Co-expression of fibroblast growth factor receptor 3 with mutant p53, and its association with worse outcome in oropharyngeal squamous cell carcinoma

Fibroblast growth factor receptor 3 (FGFR3) is expressed in squamous cell carcinoma of the head and neck (SCCHN) including oropharyngeal squamous cell carcinoma (OPSCC) and is a potential therapeutic target. However, information on its correlation with other relevant cancer related proteins stratified by p16 status and its prognostic significance in OPSCC is limited. We examined FGFR3 expression and its correlation with clinical characteristics, p16 status, and mutant p53 (mp53) among 220 retrospectively collected OPSCC cases and 40 prospectively collected SCCHN cases, including a majority of OPSCC. Correlations of FGFR3 Weighted Index (WI) with p16 status and mp53 WI as well as its association with disease-free survival (DFS) and overall survival (OS) were evaluated. FGFR3 expression was detected in 61% and 70% of cases in cohorts 1 and 2, respectively. FGFR3 level was significantly higher in p16-negative tumors in both cohorts (p<0.001 and 0.006). FGFR3 expression was highly correlated with mp53 expression in both p16 + and p16– OPSCC (p<0.0001 and p = 0.0006, respectively). In cohort 1, univariate analysis showed that FGFR3 was associated with DFS but not OS. Kaplan-Meier analysis showed that higher FGFR3 and mp53 level correlated with worse DFS (p = 0.025) and OS (p = 0.009). As expected, p16 positive status was associated with improved OS and DFS (p<0.001 for both). Our results suggest that high FGFR3 expression is associated with p16 negative status and mp53 expression in OPSCC and correlates with a worse clinical outcome. The biological relationship between FGFR3 and mp53 in OPSCC deserves further investigation.


Immunohistochemistry (IHC)
FFPE tumor tissue sections were pre-heated at 60˚C for 30 minutes and washed through a series of xylene and alcohol containers followed by antigen retrieval using 1X citrate buffer for 10 minutes in a microwave. The slides were cooled for 30 minutes at room temperature and quenched using 3% hydrogen peroxide in distilled water. The sections were washed and then blocked using 2.5% normal horse serum following the manufacturer's instructions (Vectastain Kit, Vector Laboratories, Burlingame, CA). Besides hematoxylin and eosin (H&E) staining, standard immunohistochemistry (IHC) was used to stain for p16 (1:100, Delta Biolab, Gilroy, CA), FGFR3 (PA2143, Boster Bio, Pleasanton, CA.) and mp53 (AB32049, Abcam, Cambridge, MA.) using primary antibodies followed by corresponding secondary antibodies, visualized by DAB and counterstained with hematoxylin. Positive signals were counted in five random fields under ×100 magnification and were quantified using weighted index method. WI = Intensity × % of positive staining. The intensity of staining was scored as negative (0), weak (1+), intermediate (2+), and strong (3+), respectively. p16 positivity was determined if > 70% of tumor cells stained with p16 specific antibody in the nuclei and cytoplasm [18,19]. These quantifications were determined by at least 2 individuals blindly and independently. In the case of disagreement, two pathologists had to come to a consensual agreement.

Statistical analysis
Descriptive statistics was used to summarize the characteristics for each patient. Frequency and percentage for categorical variables were presented. For numeric variables, mean, median, and standard deviation were calculated. In the univariate analysis with FGFR3, the ANOVA test was used to compare FGFR3 between different categories. Pearson correlation coefficient was calculated to examine the relationship between FGFR3 with continuous variables. Stratified analyses were conducted to assess the confounding effect of p16 on the relationship between FGFR3 and mp53, cytoplasmic mp53, and nuclear mp53 after adjusting for p16, respectively. The general linear model (GLM) was further employed in multivariate analyses to assess the adjusted relationship between FGFR3 and mp53, cytoplasmic mp53, nuclear mp53 after adjusting for p16, respectively. The univariate association of each covariate with overall survival (OS) or disease free survival (DFS) was assessed using the Cox proportional hazards model. Also, Kaplan-Meier (K-M) plot and log-rank test were presented. A multivariate model was fitted by backward selection method with removal criteria alpha = 0.20. All statistical analyses were performed using SAS 9.4 (SAS Institute, Inc., Cary, North Carolina).

FGFR3 and mp53 expression in OPSCC
Among the first cohort of 220 OPSCC cases, 62.2% (135/220) of cases were p16 positive, 78% (91/116 available tissues) were positive for mp53 expression, and 61% (134/220) were positive for FGFR3. In the second cohort, 86% (25/29 OPSCC) of cases were p16 positive, 88% (35/40) expressed mp53, and 70% (30/40) were positive for FGFR3. A higher percentage of cases expressed mp53 and FGFR3 in the second cohort, which may reflect the fact that there were 11 cases of non-OPSCC, all of which expressed mp53 (Fig 1). The negative control slide had no primary antibody in the incubation buffer. The specificity of FGFR3 antibody was supported by both positive and negative controls (S1 Fig). Membrane, cytoplasmic, and nuclear localizations of FGFR3 are consistent with the reported literature [20,21]. Expression levels of FGFR3 and total, cytoplasmic, and nuclear mp53 are summarized in S3 and S4 Tables.

Correlation of FGFR3 expression with p16 status and mp53
FGFR3 level was significantly higher in p16-negative tumors in both cohorts (p<0.001 and 0.006) (Tables 1 and 2). Furthermore, FGFR3 expression was highly associated with total, cytoplasmic, and nuclear mp53 in cohort 1 ( Table 3), with cytoplasmic mp53 being more significantly correlated (p < 0.001) than nuclear mp53 (p = 0.050). Further stratifying by p16 status, we observed that in both p16 positive and negative tumors, FGFR3 correlated with cytoplasmic mp53, but not nuclear mp53 (p = 0.0009 vs. 0.517 and 0.0007 vs. 0.319, respectively, Tables 4 and 5). In cohort 2, FGFR3 was not found to be significantly correlated with mp53 (S5 Table), which may be a result of the smaller sample size. Pearson correlation consistently showed that in both cohorts, mp53 and FGFR3 were inversely associated with p16 (S6 and S7 Tables), suggesting that FGFR3 and mp53 association is predominantly seen in p16-negative SCCHN.

Univariate and multivariate analyses of FGFR3 associated with clinical characteristics, OS, and DFS
In addition to p16 status, FGFR3 expression was highly associated with grade and smoking history in cohort 1 ( Table 1). In cohort 2, FGFR3 was highly associated with smoking history and T-stage ( Table 2).
In cohort 1, univariate analysis showed that FGFR3 was associated with DFS but not OS (Tables 6 and 7). In addition, p16 status, smoking, tumor stage, and mp53 were found to correlate with both OS and DFS. Kaplan-Meier analysis of the first cohort of patients (OPSCC) also showed that FGFR3 expression is associated DFS, not OS (S2 Fig). The factors used in the univariate analysis were further considered in the multivariate analysis which also showed that FGFR3 expression was correlated with DFS (p = 0.005), but not OS (p = 0.172) ( Table 8). Negative p16 status, current smoking, and T-stage 3 and 4 were associated with worse OS ( Table 9). The mp53 and T-stage 3 and 4 were associated with worse DFS (Table 8).
Furthermore, Kaplan-Meier analysis illustrated that mp53 and FGFR3 were significantly associated with worse OS and DFS using medians of both FGFR3 and mp53 as the cut-off values (p = 0.009 and 0.025, respectively) (Fig 2). No correlation with either OS or DFS was identified in the second cohort of samples, due to either the small sample size or short follow up time. Therefore, the results were mainly from the first cohort.

Discussion
SCCHN has been classified as an HPV-driven or non-HPV-driven disease. However, a reliable molecularly based classification that could impact clinical decision and the use of molecularly targeted agents is lacking in this disease. Despite the fact that FGFR remains one of the common genetic alterations in SCCHN, few studies have evaluated or targeted this pathway in SCCHN and none has produced data convincing enough to be practice changing. Although FGFR3 has been studied in SCCHN [10,22,23], its expression level has not been carefully correlated with common alterations in SCCHN such as HPV and p53 mutational status. In addition, there is limited and conflicting data on its prognostic significance in SCCHN [13].
Our findings are the first to confirm a strong correlation between FGFR3 and p16 negative status in SCCHN, as well as an inverse correlation between both mp53 and FGFR3 with p16 status. These findings suggest a more significant role for FGFR3 in p16-negativeSCCHN. It is of interest that FGFR3 and mp53 were both associated with a worse clinical outcome in addition to being more commonly co-expressed, raising the question of potential biological interactions between FGFR3 and mp53 and opening the door for studies to elucidate possible mechanistic interactions. A recent publication suggested that p53 IHC could be a surrogate marker for p53 mutation [24]. It is known that wild-type p53 is not stable and the previous IHC staining for p53 most likely reflected expression of mp53. In our study, we used an antibody which recognizes mp53 only. We found no nuclear expression of mp53 in some tissues using this specific antibody, but we cannot rule out whether there is or not the wild-type p53 which is not stable in tumor tissues. Other limitations in our analysis included its focus on expression levels of FGFR3 and its correlation with mp53 only but did not correlate FGFR3 mRNA and the active FGFR3. Despite this limitation, our findings raise the possibility of an improved molecular classification of mp53-altered SCCHN, particularly OPSCC, based on these findings, in addition to possibly enhancing patient selection for future clinical trials in HPV-unrelated disease.
Of further interest, and consistent with TCGA data, the FGFR3-TACC3 fusion gene was identified in two cases of HPV-positive SCCHN in our prospective second cohort [25]. This suggests that in HPV-positive SCCHN, activation of FGFR3 may be driven by genomic alterations that are different from those in HPV-negative disease. Recent studies have examined FGFR gene mutation and amplification in HPV-positive SCCHN and described the occurrence of FGFR3 [16,17]; TCGA reported two cases of FGFR3-TACC3 fusion gene only in HPV-positive tumors [6]. The FGFR3-TACC3 fusion gene was first identified in glioblastoma multiforme (GBM) by two independent groups [14,15,26,27]. Both reported that the FGFR3-TACC3 cDNA contains an open reading frame coding for a protein with 1048 amino acids. This fusion protein includes the FGFR3 N terminus (residues 1-758) and the TACC3 C terminus (residues 549-838). FGFR3 is a receptor tyrosine kinase, while TACC3 (transforming acidic coiled-coil) is a centrosomal protein involved in mitosis [28,29]. In addition to brain tumors, to date, the FGFR3-TACC3 fusion protein has been reported in several solid tumors [30], including non-small cell lung cancer [31], cervical cancer [32], esophageal cancer [33], gastric cancer [34], and SCCHN [35].
In HPV-unrelated disease, our findings suggest a proportional correlation between FGFR3 and mp53 expressions. A regulatory role of mp53 on FGFR3 expression may possibly exist, however needs to be confirmed in further investigation. In SCCHN, over 70% of p53 are mutated in SCCHN [6]. Though p53 is a tumor suppressor, mp53 may work as an oncogenic transcription activator [36][37][38]. FGFR3 and p53 play different roles in cancer development. FGFR3 initiates multiple signaling pathways to support cell proliferation, while mutation or loss of p53 function results in loss of cell cycle regulation leading to tumor progression. Two publications using mouse models of carcinogenesis reported that the FGFR3-TACC3 fusion gene could effectively induce spontaneous brain or lung tumors in mice only with p53 knockout, suggesting that abnormal FGFR3 activation in the absence of normal p53 function may enhance cancer cell growth [15,39]. Our findings that simultaneous FGFR3 and mp53 alteration is correlated with worse DFS and OS support this hypothesis and urge the need for confirmatory studies. If such a correlation is confirmed, this could have implications for patient stratification and the identification of a unique subgroup of HPV-negative SCCHN. Since mp53 may function as an oncoprotein, a possible regulatory effect on FGFR3 expression deserves further investigation.   In the current study, we used p16 as a reliable surrogate biomarker for HPV-associated OPSCC since in this disease, there is functional inactivation of RB by the viral oncoprotein E7, which is known to cause upregulation of p16 expression. In other SCCHN sub-populations, such as oral cavity and larynx, p16 is not an indicator for HPV-related cancer. However, though our second independent cohort contained 11 non-OPSCC samples, the correlation observed between FGFR3 and p16 was consistent with that in the first cohort.