In this study, we performed a targeted deep sequencing analysis of the kinome in a panel of 14 GC cell lines. Our major goals of this study were to assess the utility of targeted sequencing for accurately identifying genetic variants, and to assemble a comprehensive catalog of kinome genetic variants in the 14 commonly used cell lines or experimental models of GC. We found the targeted deep sequencing methodology to be highly robust. Virtually, all the 8,425 targeted exons in our study were captured and we achieved an overall sequencing depth of ≥20× for 90% of the sequenced kinase genes. When benchmarked against Sanger sequencing and SNP genotyping arrays, we achieved an SNV detection specificity of 93.9% and sensitivity of 95%.
We identified more than 300 novel nonsynonymous SNVs in kinase exons. The novel SNVs identified likely comprise rare germline variants and somatic mutations (32
). Both categories are important to document, as germline variants in certain kinases have been associated with an increased risk of cancer (41
). In our study, a confirmed germline variant in TNK2
(R445W) was located at a conserved residue in the kinase domain, and found in 2 of 14 GC cell lines and 1 of 48 GC tissues, both of which are mostly Asian origin. It has been shown that TNK2
is overexpressed in breast cancers where its expression is correlated with poor prognosis (42
). It would be interesting to compare the prevalence of this novel variant in the Asian healthy and GC population.
Another interesting variant we identified was a homozygous nonsense substitution (Q37X) in the STK11/LKB1
kinase in YCC16. Inactivating germline mutations in STK11/LKB1
cause Peutz–Jeghers Syndrome (PJS), an inherted disorder associated with intestinal hamartomatous polyps and frequent gastrointestinal tumors (41
). To date, 4 GC-related STK11
mutations have been reported (3757–758insT, Arg297fsX38, Leu117PhefsX46, and P324L), and of these, 3 were also associated with PJS and early onset GC (43
). Screening for STK11/LKB1
mutations in patients with early onset GC might, thus, identify additional patients with PJS.
Our study identified several interesting findings potentially related to GC tumorigenesis. We discovered that kinases related to MAPK
signaling exhibited a significantly enriched tendency to harbor “non-benign” genetic alterations, with 11 of 14 lines having at least 1 affected MAPK
gene. Abnormalities in MAPK
signaling have been shown to impinge on many phenotypes of cancer including independence from proliferation signals, evasion of apoptosis, and unlimited replication potential (47
). One of these MAP
kinases was MAP2K4
, which was altered in 2 lines. Expression of MAP2K4
suppressed cell growth in vitro
. Inactivating mutations in MAP2K4
have been found in approximately 5% of tumors from various tissues (33
) and recent evidence supports the functional role of MAP2K4
in human cancer (48
). However, to date, somatic mutations in MAP2K4
have yet to be reported in GC. Our findings suggest that it might be worthwhile to characterize MAP2K4
in an expanded panel of GC tumors.
Besides SNV detection, the kinome resequencing data allowed us to identify several kinase-related structural variants, including variants affecting GUCY2F, TYRO3, SBK2
, and the ERBB2
RTK. We demonstrated the expression of novel CDK12-ERBB2
fusion transcripts. However, as these transcripts are not expected to disrupt ERBB2 protein translation, their functional importance remains to be determined. Finally, the ability to infer genetic variation at multiple levels (SNVs, structural variants, and copy number) allowed us to perform integrative analysis utilizing data from a single deep sequencing platform. Using the example of the MET
proto-oncogene, we identified Hs746T where MET
was both amplified and associated with a splice-site point mutation. Such “dual mechanisms” of oncogene mutation and amplification have also been proposed for EGFR in lung cancer (49
). Interestingly, in lung cancer, activation of MET
by gene amplification and by splice-site mutations deleting the juxtamembrane domain appears to be mutually exclusive (50
), suggesting that either type of the alterations may confer growth advantage to lung epithelial cells. In contrast, our results reveal that in GC, a paradigm of “dual mechanisms” (amplification and activating mutation) may impinge on MET
In conclusion, we identified more than 300 novel micro- and macrogenomic alterations involving kinases, many of which may contribute to GC development. MAPK-signaling genes are identified as frequently altered kinases in our study, suggesting a causal role for dysregulated MAPK pathway in GC tumorigenesis. Our results may contribute to understanding the genetic architecture of this important subset of the human genome in GC and facilitate the usage of these GC models in the laboratory.