FGFR1 expression defines clinically distinct subtypes in pancreatic cancer

Background The clinical significance of fibroblast growth factor receptor 1 (FGFR1) protein expression in pancreatic cancer is largely unknown. In this study, we aimed investigate the clinical significance of FGFR1 expression in pancreatic cancer. Methods First, we investigated the relationship between FGFR pathway gene expression and clinicopathological data in three pancreatic cancer cohorts containing 313 cases. Subsequently, to confirm the findings from the discovery cohorts, we performed immunohistochemistry (IHC) of FGFR1 protein in a validation cohort of 205 pancreatic cancer cases. Results In discovery cohort 1, FGFR1 and Klotho beta (KLB) overexpression was associated with low tumor stage (P < 0.05), low tumor grade (P < 0.05), and better overall survival. Multivariate analysis predicted FGFR1 (P < 0.05) as a prognostic factor for better overall survival. In discovery cohorts 2 and 3, only FGFR1 overexpression was associated with better overall survival (P < 0.05). In the validation cohort, there were 15.7% and 61% strong and weak/moderate FGFR1-positive cases, respectively. FGFR1-positive cases showed better overall survival than FGFR1-negative cases (P < 0.05). Furthermore, multivariate analysis revealed FGFR1 positivity as an independent prognostic factor for better overall survival in pancreatic cancer patients (hazard ratio 0.677, 95% confidence interval 0.471–0.972, P = 0.035). Conclusions FGFR1 expression, as estimated by IHC, may be used to define clinically distinct subtypes in pancreatic cancer. Moreover, FGFR1-based subclassification of pancreatic cancer may lead to new therapeutic approaches for the FGFR1-positive subtype. Electronic supplementary material The online version of this article (10.1186/s12967-018-1743-9) contains supplementary material, which is available to authorized users.


Background
The fibroblast growth factor receptor (FGFR) pathway is one of the major carcinogenic pathways in cancer [1][2][3][4][5]. Genetic deregulation of fibroblast growth factors and their receptors plays an important role in the initiation and progression of different types of cancer [6][7][8][9].
Helsten and colleagues [10] reported that the FGFR pathway is the third most frequently altered pathway in cancer, after the p53 and KRAS pathways. Accordingly, cancer drugs targeting the FGFR pathway have been tested in multiple cancers [1,11]. Currently, phase I and phase II clinical trials of dovitinib, lucitanib, ponatinib, nintedanib, and pazopanib are underway in different solid tumors to block FGFR pathway activation [12][13][14][15][16].
In pancreatic cancer, aberrations in the FGFR pathway, particularly FGFR1 overexpression, have been reported. According to The Cancer Genome Atlas (TCGA) study on pancreatic cancer [17], FGFR1 is upregulated in approximately 5% of pancreatic cancers. Lehnen and colleagues [18] reported that FGFR1 was expressed in 4% (5/125) of pancreatic cancer cases, and FGFR1 amplification was observed in 2.6% (4/155). In contrast, Kornmann and colleagues [19] reported that 57% (4/7) of pancreatic cancer cases showed immunoreactivity for the IIIc splice variant of FGFR1 (FGFR1 IIIc). Nevertheless, the clinical significance of FGFR1 protein expression in pancreatic cancer is still largely unknown.
In this study, we aimed to investigate the clinical significance of FGFR1 overexpression in pancreatic cancer. First, we investigated the clinical significance of FGFR pathway genes using the gene expression and clinicopathological data from three pancreatic cancer cohorts containing 313 cases. Then, to confirm the findings from the discovery cohorts, we performed immunohistochemistry (IHC) targeting FGFR1 protein in a validation cohort of 205 pancreatic cancer cases.

Statistical analysis
The associations of FGF19, FGF21, FGFR1, FGFR4, and KLB expression with clinical features, including stage, grade, and survival, were calculated using χ 2 and Fisher exact tests for the three discovery cohorts. Survival analysis was performed using Kaplan-Meier curves with logrank (Mantel-Cox) P values. Cox proportional hazard regression and univariate and multivariable analyses were used to evaluate the association between gene expression and survival. Since the sample sizes of the discovery cohorts were small, the multivariable Cox regression model may have led to the overfitting of the data. Therefore, each gene was analyzed separately in combination with the clinical features in multivariable analysis. The hazard ratio (HR) and 95% confidence interval (CI) were also calculated for each factor. P values were two-sided, and P < 0.05 was considered to be statistically significant. All statistical analyses were performed with SPSS 21.0 software (IBM, Armonk, NY, USA).

Validation in 205 patients using IHC
Immunohistochemical labeling was performed in a validation cohort of 205 pancreatic cancer patients at the immunohistochemical laboratory of the Department of Pathology, Asan Medical Center. In brief, 4-μm-thick sections were deparaffinized with xylenes and hydrated in an ethanol series. Endogenous peroxidase activity was blocked by incubation in 3% H 2 O 2 for 10 min, and then heat-induced antigen retrieval was performed. Primary antibodies were used with a Benchmark autostainer (Ventana Medical Systems, Tucson, AZ, USA) in accordance with the manufacturer's protocol. Sections were incubated at room temperature for 32 min in primary antibody for FGFR1 (rabbit polyclonal, 1:100; Abnova, Taipei, Taiwan). The sections were then labeled with an automated immunostaining system and processed with an iView DAB detection kit (Benchmark XT, Ventana Medical Systems). Immunostained sections were lightly counterstained with hematoxylin, dehydrated in ethanol, and cleared in xylenes. Immunoreactivity was interpreted by light microscopic examination and independently evaluated by two pathologists, coauthors of this study (Y.N.S. and S.M.H.), who were blind to the clinicopathologic information. Cases were categorized as positive, weak positive and negative.

Expression analysis of FGFR genes in discovery cohort 1
The logR expression values of five FGFR-related genes were dichotomized according to their median expressions (Additional file 1: Figure S2). FGFR1 had the Notably, a strong correlation was observed between FGFR1 and KLB expression (Pearson's correlation = 0.60, P < 0.001). FGF21 expression was not correlated with that of any of the other four genes.

Association between FGFR genes and clinical features in discovery cohort 1
According to Fisher's exact test results, overexpression of FGFR4 (P < 0.001) and KLB (P = 0.005) were significantly associated with a low tumor grade.

Expression and survival analysis of FGFR genes in discovery cohort 2
The potential significance of the FGFR-pathway genes was further evaluated in cohort 2. Similarly, logR expression values of the five FGFR-related genes were dichotomized according to their medians (Additional file 1: Figure S3). Consistent with the results in cohort 1, the highest median expression was observed for FGFR1. Of note, of all five genes, only patients with FGFR1 expression had significantly better overall survival by Kaplan-Meier analysis (Fig. 1d).

Discussion
The main finding of this study is that FGFR1 protein expression defines clinically distinct subtypes of pancreatic cancer. FGFR1-positive cases showed better overall survival than FGFR1-negative cases. To the best of our knowledge, this study is the first to demonstrate the clinical significance of FGFR1 overexpression in pancreatic cancer.
The clinical significance of FGFR1 overexpression has not been thoroughly investigated in pancreatic cancer for two reasons: (1) KRAS mutation, which does not have any targeted solution, is almost universal in pancreatic cancer; (2) FGFR1 overexpression is not exclusive with KRAS mutation [17,20]. Although, recent genomic and transcriptomic studies identified new subtypes of PADC, but prognostic role of FGFR1 is not highlighted. For instance, Bailey and colleagues demonstrated that Kras G12D/+ ;Tr p53 fl/+ mutant subtype (containing group of genes with FGFR1 expression) showed less aggressive clinical outcome compared to Kras G12D/+ ;Trp53 fl/+ ;TAp63 fl/fl mutant subtype (containing group of genes without FGFR1 expression) in pancreatic cancer [21]. Furthermore, another transcriptomic profiling study classified PADC patients into L1-L6 subtypes. Interestingly, the L5 subtype which showed most favorable clinical outcome from the rest of the molecular subtypes was also enriched with FGFR1 pathway related genes [22]. Our finding can have immediate clinical ramifications. In pancreatic cancer, FGFR1 is the most frequently altered receptor of the four FGFR receptors, and, using readily applicable immunohistochemistry methods, FGFR1 can be used to classify pancreatic cancer into FGFR1-positive and -negative subtypes. As FGFR1-positive pancreatic cancer has better prognosis, FGFR1 can be used as an independent predictor of better overall survival in pancreatic cancer patients. Moreover, FGFR1based subclassification of pancreatic cancer may lead to new therapeutic approaches for the FGFR1-positive subtype. For example, it may be possible to target FGFR1 using antibody in strongly FGFR1-positive pancreatic cancer, in the same manner in which HER2 is targeted using anti-HER2 antibodies in HER2-positive breast cancer.
Only a few previous studies have examined FGFR1 expression in pancreas cancers, and they reported a wide range of FGFR1 positivity (4-57%) [18,19,23]. Lehnen and colleagues observed that 4% (5/125) of pancreatic cancer patients showed FGFR1 expression, and amplification was noted in 2.6% of the cases (4/155) [18]. In contrast, Kornmann et al. reported that 57% (4/7) of pancreas cancers showed immunoreactivity for FGFR1 IIIc [19]. Our study is unique in that we evaluated FGFR1 expression based on the intensity of FGFR1 labeling, and we observed strong FGFR1 positivity in 15.7% (30 cases) and moderate/weak FGFR1 positivity in 61.8% (118 cases) in a cohort of Korean pancreatic cancer patients.