Discussion
We evaluated the effects of 21 nsSNVs of 18 candidate genes for NK-AML in this study. Among 18 genes, we replicated previously reported associations of two genes (CNTNAP2 and GNAS) with AML [9, 10]. The mutation in the MUC4 gene has been reported to be associated with acute lymphoblastic leukemia [18, 19]. The MUC4 gene encodes a mucin protein and a high-molecular-weight glycoprotein in humans. This integral membrane glycoprotein, which is observed on the cell surface, plays various roles in tumor progression [20]. Particularly, MUC4, complexed with v-erb-b2 erythroblastic leukemia viral oncogene homolog 2 (ERBB2), results in the repression of apoptosis and the stimulation of proliferation; in addition, the overexpression of MUC4 in carcinoma cells promotes loss of cellular polarity by ERBB2-mediated disruption [21]. The CNTNAP2 gene is one of the largest genes in the entire human genome and occupies approximately 1.5% of chromosome 7. This gene encodes a neurexin family member that functions in the vertebrate nervous system. The GNAS gene has a highly complex imprinted expression pattern. This gene provides instructions for making one component, the stimulatory alpha subunit, of a protein complex called a guanine nucleotide-binding protein (G-protein). G-protein alpha subunit regulates the cyclic AMP (cAMP) pathway. GNAS gene mutations are known to be associated with high cAMP signaling [22]. The novel associations between 15 somatic mutations (PPIAL4G, LRRN2, OR2T33, ANKRD36, ATP13A3, TRIML1, PABPC1, MUC5B, TAS2R43, ATF7IP, KIAA1033, ATP8A2, L2HGDH, LYPD5, and SUSD2) and NK-AML need to be replicated, and their functional mechanisms in the AML should be investigated in future studies.
The GRS models that were comprised of the somatic nsSNVs showed extremely high predictive accuracy for the risk of NK-AML (86% to 100%). The combination of five nsSNVs of previously reported genes (MUC4, CNTNAP2, and GNAS) had a predictive ability of 98% for the risk of NK-AML in this study. One of the limitations in this study was the small number of study subjects (n = 10), which resulted in substantial variations in the 95% CIs of the effects (ORs) of each SNV.
In conclusion, we have highlighted 21 susceptibility nsSNVs located in 18 genes. The GRS model that comprises five candidate SNVs is highly informative in predicting the risk for NK-AML. The discovery of novel markers may provide an opportunity to develop novel diagnostic and therapeutic targets for NK-AML. Further study with a larger sample size is necessary to validate the AML-related gene mutations and will provide an opportunity to develop a powerful genetic risk prediction model for NK-AML.