Identification of FGFR4 as a Potential Therapeutic Target for Advanced-Stage, High-Grade Serous Ovarian Cancer Free

High-grade, late-stage metastatic serous carcinoma accounts for the majority of annual epithelial ovarian cancer deaths in the United States (1). More than 15,000 deaths per year result from ovarian cancer, making it the most lethal gynecologic malignancy. Researchers have made minimal advances in extending overall survival durations over the past 3 decades. Currently, the cornerstone of treatment of this cancer is surgery with the aim of reducing the tumor burden to microscopic disease (2). This is usually followed by adjuvant combination chemotherapy with platinum and taxane, which produces initial complete responses in 80% of patients and disease-free periods ranging from 12 to 65 months (3, 4). However, abdominal and pelvic recurrence rates approach 80%, and response to further chemotherapy is limited (5). Attempts at using biologic agents to improve outcomes of this disease, including trap proteins, siRNA encapsulated in nanoparticles, and humanized antibodies, are ongoing (6–8). Thus far, although only the antivascular EGF antibody bevacizumab has shown clinical activity as reported in a phase 3 clinical trial (7). PARP inhibitors also have shown promising activity in pre-phase 3 trials (9). Prognostic and predictive biomarkers that can stratify patients for treatment are still lacking.

Using a 60-mer 22K oligonucleotide-based array comparative genome hybridization platform with DNA isolated from microdissected tumor tissue samples, Birrer and colleagues have shown that a gain frequency of 5q31 to 35.3 in ovarian cancer cells was a negative prognostic indicator for advanced-stage, high-grade serous ovarian cancer, with an overall prevalence of 25% (10). Further studies showed that fibroblast growth factor 1 (FGF1) in 5q31 to 35.3 may be one of the genes that drive ovarian cancer progression (10). Fibroblast growth factor receptor 4 (FGFR4), one of the key receptors for FGF1, is located on the same chromosomal segment and has exhibited preferential binding to FGF1 (11, 12). An alteration in FGFR4 expression would suggest further activation of the FGF ligand and receptor axis, which would subsequently impact ovarian cancer progression. It would also suggest that targeting the FGF ligand and receptor axis can be a new regimen in ovarian cancer treatment.

In the present retrospective study, we evaluated the clinical significance of FGFR4 expression in patients with ovarian cancer and delineated the receptor's signaling pathways in conferring an aggressive phenotype to ovarian cancer cells. Furthermore, we evaluated the feasibility of targeting FGFR4 with siRNA delivered in nanoparticles and an FGFR4 trap protein in the treatment of ovarian cancer.

A total of 183 paraffin-embedded, high-grade, International Federation of Gynecology and Obstetrics stage IIIB to IV (advanced-stage) serous ovarian tumor samples were used in this study. They were obtained from the ovarian cancer repository at The University of Texas MD Anderson Cancer Center. The samples were collected from patients undergoing primary cytoreductive surgery for ovarian cancer from 1995 to 2006. After surgery, patients received platinum-based combination chemotherapy. Optimal surgical cytoreduction was defined by a residual tumor no more than 1 cm in diameter. The overall survival duration was measured from the date of diagnosis to the date of death or censored at the date of the last follow-up examination. Clinical data, including age, cytoreduction status (optimal vs. suboptimal), and overall survival were abstracted from the patients' medical records. Ten normal ovarian and 10 normal fallopian tube tissue samples were collected from patients with benign gynecologic diseases for use as controls. All samples and clinical data were collected with the approval of the MD Anderson Institutional Review Board.

Immunolocalization of FGFR4 was carried out on 183 serous carcinoma, 10 normal ovarian tissue, and 10 normal fallopian tube tissue samples. Slides were stained with a commercially available anti-FGFR4 antibody (1:250; sc-124, Santa Cruz Biotechnology; Supplementary Table S1). FGFR4 protein expression was visualized by 3,3′-diaminobenzidine (DAB). Normal rabbit IgG applied to high-grade serous carcinoma samples with high levels of FGFR4 expression was used as a negative control. Digital photomicrographs of representative areas of each slide were taken at ×20 magnification. Quantitative FGFR4 staining intensity was determined, and localization measurements were obtained using the Image-Pro Plus software (version 5.1; MediaCybernetics). Briefly, areas of interest were drawn around tumor regions. Color segmentation was then used to isolate DAB from background staining. A luminescence grayscale filter was then used with each pixel carrying an intensity value from 0 (black, maximum DAB saturation) to 255 (white, no DAB). The FGFR4 staining intensity score was calculated by dividing the sum of the intensity values in a tumor area of interest by the number of pixels. This score was used to group the patients according to low and high FGFR4 expression using the median score as the cut off point. Survival analyses using both a Kaplan–Meier modeling (with log-rank significance testing) and a Cox proportional hazards model were conducted to determine the effect of FGFR4 expression levels on overall survival and risk of death. A P-value less than 0.05 was considered significant.

Endogenous FGFR4 expression was first evaluated in 6 ovarian cancer cell lines (A2780, SKOV3, OVCA3, OVCA5, OVCA432, and OVCA433) and a normal ovarian surface epithelial (OSE) cell line using Western blot analysis. FGFR4 was knocked down in ovarian cancer cells by using 2 validated, commercially available FGFR4-specific siRNAs (Hs_FGFR4_5 and Hs_FGFR4_6) and a nontarget, scrambled siRNA sequence as a control (Qiagen Sciences). Transfection of ovarian cancer cells with the siRNAs was carried out using siRNA duplexes at a final concentration of 10 nmol/L and Lipofectamine RNAiMAX reagent (Life Technologies). Validation of successful FGFR4 knockdown was carried out at the protein and mRNA levels using Western blot analysis and quantitative real-time PCR analysis, respectively (Supplementary Fig. S1A).

The effect of FGFR4 expression on ovarian cell survival was determined by incubating high-grade serous ovarian carcinoma cell lines transfected with siRNAs in medium containing 2% FBS for 3 days, after which the medium was removed, and water-soluble tetrazolium salt (WST-1) reagent (Roche Applied Science) diluted in serum-free media was added to the cells. After 2 hours of incubation, the assay was read using a FLUOstar Omega plate reader (BMG LABTECH). All data were normalized to respective controls.

A real-time cell proliferation assay was conducted using the xCELLigence system (Roche Applied Science). This system uses measurement of electrical impedance, created by cells attached to the microelectrode-integrated cell culture plated to measure cell proliferation in real time (13). Cells were plated at 10,000 cells per well in serum-free Opti-MEM (Life Technologies) and allowed to attach for 12 hours. A unit-less parameter termed as the cell index was derived and used to represent the cell number based on the measured relative change in electrical impedance that occurred in the presence and absence of cells in the wells (14). The cell index was normalized to the baseline reading at time point 0 following attachment and transfection. Cell proliferation as measured according to the cell index was observed for 3.5 days.

The invasion potential of 3 high-grade serous ovarian carcinoma cell lines (SKOV3, OVCA3, and OVCA433) was assessed using the Oris cell invasion assay (Platypus Technologies) as described by the manufacturer (15). In brief, cells were seeded at a density of 30,000 per well in serum-free Opti-MEM on a 96-well plate previously coated with a collagen I layer and with silicon stoppers covering the central areas of the wells to create a cell-free gap. The cells were then transfected with Hs_FGFR4_5 and Hs_FGFR4_6 and with the nontarget, scrambled siRNA sequence for 24 hours, after which fresh culture medium containing 2% FBS was added to the cells. Twelve hours after the addition of fresh medium, the stoppers were removed, and cells were immediately overlaid with a thick layer of the same matrix. The cells were then allowed to invade into the collagen matrix at the central area of the wells for another 48 hours. The cells were then stained with calcein AM (Life Technologies). The invasion potential of the cells was assessed by measuring the fluorescence of labeled cancer cells invaded into the center of the collagen matrix with a FLUOstar Omega plate reader. Photomicrographs were taken under a fluorescent microscope for each of the wells.

Apoptosis assays were conducted with 2 high-grade serous ovarian carcinoma cell lines (SKOV3 and OVCA432) harvested after transfection with Hs_FGFR4_5 and Hs_FGFR4_6 and with the nontarget, scrambled siRNA sequence for 72 hours using the Cell Death Detection cell ELISA system (Roche Diagnostics), which quantified the relative number of apoptotic cancer cells by colorimetrically measuring cell-free nucleosomes (16).

To determine the effect of FGFR4 overexpression on cancer cell proliferation, FGFR4 was overexpressed in ovarian cancer cell lines OVCA429 and OVCA5, which have low endogenous FGFR4 expression levels. The full-length FGFR4 gene was delivered into the cells via transient transduction using FGFR4 lentiviral particles (GeneCopoeia). Forty-eight hours after transduction, cells were harvested, and overexpression of FGFR4 in transduced cancer cells was confirmed. A cell proliferation assay was conducted with FGFR4-overexpressing cells or control mock-transduced cells seeded on a 96-well plate at 2,500 cells per well. Twenty-four hours after cell seeding, cell proliferation was assayed by incubating cells with WST-1 for 1 hour, and the absorbance at 450 nm was measured.

OVCA432 cells, stably transduced with different reporter–response elements from the Cignal Lenti reporter system (SABiosciences), were transfected with Hs_FGFR4_5 and Hs_FGFR4_6 and with the nontarget, scrambled siRNA sequence. They were then plated in 96-well plates at a density of 30,000 cells per well and allowed to attach to the plate for 24 hours. After cell attachment, cell culture medium was replaced with fresh medium containing 10 ng/mL FGF1 and incubated for 6 hours. Next, the cells were lysed with cell lysis buffer and assayed according to the manufacturer's instructions (Promega) using a FLUOstar Omega plate reader. Data were normalized to the respective negative control groups.

Total RNAs were isolated from SKOV3 and OVCA432 cells transiently transfected with Hs_FGFR4_5 and Hs_FGFR4_6 and with the nontarget, scrambled siRNA sequence for 36 hours. The isolated RNA samples were then subjected to whole-genome transcriptome profiling using the GeneChip Human Genome U133 Plus 2.0 array (Affymetrix). Differentially expressed genes (>2-fold) common to both FGFR4-specific siRNAs and both cell lines were selected for further analysis using the Ingenuity Pathway Analysis program (Ingenuity Systems). Upregulation of 2 of the identified genes, CXCR4 and BNIP3, was validated in OVCA432, OVCA433, and SKOV3 cells using TaqMan gene expression assays (Life Technologies).

To identify the key canonical pathways most relevant to the set of differentially expressed genes identified in the FGFR4-silenced ovarian cancer cells, canonical pathway analysis was conducted using the Ingenuity Pathway Analysis program. Differentially expressed genes with expression levels that differed more than 2-fold from those in both OVCA432 and SKOV3 cells were input into the program for analysis. P-values less than 0.05 for key pathways were considered significant, and association of a specific pathway with the uploaded data set was unlikely to result from chance.

The FGFR4 fusion trap protein FTP-091 (FGFR4mut:Fc:R4-Trap; FGFR4:Fc) generated by Five Prime Therapeutics was used for targeting FGFR4 in the in vivo study. FGFR4mut:Fc is a chimeric fusion protein consisting of the 3 Ig-like extracellular domains of FGFR4 (D1–3) with the acid box linker between D1 and D2 replaced with the corresponding acid box region of FGFR1. The Biacore profile for the affinity of FGFR4:Fc to different FGF ligands is shown in Supplementary Table S2. Intraperitoneal tumor growth in a murine orthotopic model was used as described previously (17). Briefly, a firefly luciferase-expressing serous ovarian cancer cell line (OVCA432-Luc) was generated by transfecting the paternal cell line with commercially available lentiviral particles (GenTarget) and selected for luciferase-expressing cells using puromycin. Mice received intraperitoneal injections of 1 × 10⁶ OVCA432-Luc cells and were subsequently placed in experimental and control groups of 12 mice each. The experimental group received 20 mg/kg FTP-091 twice weekly, whereas the control group received pooled human IgGs (Sigma-Aldrich) at the same dose and schedule. Weekly bioluminescent imaging, carried out on the IVIS 100 imaging system (Caliper Life Sciences), was used to follow intraperitoneal tumor nodule development for 4 weeks, after which the animals were euthanized by CO₂ inhalation followed by cervical dislocation. All visible tumors were resected and weighed. Ex vivo bioluminescent imaging of all of the tumors in the mice was carried out, and the signal intensity of ex vivo tumors was quantified using the Living Image software (Caliper Life Sciences).

In addition, an in vivo siRNA silencing study was conducted. Specifically, 2 × 10⁶ OVCA432-Luc cells suspended in PBS were intraperitoneally injected into 30 nude mice. The mice were then randomized and placed in 3 groups for injection of dioleoyl phosphatidylcholine (DOPC) nanoliposomes containing a control siRNA (control-siRNA-DOPC), Hs_FGFR4_5 (siRNA-Hs_FGFR4_5-DOPC), or Hs_FGFR4_6 (siRNA-Hs_FGFR4_6-DOPC). DOPC nanoliposomes were prepared as described previously (18). Briefly, DOPC and siRNA were mixed and lyophilized. The lyophilized preparations were hydrated in PBS at a concentration of 5 μg/100 μL before administration. The DOPC nanoliposomes were determined to have a mean diameter of 30 nm. Injection of nanoliposomes began 1 week after the initial cancer cell injection. Each mouse was intraperitoneally injected with 5 μg of the corresponding siRNA–DOPC reconstituted in 200 μL of Ca⁺⁺- and Mg⁺⁺-free PBS. Intraperitoneal injections were conducted twice a week for 7 weeks, after which the mice were euthanized by CO₂ inhalation followed by cervical dislocation and submitted to necropsy. The mice and their tumors were weighed, and resected tumors were processed for histologic evaluation.

Ki-67 staining was conducted to visualize proliferating ovarian cancer cells using a polyclonal antihuman Ki-67 antibody (1:200; Life Technologies) with paraffin-embedded ovarian tumor sections obtained in the FGFR4 trap protein and in vivo siRNA silencing studies. Slides were incubated with the antibody and staining was visualized by incubating with fast red chromogen. For each tissue section, the Ki-67–positive cancer cells in 5 random fields under a × 20 objective were counted, and the average number of Ki-67–positive cells per unit tumor area was calculated.

To validate the differential expression of the candidate genes identified from our microarray study, immunolocalization of the 2 genes with upregulated expression identified in the FGFR4 knockdown cancer cells, CXCR4 and BNIP3, was carried out using polyclonal antihuman CXCR4 (1:200; LifeSpan BioSciences) and polyclonal antihuman BNIP3 (1:200; Sigma-Aldrich) antibodies with tumor tissue sections obtained in the in vivo FGFR4 trap protein study. The staining intensity in each section in both treatment groups was quantified and presented in a box plot.

The SPSS software (version 17; IBM Corporation) was used to conduct all statistical tests. All in vitro experiments were repeated independently in triplicate. Data were pooled with mean and SE of the mean values calculated for all end points where appropriate. The 2-tailed Student t test was used to test the significance of differences in sample means for data with normally distributed means. Alternatively, the Mann–Whitney U test was used for the analysis of nonparametric data. Kaplan–Meier survival curves were generated and compared using a 2-sided log-rank statistic. The Cox proportional hazards model was used for the multivariate analysis of patient survival. The Pearson correlation coefficient was used to test linear associations. A P value less than 0.05 was considered significant.

Previous studies showed that overexpression of FGF1, which is located on chromosome 5q31 to 35.3, was significantly associated with poor survival (10). Because FGFR4 is one of the key receptors for FGF1, we evaluated the association between FGFR4 expression and outcomes in patients with high-grade serous ovarian carcinoma. Staining of both normal ovarian tissue and ovarian tumor samples has shown cell surface expression of FGFR4. Immunolocalization of FGFR4 in the 183 high-grade serous ovarian carcinoma samples showed that FGFR4 expression was significantly higher in the tumor samples than in the normal ovarian and fallopian tube tissue samples, which are postulated to be the tissues of origin for the majority of high-grade serous ovarian carcinomas (P < 0.01; Fig. 1A–C; ref. 19). Furthermore, using the median staining intensity as a cut off, Kaplan–Meier analysis showed a significant correlation between high FGFR4 protein expression and poor overall survival (log-rank test; P < 0.001; Fig. 1D). We further confirmed the prognostic significance of FGFR4 expression using Cox regression analysis after adjusting for age and debulking status [HR, 2.1; 95% confidence interval (CI): 1.4–4.3; P < 0.01]

The association between FGFR4 expression and survival suggests that FGFR4 expression plays a key role in ovarian cancer progression. To determine the effect of FGFR4 expression on ovarian cancer cells, we first evaluated FGFR4 overexpression using Western blot analysis with the ovarian cancer cell lines A2780, SKOV3, OVCA3, OVCA5, OVCA432, and OVCA433 and compared it with that in HOSE cells. All 6 ovarian cancer cell lines exhibited higher FGFR4 expression (Fig. 2A). We confirmed successful FGFR4 silencing in all of the cell lines by the FGFR4-specific siRNAs Hs_FGFR4_5 and Hs_FGFR4_6 using quantitative PCR and Western blot analyses (Supplementary Fig. S1B and C). Cell survival analysis showed that transfection with both FGFR4-targeting siRNAs in all 6 cell lines significantly decreased cell survival in serum-reduced media more so than did transfection with the nontarget scrambled siRNA sequence and untransfected control (Fig. 2B). Furthermore, cell proliferation assays showed that all 6 ovarian cancer cell lines transfected with FGFR4-specific siRNAs have shown significant decreases in proliferation after 18 to 24 hours (Fig. 2C). Moreover, invasion assays with 3 cell lines that exhibited invasion potential in vitro have shown that the number of FGFR4-specific siRNA-transfected cells invading into the central zone of the Oris assay was significantly lower than that of the nontarget siRNA-transfected control cells (Fig. 3A and B). Finally, the number of SKOV3 and OVCA432 cells transfected with Hs_FGFR4_5 and Hs_FGFR4_6 that were apoptotic was significantly higher than that of the cells transfected with the nontarget scrambled siRNA sequence (Fig. 3C).

Our results showed that FGFR4 overexpression in OVCA429 and OVCA5 cells, which have low levels of endogenous FGFR4 expression, promoted cell proliferation. The rate of proliferation of both cell lines was more than 50% higher than that of their corresponding control cells at 24 hours after initial cell seeding (Supplementary Fig. S2A and S2B), suggesting that overexpression of FGFR4 led to an aggressive phenotype.

Researchers have shown that FGFR4 preferentially binds to FGF1 (11, 12), suggesting that FGF1-activated downstream signaling pathways are mediated by FGFR4. To confirm this in ovarian cancer cells, we first treated OVCA432 cells with exogenous FGF1 (1–10 ng/mL), which produced a significant increase in cell proliferation rates (Supplementary Fig. S2C). The cells have also shown a significant increase in reporter activity, which reflected activation of the mitogen-activated protein kinase (MAPK), NF-κB, and WNT signaling pathways (Fig. 4A). In addition, Western blot analysis has shown a significant increase in phosphorylated form of extracellular signal-regulated kinase (ERK) 1/2 and glycogen synthase kinase 3β (GSK-3β; Fig. 4B). However, when we treated OVCA432 cells transfected with FGFR4-specific siRNAs with FGF1 (10 ng/mL), their growth rates were significantly lower than those of cells transfected with the nontarget scrambled siRNA (P < 0.001; Fig. 4C). Furthermore, the reporter activities of the MAPK, NF-κB, and WNT signaling pathways activated by FGF1 were abrogated (Fig. 4D).

To further delineate the underlying molecular mechanism of FGFR4′s control of the malignant phenotypes of ovarian cancer cells, we conducted transcriptome profiling analysis on SKOV3 and OVCA432 cells transiently transfected with Hs_FGFR4_5 and Hs_FGFR4_6 and with the nontarget scrambled siRNA sequences. A total of 24 genes were identified to be altered more than 2-fold in expression, which are common for both OVCA432 and SKOV3 cells transiently transfected with the siRNAs than in those transfected with the scrambled siRNA sequence (Fig. 5A; Supplementary Table S3; Gene Expression Omnibus, GES34828). Network functions analysis showed that of these 24 genes, 13 were associated with free radical scavenging, cell death, cellular growth, and proliferation (Table 1). Two of these genes were CXCR4 and BNIP3 (Fig. 5B). We further validated upregulation of the expression of CXCR4, which regulates cell proliferation and apoptosis upon binding of its ligand CXCL12 in T cells and cancer cells (20, 21), and of BNIP3, which induces apoptosis (22), using quantitative PCR analysis of 3 ovarian cancer cell lines in which FGFR4 was silenced by siRNA (Supplementary Fig. S3A and B). In addition, pathway analyses showed that several canonical signaling pathways, including ERK/MAPK (P = 0.03), NF-κB (P = 0.05), and TNF receptor 1 (TNFR1; P = 0.04), were significantly associated with the differentially expressed gene signature identified in the transcriptome profiling study. Identification of the MAPK and NF-κB pathways in the array data was consistent with the results of the reporter assay.

The antitumor effect of silencing of FGFR4 in vitro suggests that targeting FGFR4 is a new strategy for treatment of ovarian cancer. We therefore examined the effect of blocking the binding of ligands to FGFR4 using the novel FGFR4 fusion trap protein FTP-091 (FGFR4mut:Fc:R4-Trap; FGFR4:Fc) and downregulating FGFR4 expression by injection of DOPC nanoliposomes containing FGFR4-specific siRNAs in a murine xenograft OVCA432 ovarian tumor model. The results showed that mice injected with 20 mg/kg FTP-091 had significantly lower luciferase activity in ex vivo tumors than did those given control human IgG (Fig. 6A–C), illustrating a decrease in the number of viable tumor cells in mice given the trap protein. Furthermore, mice injected with FGFR4 HS_5 and FGFR4 HS_6 in DOPC nanoliposomes had significantly smaller tumors than did those injected with scrambled siRNA (Fig. 6D). Histologic examination of the resected tumor tissue collected from the xenografts showed that the tumors had histologies consistent with serous ovarian carcinoma. Ki-67 staining of tumor sections obtained in the FGFR4 trap protein and siRNA studies has shown significantly fewer Ki-67–positive cancer cells per unit tumor area in the FGFR4 trap protein-injected mice and in the FGFR4 knockdown mice (via injection of siRNAs in DOPC nanoliposomes) than in their corresponding controls (Supplementary Fig. S4A and S4B), suggesting that targeting FGFR4 in vivo can suppress ovarian tumor growth via inhibition of cancer cell proliferation. Furthermore, immunolocalization of CXCR4 and BNIP3 protein in tumor samples collected in the FGFR4 trap protein study showed significantly lower expression of both proteins in the FGFR4 trap protein-treated mice than in their controls, suggesting that blockage of FGFR4 signaling with the trap fusion protein upregulated the expression of the CXCR4 and BNIP3 genes (Supplementary Fig. S4C and S4D), which was consistent with our findings of transcriptome profiling of FGFR4-silenced ovarian cancer cells in vitro.

In a previous study, we showed decreased survival in patients with high-grade serous carcinomas harboring amplification of 5q31 to 35.3. Specifically, we observed that overexpression of FGF1, located in this amplicon, was a poor prognostic indicator for these tumors (10). One of the key receptors for FGF1 is FGFR4, which is located on the same amplicon and preferentially binds to FGF1 (11, 12). In the present study, we have shown that FGFR4 overexpression is associated with a more aggressive high-grade serous carcinoma phenotype in vitro and in vivo, suggesting that FGF axis activation through overexpression of both FGFRs and FGF ligands may represent a targetable autocrine signaling loop associated with poor overall survival in patients with ovarian cancer.

Birrer and colleagues observed a 25% gain frequency of chromosome segment 5q31 to 35.3 harboring the FGFR4 gene in ovarian cancer cases (10). Their data also showed a significant correlation between FGFR4 copy number and overall survival. However, recently reported data from The Cancer Genome Atlas (TCGA) showed a 6% amplification/overexpression rate for FGFR4 with no significant correlation with survival in ovarian cancer cases (23). The discrepancy between these 2 sets of data may result from several factors. First, the TCGA tumor samples were bulk tissue samples with stromal cell contamination rates ranging from 5% to 50%. Tissue samples with high levels of stromal contamination will show significantly low levels of FGFR4 gene amplification and/or expression, as the FGFR4 gene and mRNA copies are diluted by the stromal DNA and mRNA. This may impact survival correlation studies. In comparison, in our analyses, we used DNA and RNA extracted from microdissected ovarian tumor cells, which had minimal stromal DNA and RNA contamination. Second, the TCGA samples were collected from multiple institutions, whereas we collected our samples at a single institution, which may have given us a more homogenous patient population. Third, we found a significant correlation between FGFR4 protein expression and survival. While FGFR4 amplification may not correlate with survival as indicated in the TCGA dataset, FGFR4 protein expression may correlate with it, as gene amplification and mRNA expression may not correlate with protein expression owing to mechanisms that regulate mRNA and/or protein expression. However, researchers have not immunohistochemically evaluated FGFR4 protein expression in the TCGA dataset.

FGFR4 differs from the other 3 members of the FGFR family in genomic structure, ligand binding, and signal transduction (24). FGFR activation, either by activating mutations (25–29) or overexpression (30–35), occurs in multiple solid tumors. FGFR2 and FGFR3 mutations are common in endometrial cancer (36) and bladder cancer (31). In comparison, FGFR1 and FGFR4 mutations are not common in carcinoma cells; instead, overexpression of FGFR1 and FGFR4 is more prevalent (37). To exclude the presence of several rare activating mutations of FGFR4, we sequenced DNA isolated from microdissected high-grade serous carcinoma samples. We did not identify any mutations in either the kinase or intermembrane domain except for the Gly388Arg polymorphism, which has exhibited no effect on cancer prognosis (38), in 6 of 43 (14%) high-grade serous carcinoma samples. Our data suggest that, similar to breast cancer (39), FGFR overexpression is the main mechanism implicated in high-grade serous ovarian carcinoma.

Unlike other FGFR family members, FGFR4 preferentially binds to acidic FGF (FGF1; refs. 11, 12). Our prior work showed that poor survival and related phenotypic changes induced by FGF1 in high-grade serous carcinoma parallel those in FGFR4. Hence, overexpression of both FGF1 and FGFR4 may provide an autocrine loop that drives high-grade serous ovarian carcinoma growth. This may result from activation of the MAPK/ERK signaling pathway by FGF1 as described previously (40) and confirmed by the pathway reporter and Western blot analysis data in the present study. In addition to activation of the MAPK/ERK pathway, we observed activation of the proliferative NF-κB and WNT signaling pathways in FGF1-treated cells, which our pathway reporter and Western blot analysis confirmed. In addition, our canonical pathway analysis of transcriptome profiling data showed a significant correlation between these pathways and the differentially expressed genes identified by manipulating the level of FGFR4 expression in ovarian cancer cells. These data suggest that multiple pathways can be activated by FGF1, most likely mediated via FGFR4, because the effect of FGF1 on pathways activation can be abrogated by downregulation of FGFR4 expression. Furthermore, our data suggest that the effects of FGFR4 on ovarian cancer cell proliferation and survival result from upregulation of expression of CXCR4, which can regulate cell proliferation and apoptosis upon binding of its ligand CXCL12 in T cells and cancer cells (20, 21), and from upregulation of expression of BNIP3, which is known to induce apoptosis (22). We confirmed upregulation of these 2 proteins in the ovarian tumor samples obtained from mice given the FGF trap protein. However, the molecular mechanisms involved by these 2 proteins in mediating the effect of the FGFR4 signaling pathway on ovarian cancer cell growth remain to be elucidated.

In the absence of activating mutations of FGFR4, downregulation of FGFR4 expression and prevention of binding of ligands to FGFR4 may be effective in the treatment of serous ovarian cancer. Using 2 strategies for targeting FGFR4 with an orthotopic mouse model of high-grade serous carcinoma, we showed that decreasing FGFR4 expression leads to a decrease in tumor growth. Although the use of FGFR4 siRNAs in vitro can completely silence FGFR4 expression, leading to decreased proliferation and survival of ovarian cancer cell lines, ovarian tumors, although small, still developed in mice given FGFR4-specific siRNAs delivered in DOPC nanoliposomes. In addition, the tumors had FGFR4 expression, suggesting that the nanoliposomes may not efficiently deliver the siRNAs to all tumor cells. In addition to the use of siRNAs, our data showed that use of the soluble fusion protein FPT-091 containing the extracellular domain of FGFR4 and an IgG Fc fragment may have significant clinical applications. Replacement of the FGFR4 acid box region with the corresponding FGFR1 region improved the bioavailability and pharmacokinetics of the fusion protein in vivo. The chimeric FGFR4/FGFR1 extracellular domain in FGFR4mut:Fc is fused to the Fc fragment of IgG1 and can bind to several members of the FGF protein family, including FGF1. The ability to sequester several FGF ligands and inhibit the pathway upstream of the FGF receptor enables the trap protein to inhibit all of the possible effects of overexpression of FGFs and FGFR4. Given the high degree of genetic heterogeneity of high-grade serous carcinomas, activation of multiple pathways and/or targets, such as Notch3 (41) and NAC1 (42), may be involved in the pathogenesis of different patients with ovarian cancer. A challenge remains in identifying patients with ovarian cancer who may benefit from targeting a specific pathway or pathways, possibly via individual tumor expression profiling.

In conclusion, the present study has shown that the overexpression of FGFR4 is an indicator of poor prognosis for high-grade serous ovarian carcinoma. It also identified mechanisms by which activation of FGFR4 by ligands such as FGF1 may lead to an aggressive cancer phenotype. Successful targeting of FGF axis activation in our orthotopic mouse model suggests that this approach is feasible in clinical trials.

No potential conflicts of interest were disclosed.

Conception and design: T. Zaid, T. Yeung, C. Leung, G. Lopez-Berestein, K. Wong, M.J. Birrer, S.C. Mok

Development of methodology: T. Zaid, M.S. Thompson, C. Leung, T. Harding, C. Rodriguez-Aguayo, G. Lopez-Berestein, A.K. Sood, M.J. Birrer, S.C. Mok

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Zaid, T. Yeung, M.S. Thompson, C. Leung, T. Harding, S. Kwan, A.K. Sood, M.J. Birrer, S.C. Mok

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Zaid, T. Yeung, C. Leung, N. Co, G. Lopez-Berestein, A.K. Sood, K. Wong, S.C. Mok

Writing, review, and/or revision of the manuscript: T. Zaid, T. Yeung, M.S. Thompson, C. Leung, R.E. Schmandt, G. Lopez-Berestein, A.K. Sood, K. Wong, M.J. Birrer, S.C. Mok

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T. Zaid, C. Leung, N. Co, G. Lopez-Berestein, M.J. Birrer, S.C. Mok

Study supervision: G. Lopez-Berestein, K. Wong, S.C. Mok

This study was supported in part by grants R01 CA133057 (S.C. Mok), RC4 CA156551 (M.J. Birrer), and U54 CA151668 (A.K. Sood); the Gilder Foundation (A.K. Sood); and MD Anderson Ovarian Cancer SPORE grant P50 CA083639 (S.C. Mok) from the National Institutes of Health. This research is supported in part by the MD Anderson Cancer Center Support Grant CA016672.

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