Targeting FGFR in non-small cell lung cancer: implications from the landscape of clinically actionable aberrations of FGFR kinases

Zhen Zhou＊, Zichuan Liu＊, Qiuxiang Ou, Xue Wu, Xiaonan Wang, Yang Shao,5, Hongyan Liu, Yu Yang

1Shanghai Lung Cancer Center, Shanghai Chest Hospital, Shanghai Jiao Tong University, Shanghai 200030, China; 2Section No. 2 Internal Medicine, Cancer Center of Guangzhou Medical University, Guangzhou 511436, China; 3Translational Medicine Research Institute, Geneseeq Technology Inc., Toronto M5G1L7, Canada; 4Nanjing Geneseeq Technology Inc., Nanjing 211500,China; 5School of Public Health, Nanjing Medical University, Nanjing 211166, China; 6Department of Respiratory Medicine,The Second Hospital of Anhui Medical University, Hefei 230031, China; 7Department of Oncology, The Second Affiliated Hospital of Harbin Medical University, Harbin Medical University, Harbin 150086, China

ABSTRACT Objective: Dysfunction in fibroblast growth factor receptor (FGFR) signaling has been reported in diverse cancer types, including non-small cell lung cancer (NSCLC). The frequency of FGFR aberrations in Chinese NSCLC patients is therefore of great clinical significance.

KEYWORDS FGFR; oncogenic mutation; fusion; gene amplification; targeted therapy

Introduction

The fibroblast growth factor/fibroblast growth factor receptor(FGF/FGFR) signaling pathway plays important roles in a variety of biological processes, including development, differentiation, cell proliferation, migration, angiogenesis, and carcinogenesisviaseveral intracellular pathways, including the Ras/Raf/MEK and the phosphatidylinositol 3-kinase (PI3K)-AKT pathways1. The FGF family contains 22 members, which are usually divided into 7 subfamilies according to their sequence similarities, biochemical functions, and evolutionary relationships2. All 4 FGFRs, including FGFR1, FGFR2, FGFR3, and FGFR4 are structurally homologous to vascular endothelial growth factor receptors (VEGFRs), platelet-derived growth factor receptor (PDGFR), and other tyrosine kinase receptors3,and represent therapeutic targets of great potential.

Previous studies have shown thatFGFR2/3gene alterations,includingFGFR3activating mutations that affect either the extracellular (R248C and S249C) or transmembrane (G370C,S371C, Y373C, and G380R) domains of the protein, and gene fusions such asFGFR3-TACC3, are common in patients with urothelial carcinoma and cause constitutively activated FGF signaling, resulting in carcinogenesis4. Multiple FGFR inhibitors5, including erdafitinib6,7have shown antitumor activities in preclinical models and in early phase clinical trials involving patients withFGFRalterations. A recent study by Loriot et al.8reported that the use of erdafitinib was associated with an objective tumor response in 40% of previously treated patients who had locally advanced and unresectable or metastaticFGFRalteration-positive urothelial carcinomas. Such findings were superior to prior observations of an objective response rate of approximately 10% using second-line, single agent chemotherapy in an advanced urothelial carcinoma population9-11.

Activation of FGF signaling has also been described in lung cancer, including non-small cell lung cancer (NSCLC). As previously described, the incidence ofFGFRalterations, particularlyFGFR1amplification, was higher in squamous cell carcinoma (SCC) of the lung than in adenocarcinoma12. Moreover,FGFR2mutations were also reported in NSCLC patients,including the extracellular domain mutations, W290C and S320C, and the kinase domain mutation, K660E/N13. In this study, we investigated the landscape ofFGFRaberrations in a large Chinese NSCLC population by comprehensive genomic profiling using next-generation sequencing (NGS), to identify potential therapeutic options forFGFR-mutatedNSCLC patients.

Materials and methods

Patients

A total of 15,150 consecutive clinical lung cancer patients were analyzed using comprehensive genomic profiling targeting 400+ cancer-relevant genes, including all the exons ofFGFRgenes (FGFR1-4), as well as flanking intronic regions, and other introns selected by a Clinical Laboratory Improvement Amendments-certified, and College of American Pathologistsaccredited laboratory (Nanjing Geneseeq Technology, Jiangsu,China), as previously described14. We identified patients withFGFRalterations using a natural language search tool in the laboratory information management system database.Relevant demographic and clinical data were extracted from the database, including age, gender, date of diagnosis, histology, pathological stage, and evaluation of treatment response based on reports by clinical investigators.

For tumor tissue samples, the pathological diagnosis and tumor content of each case was confirmed by pathologists.Peripheral blood (8-10 mL) was collected in EDTA-coated tubes (BD Biosciences, San Jose, CA, USA) and centrifuged at 1,800 ×gfor 10 min within 2 h of collection to isolate the plasma for circulating tumor DNA (ctDNA) extraction, and white blood cells for genomic DNA extraction as the germline control.

DNA extraction and targeted enrichment

The ctDNA from plasma was purified using a Circulating Nucleic Acid Kit (Qiagen, Hilden, Germany) following the manufacturer’s protocol. Genomic DNA from white blood cells was extracted using the DNeasy Blood and Tissue Kit(Qiagen), while genomic DNA from formalin-fixed paraffin-embedded (FFPE) samples was purified using the QIAamp DNA FFPE Tissue Kit (Qiagen). All DNA was quantified using the dsDNA HS Assay Kit using a Qubit Fluorometer(Life Technologies, Carlsbad, CA, USA). Sequencing libraries were prepared using the KAPA Hyper Prep Kit (Roche, Basel,Switzerland), as described previously14. Indexed DNA libraries were pooled for probe-based hybridization capture of the targeted gene regions covering over 400 cancer-related genes for all solid tumors; all of which contained all exons ofFGFRgenes and selected introns for the detection ofFGFRfusions.

Sequencing data processing

Sequencing was performed using the Illumina HiSeq4000 platform (Illumina, San Diego, CA, USA), followed by data analysis as previously described15. In brief, sequencing data were analyzed by Trimmomatic16to remove low quality (quality ＜ 15) ornbases, and were then mapped to the human reference genome, hg19, using the Burrows-Wheeler Aligner(https://github.com/lh3/bwa/tree/master/bwakit). PCR duplicates were removed by Picard (https://broadinstitute.github.io/picard/). The Genome Analysis Toolkit (GATK) (https://software.broadinstitute.org/gatk/) was used to perform local realignments around indels and for base quality reassurance.Single nucleotide polymorphisms (SNPs) and indels were analyzed by VarScan217and HaplotypeCaller/UnifiedGenotyper in GATK, with the mutant allele frequency cutoff at 0.5% for tissue samples, 0.1% for cfDNA samples, and a minimum of three unique mutant reads. Common SNPs were excluded if they were present in ＞ 1% population frequency in the 1,000 Genomes Project or the Exome Aggregation Consortium(ExAC) 65,000 exome database. The resulting mutation list was further filtered using an in-house list of recurrent artifacts based on a normal pool of whole blood samples. Gene fusions were identified by FACTERA18.

Ethical approval

The study was approved by the Ethics Committee of Guangdong General Hospital, China (Approval No. GDREC2016262H).Shanghai Chest Hospital served as one of the hospitals participating in the research project. The study was conducted in accordance with the tenets of the Declaration of Helsinki,and written informed consent was collected from each patient prior to sample collection.

Results

The incidence of FGFR aberrations in NSCLC patients

From December 2016 to February 2019, a total of 15,150 individual clinical lung cancers were successfully evaluated by comprehensive genomic profiling using hybrid capture-based NGS. This work was based on the validated dataset for a total of 10,966 patients in our database system.Lung cancer tumor samples and liquid biopsies, if applicable,were compared to matched normal whole blood controls. A total of 87% of NSCLC samples examined were lung adenocarcinomas [lung adenocarcinoma (LUAC),n= 9,596], 9%were lung squamous cell carcinoma (LUSC,n= 954), and the remainder (4%) were of either mixed adenocarcinomas and squamous cell carcinomas or were missing sub-histological information in the database. Approximately 40% of the entire study population had only liquid biopsy specimens for genetic testing. A total of 210 patients (1.9%, 210/10,966)were identified with somatic aberrations ofFGFRs(FGFR1-4), including mutations, gene rearrangements, and gene amplifications (Figure 1A). Fifty-one patients (roughly 24%) had liquid biopsy samples including only plasma and pleural effusion samples. The median age of the cohort was 62 years of age (range: 34-84 years of age). Approximately 72% (152/210) of the patients were male. Approximately 61% ofFGFR-positive patients were LUAC (n= 128), 31%were LUSC (n= 65), and the remaining 7 cases were of either mixed or unknown histology. Thus,FGFRalterations were more frequent in LUSC patients (6.8%, 65/954) than in LUAC patients (1.3%, 128/9,596). The majority of theFGFRaberrations were gene mutations (75%) with gene amplification and gene rearrangements being observed in similar frequencies (10% and 15%, respectively) (Figure 1A).FGFR1alterations were slightly more abundant than alterations inFGFR2-4(Figure 1B). Notably, we observed more amplification events inFGFR1sthan in otherFGFRs, and over 90% ofFGFR4alterations were mutations (Figure 1C).

Enrichment of the activated PI3K pathway in the FGFR mutant cohort

We identified a total of 187 patients with somatic point mutations and indelsinFGFRs. The most frequent amino acid replacements across all FGFRs were FGFR3 S249C and R248C(Supplementary Figure S1). In particular, 19 patients representing ～0.17% (19/10,966) of the NSCLC population were identified withFGFR1-4oncogenic or likely oncogenic mutations according to the OncoKB database19(Figure 1D, Table 1,and Supplementary Table S1). The majority of these patients(68%, 13/19) had lung squamous cell carcinoma, and twothirds were male. Intriguingly, more than half of the 19 patients(58%, 11/19) had co-occurringPIK3CAaberrations, includingPIK3CAE545K (n= 3), E453K (n= 1), H1049R (n= 1),A1035T (n= 1),PIK3CAamplifications (n= 4), andPIK3R2G373R (n= 1) mutations. One patient had a concurrent activatingEGFRex19del, 4 patients hadKRASG12D/V or Q61L mutations, and the remaining 6 patients had no other known driver mutations (Table 1). A majority of the 19 patients withFGFR1-4oncogenic mutations (68%, 13/19) were systemic treatment-naïve, with the exception that 1 patient progressed on multiple lines of EGFR tyrosine kinase inhibitors 9TKIs0, including gefitinib, osimertinib, and afatinib, and 5 patients either received multiple lines of chemotherapy or chemotherapy in combination with radiotherapy or VEGFR antibody therapy (Table 1). Notably, the patient (P2) who received multiple EGFR TKIs likely acquiredFGFR3R248C and/or G380R to overcome the anti-tumor activity of TKIs,including osimertinib and afatinib, although pretreatment samples were unfortunately not available (Table 1).

Figure 1 Distribution of FGFR aberrations in a large population of Chinese patients with non-small cell lung cancer. (A) The frequency of FGFR aberrations among all cases and (B) the relative proportion of FGFR aberrations of FGFR genes among all cases, with the breakdown of FGFR alterations (C). (D) Co-mutation plot showing patients who carried FGFR oncogenic mutations, fusions, and gene amplifications, as well as concomitant aberrations of genes, including EGFR, RAS, and components of the PI3K pathway. An additional 9 patients with FGF19 amplifications were also plotted. The asterisk indicates mutations or fusions in FGFRs other than FGFR3. The triangle indicates non-FGFR1 amplifications.

The identification of novel FGFR fusions in NSCLC patients

FGFRfusions retaining the intact kinase domain were identified in 0.11% (12/10,966) of NSCLC patients examined(Figure 1D and Table 2). A majority of these patients (75%,

9/12) were positive forFGFR3-transforming acidic coiled-coil containing protein 3 gene (TACC3) fusions (FGFR3-TACC3),which were mostly reported in solid tumors20. Four of the 9(45%) patients withFGFR3-TACC3fusions had 5′ breakpoints inFGFR3exon 17 and the remaining 55% were in exon 18, whileTACC3exons 10 and 11 were the most common 3′ breakpoint locations (Figure 2A). We observed 1 case ofFGFR3exon 17 fused toTACC3exon 14 that may have resulted in a fusion protein with compromised dimerization capacity due to a truncated coiled-coil domain (Figure 2A).

Table 1 The demographical and clinicopathological characteristics of patients who had FGFR oncogenic mutations

Table 2 The demographical and clinicopathological characteristics of patients who carried FGFR fusions encoding intact kinase domains

We also observed 1 gene rearrangement event involvingFGFR2and an internexin neuronal intermediate filament protein α gene (INA) fusion (FGFR2 F17: INA I2) in a patient(P16) with stage IV lung adenocarcinoma (Figure 2B). TheFGFR2-INAfusion was previously reported in low grade gliomas that drove oncogenesisviaMAPK and PI3K/mTOR pathway activation21. Our observations represented the first case of aFGFR2-INAfusion in NSCLC, in particular, lung adenocarcinoma. Furthermore, 1 gene fusion event involving fibroblast growth factor receptor 4 (FGFR4) and the Rap guanine nucleotide exchange factor like 1 gene (RAPGEFL1)(FGFR4 F17: RAPGEFL1 R4) was detected in a lung adenocarcinoma patient (P26) (Figure 2C), which has not been previously documented, and therefore further validation of its function is necessary in future research. Notably, a concurrent activatingEGFRex19del mutation was also detected at an allele frequency of 21.71% in this patient. In addition, we observed 1 patient with a 5′-untranslated region of the Solute Carrier Family 20 Member 2 gene (SLC20A2) fused toFGFR1exon 17 (Figure 2D).

Figure 2 Visualization of FGFR fusions, including fusion partners, using the Integrative Genomics Viewer Browser. (A) The frequency of FGFR3-TACC3 fusions in the cohort. (B-D) The IGV screenshots display the reads from next generation sequencing and reveal FGFR fusions of(B) FGFR2-INA (F17:I2), (C) FGFR4-RAPGEFL1 (F17:R4), and (D) SLC20A1-FGFR1.

Of note, concomitantEGFRmutations orEGFRamplifications were observed in 6 of the 12FGFRfusion patients(Table 2), 4 of which were previously treated with EGFR TKIs,but the disease had progressed prior to NGS tests. Although half the patients (n= 2) did not have pretreatment samples, the remaining 2 patients (P16 and P17) likely acquiredFGFRfusions as alternative mechanisms to combat the anti-tumor activity of EGFR TKIs (Table 2). Furthermore, a concurrentPIK3CAH1047R mutation was observed in 1 patient (P22) and may also have acted as a mechanism of acquired resistance to prior therapies including TKIs (Table 2). No other known dominant driver mutations were detected in the remaining 6 patients (Table 2).

Amplification of the FGF19 and FGFR genes in NSCLC patients

As previously mentioned, we observed more amplification events inFGFR1than otherFGFRs(Figure 1B).FGFRamplification was detected in a total of 24 patients, a majority of which (87.5%, 21/24) wereFGFR1amplifications (Figure 1D).Similarly, the majority ofFGFR-amplified patients (67%) were LUSC and 92% were male (Table 3). Notably, 25 patients (12%,25/210) had multiple alterations inFGFRgenes, but oncogenicFGFRmutations, fusions, or gene amplifications were almost mutually exclusive events, with the exception that 4FGFR3-mutant patients had concurrentFGFR1amplifications(Figure 1D). Two patients had concurrentEGFRactivating mutations and received prior EGFR-TKI treatments. However,no pretreatment samples were available for mutation profiling for these patients. The remaining patients (92%, 22/24) had no other dominant driver mutations and were either chemotherapy-refractory or treatment naïve (Table 3).

We also identified 9 patients (0.08%, 9/10,966) who had amplifications ofFGF19(Figure 1D), which encodes a unique,high affinity ligand that specifically binds to FGFR4 in a heparin-dependent manner. Our observations were consistent with previous studies reporting on the role of the FGF19-FGFR4 signaling axis in human cancers, including hepatocellular carcinoma22and lung squamous cell carcinoma23. Two patients had concomitant aberrations of the PI3K signaling pathway,includingPIK3CAamplification and thePIK3R2G373R missense mutation (Table 3). All patients were either chemotherapy-refractory or treatment naïve.

Discussion

This study represented the first comprehensive survey ofFGFRaberrations in a large population of Chinese patients with NSCLC. Approximately 1.9% of the population hadFGFRaberrations, including point mutations, gene rearrangements,and amplifications, with the most common abnormality being FGFR point mutations. The prevalence of FGFR alterations in this Chinese NSCLC population was relatively lower than that of a prior study (5.7%), as reported by Helsten et al.24in which the study population was unlikely to be only Chinese.Currently, there are a number of FGFR inhibitors approved by the Federal Drug Administration (FDA), including ponatinib, regorafenib, pazopanib, lenvatinib, and nintedanib,which were included in a trial specifically targeting NSCLC patients25. All these FGFR inhibitors are multi-kinase inhibitors that also exhibit nonspecific anti-tumor activities against other tyrosine kinases, including VEGFR, PDGFR, ROS1, and/or RET. However, there are also specific FGFR inhibitors in clinical development. Notably, erdafitinib, a functionally selective pan-FGFR inhibitor, has been approved by the FDA to treat advanced metastatic urothelial cancers6,8. DifferentFGFRabnormalities responded differently to erdafitinib, with the highest response rate seen for patients with FGFR point mutations8. Another selective FGFR inhibitor, pemigatinib,was also recently granted accelerated approval for treatment of late stageFGFR2+ cholangiocarcinoma patients26. It is definitely of great clinical interest to study these FGFR inhibitors in NSCLC patients, so future trials may be warranted.

Unlike lung adenocarcinomas, no targeted molecular therapies have been developed for squamous cell lung cancers because targetable oncogenic aberrations are scarce in this tumor type. Here, we report thatFGFRaberrations were present in approximately 6.8% of the LUSC cohort of this study,which was higher than the frequency (1.3%) in LUAC patients.Notably, over 75% ofFGFR1amplification events were observed in LUSC patients, which is consistent with previous findings24,27. More than half of the patients who carriedFGFRactivating/transforming mutations had concurrent dominant mutations in PI3K pathway genes, includingPIK3CAandPIK3R2, consistent with previous reports28-30. Furthermore,we reported the overlapping of activatedFGFRgenes and genetic alterations of the PI3K pathway in NSCLC, including both LUAC and LUSC. A prior study by Packer et al.31revealed that PI3K inhibitors enhanced the anti-tumor efficacies of anti-FGFR inhibitorsin vitroin endometrial cancers in which the activation of the PI3K pathway was observed in ＞ 90% ofFGFR2-mutated cases. The activation of the PI3K pathway was also reported to be enriched in breast cancer patients with activatedFGFR/FGFsignaling32. Together, our findings highlighted an intriguing molecular feature and potential therapeutic target for combination therapies targeting the FGFR and PI3K pathways in FGFR-positive NSCLC patients exhibiting activated PI3K and MAPK pathways.

Table 3 The demographical and clinicopathological characteristics of patients who had FGFR and FGF19 amplifications

Furthermore, we identified a total of 12FGFRgene rearrangements in the NSCLC population that maintained intact FGFR kinase domains.FGFRfusions did not segregate well by histology or sex, as was previously reported by Wang et al.33which was likely due to the restricted cohort size.The majority of these patients wereFGFR3-TACC3positive,but we also observed 1 case of aFGFR2-INAfusion that was originally described in gliomas, and 2 novelFGFRfusions,includingSLC20A2-FGFR1andFGFR4-GAPGEFL1. A prior study by Wu et al.34reported a case of prostate cancer with theSLC45A3non-coding exon 1 fused to the intact coding region ofFGFR2,in which theSLC45A3-FGFR2fusion was predicted to drive the overexpression of wildtypeFGFR2.Thus, theSLC20A2-FGFR1fusion observed in the current study may also have been able to drive the overexpression of wildtypeFGFR1, although additional studies are needed to test this possibility. It is worth noting that half (n= 6) of theFGFRfusion patients carriedEGFRaberrations, includingEGFRex19del, T790M, C797S, andEGFRamplifications.Two-thirds of those patients received prior EGFR TKI therapies. Reminiscent of a prior report by Ou et al.35, this observation suggested thatFGFRfusions may act as a mechanism of acquired resistance to EGFR inhibitors in patients (P16,P17, P21, and P22) who were previously treated with EGFR TKIs.

Aside from point mutations and gene rearrangements,approximately 15% of allFGFRaberrations were amplifications, withFGFR1amplifications being the most common anomalies.FGFRamplifications predominated in LUSC patients at a prevalence of 1.6%, in contrast to that of ＜ 0.1%in the LUAC population. These frequencies were relatively lower than those reported by Helsten et al.24(9% and 4%,respectively), which could be attributed to a number of reasons including the ethnic differences underlying these two study populations, the restricted NSCLC cohort size of Helstenet al.,as well as the inclusion of cases who had only liquid biopsy ctDNA samples in this work.

Previous studies have shown thatFGFR1amplification was common in breast cancer patients with early relapses and poor clinical outcomes36. Therefore, antibodies targeting FGFR represent a valid therapeutic strategy to treat breast cancer or other cancer histologies, including NSCLC. In addition,we also observed a low frequency ofFGF19amplifications in our NSCLC population. FGF19 encodes the ligand for FGFR4,and it was previously shown thatFGF19amplifications corresponded with constitutive activation of FGF receptor 4(FGFR4)-dependent ERK/AKT-p70S6K-S6 signaling activation in head and neck squamous carcinoma cells37; thus, raising the question as to whether the FGF19/FGFR4 axis also acts as an oncogenic driver in these NSCLC patients and represents a therapeutic target.

Table 3 Continued

Conclusions

This study reported the frequency ofFGFRaberrations,including activating mutations, gene rearrangements, and gene amplifications in a large population of Chinese NSCLC patients, and revealed the potential clinical utility of targetingFGFRaberrations with FGFR inhibitors in NSCLC patients.We also reported novelFGFRfusion events in NSCLC patients,includingSLC20A2-FGFR1,FGFR2-INA, andFGFR4-GAPGEFL1; thus, highlighting potential therapeutic targets for the management of such patients.

Acknowledgements

We thank the patients and their family members who provided consent to present their data in this study, as well as the investigators and research staff at all research sites involved. Sincere thanks to Dr. Ryan Lamers for his professional editing and proofreading of the manuscript.

Grant support

This work was supported by the National Key R&D Program of China (Grant No. 2016YFC1303800).

Conflict of interest statement

Qiuxiang Ou, Xue Wu, and Yang Shao are employees of Geneseeq Technology Inc. Canada. Xiaonan Wang is an employee of Nanjing Geneseeq Technology Inc. China. The remaining authors have no conflicts of interest to declare.