Analysis of the DPYD Gene Implicated in 5-Fluorouracil Catabolism in a Cohort of Caucasian Individuals

Dihydropyrimidine dehydrogenase (DPD; EC 1.3.1.2) is the initial and rate-limiting enzyme in the catabolism of pyrimidines. The homodimeric protein catalyzes the reduction of uracil and thymine in an NADPH-dependent manner and, furthermore, plays a critical role in the pharmacokinetics of fluoropyrimidine-based anticancer drugs. Eighty percent to 85% of administered standard doses of the common chemotherapeutic agent 5-fluorouracil (5-FU) are rapidly degraded by DPD to inactive compounds followed by excretion of α-fluoro-β-alanine within 24 hours (1–3). The rationally designed orally administered fluoropyrimidine drug Xeloda (capecitabine) is converted in situ into 5-FU due to intracellular thymidine phosphorylase activity. Capecitabine has been shown to be safer and more effective than 5-FU and has greater patient convenience (4, 5).

A high variation of DPD enzyme activity has been observed in Caucasian populations. This variability is likely to influence the individual response of cancer patients to fluoropyrimidine drugs with respect to resistance or increased toxicity. In particular, loss of DPD activity may be associated with severe unexpected toxicity following a fluoropyrimidine-based therapy (6–9). Accordingly, a relationship between increased half-life of 5-FU in patients with DPD deficiency and life-threatening drug-adverse effects has been proposed. The risk of increased DPD deficiency–related toxicity may become particularly evident when the fluoropyrimidine is itself associated with a DPD inhibitor, such as with UFT/LV (uracil + tegafur/leucovorin; ref. 10).

Clinically apparent DPD deficiency is observed with variable phenotypes and genotypes initially having been described in pediatric patients (11). Most of the mutations associated with this inborn error of pyrimidine metabolism have been reported in these patients. Meanwhile, 39 different alleles in the DPYD gene encoding DPD protein have been identified, but no clear correlation to the phenotype of fluoropyrimidine-associated toxicity could be established thus far (7, 12, 13). In this context, a splice site mutation affecting the splicing donor consensus sequence of intron 14 has been reported as the most prominent mutation in the DPD-deficient patient cohort, resulting in the deletion of 55 amino acids in the native protein (14–17). In addition, a number of frameshift mutations (e.g., 295-298delTCAT) lead to an altered and truncated protein with subsequent loss of function (11, 18). For most of the missense mutations, however, genotype-phenotype correlations are still lacking. Due to these uncertainties accompanying the interpretation of genetic data and due to the rarity of complete DPD deficiency, routine DPYD testing has not been recommended by any health authority to this end. Hence, following the characterization of the DPYD gene (19, 20), and subsequently of its promoter (21), the classification of DPYD sequence variants as common polymorphisms or toxicity-related mutations will be of major importance to support a clinically based genetic test that could identify patients at risk for intolerance to fluoropyrimidine drugs.

We conducted a population study to obtain detailed information regarding genetic variations in the DPYD gene of Caucasian individuals and the relation of genotypes to function; rapid genotyping was made possible by using denaturing high-performance liquid chromatography (DHPLC) as a fast and sensitive method for screening large numbers of individual samples.

Blood donors. Blood samples of Caucasian individuals of different age and gender were randomly collected for testing. Written informed consent was obtained from all tested subjects on an institutionally approved protocol. Individual blood samples were further processed in the three research centers participating in this study. In detail, 157 samples were genotyped; 132 samples were processed for the DPD enzyme assay; and 182 blood samples were obtained for DPD mRNA quantification measurements. The pool of study subjects who were both genotyped and phenotyped is composed of 100 individuals.

PCR amplification of DPYD exons. DNA was prepared from frozen EDTA-blood samples (10 mL) using standard techniques. The entire coding region of DPYD was amplified with 23 primer pairs corresponding to 23 exons and the exon-intron boundaries as previously described (22). Amplicons were generated with the Expand High Fidelity PCR System supplied by Roche GmbH (Mannheim, Germany). PCR conditions are listed in Table 1.

Mutation analysis by denaturing high-performance liquid chromatography. The detection of DPYD sequence variants was carried out by DHPLC analysis (22). In brief, the mutation analysis was done on a Wave DNA fragment analysis system (Transgenomic, Omaha, NE) under partially denaturing conditions. PCR products were heated up to 95°C and slowly cooled down to generate heteroduplex molecules in the case of heterozygous DNA samples. The separation of heteroduplexes from intact double-stranded counterparts is mediated by ion-pairing HPLC using a hydrophobic poly(styrene-divinylbenzene)–based column matrix (DNA-Sep cartridge, Transgenomic) and a triethylammoniumacetate/acetonitrile eluent system. Optimal heteroduplex formation on this column occurs at a critical temperature that was predicted for each exon by melt software available either through the web site http://insertion.stanford.edu/cgi-bin/melt.pl (23) or by the Wavemaker calculation software (Transgenomic). To avoid missing any mutation in the course of establishing the method for the detection of DPYD variants, additional runs were carried out at one degree below and above the calculated analysis temperature(s). Because each sequence variation generates a specific peak pattern in DHPLC analysis, the respective chromatographic profile enabled the detection of identical variants in unknown samples by overlaying the chromatograms. Furthermore, the above-mentioned temperature profiling of each DNA fragment allowed the differentiation between similar peak patterns to identify the specific genetic variants (24, 25). The optimal system performance was checked by control samples (Transgenomic). Homozygous sequence alterations were detected by adding a known wild-type PCR sample to an unknown sample before the renaturation step was carried out. The DHPLC conditions used in this study are illustrated in Table 1.

Sequence analysis. DHPLC-screened samples indicating the presence of a new or rare mutation were once more amplified and the new PCR product was further analyzed on an ABI PRISM 3100 sequencing system (Applied Biosystems, Weiterstadt, Germany) using the Big Dye terminator technology according to the instructions provided by the manufacturer. The analysis was done by Sequencing Analysis 3.7 software and optionally by Factura 3.0 (Applied Biosystems) to facilitate the detection of heterozygote genotypes.

Hardy-Weinberg equilibrium. Genotype frequencies in the analyzed population were tested for deviation from Hardy-Weinberg equilibrium using a standard χ² test (26).

Haplotype analysis. Haplotypes were estimated from 23 loci with at least one variant allele using the program Hplus (27). The pairwise D′ and r² measures of linkage disequilibrium were estimated using the program LDA (28).

Determination of dihydropyrimidine dehydrogenase activity. Fresh heparinized blood samples (60 mL) were used for the preparation of intact peripheral blood mononuclear cells. DPD enzyme activity was determined from frozen cell pellets as described in ref. (1) using the HPLC radioassay developed by Johnson et al. (29).

Quantification of dihydropyrimidine dehydrogenase mRNA. Total RNA was isolated from freshly drawn EDTA-blood (0.5 mL) according to the High Pure RNA Isolation Kit purchased from Roche Diagnostics (Mannheim, Germany). The concentrations of DPD mRNA were determined by the DPD mRNA Quantitation Kit (Roche Diagnostics) following the instructions of the manufacturer and normalized to glucose-6-phosphate dehydrogenase expression.

Statistical methods. DPD enzyme activity and DPD mRNA expression data were correlated according to Pearson using the program SPSS 12.0.

For comparison of genotypes or haplotypes with functional data, DPD activity (nmol/min/mg protein) and DPD mRNA expression (normalized to glucose-6-phosphate dehydrogenase expression) were treated as dependent variables. All statistical testing (one-sided, 95% confidence) was done by bootstrap analysis with 10,000 replications. The replications included individuals with “missing” data so that different numbers of missing samples were present in different replications.

Distribution of DPYD mutations in a normal population of Caucasian individuals. A mutation analysis of the entire coding region of the human DPYD gene was done by DHPLC. Twenty-three different variant alleles in the coding and flanking intronic regions of DPYD were discovered in 157 Caucasian individuals (Table 2). The allelic frequency of the sequence alterations ranged between 0.3% and 19.4%, reflecting the presence of rare variants and common polymorphisms. The genotype frequencies in the study subjects did not differ significantly from those expected under Hardy-Weinberg equilibrium (Table 2). Some of the observed sequence variations have already been described as common polymorphisms in previous studies, e.g., 85T>C (DPYD*9A, Cys²⁹Arg), 1627A>G (DPYD*5, Ile⁵⁴³Val), 1896T>C (Phe⁶³²Phe), and 2194G>A (DPYD*6, Val⁷³²Ile; refs. 7, 12, 18, 30). Three rare exonic alterations have not been reported before: The novel mutation 1218G>A (Met⁴⁰⁶Ile) was present in two individuals. One of these individuals also harbored a deletion of TCAT in exon 4 (295-298delTCAT, DPYD*7); the other subject showed an additional undescribed missense mutation: 3067C>T (Pro¹⁰²³Ser). The third undescribed bp change, 1236G>A, occurred in one study subject and results in a silent mutation without amino acid change (Glu⁴¹²Glu).

Haplotype frequencies and linkage disequilibrium. We estimated the haplotype frequencies of the 23 loci consisting of at least one variant allele. Sixty-two possible haplotypes were found, 15 of them are common with a frequency of >1% (Table 3). No evidence for strong linkage disequilibrium was found for the common variants present in these haplotypes that was reflected by differences in the D′ and r² measures. Moderate association was estimated for the loci 496A>G (Met¹⁶⁶Val) and IVS10-15T>C (D′ = 0.773; r² = 0.387).

To classify the different mutations as common polymorphisms or alterations causing dysfunction, we compared the genotypes/haplotypes with functional data consisting of DPD enzyme activity and DPD mRNA expression measurements.

Determination of dihydropyrimidine dehydrogenase activity and DPYD gene expression. One hundred thirty-two healthy individuals were analyzed for DPD activity in peripheral blood mononuclear cells; 100 cases of these individuals had been both phenotyped and genotyped (see Materials and Methods). The activity values in this population ranged between 0.08 and 0.25 nmol/min/mg protein with a mean value of 0.143 nmol/min/mg protein. Regarding skew and kurtosis of the distribution curve (Fig. 1A), the DPD activities slightly differed from Gaussian Normal. None of the samples yielded values lower than 0.06 nmol/min/mg, which would point to a DPD-deficient phenotype according to Johnson et al. (29). The carrier of the frameshift mutation 295-298del TCAT in exon 4 (DPYD*7) had not been available for an additional DPD activity test. However, the well-described deletion of TCAT (11) in the genomic sequence destroys the regular protein structure of DPD and consequently there might be a partial loss of function regarding the heterozygous state of the mutation. Nevertheless, the DPD mRNA level of this individual was within the reference range (relative amount, 23.04). The relative gene expression levels in our population (n = 182) showed great variation with levels between 10.7 and 53.0 (m = 27.9); the distribution deviates substantially from Gaussian Normal (Fig. 1B). Gene expression data correlated with DPD activity data (P = 0.014; Fig. 2).

Comparison of dihydropyrimidine dehydrogenase activity data and genotype. The comparison of the activity data with genetic data (n = 100) supported the results of previous studies that the common nonsynonymous amino acid exchanges Cys²⁹Arg (nucleotide 85T>C), Ile⁵⁴³Val (1627A>G), and Val⁷³²Ile (2194G>A) do not affect DPD function (12, 13, 18, 30). Five homozygous carriers of Arg²⁹ and three homozygous carriers of Val⁵⁴³ showed almost normal enzyme activity ranges with m = 0.13 nmol/min/mg and m = 0.14 nmol/min/mg, respectively. The heterozygous amino acid exchange Val⁷³²Ile was associated with a mean (normal) activity value of m = 0.14 nmol/min/mg in our population. In line with these findings is the alignment of DPD protein sequences of various species showing nonconservation of the three mutation sites during evolution (31).

The novel mutations discovered in our study were not related to a reduced DPD activity as well. The variants 1218G>A (Met⁴⁰⁶Ile) and 3067C>T (Pro¹⁰²³Ser) were present in one individual with a normal activity value of 0.14 nmol/min/mg. No conservation of the corresponding amino acids was obvious among various species listed in ref. (31). The third novel point mutation, 1236G>A, resulting in a silent mutation without amino acid change (Glu⁴¹²Glu), also displayed a normal DPD activity value of 0.17 nmol/min/mg.

Correlation of genotype and functional data. The objective of the statistical analysis was to identify a possible association of the phenotype (lower/higher DPD activity) with functional DPYD variants. We also considered haplotypes in the statistical testing. Therefore, the common haplotypes with suitable frequency (>1%, as listed in Table 3) were implicated in the statistical analysis. According to bootstrap analysis (Table 4), two haplotypes carrying the variant C-allele of nucleotide 85 (hap 3) or the C-allele of nucleotide 1896 (hap 13) were associated with higher median activity values; the latter haplotype was also accompanied by higher median transcript levels of DPD (median difference: 5.371, 5% confidence: 2.538, and 95% confidence: 8.551). Two other haplotypes carrying the noncoding alleles IVS15+75 G and IVS11-106 A (hap 4 and hap 8, respectively) seemed to be associated with lower median DPD activity. These data were statistically significant, although the number of cases was low (Table 4).

The G-allele of the variant 496A>G (Met¹⁶⁶Val), which has been described in six patients with 5-FU–related toxicity (18, 30, 32), is contained in two common haplotypes. In our study, carriers of the 496 G-allele did not show any significant deviation from the median DPD activity of the population. Similarly, the corresponding haplotypes (hap 5/hap 11) were not significantly associated with reduced DPD function (Table 4). These data are in agreement with Johnson et al. (18) who proposed that the variant G-allele leading to the amino acid substitution Val¹⁶⁶ might be unrelated to the DPD-deficient phenotype as they identified the alteration in a cancer family member with normal DPD enzyme activity.

Rare variants of an allelic frequency below 1% (which did not occur in the 15 common haplotypes) particularly included the alterations 1601G>A (DPYD*4, Ser⁵³⁴Asn) and 2846A>T (Asp⁹⁴⁹Val), which have been related to a decreased DPD activity in previous studies (12, 33, 32). Interestingly, a modest but significant deviation of the median DPD activity values of the population was evident according to our statistical analysis, comparing either carriers of Ser⁵³⁴Asn or Asp⁹⁴⁹Val to noncarriers (Table 4). In contrast to the enzyme activity data, no statistically significant deviation from median DPD gene expression values was observed with these variants (not shown), suggesting that the effect is not due to altered gene expression.

The results obtained with Ser⁵³⁴Asn and Asp⁹⁴⁹Val by statistical analysis suggest a slight but significant impact of these amino acid substitutions on DPD function. The nonconservative amino acid substitution of Asp⁹⁴⁹ by valine occurs in a highly conserved region (31) near the iron-sulfur clusters of the enzyme, which play an important role in electron transfer during the catalytic reaction of DPD (34). The mutation Asp⁹⁴⁹Val located in exon 22 of the DPYD gene was detected in the heterozygous state in two of our study subjects who displayed enzyme activity levels ranging in the lower activity group (0.11 nmol/min/mg protein, both individuals). Furthermore, this mutation has also been reported in two patients with severe 5-FU–related toxicity, one of them with lethal outcome. The latter patient displayed additionally the amino acid exchanges Met¹⁶⁶Val and Ile⁵⁴³Val, and the truncating splice variant IVS14+1G>A (35). In the other individual, Met¹⁶⁶Val was present together with Asp⁹⁴⁹Val (32). In our study, Asp⁹⁴⁹Val occurred either as single mutation or in combination with the common polymorphism Ile⁵⁴³Val.

Compared to the mean activity values of all blood donors (m = 0.143 nmol/min/mg), the DPD activity values of the Ser⁵³⁴Asn carriers were in the lower group of values as well (m = 0.12 nmol/min/mg) including two individuals with even lower values, but within reference range (0.11 nmol/min/mg). Whereas the position 534 is—like codon 949—highly conserved during evolution (31), the Ser⁵³⁴Asn exchange has been controversially discussed up to now either as a polymorphism or as a variant implicated in DPD insufficiency (12, 13, 33). Accordingly, detailed analysis in homozygous genotypes would be required to confirm a clinically significant influence of Ser⁵³⁴Asn and Asp⁹⁴⁹Val on DPD function.

In the light of widespread use of 5-FU and the expected increase in use of the orally administered fluoropyrimidine capecitabine in cancer treatment (4), there is a debate as to whether individuals should be genotyped for DPD deficiency before treating them with a fluoropyrimidine-containing chemotherapy to avoid potentially life-threatening side effects. Testing 100 normal persons for both DPYD mutations and DPD function we did not find evidence for clinically significant DPD deficiency in our population of Caucasian origin. Altogether, our cohort presented with 11 different exonic mutations and 12 variations in the flanking intronic regions of the DPYD gene. No dramatic decrease of DPD enzyme activity was observed in the analyzed individuals showing that profound DPD deficiency is rare in Caucasians. However, statistical analysis of distinct missense mutations indicated a slight but significant reduction of the median enzyme activity due to 1601G>A (Ser⁵³⁴Asn) and 2846A>T (Asp⁹⁴⁹Val). The finding of significant shifts is all the more remarkable in that the genotypes were of heterozygous nature. Four haplotypes were also associated with altered median DPD activity including two intronic variations with no evidence for a pathogenic function. These effects may be related to loci elsewhere on the same chromosome being in linkage disequilibrium with the tested sites. Overall, the present results point toward a potential for better understanding the variability of individual tolerance of fluoropyrimidine drugs. The full biological effects of the observed amino acid changes need further evaluation by a detailed analysis of homozygous genotypes in vivo and in vitro studies on expression of distinct mutations.

We thank Paul D.P. Pharoah and Bettina Kuschel for the professional support in haplotype analysis and Steffi Neubauer for excellent technical assistance.