Mutational Landscape of DDR2 Gene in Lung Squamous Cell Carcinoma Using Next‐generation Sequencing

Micro‐Abstract Discoidin domain receptor 2 (DDR2) alterations were identified as a promising therapeutic target in lung squamous cell carcinoma. Using a large cohort, DDR2 genetic landscape was assessed by next‐generation sequencing. DDR2 mutations are present in 4% of squamous cell carcinoma, are not exclusive with driver genes alterations, and have no prognosis impact. This study emphasizes the need for a better knowledge of DDR2 biology before developing dedicated targeted therapies. Background: Lung cancer represents the leading cause of cancer‐related death worldwide. Despite great advances in lung cancer management with the recent emergence of molecular targeted therapies for non‐squamous non–small‐cell lung cancer, no dramatic improvements have been achieved in lung squamous cell carcinoma (SCC). Mutations in discoidin domain receptor 2 (DDR2) gene were recently identified as promising molecular targets in this histology. The aim of this study is to describe the DDR2 mutational landscape of lung SCC and investigate the associated clinical factors. Methods: Next‐generation sequencing of the DDR2 gene was performed on 271 samples of lung SCC. Patients followed in our institution from January 2011 to August 2014 were retrospectively selected for data collection. Other driver gene alterations (EGFR, KRAS, BRAF, HER2, and PI3KCA) were analyzed using pyrosequencing. Results: A total of 11 patients harboring a DDR2 mutation was detected among the 271 sequenced lung SCC samples (4%). We describe 10 unreported mutations, comprising a novel DDR2 exon 7 splice mutant. DDR2 mutations were not mutually exclusive with other driver gene alterations. One hundred thirty‐six patients were included for clinical comparison and logistic regression analysis. No difference was detected between DDR2‐mutant and DDR2 wild‐type lung SCC regarding clinical characteristics or survival. Conclusion: DDR2 mutations were observed in 4% of cases of lung SCC of European descent. DDR2‐mutated tumors can exhibit other driver gene alterations. No clinical characteristics were significantly associated with DDR2 mutation.


Introduction
Lung cancer is the leading cause of cancer-related death worldwide, accounting for more than 1.5 million deaths in 2012. 1 Non-small cell lung cancers (NSCLCs) represent approximately 85% of lung neoplasms; among these, squamous cell carcinomas (SCC) account for approximately 30% of cases. The landscape of thoracic oncology has greatly changed during the last decade with the discovery of tumor driver mutations predicting response to targeted therapies. 2 Recent clinical trials demonstrated an unprecedented improvement in progression-free survival (PFS) and quality of life, especially in molecularly selected lung adenocarcinoma. [3][4][5] Lung SCC lacks available molecular targets, and tyrosine kinase inhibitors are mostly ineffective in unselected populations 6 despite many efforts made to understand its genomics. 7 Among all of the somatic alterations explored to date in lung squamous histology, Discoidin Domain Receptor 2 (DDR2) gene mutations are reported in approximately 3% of cases. 8,9 DDR2 is a tyrosine kinase receptor that belongs to the same family as Epidermal Growth Factor Receptor (EGFR). 10 Numerous in vitro studies on various tumoral cell lines demonstrated the important role of DDR2 in the regulation of cellular proliferation, migration, metastasis, and secretion of matrix metalloproteinase. 11 In 2011, Hammerman et al showed that a subset of DDR2 mutants are oncogenic in SCC cell lines in vitro, delivering a strong rational for targeting DDR2 mutations. 8 In addition, two patients with DDR2-mutated lung SCC were reported in the literature to experience a dramatic response to dasatinib, a potent SrC inhibitor. 8,12 Altogether, these data establish DDR2 mutation as a promising molecular target of tyrosine kinase inhibitors in lung SCC.
Few studies are available concerning the clinical characteristics of patients harboring DDR2 mutations. The first aim of this study was to describe DDR2 gene mutations in a large monocentric cohort of SCC. Further objectives of the study were to compare the clinical characteristics of DDR2mutant to DDR2 wild-type (WT) lung SCC and to investigate factors associated with DDR2 mutations.

Tissues and patients
Samples were obtained from patients who were diagnosed with SCC by tissue biopsy or primary surgical resection from January 2011 to January 2015 in five hospital centers in Brittany (north-west of France). A total of 271 lung SCC samples were retrospectively enrolled for DDR2 gene sequencing on a centralized platform. For demographics, clinical and survival comparisons, we chose to only select patients diagnosed and followed from January 2011 to August 2014 in Rennes University Hospital to avoid site-dependent care bias and limit retrospective data missing in patient's record. The choice of this recruiting period allows a minimal follow-up of at least 12 months for all patients.
Subsequently, a total of 136 patients with lung SCC composed the investigation cohort (appendix 1).
Central localization of the tumor was defined as proximal to the third bronchial division and advanced disease as unresectable stage IIIB or stage IV tumors according to the 7 th IASLC TNM classification for lung cancer. 13 Concerning the smoking exposure, patients were classified as active smoker, former smoker (no exposition during the last 12 months at diagnosis), never-smoker or unknown. The study was approved by the Rennes University Hospital Ethics Committee (n° 15.87).
DDR2 gene sequencing, sequence processing, variant calling and analysis of DDR2 mutations All tumor tissues were subjected to review by a pathologist to estimate the tumor cell content and 10 µM-thick formalin-fixed paraffin-embedded tissue sections were sampled. A tumor cell cut-off of at least 10% was used to proceed further with the sample. Normal tissue was also sampled when matched non-tumoral sequencing was indicated. Genomic DNA was extracted from routine FFPE samples using the MagDEA DNA 200 kit (Precision System Science®, Japan). Genomic DNAs from all samples were quantified with the Quan-iT PicoGreen dsDNA assay kit (Invitrogen). A panel of 36 specific primers targeting the DDR2 gene was created using AmplifX (see appendix 4 for sequence). Next, libraries were pooled together to create one PCR product library. The resulting sequencing-ready amplicon libraries were sequenced on a MiSeq sequencer (Illumina) using MiSeq Reagent Kit v2 (300 cycles). Each patient was sequenced twice with independent amplicon library preparation. The bioinformatics pipeline to detect low frequency variants developed for molecular testing in a diagnosis purpose was used to detected alterations in the DDR2 gene. Briefly, reads were aligned to the human reference genome GRCh37 (hg19) using BWA software (bwa-sw 0.5.9), recalibrated and realigned using GATK (GATK-2.1-11). NGS amplicon reads were split into non-overlapping bam files. These The methodology was based on pyrosequencing assays, as previously described. 17 ALK and FGFR genes fusions were not include in the analysis as mainly not available in our cohort (not routinely assessed at the time of inclusion).

Statistical analysis
Demographic and descriptive data are given as the median with the range. Categorical variables were compared with the Fischer exact test or Pearson chi-squared test, and quantitative variables were compared with the Mann-Whitney U test when appropriate. Overall survival was calculated in months from the date of diagnosis to the date of death (failure) or the date the patient was last known to be alive (censored). Two-tailed p-values were reported, with p<0.05 considered as statistically significant. The Kaplan-Meier method with log-rank test were used to perform survival analysis.
Finally, a binomial logistic regression model was built with the DDR2 status (mutant or wild-type) as the dependent variable. GraphPad Prism® for Windows software (version 5.03; Graph Pad Software Inc., Los Angeles) was used for the statistical and survival analyses.

Results
Frequency, distribution and characteristics of lung SCC DDR2 mutation Next generation sequencing was performed on 271 samples of lung SCC. Percentage of DDR2 gene coding sequence coverage was 100%, with a minimum of 600X depth for each sample (see appendix 5). DDR2 mutations were found in 4% (11/271) of the cases. Table 1 summarizes DDR2 mutations characteristics. The majority of DDR2 point mutations observed were nucleotide transversions (73%).
Prediction of the functional impact on the DDR2 protein reveals that four mutations had a neutral effect whereas seven showed a deleterious effect. DDR2 mutations predominantly map onto the discoidin homology domains and the kinase domain, each comprising 36% of the mutations ( Figure   1A). Only one DDR2 mutation (c.1857G>C, p.R619S) described in this study was previously reported in lung SCC. 18 Table 3 and clustered genomic profile of the investigation cohort is presented in figure 2. The two groups were composed of male heavy smokers (81% and 93% tobacco exposure in the DDR2-mutant and the DDR2-WT groups, respectively) with a high level of comorbidities, mainly cardiovascular disease, in approximately two-thirds of the cases in the entire cohort. No differences in age, gender, smoking history, pack-year, comorbidities, Eastern Cooperative Oncology Group (ECOG) performance status, disease stage or follow-up were observed between the two groups.
Factors associated with DDR2 mutation and prognosis impact of DDR2 mutations Performing logistic regression, we investigated factors associated with DDR2 gene mutational status.
In univariate analysis, no variables were significantly associated with DDR2 mutant lung SCC (see

Discussion
In this study, we provide one of the largest description of DDR2 genetic landscape in lung SCC to date. We identified eleven DDR2 mutations (4%) and report for the first time, a splicing mutant c.566-1G>C of the DDR2 gene in lung cancer. The following rates of mutation were previously observed: 2% (2/100), 3.2% (9/277), 4.6% (4/86), 1.1% (2/178), and 1.3% (3/178)  mutation in their Japanese cohort comprising more than one hundred cases. Of note, sequencing coverage of DDR2 gene is very different between these studies and may contribute to such prevalence inconsistency. Some authors have suggested that DDR2 mutations may be related to tobacco exposure. 23 To support this notion, the largest percentage of DDR2 mutations described in this report are transversions, known to be smoking-related. 24 Epidemiologic studies showing that East Asian populations are less susceptible to smoking-related lung cancer 25 might also partially explain these geographic variations.
We identified a new DDR2 truncating mutation p.E85X that is predicted to have a major impact on the receptor function. Nevertheless, to date, some discrepancies exist in the literature concerning the oncogenic potential induced by DDR2 aberrant signaling. Functional validations assays are needed to assess the oncogenic properties of this variant. Moreover, DDR2 mutations appear not to be exclusive from other driver gene alterations. This is supported by the observation of co-occurrence of such mutations with a KRAS p.G12D mutation in our cohort. However, known-driver mutations frequencies appear lower in our cohort than expected, comparing to TCGA 7 . This could be explained conditional loss of TP53 is oncogenic in a murine model. 26 However, a recent publication showed the presence of DDR2 p.L63V and p.G505S alterations not only in tumor specimens but also in matched normal tissues, suggesting that these variants are inherited. 27 Some of these variants, also identified in our cohort, have already been reported in the NHLBI Exome Sequencing Project 28  Some limitations to our study should be noted. First, the sample size of our study was limited.
As DDR2 mutations are rare in lung SCC, it is possible that the power of our study was insufficient to detect potential associated factors. Second, a significant proportion of the 271 sequenced patients had to be withdrawn from the data collection to avoid missing information. Consequently, potential selection bias could have influenced our findings. Third, this study was monocentric, and reproducibility of the molecular results could not be assessed on multiple genomic platforms. Fourth, we cannot definitively confirmed the non-germinal status of one of the new p.P492S DDR2 variant described in this cohort, as samples with healthy tissue were not available for this patient.

Conclusion
In conclusion, DDR2 mutations were detected in 4% of cases in our lung SCC cohort. Our study identified ten new DDR2 mutations that have not been described to date, including a unique DDR2 splice mutant and shows that DDR2-mutated tumors can exhibit other driver gene alterations.
No clinical characteristics or survival difference was detected between DDR2-mutant and DDR2-WT lung SCC, and no clinical factors were significantly associated with DDR2 mutation. A better understanding of DDR2 biology and its mutants will be critical for developing future dedicated therapy in molecularly selected DDR2-driven tumors.