This study investigated the mutational landscape of ESCC in a Korean cohort. We identified several known and novel variants in ESCC. The observed median tumor mutation burden of non-silent variants (2.47/MB) was comparable to the reported mutation load of 1.9–3/MB.5–9,15 We identified several frequently mutated genes, such
as TP53, TTN, NFE2L2, ZNF750, NOTCH1, FSIP2, and KMT2D, which have been reported in ESCC. Mutated genes commonly found in over 10% of the ESCC samples in previous reports and this study are described in Supplementary Table S2. Except for TTN, NFE2L2, and FSIP2, most of these genes had nonsense, frameshift, or splicing variants, suggesting their tumor suppressor roles (Fig. 1).
The NRF2 pathway consisting of NFE2L2, KEAP1, and CUL3 was genetically deregulated in 37.2% of our cohort (Supplementary Fig. S1). NFE2L2 encodes a transcription factor that induces the production of a cytoprotective enzyme in response to oxidative stress, whereas the KEAP1/CUL-dependent proteasomal mechanism degrades the NFE2L2 protein under non-stressed circumstances.16 All NFE2L2 variants in our cohort involved two hot spots within the DLG and ETGE motifs that bind to KEAP1, as observed in a Japanese cohort.8 These mutations are thought to contribute to tumor development by stabilizing the NFE2L2 protein.17 There were NFE2L2 mutations in 9.6–16.7% of the cases in previous reports.6,8,18 Recently, in a Chinese cohort of 508 patients, NFE2L2 mutations were reported to be significantly associated with a worse prognosis of ESCC.19 But there was no significant association between the prognosis of ESCC and NFE2L2 mutations based on our data.
ZNF750, an epidermal differentiation regulator, is thought to be an ESCC tumor suppressor gene,7,8,15 as supported by our data, which showed that most ZNF750 variants were null variants (Fig. 1). ZNF750 is mutated in 3.9–16.7% of ESCC cases.6,18,20 In Japanese ESCC datasets, ZNF750 variants were positively associated with the APOBEC signature,8 but not in our study (p = 0.069).
We observed a mutually exclusive mutation pattern in NFE2L2–ZNF750 gene pairs in ESCC, although this mutual exclusivity has not been implicated in any type of cancer.21 Mutual exclusivity has been widely observed in cancer genomes and there are two major hypotheses associated with mutually exclusive mutated genes in cancers: the functional redundancy in downstream pathways and synthetic lethality hypotheses.22 Because these two genes do not share the same pathway, the mutually exclusive mutation in NFE2L2–ZNF750 gene pairs supports the synthetic lethality hypothesis.
NOTCH1 is frequently disrupted by loss-of-function mutations, implying that the loss of NOTCH pathway activity is critical for the growth of tumor cells with squamous differentiation characteristics.23 NOTCH1 is mutated in about 16% of cases.6,8,18 There are several reports that NOTCH1 is associated with a poor prognosis, and its mutations are mutually exclusive with PIK3CA mutations.5,10,24 However, there was no significant association between the prognosis of ESCC and NOTCH1 mutations based on our data. Nevertheless, the analysis to identify potentially druggable genes revealed NOTCH1 as a potential therapeutic target.
KMT2D is tumor suppressor gene that encodes histone methyltransferase and promotes the transcriptional activation of target genes by modifying histone H3 lysine 4 trimethylation (H3K4me3).25 KMT2D is mutated in 11–19% of ESCC cases,6,8,18 and Kaplan–Meier survival analysis showed that patients with non-silent variants of KMT2D had poor overall survival. Abulajiang et al. reported an association between KMT2D expression and the prognosis of ESCC, with KMT2D overexpression predicting poor clinical outcomes and facilitating ESCC tumor progression.26
LRP1B is putative tumor suppressor gene that encodes low-density lipoprotein receptor-related protein 1b27. LRP1B is frequently deleted in various tumors and was deleted in 20.8% of a Japanese cohort8 and mutated in 25% of an Indian cohort.5 A study of LRP1B gene expression in ESCC showed that 42.9% of primary esophageal cancer cases have homozygous LRP1B deletions, and LRP1B mRNA expression was frequently silenced in cell lines without homozygous deletions (37.8%).28 Bisulfite-PCR analysis and sequencing showed that LRP1B-nonexpressing cells without homozygous deletions were highly methylated at a LRP1B CpG island. Recently, Brown et al. reported better immune checkpoint inhibitor treatment responses in a group with likely pathogenic/pathogenic variants in LRP1B genes compared with a group carrying variant of unknown significance, indicating that mutations in the LRP1B gene have significant implications for the prognosis and treatment of multiple cancer types.27 All LRP1B variants in our study were missense variants. Further studies of their expression and methylation are needed to clarify the roles of these LRP1B missense mutations.
Kiran et al. showed that tobacco chewers had a higher frequency of mutation signature 4 than did smokers and non-users of tobacco in an Indian ESCC cohort.5 There were no tobacco chewers in our study and no distinct signatures associated with ESCC in patients with a smoking history.
Our study is the first report of the mutational landscape of ESCC in a Korean cohort. It provides insight into molecular alterations in ESCC in Koreans and reveals potential candidates for therapeutic targeting. Our data suggest that KMT2D is a molecular prognostic marker. Further studies need to examine the role of KMT2D in ESCC prognosis due to the small size of our cohort. Additionally, copy number and expression studies of ESCC in Koreans are needed, as none were conducted here.