A comprehensive population-based study comparing the phenotype and genotype in a pretherapeutic screen of dihydropyrimidine dehydrogenase deficiency

Pretherapeutic screening for dihydropyrimidine dehydrogenase (DPD) deficiency is recommended or required prior to the administration of fluoropyrimidine-based chemotherapy. However, the best strategy to identify DPD-deficient patients remains elusive. Among a nationwide cohort of 5886 phenotyped patients with cancer who were screened for DPD deficiency over a 3 years period, we assessed the characteristics of both DPD phenotypes and DPYD genotypes in a subgroup of 3680 patients who had completed the two tests. The extent to which defective allelic variants of DPYD predict DPD activity as estimated by the plasma concentrations of uracil [U] and its product dihydrouracil [UH2] was evaluated. When [U] was used to monitor DPD activity, 6.8% of the patients were classified as having DPD deficiency ([U] > 16 ng/ml), while the [UH2]:[U] ratio identified 11.5% of the patients as having DPD deficiency (UH2]:[U] < 10). [U] classified two patients (0.05%) with complete DPD deficiency (> 150 ng/ml), and [UH2]:[U] < 1 identified three patients (0.08%) with a complete DPD deficiency. A defective DPYD variant was present in 4.5% of the patients, and two patients (0.05%) carrying 2 defective variants of DPYD were predicted to have low metabolism. The mutation status of DPYD displayed a very low positive predictive value in identifying individuals with DPD deficiency, although a higher predictive value was observed when [UH2]:[U] was used to measure DPD activity. Whole exon sequencing of the DPYD gene in 111 patients with DPD deficiency and a “wild-type” genotype (based on the four most common variants) identified seven heterozygous carriers of a defective allelic variant. Frequent genetic DPYD variants have low performances in predicting partial DPD deficiency when evaluated by [U] alone, and [UH2]:[U] might better reflect the impact of genetic variants on DPD activity. A clinical trial comparing toxicity rates after dose adjustment according to the results of genotyping or phenotyping testing to detect DPD deficiency will provide critical information on the best strategy to identify DPD deficiency.

BACKGROUND Chemotherapy with fluoropyrimidine drugs, such as fluorouracil, capecitabine and tegafur, is the standard treatment for many types of advanced cancer and is used by millions of patients with cancer worldwide. 1 However, 15-30% of patients will experience severe treatment-related toxicity, which is lethal in 0.5-1% of patients. [2][3][4] Capecitabine and tegafur are metabolised to fluorouracil, an anti-metabolite and pyrimidine analog that plays a pivotal role in the occurrence of this toxicity. The most well-known biochemical cause of intolerance to fluoropyrimidines is a deficiency in the catabolic enzyme dihydropyrimidine dehydrogenase (DPD). 5 Patients with partial or complete DPD deficiency have a reduced capacity to degrade fluorouracil and are at risk of developing severe fluoropyrimidine-associated toxicity.
Genetic polymorphisms in DPYD, the gene encoding DPD, predict fluoropyrimidine-associated toxicity. 1,[6][7][8][9] Pre-treatment screening for the most clinically relevant defective variants, c.1679T>G, c.1905 + 1G>A, c.2846A>T, and Haplotype B3 (c.1236G>A or c.1129-5923C>G), and a dose adjustment according to the DPYD genotype improves the safety of chemotherapy regimens based on fluorouracil. 10,11 International recommendations provide indications for drug-related genetic tests and DPYD genotype-guided dosing to improve their integration in routine clinical practice. 6,12 The U.S. Food and Drug Administration and the Health Candida Santé Canada have added statements to the drug labels for fluorouracil and capecitabine that warn against use in patients with DPD deficiency, and genotype-guided prescribing recommendations for fluorouracil, capecitabine, and tegafur are also available from the Dutch Pharmacogenetics Working Group. 12 The Clinical Pharmacogenetics Implementation Consortium (CPIC) has also proposed guidelines for the genotype-guided dosing of fluoropyrimidines. 6 Another approach to identify DPD-deficient patients is to assess DPD enzyme activity by determining the plasma concentrations of uracil ([U]), the endogenous substrate for DPD, or its product dihydrouracil (UH 2 ) to calculate the [UH 2 ]:[U] ratio. 1 [U] and the [UH 2 ]:[U] ratios are highly correlated with DPD activity in peripheral blood mononuclear cells (PBMCs). 1 Although debate exists regarding whether [U] or the [UH 2 ]:[U] ratio correlates well with the clearance of fluorouracil, 13,14 numerous studies have identified a relationship between fluoropyrimidine-induced toxicity and a DPD phenotype characterised by a high [U] or a low [UH 2 ]:[U] ratio. 1,[13][14][15] Measuring the DPD phenotype has the potential to improve the performance of the prechemotherapy tests designed to identify patients at risk of fluoropyrimidineassociated toxicity. 1,5,7,14,16 Additionally, dose individualisation in patients with DPD deficiency as evaluated by measuring [UH 2 ] and [U] may improve the safety of these patients. [17][18][19] A fast and reliable quantitative analysis of the DPD phenotype is performed with an accurate, precise, robust and sensitive liquid chromatography tandem mass spectrometry assay. 20 1,14 To date, the most appropriate strategy of screening for DPD deficiency is a highly debated topic, and data from large-scale studies designed to establish the respective performances of phenotyping, genotyping and combined approaches are lacking. An important and yet unresolved issue is to what extent DPYD genotypes reflect pre-treatment DPD activity estimated by measuring [U] or [UH 2 ]:[U] ratio. 1 We took advantage of the extensive experience of the pharmacogenetics unit of a French university hospital in testing for pretherapeutic screening of DPD deficiency to address this issue. 21 We performed a cross-sectional observational study that comprehensively describes the DPD phenotypic characteristics among 3680 patients. Complementary genetic testing has been performed for common DPYD variants in all patients and rare variants in selected cases, and the relationship between the DPYD genotype and DPD phenotype has been analysed.

METHODS
We retrospectively reviewed the laboratory database of the Hôpital Européen Georges Pompidou (Assistance Publique-Hôpitaux de Paris). All patients referred to the pharmacogenetics unit (2016-2019) for pre-treatment testing for DPD deficiency were considered (Fig. S1). Screening for DPD deficiency was performed at the discretion of the physician. Genotyping for the four common genetic variants of DPYD (c. 1679T>G Fig. S2. Overall, 3680 patients were screened for DPD deficiency by both genotyping and phenotyping (Fig. S1). Whole-exome sequencing of the DPYD gene was performed in 111 individuals with a wildtype genotype, and DPD deficiency was defined as [U] > 16 ng/ ml 7,14 and/or [UH 2 ]:[U] < 10 14 (see the "Phenotyping" section). This cross-sectional study is reported according the STROBE statement.

Phenotyping
[U] and [UH 2 ] were measured using a Waters Acquity® TQD LC®/ MS/MS System consisting of an Acquity® ultra high-performance liquid chromatography system (Waters; Milford, MA) coupled with an Acquity® triple-quadrupole tandem mass spectrometer with an electrospray ionisation interface. Complete validation was performed according to the Food and Drug Administration guidelines on bioanalytical method validation. All data were acquired and processed using MassLynx™ 4.1 software with the QuanLynx™ program. The detailed technical protocol is described elsewhere. 21 Based on the published data available to define a uracil threshold discriminating normal and deficient activity, a clinically relevant increase in the risk of severe toxicity occurs when [U] is >16 ng/ ml. 1,7,11,14

Genotyping
Genotyping was restricted to patients who consented to genetic analyses. Conventional PCR-based assays were used to detect the four DPYD polymorphisms with strong evidence supporting defective function (c.1679T>G, c.1905 + 1G>A, c.2846A>T, and c.1129-5923C>G) using the TaqMan® DME Assay (Applied Biosystems, France) for allelic discrimination. Based on the genotype result, an activity score is calculated to predict whether the patient has normal, intermediate or low metabolism. 6 An activity score of 2 indicates that a patient has two fully functional alleles (activity score: 1 + 1), predicting normal metabolism. An activity score of 1 to 1.5 signifies that a patient is a carrier of one fully functional allele (=1) and one allele with decreased activity (=0.5), 2 alleles with decreased activity (0.5 + 0.5), or one fully functional allele (=1) and one non-functional allele (=0), predicting intermediate metabolism. An activity score of 0.5 or 0 signifies that a patient has 1 non-functional allele (=0) and one allele with decreased activity (=0.5) or 2 non-functional alleles (0), predicting low metabolism. Among the four common genetic variants detected by genotyping, c.1679T>G and c.1905 + 1G>A are considered "non-functional" alleles (activity score 0), and c.2846A>T and c.1129-5923C>G are considered "decreased activity" alleles (activity score 0.5).
Next-generation sequencing Samples were characterised for molecular alterations using targeted Next-Generation Sequencing (Ion AmpliSeq™ Custom 5FUIRI IAD68279, Life Technologies™, Carlsbad, CA). Briefly, the multiplex barcoded libraries were generated from 10 ng of DNA according to the manufacturer's recommendations (Ion ampliseq  (Table 2), and these patients were predicted to have intermediate metabolism according to the activity score of DPYD alleles (1 or 1.5) (see the "Methods"). 6 The presence of a mutation in the DPYD gene has a low positive predictive value (PPV) to identify individuals with partial DPD deficiency (  (Fig. 2). However, the negative predictive value (NPV) of DPYD genotyping averaged 90% for [U] or [UH 2 ]:[U] (Table 3). Receiver operating characteristic (ROC) curves (Fig. S4) (16). Finally, two patients with [UH 2 ]:[U] < 1 (also corresponding to complete deficiency of DPD activity) did not carry defective alleles, as assessed using genotyping or NGS. Based on these results, DPYD allelic variants may be poor predictors of complete DPD deficiency, and the absence of a mutation does not exclude the possibility of complete DPD deficiency.
Next-generation sequencing to identify rare genetic DPYD variants DPYD gene sequencing was performed in 111 individuals with discordant genotype-phenotype correlations ([U] > 16 ng/ml or [UH 2 /][U] < 10 and DPYD wild-type genotype). Eighteen genetic variants were identified; all were missense mutations (Table S1) and the variant c.557A>G (p.Y186C; rs11523289) was the only variant to achieve a consensus regarding pathogenicity (activity score of 0.5, PharmGKB evidence level 1A for toxicity/ adverse drug event (https://www.pharmgkb.org/clinicalAnnotation/ 1183703784)). Seven (7) patients (6%) were carriers of this variant and were predicted as having intermediate metabolism. 6 Thus, in the subgroup of 111 patients with a genotype-phenotype discordance, whole-exome sequencing of the DPYD gene allowed us to correctly reclassify 6% (7/111) of the patients with DPD deficiency.

DISCUSSION
We describe here the largest series of pre-treatment DPD phenotyping combined with DPYD genotyping in non-selected patients with cancer. This timely resource is clinically relevant and consistent with the recent regulation in France related to the requirement for DPD phenotyping before the administration of fluoropyrimidines. The generalisation of a pre-treatment screen for DPD deficiency will exert a substantial effect on the volume of prescriptions of this test and will constitute a challenge in terms of health-care resources. 20 Therefore, a comprehensive and accurate knowledge of the distribution of DPD activities and their relationships with DPYD mutations in a large and non-selected population is critical.  , we conclude that DPD phenotyping is a more appropriate approach to screen for DPD deficiency. However, our study provides information on the comparative performances of DPD phenotyping and DPYD genotyping in predicting fluoropyrimidine toxicity. Pre-treatment screening for the most frequently observed DPYD genetic variants and dose individualisation improve the safety of the patients, 10,25 [U] and [UH 2 ]:[U] might more comprehensively reflect DPD enzymatic activity than DPYD variants alone. A combined DPD phenotype-genotype testing to screen for DPD deficiency is a potentially clinically relevant approach, 15 but it remains to be demonstrated that its yield a significantly higher predictive value. 26  Although DPYD mutations are poor predictors of DPD activity, they are better predictors of partial DPD deficiency when it is estimated by measuring the [UH 2 ]:[U] ratio compared with [U] alone. This assumption is supported by the fact that [UH 2 ]:[U] is the ratio of the product to the substrate of DPD, and therefore more accurately reflects the enzymatic activity of DPD than [U] alone. Consistent with this finding, several studies have shown a significant correlation between the [UH 2 ]:[U] ratio and systemic fluorouracil exposure and the degree of toxicity. 23 7 The complementary and additional information provided by the concomitant assessment of [U] and the [UH 2 ]:[U] ratio remains to be examined in specific settings. For example, the rapid or ultrarapid DPD phenotype is poorly described, and its clinical impact is likely underestimated. 28 However, it could be considered a potent predictor of critical end-points, such as the response to the chemotherapy or patient survival. In this case, DPD phenotyping becomes critically relevant based on the findings of   pharmacokinetic studies aiming to balance therapeutic effectiveness with safety that have proposed target ranges using the area under the curve (AUC) and that have shown that in phase III trials, > 65% of the patients are below the target AUC, less than 20% are above the target range, and only 15% of the patients have an AUC for the drug within the target range. 29 The implementation of more sensitive tests for DPD deficiency, which appears to be the case for DPD phenotyping instead of DPYD genotyping, would theoretically expose a greater proportion of patients to the risk of being undertreated with fluoropyrimidine and, consequently, of being less responsive to the chemotherapy. Dose escalation after the first cycle to achieve maximal safe exposure is probably the most reasonable therapeutic strategy in these cases. The sensitivity of DPYD genetic testing depends on the number of variants investigated, and a genetic analysis investigating only a selected panel of variants with decreased or no function will not exclude DPD defects due to the presence of rare genetic variants. Alternatively, investigating the complete coding sequence of DPYD would theoretically improve the predictive value of the screening test. The results of our NGS analysis do not reveal a substantial enrichment of rare DPYD variants in mutations predicted to cause functional alterations, since only one pathogenic variant (c.557A>G, p.Y186C) was identified in seven patients (6%), indicating that as a complementary approach to genotyping for the detection of rare mutations, NGS does not seem to increase the predictive performances of the genetic approach, at least in European populations. However, complementary genetic testing might help to confirm the genetic origin of the enzymatic deficiency, particularly in patients with complete DPD deficiency, and to justify family counselling.
A limitation of this study is the lack of clinical outcomes, including the rate of grade 3/4 toxicities. The retrospective design of the study did not allow us to evaluate the impact of DPD deficiency screening on the occurrence of toxicities during fluoropyrimidine treatment. These patients were followed in numerous oncology departments throughout France and we could not generate a data collection gathering biological and clinical variables of the patients screened. The primary end point of our study was to examine the relationship between DPD phenotypes and DPYD genotypes, and to what extent DPYD genotypes reflect pre-treatment DPD activity estimated by measuring [U] or [UH 2 ]:[U] in a pre-treatment screening strategy. As such, our aim was not to compare the performances of two methods of DPD deficiency screening for fluoropyrimidines dose adjustment strategy, which would have required a prospective study to collect clinical outcomes. However, a clinical trial comparing toxicity rates after dose adjustment according to the results of genotyping or phenotyping testing to detect DPD deficiency will provide critical information on the best strategy.
In conclusion, common DPYD variants alone are moderately predictive of DPD deficiency and DPD phenotyping is considered the most appropriate method to screen for DPD deficiency. An assessment of [UH 2 ]:[U] ratio can easily be implemented for routine screening because it rapidly provides reliable results, while allowing a more exhaustive identification of patients at risk of severe toxicity due to DPD deficiency.

ADDITIONAL INFORMATION
Ethics approval and consent to participate This study received approval (N°0 0011928) from the hospital institutional review board (CERAPHP.5). All patients provided written informed consent for study participation and for genetic analyses prior to inclusion. The study has been performed in accordance with the Declaration of Helsinki.
Data availability Data supporting the results reported in the article can be obtained upon request to the corresponding author. A comprehensive population-based study comparing the phenotype and. . . N Pallet et al.