Pharmacogenetics focuses on identifying the role of a gene of interest that mediates drug-dependent mechanisms or triggers adverse effects. Therefore, dealing with the gene of interest is important to predict individual drug responses and toxicities. Genetic variations in genes of interest that interact with a drug may contribute to inter-individual differences in drug responses and play an important role in the designing of drugs that act on individuals with risk alleles. Recent pharmacogenetic studies have highlighted the role of genetic variations in several genes such as the UGT family, CYP family, EGFR, and Dihydropyrimidine dehydrogenase (DPYD) (1, 2).

DPYD is the initial and rate-limiting enzyme of the pyrimidine bases metabolic pathways, particularly fluorouracil (5-FU) catabolism. Recent studies have revealed that more than 80% of the medicated 5-FU, a commonly used chemotherapeutic agent for solid carcinoma, is rapidly degraded through the catabolism pathway (3, 4). Genetic variations of DPYD can cause an enzyme deficiency state, which results in severe toxicity or other adverse side effects such as DNA damage or RNA damage caused by imbalance of the nucleotide pool (5-7). Several genetic variations have been reported in previous studies. DPYD^*2A (rs3918290), located in the intron site, plays the role of an alternative splicing variant, and other polymorphisms such as DPYD^*9A (rs1801265), DPYD^*7 (rs72549309), DPYD^*8 (rs1801266), DPYD^*9B (rs1801267), and DPYD^*10 (rs1801268), affect DPYD enzyme activity in other ways (8-14).

In this study, we directly sequenced the DPYD whole gene in 288 subjects (96 Korean, 48 Japanese, 48 Chinese, 48 African Americans, and 48 European Americans). We also analyzed the linkage disequilibrium (LD) structures and minor allele frequencies (MAFs) of the discovered single nucleotide polymorphisms (SNPs) for the gene among the different ethnicities.

In order to discover DPYD SNPs, we directly sequenced 288 samples from five ethnic groups (Korean, Han Chinese, Japanese, African American and European American). As a result, 56 SNPs were found, including 18 novel SNPs. Among the novel SNPs, five (+199381A > G, Asn151Asp; +199404T > C, Phe158Phe; +221378A > G, Val162Val; +221531C > T, Asp213Asp; and +841847T > C, Leu993Arg) were located in coding regions (Table 1).

MAFs and relative physical coordinates of all SNPs are shown in Table 1 and Supplementary Fig. 1. Allele frequencies were nearly the same among the Korean, Han Chinese, and Japanese samples, whereas several SNPs showed different genetic distributions between Asians and other ethnic populations (African American and European American). Among those SNPs, the frequency of a core marker, ^*9A (rs1801265), in Asian populations was somewhat lower than in the African American and European American samples (MAF: Korean = 0.016, Han Chinese = 0.043, Japanese = 0.065, African American = 0.490, European American = 0.177). In contrast, other core markers, ^*7 (rs72549309), ^*8 (rs1801266), ^*2A (rs3918290), ^*9B (rs1801267), and ^*10 (rs1801268), were monomorphic in all the studied populations.

In order to find significant differences in allele frequencies between Korean and other ethnic groups, Fisher's exact test was additionally conducted (Supplementary Fig. 1). The test results indicated that there were significant differences between Asians and other ethnic populations (African American and European American) in the six SNPs (^*9A, rs668296, rs2811178, rs56160474, rs291592, and rs291593). Among them, a core marker ^*9A (rs1801265) showed the most significant differences (P = 6.61 × 10^-19 and 2.47 × 10^-6 for Korean vs African American and Korean vs European American samples, respectively). Moreover, the reversal of major and minor alleles was observed in rs291592 (C allele is major in Asians, but minor in African American and European American). Also, genetic difference was also observed within the Asians in rs291593 (T allele is common in Korean and Han Chinese, whereas it is rarely found in Japanese, African American, and European American). Detailed information about core markers such as star allele nomenclature, position, allele change, amino acid change, and any known roles in pharmacogenetics is presented in Table 1.

DPYD is an enzyme that takes part in a rate-limiting step of 5-FU catabolism. Previous studies have shown that the enzyme deficiency state of the 5-FU degradation pathway causes damage and degeneration of the central nervous system (8, 14, 16). Thus DPYD is known as a biomarker of severe toxicity in chemotherapeutic agents. Several DPYD polymorphisms have been reported as clinical loci associated with a reduced level of enzyme activity and severe 5-FU toxicity, and these polymorphisms are called "core markers". The most studied core markers are DPYD^*9A (rs1801265) and DPYD^*2A (rs3918290) (17-19).

The core marker ^*9A (rs1801265) is located in the coding region and induces amino acid change (cysteine to arginine) that may affect enzyme activity. It is relatively common in Caucasian populations (MAF > 10%), although DPYD enzyme activity is not affected by the polymorphism (20-23). This polymorphism is rare in Asian populations, but previous studies have reported that the incidence of clinical presentation of enzyme deficiency caused by heterozygous ^*9A (rs1801265) is significantly higher than the wild type (P < 0.05) in the Chinese population (23-26). In this study, frequency of ^*9A (rs1801265) showed a similar trend to that of previous studies (Supplementary Table 1).

Another important clinical locus, ^*2A (rs3918290), is in the splicing recognition sequence of the intron region. Genetic variation on this site can lead to deletion of the 165 base pair corresponding to the nearby coding region (exon 14), and consequently, the ceasing of enzyme activity (14). However, enzyme deficiency caused by ^*2A (rs3918290) is known to be rare in both Caucasians and Asians (Supplementary Table 1) (22, 24, 27, 28).

In addition, direct sequencing of DPYD in our study led to the discovery of a number of novel SNPs. In order to predict the function of the novel SNPs, in silico analyses were conducted according to the position of the polymorphisms. As a result, +842533C > T, located in the 3'-untranslated region (3'UTR), was predicted to introduce an internal ribosome entry site (IRES) motif. IRES allows the ribosome to bind to the mRNA internally and translate it rather than binding to the 5' cap as normally occurs (29-31). Although IRES is involved in 5'UTR mediated translation initiation, conserved secondary structures of IRES can influence the 3'UTR stability (32, 33). Moreover, a recent study in the Japanese population discovered a number of 3'UTR novel SNPs which were located near microRNA target sites (34). In addition, a previous study reported that binding of microRNA to 3'UTR negatively regulates the mRNA of target gene (35). Moreover, two nonsynonymous SNPs (+199381A > G, Asn151Asp and +841847T > G, Leu993Arg) were also found in our study. Polarity of the amino acid was affected by the charge of its side-chain (36). Polarity alteration between amino acids was predicted that Asn151Asp may affect protein structure or function (Asn, neutral; Asp, negative). Although frequency of Asn151Asp was low or monomorphic in our study subjects, it can be associated with the enzyme deficiency state.

Although PCR direct sequencing method was adopted for estimating the frequency differences across different ethnic groups and 48 samples per population is large enough to discover novel SNPs, a larger sample would be required to achieve more detailed screening result. Also, no functional study was conducted to further confirm the SNPs' role, although in silico analyses were performed to compensate for the lack of functional analysis to an extent.

In summary, the present study analyzed DPYD by directly sequencing 288 subjects from five ethnic groups. This yielded 56 SNPs, including 9 core SNPs and 18 novel SNPs. Moreover, we predicted the function of novel SNPs using in silico analyses. Although a lack of functional studies might be a limitation of the study, the results could make a valuable contribution to further research, especially pharmacogenetic studies of drug responses.