FGFR genes mutation is an independent prognostic factor and associated with lymph node metastasis in squamous non-small cell lung cancer

ABSTRACT

Targeting FGFRs is one of the most promising therapeutic strategies in squamous non-small cell lung cancer (SQCC). However, different FGFR genomic aberrations can be associated with distinct biological characteristics that result in different clinical outcomes or therapeutic consequences. Currently, the full spectrum of FGFR gene aberrations and their clinical significance in SQCC have not been comprehensively studied. Here, we used Next-generation sequencing to investigate the presence of FGFR gene mutations in 143 tumors from patients with stage I, II or III SQCC and who had not been treated with chemotherapy or radiotherapy prior to surgery. FGFR gene mutations were identified in 24 cases, resulting in an overall frequency of 16.9%. Among the mutations, 7% (10/143) were somatic mutations, and 9.8% (14/143) germline mutations. FGFR mutations were significantly associated with an increased risk of lymph node metastasis. SQCC patients with a FGFR somatic mutation had shorter OS (overall survival, log rank P = 0.005) and DFS (disease-free survival，log rank P = 0.004) compared with those without an FGFR mutation. The multivariate analysis confirmed that a somatic mutation was an independent poor prognostic factor for OS (HR: 4.26, 95% CI: 1.49–12.16, P = 0.007) and DFS (HR: 3.16, 95% CI: 1.20–8.35, P = 0.020). Our data indicate that FGFR genes mutation is an independent prognostic factor and associated with lymph node metastasis in stage I to III Chinese SQCC patients.

Keywords:

• FGFR mutation
• lung squamous cell carcinoma
• lymph node metastasis
• prognostic markers
• stage I-III

Introduction

Lung cancer is the leading cause of cancer-related death worldwide.¹ Although squamous-cell carcinoma (SQCC) and adenocarcinoma constitute the vast majority of diagnosed cases of lung cancer,² they exhibit different somatic genomic profiles.^3,4 Activated oncogenes, such as mutant alleles of EGFR, KRAS, and BRAF, or translocations involving ALK, RET, and ROS1 are commonly observed in lung adenocarcinoma cancer (Ad-NSCLC), whereas these mutations are rarely detected in SQCC.^4,5 Therapies targeted against specific molecular and genomic aberrations have resulted in improved survival of some patients.^6–8 However, these therapies are unlikely to be of benefit in patients with SQCC, since the most frequent aberrations among SQCC case known to date are TP53, PTEN, NFE2L2, and KEAP1 mutations and changes in SOX2 and CDKN2A copy number, all of which are not clinically actionable. Among the genes associated with SQCC genetic abnormalities, fibroblast growth factor receptor (FGFR) appears to be one of the most promising therapeutic targets.^5,9–11

The FGFR tyrosine kinases are encoded by 4 genes (FGFR1, FGFR2, FGFR3, and FGFR4) and composed of an extracellular component with 3 immunoglobulin-like domains, a transmembrane domain and an intracellular tyrosine kinase domain. FGFR activation triggers a cascade that leads to the activation of multiple signal transduction pathways, including the four primary downstream pathways Ras/Raf/MapK, PI3K/Akt, Stats and PLCγ, which can promote cell survival, motility and invasiveness, cell proliferation, epithelial-to-mesenchymal transition (EMT) and angiogenesis.¹² FGFR genes amplifications, somatic missense mutations and chromosomal translocations can promote aberrant receptor tyrosine kinase activation.^13–17 Germline mutations in the FGFR genes family have been described in craniofacial and skeletal syndromes,¹⁸ and subsequent studies have reported that some known FGFR SNPs are also observed in tumor samples as somatic mutation which might contribute to tumorigenesis.^16,19,20

Previous studies have demonstrated that FGFR genes aberrations are associated with transformative potential and mediate inhibitor sensitivity in SQCC.^15,16,21,22 Preliminary data from early-phase trials evaluating FGFR inhibitors have provided proof-of-concept support for the effective inhibition of FGFR signaling by TKIs in patients with lung SQCC.²³ However, the underlying FGFR mutations in SQCC patients are not well known. In order to help identify patients who may benefit from therapies targeted against FGFR activation, the full spectrum of somatic FGFR mutations and their clinical significances remains to be characterized. In this study, we characterized FGFR genes mutations completely in SQCC from Chinese patients and investigated the correlation between various types of FGFR aberrations and clinical outcomes.

Materials and methods

Patients and samples

Patients who had been diagnosed with SQCC and treated in Peking University Cancer Hospital (Beijing, China) between 2003 and 2013 were recruited for this study. All of the recruited patients were of Chinese ethnicity, and the clinic stage of the tumors diagnosed were stage I to III, with no patients with stage IV disease included in this study. All of the patients received anatomical resection, including lobectomy, sleeve lobectomy and pneumonectomy,except for one who received segmentectomy. All patients received SMLD (systematic mediastinal lymph node dissection). Surgeons systematically resected the mediastinal lymph nodes. Stations resected included no. 2, 4, 7, and 9 on the right side, and no. 4, 5, 6, 7, and 9 on the left side. Tumor samples were obtained by surgical resection, and patients treated with chemotherapy or radiotherapy prior to surgery were excluded from the study. Among the recruited patients, 51 of them received adjuvant treatment. 46 patients received adjuvant chemotherapy only, 1 patient received adjuvant radiotherapy, and 4 patients received both chemotherapy and radiotherapy. All of the patients with stage IIA-IIIB were suggested to accept adjuvant chemotherapy, while patients with N2 lesion were advised to receive adjuvant radiotherapy. Some patients with stage IB were also suggested to receive chemotherapy due to the big size of resected tumor or other risk factors of recurrence. However, a portion of patients could not afford the side effect, or they refused adjuvant treatment from their own will. Written informed consent from each patient was obtained prior to sample collection. This study was approved by the medical ethics committee of Peking University Cancer Hospital (#2013KT31).

Genomic DNA was isolated using a DNeasy Blood and Tissue Kit (QIAGEN). DNA quantification and purity estimation were performed by NanoDrop ND 1000 spectrophotometer (Thermofisher).Baseline demographic and clinical characteristics, including smoking, age, gender, family history, pathological subtype, tumor node metastasis (TNM) classification, tumor stage, lymph node stage, differentiation status, adjuvant therapy after surgery and duration of survival after surgery, were obtained from medical records. Tumors were staged according to the UICC/AJCC (7th edition) for the lung.²⁴

Mutation analysis

An oncogene panel targeting 108 SCC-related genes as determined by a custom SeqCap EZ system (Roche NimbleGen, Madison, WI, USA) was designed in our previous study (data not published). The full FGFR (1–4) coding sequences (CDS) were included in this panel (supplementary Table S1) and the FGFR sequence data were used in this study. Approximately 1microgram of quantified genomic DNA from each SCC sample was used to prepare the DNA library, and the indexed samples were pooled for sequencing on a Hiseq2500 machine (Illumina, San Diego, CA, USA) with 2 × 150 bp pair-end reads. BWA, SAMtools and GATK methods were conducted to detect the SNVs. Based on data derived from NCBI dbSNP build 134, major germline SNVs were excluded. In addition, rare germline SNVs were excluded using the VN filtering method,²⁵ which is available as a tool from the Galaxy platform (http://galaxy-demo.ctmm-trait.nl/u/saskia-hiltemann/p/virtual-normal-analysis).

SQCC samples with FGFR genes mutations and the corresponding adjacent tissues samples were selected for further validation. PCR primers (primers are listed in Table S2) providing coverage of the mutant loci were designed using primer3 online software.²⁶ PCR fragments were amplified using standard procedures with the Q5® High-Fidelity DNA Polymerase (NEB, Ipswich, England), and the PCR products were sequenced using the ABI 3730 Genetic Analyzer (Life Technologies, Carlsbad, CA).

Each FGFR missense mutation was computationally analyzed for a predicted effect on protein function. Sorting Intolerant From Tolerant was used to calculate a SIFT probability score for the likelihood of the mutation to affect protein function.^27,28 Scores of 0.05 or less were predicted to affect protein function. Polymorphism phenotyping was also used, predicting either unknown (insufficient data for a prediction), benign, possibly damaging, or probably damaging mutations based upon characterization of the substitution site.²⁹

Statistical analysis

Spearman correlation coefficients and the corresponding P values were used to evaluate the association between two continuous variables. Associations between qualitative variables were tested using the chi-square test or Fisher’s exact test, contingent on distributional assumptions. Continuous variables were analyzed using the non-parametric Wilcoxon’s rank sum test. Univariate and multivariate regress analysis were employed to evaluate the risk ratio of FGFR mutation to clinicopathological features. Age (categorical, >60 years), sex, tumor differentiation, family history of malignant tumor and smoking history were used in the multivariate model. When evaluate the FGFR germline mutation, FGFR somatic mutation cases were censored in the control group; the same way was used to evaluate the FGFR somatic mutation. Overall survival (OS; date of operation until the date of death from any cause or last date of follow-up) and disease-free survival (DFS; date of operation until date of any sign of tumor relapse) were the primary endpoints. Patients without the event were censored at the date of the last follow-up. Survival curves were plotted using the Kaplan-Meier method, and significance was determined using the log-rank test. Multivariate analysis was performed using the Cox proportional hazard regression model with stepwise model selection. The hazard ratio (HR) and 95% confidence interval (CI) were calculated for each factor evaluated. All P values were 2-sided with a significance level of 0.05 Statistical calculations were performed using Stata 12.1 (StataCorp LP).

Results

Spectrum of FGFR mutations identified in 143 SCC patients without preoperative therapy

Of the 143 tested samples, we observed 22 different FGFR mutations among 24 tumors, amounting to an overall FGFR mutation rate of 16.8% (24/143). Twenty-one of the identified aberrations were missense mutations, and only one of them was a cds-indel (FGFR3- E135del). FGFR3-A500T, FGFR2-W290C and FGFR3-S249C were observed in the COSMIC database (http://cancer.sanger.ac.uk/cancergenome/projects/cosmic/), whereas the other mutations appeared to be novel. FGFR2-W290C and FGFR3-S249C were also found in TCGA SQCC samples.³

Eight distinct somatic mutations were identified by confirming the absence of the mutation in the paired adjacent lung cancer tissues. The remaining were also present in the apparently normal adjacent tissue so we defined them as potential germline mutations (hereafter referred to as germline mutation). The frequency of somatic mutations was 7% (10/143), and the frequency of FGFR germline mutation was 9.8% (14/143). Among the different FGFR genes, FGFR1 was associated with 3 somatic mutations and 3 germline mutations, FGFR2 was associated with 2 somatic mutations and 2 germline mutations, FGFR3 was associated with 2 somatic mutations and 6 germline mutations, and FGFR4 was associated with 1 somatic mutation and 3 germline mutations. Most mutations were unique, with the exception of FGFR3-S249C and FGFR3-A500T, which were observed in two cases each (Table 1). Although FGFR3-S249C is predicted to be damaging by Polyphen and SIFT,^27,29 it was also present in Exac and in dbSNP, as was FGFR3 which encodes A500T predicted to be tolerated by Polyphen and SIFT. Therefore we cannot exclude that they may represent bystander mutations.

Table 1. FGFR mutations identified in 143 patients with squamous lung cell cancer.

Sample	Gene	Exon	Mutations (CDS)^a	Mutations (Amino Acid)^b	Type	Polyphen	SIFT	MAF	Mutated ID^c
1	FGFR1	Exon 5	c.509A > G	E137G (ECD)	SM	PD	Tol	0.26	NA
2	FGFR1	Exon 8	c.842G > C	R250P (ECD)	SM	PD	Dam	0.32	NA
3	FGFR1	Exon 10	c.1200G > T	M369I (ECD)	GM	Benign	Tol	0.58	NA
4	FGFR1	Exon 11	c.1459A > G	M456V (CKD)	SM	Benign	Tol	0.37	NA
5	FGFR1	Exon 13	c.1693A > G	M534V (CKD)	GM	PD	Tol	0.33	NA
6	FGFR1	Exon 19	c.2476G > A	V795I (CKD)	GM	PD	Dam	0.50	NA
7	FGFR2	Exon 7	c.870G > T	W290C (ECD)	SM	PD	Dam	0.25	COSM41286 (5)^f
8	FGFR2	Exon 9	c.1216A > G	K405E (CKD)	GM	Benign	Dam	0.43	NA
9	FGFR2	Exon 11	c.1534G > A	A511T (CKD)	GM	Benign	Tol	0.53	NA
10	FGFR2	Exon 18	c.2390C > A	S796Y (CKD)	SM	PD	Dam	0.50	NA
11	FGFR3	Exon 2	c.87G > C	Q29H (ECD)	GM	PD	Dam	0.52	NA
12	FGFR3	Exon 4	c.403_405delGAA	E135del (ECD)	GM	NA	Dam	0.84	NA
13	FGFR3	Exon 5	c.490C > G	L164V (ECD)	GM	Benign	Tol	0.43	NA
14	FGFR3	Exon 7	c.746C > G	S249C^d (ECD)	SM	PD	Dam	0.34	COSM715 (1312)^f
15	FGFR3	Exon 7	c.746C > G	S249C^d (ECD)	SM	PD	Dam	0.22	COSM715 (1312)^f
16	FGFR3	Exon 11	c.1504G > A	A500T^e (CKD)	SM	Benign	Tol	0.74	COSM88800 (1)
17	FGFR3	Exon 11	c.1504G > A	A500T^e (CKD)	SM	Benign	Tol	0.53	COSM88800 (1)
18	FGFR3	Exon 12	c.1593G > A	M529I (CKD)	GM	PD	Dam	0.38	NA
19	FGFR3	Exon 13	c.1754A > T	K583M (CKD)	GM	PD	Dam	0.52	NA
20	FGFR3	Exon 18	c.2386C > G	L794V (CKD)	GM	PD	Dam	0.48	NA
21	FGFR4	Exon 3	c.146T > A	L49H (ECD)	GM	PD	Dam	0.66	NA
22	FGFR4	Exon 6	c.686C > T	A229V (ECD)	GM	Benign	Tol	0.51	NA
23	FGFR4	Exon 13	c.1727G > A	G576D (CKD)	GM	PD	Tol	0.47	NA
24	FGFR4	Exon 18	c.2266G > T	D756Y (CKD)	SM	PD	Dam	0.87	NA

^a mRNA Reference transcript: FGFR1, NM_001174067.1; FGFR2, NM_022970.3NM_022970; FGFR3, NM_001163213.1; FGFR4, NM_213647.1.

^b Amino acid numbers are derived from the following reference transcripts: FGFR1, NP_001167535; FGFR2, NP_001138387; FGFR3, NP_075254; FGFR4, NP_075252.

^c COSMIC mutation ID indicates that this mutation was included in the COSMIC database, and numbers inside the bracket represent the number of the mutated samples in the COSMIC database.

^d Two samples harbored the FGFR3 somatic mutation S249C.

^e The FGFR3 mutation A500T was detected in paired tumor and adjacent tissues of sample 13 but only in the tumor tissue of sample 14.

^f FGFR2-W290C and FGFR3-S249C were also found in TCGA SQCC samples.

Abbreviations: ECD: extracellular domain; CKD: cytoplasmic domain; SM: somatic mutation (only present in the tumor tissue); GM: germline mutation (present in both tumor and surrounding apparently normal tissue); MAF: mutant allele frequency; PD: Probably damaging; Tol: Tolerate; Dam: Damaging.

According to the protein topology characterization, the observed mutations resided in both the extracellular and cytoplasmic domains. Six germline mutations and one somatic mutation were located in the kinase domain (Figure 1). The somatic FGFR mutations tended to occur in an alternative exon or in an Ig-like domain (Figure 1), but no significant difference was observed in the distribution between somatic mutations and germline mutations (P = 0.193).

Figure 1. Alignment of altered amino acids encoded by the mutations observed in FGFRs.

Somatic mutations are represented by red circles, germline mutations are represented by green circles, and the purple circle indicates the FGFR3-A500 mutation detected in the paired tumor and adjacent tissues of sample 13 but only observed in the tumor tissues of sample 14. Mutations marked with an asterisk are potentially pathogenic sites (predicted as damaging sites by both SIFT and Polyphen2). The vertical axis represents the frequency of the mutation. FGFR2-S290C and FGFR3-S249C are hotspot mutation in squamous lung cancer samples which also be found in TCGA SQCC samples.

All of the listed FGFR aberrations were associated with a mutant allele frequency (MAF) greater than 0.22 (Table 1), which indicating that the Sanger sequencing method is sufficiently sensitive to detect FGFR mutant alleles for validation.

FGFRs mutation is associated with lymph node metastasis

To further evaluate the impact of the various types of FGFR mutations on the clinicopathological features of patients with SQCC, we analyzed FGFR somatic mutations, germline mutations and the overall FGFR mutation status separately.

Non-smokers were more likely to harbor a somatic FGFR mutation (P = 0.037) compared with smokers, whereas the frequency of FGFR germline mutation in smokers was not significantly greater compared with non-smokers (P = 0.425). TNM stage was correlated with the presence of FGFR overall mutations (P = 0.037, Fisher’s exact test). Patients at more advanced TNM stages exhibited a higher frequency of somatic or germline FGFR mutations. When FGFR somatic and germline mutation were considered separately, the correlation with TNM stage was no longer significant (Table 2).

Table 2. Clinicopathological features of lung squamous cell carcinoma patients harboring FGFR mutations.

	Total		FGFR somatic mutation (n = 10)			FGFR germline mutation (n = 14)			Any FGFR mutation (n = 24)
	No.	%	No.	%	P	No.	%^b	P	No.	%	P
Age					0.198			0.405			1
≤ 60	56	38.4%	6	10.7%		4	7.1%		10	18%
> 60	84	57.5%	4	4.8%		10	11.9%		14	17%
Gender					0.155			0.423			0.096
Male	120	82.2%	7	5.8%		11	9.2%		18	15%
Female	20	13.7%	3	15.0%		3	2.5%		6	30%
Smoking					0.037			0.423			0.048
Ever	120	82.2%	6	5.0%		11	9.2%		17	14%
Never	20	13.7%	4	20.0%		3	15.0%		7	35%
Stage					0.169			0.208			0.03
I	79	54.1%	3	3.8%		5	6.3%		8	10%
II	38	26.0%	4	10.5%		5	13.2%		9	24%
III	23	15.8%	3	13.0%		4	17.4%		7	30%
Ptrend^a					0.092			0.093			0.013
Differentiation					0.176			1			0.626
Well	4	2.7%	1	25.0%		0	0.0%	0.862	1	25%
Moderate	74	50.7%	6	8.1%		8	10.8%		14	19%
Poor	62	42.5%	3	4.8%		6	9.7%		9	15%
Ptrend^a					0.206			0.918			0.428
Lymph node metastasis					0.052			0.012			0.001
N0	96	65.8%	4	4.2%		5	5.2%		9	9%
N+	44	30.1%	6	13.6%		9	20.5%		15	34%

^a P_trend was derived by logistic regression analyses treating categorical variables as continuous variables.

The bold values indicate that the differences are significant.

Univariate logistic regression analysis revealed that the presence of any FGFR mutation in SQCC was associated with an increased risk of lymph node metastasis [odds ratio (OR) = 5.00, 95% confidence interval (CI) = 1.98–12.64, P = 0.001). When analyzing the data separately, a similar association was observed between FGFR germline mutation and lymph node metastasis (OR = 4.85, 95% CI = 1.52–15.55, P = 0.008). Association between FGFR somatic mutations and lymph node metastasis was also significant (OR = 4.05, 95% CI = 1.07–15.25, P = 0.039) (Table 2). After adjusting for age (categorical, >60 years), sex, tumor differentiation, family history of malignant tumor and smoking history, FGFR overall mutation (OR = 4.75, 95% CI = 1.78–12.7, P = 0.002) and germline mutation (OR = 5.56, 95% CI = 1.63–18.95, P = 0.006) remained positively associated with lymph node metastasis. TNM stage was not included in the multivariate unconditional logistic analysis because it was strongly correlated with lymph node metastasis (Spearman’s rho = 0.812, P < 0.001).

Survival outcome according to FGFR somatic mutation status

OS and DFS data were available for 140 patients (three patients were lost to follow-up). The log-rank test of the Kaplan-Meier survival analysis demonstrated that SQCC patients without an FGFR somatic mutation experienced longer OS (log rank P = 0.005) and DFS (log rank P = 0.004) than those with the FGFR somatic mutation (Figure 2). Patients with any type of FGFR mutation also experienced significantly shorter DFS (log rank P = 0.036) and OS (log rank P = 0.070). In contrast, no association was observed between FGFR germline mutation and patient survival.

Figure 2. FGFR genes mutation status and overall (OS) and disease-free survival (DFS) in SCC patients.

(a)/(c) Patients (10/140) with an FGFR somatic mutation (red curve) had significantly worse DFS and OS compared with patients with no FGFR somatic mutation (green curve). (b)/(d) Patients (24/140) with any FGFR mutation (red curve) had significantly worse OS and FGFR mutations exhibited a trend in OS compared with patients with no FGFR mutation (green curve).

To eliminate potential confounding effects, a multivariate Cox regression model was applied to adjust for confounding factors, such as sex, age, family history of malignant tumor, smoking history, tumor differentiation, TNM stage and postoperative chemotherapy/radiotherapy. The multivariate analysis confirmed that the presence of an FGFR somatic mutation was an independent poor prognostic factor for OS (HR: 4.26, 95% CI: 1.49–12.16, P = 0.007). Patients with FGFR somatic mutations also tended to have shorter DFS (HR: 3.16, 95% CI: 1.20–8.35, P = 0.020). In addition, higher TNM stages were also associated with poorer survival in DFS and OS, as expected (Table S3).

Discussion

Accumulating in vivo evidence has shown that tumors with FGFR aberrations frequently respond to FGFR inhibitors,^{15,21,30–34} and these findings provide a rationale for conducting clinical trials with these agents. Many FGFR inhibitors have been developed, such as JNJ-42756493 (Johnson & Johnson), AZD4547 (AstraZeneca) and BGJ398 (Novartis). In addition, some have been approved by the U.S. Food and Drug Administration (FDA), for example lenvatinib, which is the most recently FDA-approved FGFR-inhibiting drug for iodine-refractory, well-differentiated thyroid carcinoma. However, genomic FGFR aberrations are diverse, and patients with different FGFR aberrations may respond differently to different therapies. Therefore, it is important to both identify the spectrum and determine the clinical implications of different FGFR aberrations.

The functional effects of some known FGFR somatic mutations have been identified in vivo, the most frequent mutations are FGFR2-W290C and FGRF3-S249C,^35,36 which are implicated in driving cell transformation and mediating sensitivity to FGFR inhibitors in lung cancer cell line and mouse model.¹⁶ Two FGFR2 extracellular domain insertion mutations, FGFR2-A266_S267ins and FGFR2-290_291WI>C were proved to be both oncogenic and sensitive to inhibition by FGFR kinase inhibitors in NSCLS recently.³⁷ In other cancer such as endometrial carcinoma and human rhabdomyosarcomas, multiple FGFR2 mutations exhibit oncogenic potential and suggested to be novel therapeutic target.^14,34 FGFR2-S290C is in the immunoglobulin-like domain in the extracellular domain (ECD) that mediates the activation of the FGFR2 kinase by forming covalently bound receptor dimers.^16,37 The FGFR3-S249C mutation resides in the receptor’s extracellular linker region between the two Ig-like domains, a key site for ligand binding. In our dataset, among the 8 identified FGFR somatic mutations, 4 represented ECD mutations (Table 1) and we hypothesise that these mutations might exert effects similar to FGFR2-W290C or FGFR3-S249C in SCC. The remaining somatic mutations identified were located either in the TK domain or in an alternative exon of FGFR (Figure 1).

It is intriguing that FGFR2-W290C and FGRF3-S249C³¹ have also been detected in the germline causing the congenital developmental disorder Apert Syndrome and the related Pfeiffer and Crouzon syndromes.^18,38 Furthermore, functional germline mutation such as FGFR4-G388R has been described in lung cancer and predicted poor survival in lung adenocarcinoma patients^39,40 and recent genome-wide association studies have suggested that FGFR genes polymorphisms are associated with an increased risk of developing breast cancer.^41,42 The germline mutations that we identified were mostly localized to the tyrosine kinase and extracellular ligand binding domains, indicating likely pathogenicity as has been observed for other germline mutation of proto-oncogenes, for example RET mutations causing multiple endocrine neoplasia type 2.⁴³

To date, few studies have comprehensively evaluated the relationships of FGFR mutations with patient features and clinical outcomes in SQCC. Since different FGFR members have homogeneous protein structures and exerting similar biological functions in cellular signal transduction, we considered it reasonable to define them together as a whole. For example, both FGFR2-W290C and FGRF3-S249C mutants are sensitivity to FGFR TKIs through the same mechanism. In the meanwhile, the incidence of each mutation of FGFR genes is so low that it may not be helpful or meaningful to analyze them separately. We found that non-smokers were more likely to have somatic mutations than smokers. To our knowledge, this is the first study investigating a possible correlation between FGFR mutation status and smoking in lung cancer although such a correlation has been previously demonstrated for other tyrosine kinase receptors, such as EGFR, in lung adenocarcinoma.⁴⁴ However, the mechanisms by which smoking status contribute to tyrosine kinase receptors mutation is still largely unknown. Tumors with more advanced TNM stages or with lymph node metastasis were found to exhibit a higher frequency of either FGFR somatic or germline mutation. TNM stage is determined by tumor size and the presence of lymph node and distal metastases, and the strong correlation between TNM and lymph node metastasis was proven in our dataset (Spearman’s rho = 0.812, P < 0.001). Therefore, we presumed that an actual correlation exists between lymph node metastasis and the presence of an FGFR mutation, and the association with TNM stage might simply be a secondary consequence of the association with lymph node metastasis. FGFR activating mutations promote rhabdomyosarcomas metastasis in xenotransplant models,¹⁴ and our data suggests that both germline and somatic FGFR mutation may contribute to lymph node metastasis in SQCC. The association between somatic mutation and lymph node metastasis was not significant statistically in multivariate analysis, which may be due to inadequate statistical power resulting from the small number of FGFR somatic mutation cases. Therefore, we hypothesize that there are functional somatic and germline mutations among these FGFR mutations we found in this study, and the impact of each FGFR mutation on tumor metastasis merits further investigation.

The prognostic role of FGFR mutations in lung cancer was recently investigated. Wang et al. evaluated FGFR fusions and mutations, and no correlation between FGFR aberrations and survival in SQCC was observed.³⁶ However, this previous study had two significant limitations. First, the full coding regions of FGFR2 and FGFR3 were not evaluated and secondly, FGFR mutations and the FGFR fusions were combined for the survival analysis rather than evaluated individually. Evaluating every FGFR aberration independently is not feasible because the number of cases associated with each is limited, however; stratifying FGFR abnormalities by type is reasonable for association studies. We evaluated the prognostic roles of FGFR somatic mutations, germline mutation and the overall set of FGFR mutations in SQCC separately and found that patients harboring an FGFR somatic mutation experienced significantly worse DFS and OS. We also observed that more advanced TNM stage was a predictor of worse survival, as expected. Poorer rates of OS and DFS in SQCC patients with FGFR somatic mutations is consistent with our finding that tumors with somatic FGFR mutations had a greater frequency of lymph node metastasis, which is regarded to be associated with a poor prognosis. However, no association between FGFR germline mutation and survival was observed, suggesting that somatic mutations may have a greater impact on survival compared with germline mutations.

FGFR inhibitors have been developed as potential new therapeutic options in SCC with preclinical and early clinical data suggest that targeting certain FGFR aberrations with cognate inhibitors exerts beneficial therapeutic effects.⁴⁵ Therefore characterizing the landscape of FGFR genes aberrations and determining which mutations are clinically relevant is important to identify patients most likely to benefit from FGFR-targeted therapies and optimizing treatment outcomes.

The main limitation of our study resides in the relatively small sample size and patient selection. The patients in this study were from a Chinese population and the study did not include those patients with stage IV SQCC. Therefore, more studies will be warranted to validate our findings in an independent cohort, particularly in a Western population. Furthermore, the presumed germline mutations identified in this study were based on the finding of the FGFR mutations detected to be also present in the surrounding normal lung tissue (no paired blood samples were available for analysis) and so we were unable to exclude the possibility that they may represent somatic mutation. Although we verified the FGFR mutations identified against publically available databases we cannot exclude that some of the variants identified may represent previously unrecognized polymorphisms.

Our observed association between FGFR mutation and clinical outcome in SQCC, along with preclinical and early clnical data already reported suggest that targeting FGFR mutation with cognate inhibitors has therapeutic potential. There is also evidence that there are FGFR alterations that confer resistance to other types of cancer treatment and that some specific FGFR aberrations may demonstrate differential sensitivity/resistance to distinct FGFR inhibitors. FGFR mutation also have prognostic value. Because many FGFR changes appear to activate signaling, it is also important to characterize the clinically relevant effects of the many potent FGFR inhibitors that are currently in clinical trials. Molecular interrogation of patients for FGFR muations in the clinical research and practice setting may be warranted.

In this comprehensive analysis of FGFR genes mutations in SQCC, we demonstrated that FGFR mutation may increase the risk of lymph node metastasis in SQCC, and the presence of an FGFR somatic mutation specifically is a predictive independent factor for poor prognosis. Thus the findings from this study will help identify patients with SQCC most likely to benefit from FGFR-targeted therapies and so facilitate future clinical studies.

Abbreviations

Table

CKD	cytoplasmic domain
DFS	disease-free survival
ECD	extracellular domain
FGFR	fibroblast growth receptor
GM	germline mutation
MAF	mutant allele frequency
NSCLC	non-small cell lung cancer
OS	overall survival
PCR	polymerase chain reaction
SM	somatic mutation
SQCC	squamous cell carcinoma

Acknowledgments

The authors acknowledge the technical contributions of Jia Jian (Shandong Mental Hospital) in DNA preparation. It thanked very specically to the Professors Hong Cai and Yang Ke who contributed to the study conecept and design.

Disclosure Statement

The authors have no conflict of interest.

Supplementary data

Supplemental data for this article can be accessed on the publisher's website.

Additional information

Funding

This study was funded by Clinical and Basic Science Training Program of Peking University/Tsinghua University Life Science United Center, and Peking University (PKU) 985 Special Funding for collaborative Research with PKU hospitals;Clinical and Basic Science training program of Peking Universtiy/Tsinghua University Life Science United center;Peking University (PKU) 985 special funding for collaborative research with PKU hospitals;