DPYD and TYMS polymorphisms as predictors of 5 fluorouracil toxicity in colorectal cancer patients

Abstract Colorectal cancer (CRC) is the third most common cancer and the fourth leading cause of cancer death. 5-Fluorouracil (5-FU) is an essential component of systemic chemotherapy for CRC. Our objective was to determine the genotypic frequency of polymorphisms affecting dihydropyrimidine dehydrogenase (DPYD) and thymidylate synthetase (TYMS) genes and to correlate the genetic profile with the toxicity due to 5-FU, also considering nongenetic factors. This is a prospective study that involved 66 patients. We extracted DNA by salting out methods. We carried out the genotyping of the different polymorphisms by simple PCR for the TYMS 5'UTR and by PCR-RFLP for DPYD: 1905 + 1 G > A, 85 T > C, 496 A > G, 1679 T > G, c.483 + 18G > A and the TYMS: 5'UTR VNTR, 5'UTR G > C and 3'UTR. The study of the association of DPYD and TYMS polymorphisms with the various signs of toxicity under 5-FU revealed that the polymorphisms 496 A > G were significantly associated with hepatotoxicity: OR = 3.85 (p = 0.04). In addition, 85 T > C was significantly associated with mucositis and neurotoxicity: OR = 4.35 (p = 0.03), OR = 3.79 (p = 0.02). For TYMS, the only significant association we observed for 5'UTR with vomiting: OR = 3.34 (p = 0.04). The incidence of adverse reactions related to 5-FU appears to be influenced in patients with CRC by the identified DPYD and TYMS gene polymorphisms in the Tunisian population.


Introduction
Colorectal cancer is the most frequent malignant disease of the gastrointestinal tract, the third most common cancer in men (746,000 cases, 10% of all cancers), the second most common cancer in women (614,000 cases, 9.2% of all cancers), and is responsible for 600,000 deaths annually worldwide [1]. 5-Fluorouracil (5-FU) remains the cornerstone and the most critical component of all currently applied chemotherapy regimens. 5-FU has a narrow therapeutic index, and 10%-26% of patients treated with fluoropyrimidine-based regimens develop early-onset, severe, or life-threatening toxicity [2].
Dihydropyrimidine dehydrogenase (DPD) deficiency is a significant determinant of severe 5-FU-associated toxicity [3]. DPD is the initial and rate-limiting enzyme in the degradation of 5-FU. Because DPD catabolizes more than 80% of the administered 5-FU, patients with a partial or complete DPD deficiency have a strongly reduced capacity to degrade 5-FU and, therefore, an increased likelihood of suffering from severe and sometimes fatal multi-organ toxicity [3,4]. Many mutations and polymorphisms have been described in the gene (DPYD) encoding DPD, and ample evidence shows that carriers of the variant allele have an increased risk of developing toxicity [5,6]. In principle, pharmacogenetics-guided dosing will enable the identification of patients at risk of developing severe toxicity before the start of fluoropyrimidinebased chemotherapy [7].
DPYD genotyping is based on the search for four variants, DPYD Ã 2A, DPYD Ã 13, c.2846A > T, and HapB3, all of which have a very low population frequency (0.1% to 2.4%). When the variant is present in the homozygous state, the functional impact data show that they cause a decrease in the activity of the enzyme DPD, which can be complete (DPYD Ã 2A), almost complete (DPYD Ã 13), or partial (c.2846A > T, HapB3).
The EMA (European Medicines Agency) recommendations are for patients with partial or complete DPD deficiency with an increased risk of severe toxicity during fluoropyrimidine treatment. Before starting fluoropyrimidine treatment, phenotype testing, genotype testing, or both is recommended in patients with known complete DPD deficiency. A lower starting dose should be considered in patients with identified partial DPD deficiency [8].
A second gene is TYMS coding for the enzyme thymidylate synthase (TS), which catalyzes the methylation of deoxyuracil monophosphate, one of the three active metabolites of 5-FU of deoxythymine monophosphate. TYMS is an important target for 5-FU in patients with CRC. The three main polymorphisms that affect the TYMS gene have been associated with underexpression or overexpression of TYMS, and some authors have linked them to 5-FU resistance in patients with metastatic CRC [9]. Researchers have previously described polymorphisms in the TYMS gene's 5 and 3 untranslated regions as being both predictive of toxicity and prognostic of efficacy with fluoropyrimidines [10,11]. However, the data are inconsistent.

Study and objective
We conducted this prospective study within the Carcinology Department of the Farhat Hached University Hospital in Sousse and the research laboratory (LR12SP11) of the Biochemistry Department of the Sahloul University Hospital in Sousse.
We chose these polymorphisms after extensive research with what the EMA and CPIC (Clinical Pharmacogenetics Implementation Consortium) have already recommended. We also chose polymorphisms (496 A > G and 85 T > C) that generated inconsistent results in other studies. Researchers have never studied the polymorphisms associated with 5-FU chemotherapy treatment in the Tunisian population. Therefore, our study is unique, as we specifically studied the correlation between DPYD and TYMS polymorphisms and 5-FU toxicity.
This study obtained authorization from the Ethics Committee at Farhat Hached University Hospital in Sousse.

Population
The study population included 66 colorectal cancer patients treated with a 5-FU-based chemotherapy protocol. We dropped 28 patients from the study because of the lack of medical records.
We recruited the patients from the Carcinology Department at the Farhat Hached University Hospital in Sousse, where we obtained written consent from them.

Data collection
We used an information sheet to collect clinical and non-clinical information by interviewing patients and looking at their medical records.

DNA extraction and genotyping
The study of genetic polymorphisms focused on the DPYD and TYMS genes. We carried out DNA extraction from whole blood leukocytes taken from EDTA using the salting-out method and genotyping by direct PCR for TYMS 5 0 UTR VNTR and PCR-RFLP (polymerase chain reaction-restriction fragment length polymorphism) method, using specific restriction enzymes (Table A.1). We used the following PCR protocols for all 8 SNPs: first denaturation, 5 min; at 95 C, denaturation 45 s; at 95 C, hybridization 45 s (see Table A.1 for temperature), elongation 1 min; at 72 C, final elongation 7 min; and at 72 C for 32 cycles. We used DreamTaq Green DNA Polymerase (5 U/mL, Thermofisher; EP0712) following the recommended PCR protocol. The digestion reactions with restriction endonuclease contained 1X buffer, 5 U of RE (restriction enzyme), and 3 ml of PCR-amplified DNA. We incubated the restriction endonuclease digestions at the temperature recommended by the manufacturer (Table A.1).

Toxicity assessment
We based the toxicity assessment on the criteria of common terminology of the National Cancer Institute for Adverse Events (AENCI-CTC criteria) version 3.0 (https://webapps.ctep.nci.nih.gov/webobjs/ctc/webhelp/Welcome_to_CTCAE_1.htm). We excluded paresthesias and dysesthesias from the analysis in patients treated with the FOLFOX regimen, as these symptoms could be attributed to oxaliplatin [12,13].

Statistical analysis
We performed statistical analyses with SPSS version 20 software (IBM, NY, USA) [14]. The Hardy Weinberg equilibrium was tested for the different genotypic frequencies by the v2 test. We used the chi-square test to compare qualitative variables reported in number (n) and percentage (%). We presented the quantitative results (age, biological parameters) as means ± standard deviations (SD) and compared them using the ANOVA test or the Student's t-test (if Gaussian distribution). We estimated the risk of toxicity with genotyping by calculating the odds ratios (ORs) with 95% confidence intervals (CI). We adjusted ORs for potential confounding factors (p < 0.25) by binary logistic regression. We considered a p-value < 0.05 as statistically significant.

Population characteristics
Our study involved 66 colorectal cancer patients receiving 5-fluorouracil treatment; the average age was 55.34 ± 13.06, and the sex ratio was 1.4. The tumor location was heterogeneous in our study population. The tumor affected the colon in 65% of cases and the rectum in 35%. The sigmoid colon tumor was predominant and found in 43% of patients, followed by the tumor of the lower and middle rectum, found in 29% of patients. In our population, 50.8% of patients had already reached the metastatic stage. Regarding the level of differentiation, the tumor was moderately differentiated in 66.2%, well-differentiated in 23.1%, and poorly differentiated in 9.2% of cases.
The most frequently used treatment regimen in the first cycles was FOLFOX, administered to 87.5% of patients (19.5% subsequently changed protocols because of protocol ineffectiveness), followed by FOLFIRI (11%) and XELOX (1.5%).

Genotypic frequency
The genotypic frequencies of the eight studied polymorphisms were in the Hardy Weinberg equilibrium (Table 1). Table 1. Genotypic frequency and Hardy-Weinberg equilibrium. We noted that for DPYD 1905 þ 1 G > A, the most common genotype in our population was G/G, while G/A and A/A genotypes were absent.
For DPYD 85 T > C, the most common genotype was the homozygote (T/T) (68.2%), followed by the heterozygote genotype (28.8%), while the least present genotype was the homozygote variant C/C with a frequency of 3%. Regarding DPYD 496 A > G, the frequencies of the observed genotypes were 80.3% for (A/A), 16.4% for (A/G), and 3.3% for (G/G).
Regarding the TYMS 5 0 UTR VNTR, the genotypic frequency was 31.2% for 2R2R, 39.1% for 2R3R, and 29.7% for 3R3R. As for TSER Ã 3G > C polymorphism, the most common genotype was the normal homozygote (G/G) with a frequency of 52.3%. Lastly, for the TYMS 3 0 UTR6pb ins/del, we noted that the most common genotype in the general population was the insertion/insertion homozygote, with a frequency of 43.8%. The least present genotype was the homozygous deletion (-6bp/-6bp), with a frequency of 15.6%.

Toxicity related to 5-FU
In our population, 22.64%, 16.98%, 7.54%, and 12.57% of patients developed vomiting, hematotoxicity, mucositis, and diarrhea, respectively ( Table 2). Most patients developed grade 0 and 1 toxicity. Patients mainly developed grade 3 and 4 toxicity in the form of hematotoxicity, mucositis, and digestive toxicity. Regarding neurotoxicity and allergy, 18.86% and 6.91% of patients reported them, respectively; according to reports from clinicians, they appeared to be due to oxaliplatin.
According to the literature, the incidence of diarrhea can reach 50%-80%, with fluoropyrimidines including 30% of grades 3-4. Fluoropyrimidines induce lesions of the small mucosa by their cytotoxic and antiproliferative effects. These lesions cause dysfunction of the cells in the crypts and goblet cells with the loss of villi and enzymatic disturbances, resulting in an imbalance between absorption and secretion [15].
Regarding mucositis, 5-FU seems to be associated with exceptionally high levels of mucositis [16]. It has been reported that patients with pre-existing oral problems and poor oral hygiene have a higher incidence of oral problems after chemotherapy [17].
The exact pathophysiology of mucositis seems to be due to direct and indirect mechanisms. On the one hand, the mechanism of direct toxicity can be explained by the fact that the epithelial cells of the oral mucosa are rapidly dividing cells, making them sensitive to cytotoxic therapy's effects. On the other hand, the indirect mechanism is explained in neutropenic patients by an increased risk of oral infections due to infections by gram-negative bacteria and fungal species [18].
Hematological toxicity is also important; we reported it in our study in 16.98% of the cases. Boisdron-Celle and collaborators reported 19% of hematological toxicity [19].
As the cell toxicity by which the anti-tumor effects of chemotherapy manifest themselves is not specific to cancerous tissues, any drug administration will be accompanied by manifestations linked to the concomitant damage to the healthy tissues of the organism [20]. Allergy and neurotoxicity were attributed to oxaliplatin [21].

Study of the toxicity association with nongenetic factors
In our population, we noticed that men were more likely to develop diarrhea than women. Diarrhea was also significantly more frequent in subjects who had already developed metastases than those who had not. However, diabetic subjects were more likely to develop constipation than non-diabetic subjects.
The tumor location was significantly associated with the appearance of vomiting (p ¼ 0.04). Hematological toxicity was significantly higher in smokers (p ¼ 0.01), diabetics (p ¼ 0.01), subjects who had already developed metastases (p ¼ 0.01), and patients with sigmoid and right colon tumors (p ¼ 0.03).
Our results showed that the severity of the toxicity was significantly associated with the location of the tumor (p ¼ 0.03) because the sigmoid colon tumors seemed to be the most associated with severity. Severe toxicity was present in 35.7% of cases, and moderate toxicity was present in 50%. Metastases appeared to be significantly linked to the development of severe toxicity (p ¼ 0.01). In fact, 28.1% of subjects who developed metastases reached grade 3 or 4 toxicity. Conversely, 10.7% of subjects with a localized tumor reached severe toxicity (grade 3 or grade 4), while 42.9% had no toxicity.
We noted significant differences related to sex, smoking, diabetes, treatment regiment, tumor localization, and metastasis stage (p < 0.001). We considered all these parameters' potential confounding factors and used them to adjust the association between genetic polymorphisms and signs of toxicity linked to 5-FU.

Correlation of DPYD gene polymorphisms with signs of toxicity linked to 5-FU
After adjusting for potential confounding factors previously described, using the dominant models, we calculated the ORs for the occurrence of toxicity signs associated with carrying the variant genotype compared to the common genotype ( Table 3).
The DPYD 496 A > G polymorphism was significantly associated with the 5-FU toxicity sign. Indeed, a carrier of the allele G showed an increased risk of hepatotoxicity; (OR ¼ 3.85; P ¼ 0.04) ( Table 3). The 85 T > C also appeared to increase the risk of mucositis and neurotoxicity with an OR ¼ 4.35 (p ¼ 0.03) and OR ¼ 3.79 (p ¼ 0.02), respectively ( Table 3).

Correlation of TYMS gene polymorphisms with signs of toxicity linked to 5-FU
The only significant association we observed was between the 5 0 UTR VNTR and vomiting ( Table 4). The OR of toxicity associated with the 3 R allele was 3.35 (p ¼ 0.04).

Genotypic frequency
DPYD 1905 þ 1 G > A consists of G to A at a splice donor site, leading to the loss of exon 14 [22][23][24], resulting in a truncated protein that is catalytically inactive. The absence of the AA homozygous mutation of 1905 þ 1 G > A has also been reported in a Tunisian population by Fredj and colleagues [25] and in a French study [26].
DPYD 85 T > C results in cysteine to arginine substitution at position 29 of the protein. Another Tunisian study reported similar frequencies with the heterozygous genotype at 25.8% and the homozygote variant at 0.9% [25].
DPYD 496 A > G is located in exon 6, the result of methionine to valine transition at position 166 in the protein. Another Tunisian study reported a heterozygous genotype frequency of 11.3% but with no homozygous mutated genotype (G/G) [25].
DPYD 1679 T > G results in an isoleucine to serine substitution at position 560 of the protein. This polymorphism is linked to the reduced enzymatic activity of DPD [27,28]. In an in vitro study, Offer and colleagues found that homozygous expression of this variation resulted in a 75% reduction in DPD enzyme activity relative to wild type, implying that this mutation almost entirely inactivates the protein. Therefore, heterozygous carriers would have a 50% reduction in DPD enzyme activity [29]. According to phase 3 of the 1000 Genome project, the mutation is absent in its heterozygous and homozygous state in African and American populations, and it has been detected in the Finnish population only in the heterozygous state (T/T: 99%, T/G: 1%) [30].
DPYD c.483 þ 18G > A is an intron variant, described as one of the variants of HapB3. This DPD enzyme's activity has been reported to be decreased in c.483 þ 18G > A (HapB3) carriers [31]. Because DPD activity is not entirely absent in homozygous carriers of this DPYD mutation, a 25% dosage decrease for heterozygous genotype is expected [32].
Concerning the TYMS gene polymorphisms, the TYMS 5 0 UTR VNTR consists of a 28 bp repeat in variable number tandem (VNTR) located in the 5 0 UTR region at the activator site of the TYMS promoter thymidylate synthase activation region (TSER) [33]. A Tunisian study reported similar frequencies of 68.5% for the 2 R/2R genotype and 17% for the homozygous 3 R/3R [34].
The TSER Ã 3G > C polymorphism in the 12th nucleotide of TSER Ã 3's second repeat (rs2853542) changes TYMS expression by removing a transcription factor binding site. Patients with the TSER Ã 3 G > C polymorphism (TSER Ã 3C, 3 C) have a higher risk of toxicity due to a decrease in TYMS expression [35,36]. The frequency in our population is almost identical to that reported for the general Caucasian population (58.8%).
The 3 0 UTR (ins/del) is an insertion-deletion polymorphism of 6 pairs; it affects position 1494 of the 3 0 UTR within the TYMS gene [37]. The 3 0 UTR 6pb ins/del modulates the regulation of genes at a posttranscriptional level by controlling the stability of the resulting mRNA. The frequency of this SNP in our population is close to the Caucasian (26-29%) population but different from the African (50%) and Asian (76%) populations [37,38].

Correlation of DPYD gene polymorphisms with signs of toxicity linked to 5-FU
For several decades, studies on the genetic biomarkers of toxicity have linked them to 5-FU. However, few rare genetic polymorphisms associated with toxicity have been identified with great confidence. These include rare functionally deleterious DPYD variants, mainly 1905 þ 1 G > A and 2846 A > T.
Among the polymorphisms we studied, only 496 A > G was statistically significant. Indeed, carrying the variant allele seemed to increase the risk of hepatotoxicity; (OR ¼ 3.85; p ¼ 0.04).
Similarly, Falvella and colleagues reported that the G allele of the DPYD 496 A > G polymorphism was significantly associated with grades 3-4 toxicity (p ¼ 0.021), with an OR of 4.93 (95% CI 1.29-18.87) [39]. Additional studies have reported a significant association between DPYD 496 A > G and toxicity under 5-FU for other types of cancer [40].
According to our results, DPYD 1679 T > G did not appear strongly associated with toxicity to 5-FU, but our results concerning this polymorphism were inconclusive given the rarity of the mutated G allele: 0% G/G and 1.5% G/T. The mutated allele was only in one patient under 5-FU treatment. This patient only exhibited a few signs of moderate toxicity, with the highest being grade 3 hematological toxicity. However, in other studies, patients with the DPYD 1679 T > G variant have shown signs of severe toxicity [28,41,42]. Similarly, a meta-analysis by Meulendijks and colleagues [6] showed that the risk of severe global toxicity was approximately four times higher in patients with the DPYD 1679 T > G polymorphism. The risk of hematological and gastrointestinal toxicities also increased [6].
Based on functional data available for the 1679 T > G variant, this possible association with toxicity could be explained by the fact that a heterozygous genotype should result in a 40 to 50% decrease in DPD activity, similar to the well-known effect of 1905 þ 1 G > A [43][44][45].
Given that DPD contributes to 80%-90% of 5-FU metabolism [46], a 40%-50% decrease in DPD activity should translate into a 50%-100% increase in tissue exposure to 5-FU. Indeed, systemic exposure to 5-FU is 50% higher in 1905 þ 1 G > A carriers [43]. Based on available functional data, the recommendation is a 50% dose reduction in patients with DPYD 1679 T > G polymorphism [45]. In addition, different studies have described dose adjustments [28,47]. Morel and colleagues describe a patient with the heterozygous genotype who suffered from severe grade 4 toxicity. After a 6-week treatment interruption, they safely reintroduced 5-FU with individual pharmacokinetic adjustments [28]. For the carrier of this polymorphism, toxicity is likely to develop without reducing the dose.
According to our results, the polymorphism DPYD 85 C > T seemed to increase the risk of mucositis and neurotoxicity. The frequencies of toxicity were higher in C/T and T/T of 85 C > T OR¼ 4.35 (p ¼ 0.03) and OR¼ 3.79 (p ¼ 0.02) respectively. From a functional point of view, we initially reported that the DPYD 85 T > C polymorphism, leading to the change of tyrosine to cysteine at position 186, would have an inhibitory effect on the enzymatic activity of DPD by affecting the formation of the DPD dimer. Biochemical studies of the mutation of tyrosine to cysteine in another protein have suggested that this type of substitution can cause aberrant cross-linking of the dimer [48].
Our results concerning c.483 þ 18G > A were inconclusive given the rarity of the mutated T allele: 0% G/G and 1.5% G/T. Only one of these allele carriers was under 5-FU treatment. The c.483 þ 18 G > A variant is a part of HapB3 [49]. One study identified a deep intronic variant (c.1129-5923C > G) in tight linkage with HapB3 that was shown to affect pre-mRNA. This variant creates a cryptic splice donor site, which causes a premature stop codon and a truncated protein [31]. Although evidence suggests a correlation between c.1129-5923C > G/HapB3 and severe 5-FU toxicity [4], other studies have shown no significant associations [40,50].
Correlation of TYMS gene polymorphisms with signs of toxicity linked to 5-FU We did not report any significance in the association of the 5 0 UTR VNTR and 3 0 UTR polymorphism with the occurrence of 5-FU-related toxicities. The only exception was the 3 R allele of 5'UTR VNTR, with a vomiting risk of OR ¼ 3.34 (p ¼ 0.04). Our results are inconsistent with those reported in other studies, where the risk of toxicity increased significantly with the number 2 R [21]. The TYMS promoter and TYMS 3 0 UTR polymorphisms were in linkage disequilibrium, and the haplotype 2 R/ins6-bp was significantly associated with a high risk of severe side effects to 5-FU [21]. Although the complete mechanism is unknown, the presence of at least one copy of the repeat is necessary for transcription to take place. In addition, recent data suggest that a USF1(upstream transcription factor 1) binding site is present within these repeats. Therefore, increasing the number of repeats increases the number of transcription factor USF1 recognition sites, leading to increased transcription of TYMS [33].
Regarding the additional polymorphism TYMS TSER Ã 3G > C, there are few studies concerning this polymorphism and its toxicity to 5-FU. This polymorphism occurs on the 3 R allele, located at the 12th nucleotide of the second repeat [51]. This nucleotide is critical for the E-box sequence [36]. The presence of cytosine (3RC) alters the E-box consensus. Therefore, the transcription factors USF1 no longer recognize the sequence, which decreases in transcription compared to the presence of guanine (3RG) [36].
We included only 66 patients in our study for various reasons, which impeded us from drawing a clear conclusion about the effects of this polymorphism for either DPYD or TYMS.
We are still enrolling more patients to confirm our results. It will be more population specific to add more SNPs to the already established panel from EMA and CPIC, which should assist clinicians in requesting additional screening for these genes before starting a chemotherapy regimen based on fluoropyrimidine.

Conclusion
Our study reports preliminary data specific to the Tunisian population. The frequency of the different SNPs of DPYD and TYMS genes is similar to that reported in another Tunisian study. However, due to a lack of systematic screening for these polymorphisms in our country, our study presents a novel view of the spectrum of the SNPs frequency to verify whether we can implement a screening at the national level for all cancer patients who will receive fluoropyrimidine chemotherapy regimens. Our study confirmed the influence of DPYD and TYMS polymorphisms in CRC patients on the occurrence of adverse reactions related to 5-FU. Nonetheless, much conflicting evidence forces us to take a new approach to do a deep investigation of haplotype-phenotype-response correlation and to look for other polymorphisms that affect these genes to introduce to the panel already in place and that EMA and CPIC have recommended.

Author contributions
This study was supervised by Asma Omezzine, Ali Bouslama, and Slim ben Ahmed. Yassine Khalij designed the study. Yassine Khalij and Sana Chouchene recruited the patients. Yassine Khalij and Imtinen Belaid obtained the data. Yassine Khalij performed the genotyping. Asma Omezzine and Yassine Khalij performed the statistical analysis and the interpretation of the data. Yassine Khalij wrote the manuscript. Asma Omezzine, Dorra Amor, and Nabila Ben Rejeb provided a critical review of the manuscript

Ethics approval
The study protocol was approved by the local ethics committee of the Farhat Hached University Hospital, Sousse, Tunisia.
Consent to participate: all patients signed specific written informed consent for the relevant pharmacogenetic analysis.