis or bile acid accumulation. Because chronic liver cirrhosis and cholestasis often trigger compensatory regeneration and HCC, FGF19 might present a missing link between liver injury, regeneration, and cancer.
HCC represents an important unmet medical need (38). Given the potential role of FGF19 in HCC development, efforts have been devoted to developing therapies by targeting FGF19. Indeed, a neutralizing anti-FGF19 monoclonal antibody was developed and demonstrated antitumor activity in xenograft models (22). However, such strategy has suffered setback due to serious adverse effects. Administration of this antibody to cynomolgus monkeys led to dose-related liver toxicity accompanied by severe diarrhea (14). This adverse effect is apparently due to on-target inhibition of endogenous FGF19, resulting in increased hepatic bile acid synthesis, elevated serum bile acid, perturbation of enterohepatic circulation, and the development of diarrhea and liver toxicity (14). Thus, alternative approaches must be developed.
Through a high-throughput in vivo screen, we identified an engineered FGF19 variant M70 that does not cause any liver tumors even after prolonged exposure at supraphysiologic levels in mice. Although there were previous reports of generating tumor-free FGF19 variants (9, 32), those variants were specifically designed to eliminate FGFR4 binding and, therefore, impaired in regulating bile acid metabolism. In contrast, M70 retains the ability to maintain bile acid homeostasis. Importantly, we show that not only does M70 lack tumorigenic potential, but that it can effectively block the tumorigenic effects associated with wild-type FGF19.
The major difference between M70 and FGF19 resides in the N-terminus of the protein. Each FGF family protein consists of the structurally conserved central globular domain, and the flanking N-terminal and C-terminal segments that are structurally flexible and are divergent in primary sequences (1). In X-ray crystal structures of multiple FGF/FGFR complexes, the N-terminal segment of the FGF molecule makes specific contact with the FGFR and is believed to play an important role determining the specificity of the FGF–FGFR interaction (1). Through systematic efforts using an in vivo screen, we found that changing three amino acids at the N-terminus coupled with a 5-amino acid deletion eliminated tumorigenicity without impairing its ability to activate FGFR4-dependent process such as bile acid regulation.
To elucidate the molecular basis for the lack of tumorigenic potential of M70, we carried out in vivo analyses to assess the activation of key signaling proteins involved in tumorigenesis, including ERK, PI3K/AKT, STATs, and WNT/β-catenin pathways. We found that FGF19, but not M70, activates STAT3 in mouse liver. STAT3 is a major player in hepatocellular oncogenesis (34). Phosphorylated (i.e., activated) STAT3 is found in approximately 60% of HCC in humans (39), and correlates with poor prognosis in patients with HCC (30). Constitutively active STAT3 acts as an oncogene in cellular transformation (40). Hepatocyte-specific ablation of STAT3 prevented HCC development in mice (39). Inhibitors of STAT3 activation block the growth of human cancer cells and are being tested in the clinic for treating various cancers, including HCC (41–43). However, the events that lead to STAT3 activation in human HCC are not known. IL-6, among other inflammatory cytokines, is postulated to be the major STAT3 activator in the liver (39). Here, we show that FGF19 also activates STAT3 signaling in vivo, an effect that could be directly mediated by the FGFR4 receptor complex, or indirectly through induction of cytokines or growth factors. The precise mechanism for FGF19-induced STAT3 activation remains to be investigated. Although previous studies reported that FGF19 seems to have a strong effect on β-catenin activation in cultured cell lines (10, 12), we did not observe an impact of FGF19 on levels of dephosphorylated (i.e., activated) β-catenin or P-GSK3β in an acute in vivo setting in mice. At present, the basis for these disparate observations is unclear.
Our data demonstrate that M70 exhibits the pharmacologic characteristics of a “biased ligand” or a selective modulator. M70 exhibits bias toward certain FGFR4 signaling pathways (e.g., CYP7A1, pERK) to the relative exclusion of others (e.g., tumor, pSTAT3). Studies of G protein-coupled receptors and ion channels have demonstrated that selective receptor modulators offer opportunities for fine-tuning biologic responses in a manner that is not attainable via classic orthosteric mechanisms (44). As a selective FGFR4 modulator, M70 antagonizes the oncogenic activity of FGF19 in HCC.
FGF19 demonstrates an array of biologic effects. The therapeutic potential for FGF19 includes the treatment of chronic liver diseases, as well as obesity and diabetes (45). However, the carcinogenicity of FGF19 challenges the development of an FGF19 therapeutic for chronic use. With the identification of M70, an engineered FGF19 devoid of tumorigenicity with potent metabolic properties, therapeutic benefits could be achieved without unwanted side effects. Our study opens new avenues for modulating the FGF19 pathway to treat cancer, diseases with bile acid dysregulation, type II diabetes, and other metabolic disorders.
Disclosure of Potential Conflicts of Interest
The authors have ownership interest (including patents) in NGM Biopharmaceuticals.
Conception and design: X. Wang, D.A. Lindhout, K. Mondal, X. Ding, A.M. DePaoli, H. Tian, L. Ling
Development of methodology: M. Zhou, X. Wang, D.A. Lindhout, J.-Y. Hsu, X. Ding, T. Arora
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Zhou, X. Wang, V. Phung, K. Mondal, M. Humphrey, T. Arora
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Zhou, X. Wang, V. Phung, K. Mondal, A.M. DePaoli, L. Ling
Writing, review, and/or revision of the manuscript: M. Zhou, X. Wang, D.A. Lindhout, T. Arora, R.M. Learned, A.M. DePaoli, L. Ling
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Wang, H. Yang, R.M. Learned
Study supervision: X. Wang, R.M. Learned, H. Tian, L. Ling
Performed experiments and reviewed the manuscript: X. Ding
The authors thank Jin-Long Chen and Thomas Parsons for advice and insightful discussion, Suzanne Crawley, Sara Zong, and Hadas Galon-Tilleman for technical support, Krishna Allemneni for coordinating histology and pathology assessments of tissue sections, and the NGM vivarium staff for the care of the animals used in the studies.