Amy Carpenter
January 2025—A study published last year found variability in the variants tested for in the commercial lab DPYD genotyping assays available at the time of the study, underscoring “the importance of comprehensive DPYD genotyping to accurately identify patients with DPD deficiency,” the authors said. Compromised dihydropyrimidine dehydrogenase deficiency raises a cancer patient’s risk of fluorouracil toxicity (Nguyen DG, et al. J Natl Compr Canc Netw. 2024;22[4]:e247022).
Second of two parts on DPYD genotyping.
Part one in the December issue.
“This is where our recommendations come in, which is to address this need to standardize testing for DPYD,” said Reynold Ly, PhD, of the joint consensus recommendation of the Association for Molecular Pathology, the CAP, and other groups to guide clinical labs and assay manufacturers that develop, validate, and/or offer DPYD pharmacogenomic testing (Pratt VM, et al. J Mol Diagn. 2024;26[10]:851–863). DPYD encodes DPD, the initial rate-limiting enzyme that’s important in breaking down the pyrimidine bases, thymine and uracil, and 5-FU.
Dr. Ly is a member of the AMP pharmacogenomics working group that developed the consensus recommendation. He is clinical laboratory director of the Steve and Cindy Rasmussen Institute for Genomic Medicine, Nationwide Children’s Hospital, Columbus, Ohio, and clinical assistant professor of pathology, Ohio State University College of Medicine. He explained in an AMP webinar last fall the recommendations for clinical DPYD genotyping allele selection.
In the metabolism breakdown of capecitabine and 5-FU, more than 80 percent of the drug is metabolized by DPD and then excreted, leaving less than 20 percent of the drug to be incorporated into cells to assert its anticancer effect, which damages DNA and RNA. “Imagine if your DPD enzyme has reduced or loss of function,” Dr. Ly said. “The balance will shift,” with less of the drug metabolized and excreted and more of the drug incorporated into cells, causing more damage and increasing the likelihood of an adverse event or severe toxicity.
The DPYD gene is located on chromosome 1p21.3, spans 843 kilobases in length, contains 23 coding exons, and is expressed in many tissues, including the liver and peripheral blood. The five well-studied DPYD toxicity risk variants listed in the Clinical Pharmacogenetics Implementation Consortium 2017 guideline for DPYD genotype and fluoropyrimidine dosing were included in the AMP PGx working group’s recommended tier one (must-test) panel. They are NM_000110.4:c.1905+1G>A (*2A), c.1679T>G (*13), c.1129-5923C>G (HapB3), c.2846A>T, and c.557A>G. Two additional CPIC-defined decreased-function variants are in tier one: c.868A>G and c.2279C>T.
Large-scale sequencing efforts, such as the 1000 Genomes Project and Genome Aggregation Database, have uncovered additional variants in the coding region, Dr. Ly said. In the Genome Aggregation Database v4.1.0, there are more than 1,000 missense variants. “However, there is limited knowledge on the function of or potential clinical relevance to 5-FU toxicity, so more studies are needed to characterize these variants and understand their effect on DPD enzyme activity and their association and relevance to 5-FU toxicity.”
In the previously cited study that evaluated which DPYD variants are detected by commercial lab assays, five of the 13 tests performed by CLIA-certified clinical labs as of November 2023 were found to not test for all five clinically actionable variants associated with DPD deficiency (Nguyen DG, et al. J Natl Compr Canc Netw. 2024;22[4]:e247022). (The U.S. labs were identified through a search of the NIH Genetic Testing Registry and the authors’ knowledge and that of other pharmacogenomic experts in clinical labs that were performing DPYD genotyping, and the authors note that lab submission of testing information to the registry is voluntary. The variants detected on each commercial test were determined by information on the labs’ websites.)
Of the 13 tests, eight detected at least the five variants. The remaining five tests detected a few of those variants. “The *2A variant was included on all 13 tests,” the authors write, “followed by *13 (n=11), c.2846A>T (n=11), HapB3 (n=9), and c.557A>G (n=9).”
Pathogenic variants in DPYD are inherited in an autosomal recessive manner known as DPD deficiency.
“This is also known as an inborn error of metabolism called thymine-uraciluria,” Dr. Ly said, which is complete DPD deficiency and a more severe, rare disorder with variable expressivity. Its features can include failure to thrive, microcephaly, seizures, and motor impairment and intellectual disability.
Partial DPD deficiency varies among asymptomatic populations (three to eight percent). “In these individuals you have a normal allele with one no-function allele or two decreased-function alleles,” Dr. Ly said. “With partial DPD deficiency you’re typically asymptomatic, but if you have two pathogenic variants in DPYD you would have complete DPD deficiency.”
Both partial and complete DPD-deficient subjects are susceptible to severe 5-FU toxicity, he said.
Some DPYD variants were assigned star (*) alleles by authors when first published, Dr. Ly said, whereas others are described by their descriptive name or single nucleotide polymorphism database ID (dbSNP ID). For example, DPYD*2–*13 describes individual variants rather than haplotypes, unlike other pharmacogenomic genes.
In contrast, descriptive names are used for DPYD HapB3. “That includes the deep intronic variant as well as the synonymous variant found in linkage disequilibrium with it typically,” he said.
The Pharmacogene Variation (PharmVar) Consortium DPYD gene expert panel deemed haplotype-based star allele nomenclature for DPYD impractical. “The gene’s large size and potential recombination between exons makes haplotype phasing across all exons extremely difficult,” Dr. Ly said. “Most of the functionally relevant variants are rare or extremely rare, and clinicians may act on their presence regardless of whether the haplotype has other variants.”
PharmVar is a central repository for pharmacogenomic genes and alleles and lists DPYD alleles by reference SNP-cluster identifications (rsIDs) instead of star alleles. PharmVar and the AMP recommend using standard Human Genome Variation Society nomenclature to describe DPYD variants (https://www.pharmvar.org/gene/dpyd). “If no other variants are detected, DPYD should be reported as no variant detected and the phenotype as normal metabolizer,” he said.
Four of the five well-studied toxicity-linked variants listed in the 2017 CPIC guideline for DPYD genotype-guided therapy were characterized primarily in the Caucasian population, Dr. Ly said.
The exception was the c.557A>G decreased-function variant, found in individuals of African genetic ancestry. An activity score system is used for genotype-phenotype translation and linked to prescribed recommendations (Amstutz U, et al. Clin Pharmacol Ther. 2018;103[2]:210–216).
One of the challenges the AMP working group faced is that information on DPYD variant phasing is not well documented, he said. “And we found that DPYD testing can be different depending on the clinical indication,” whether it’s a pharmacogenetic indication or an indication for diagnostic testing for the autosomal recessive DPD deficiency. The group also found limited reference materials for DPYD, but the Genetic Testing Reference Materials Coordination Program of the CDC Division of Laboratory Systems, the Coriell Institute for Medical Research, and the genetic testing community collaborated to characterize genomic DNA samples from 33 publicly available cell lines for use as DPYD variant reference materials in clinical testing (Gaedigk A, et al. J Mol Diagn. 2024;26[10]:864–875).
Existing guidelines and recommendations for genotype-guided therapy for fluoropyrimidines and DPYD are available not only from the CPIC but also the Dutch Pharmacogenomics Working Group, the French National Network of Pharmacogenetics, and others. The AMP’s recommendations are intended to complement other guidelines.
The AMP uses a two-tier framework to develop its recommendations. Tier one must-test variants have a well-characterized effect on protein function or gene expression or both, have an appreciable minor allele frequency (0.1 percent) in at least one subpopulation, have publicly available reference materials, and must be technically feasible to detect using standard molecular testing. Tier two recommended variants meet at least one but not all tier one criteria.
For its tier one panel, the working group included the five well-studied DPYD variants and two additional CPIC-defined function variants, c.868A>G and c.2279C>T. “These are decreased-function alleles, similar to the c.2846A>T change,” Dr. Ly said. All tier one variants have activity values that can be used to determine genotype-phenotype translation.
Tier two variants are c.299_302del, c.703C>T, c.1314T>G, c.1475C>T, c.1774C>T, and c.2639G>T. These tier two variants are either no- or decreased-function variants.
“Most of the recommended alleles that we found are rare in a population,” Dr. Ly said. “And because of the extreme toxicity associated with DPD deficiency, variants that have an allele frequency of 0.1 percent or more in any subpopulation are recommended as tier one.” The no-function variant c.1679T>G (legacy name DPYD*13) did not meet the tier one population allele frequency cutoff but was included in tier one because of its association with extreme toxicity and the European Medicines Agency drug label recommendations.
“We found that the overall detection rate of our recommended panel to identify individuals with impaired DPD function cannot be reliably determined at this time because the overall incidence of partial or complete DPD deficiency is not well defined,” Dr. Ly said. “A large percentage of deleterious variants are rare or novel.”
Additional variants present in the Genome Aggregation Database may be associated with DPD deficiency or 5-FU toxicity or both, Dr. Ly said, but were not included in a tier. The c.2043_2058 deletion variant (rs773499329) that causes a frameshift was identified in the DPYD CDC’s reference materials GeT-RM study and has an overall minor allele frequency of about 0.006 percent but is observed predominantly in the South Asian population at 0.1 percent. “This doesn’t meet tier two inclusion as it was not included in the list of variants previously curated by CPIC, but laboratories may choose to include these additional variants as they are identified,” he said.
Clinical laboratories may choose to do full gene sequencing rather than genotyping because of the large number of rare variants and potential severe toxicities, Dr. Ly said.
“But keep in mind that current ACMG/AMP guidelines for interpreting sequence variants are not designed for interpreting pharmacogenomic variants. Many rare variants encountered in clinical sequencing may ultimately be classified as variants of unknown significance unless they’re frameshift or nonsense variants, which are classified as likely pathogenic or pathogenic.”
Sequencing may detect both common and rare variants, Dr. Ly said, but Sanger or short-read sequencing will not resolve phasing when more than one variant is detected.
Another consideration is the DPYD HapB3 haplotype. “HapB3 consists of a deep intronic splicing variant, the c.1129-5923C>G change, in cis with the synonymous variant c.1236G>A. The deep intronic variant and the c.1236G>A have been assumed to be in perfect linkage disequilibrium.” Some laboratories test for c.1236G>A and not the deep intronic variant to predict risk of severe 5-FU toxicity. However, Turner and coauthors reported that the c.1236G>A and deep intronic variant are not in perfect linkage disequilibrium because in rare cases the c.1236G>A variant has been found without the deep intronic variant (Turner AJ, et al. Clin Transl Sci. 2024;17[1]:e13699).
Thus, using c.1236G>A as a tag variant may not predict an accurate phenotype in rare cases and can result in a false-positive call of decreased function, Dr. Ly said. “So we recommend testing for the deep intronic variant and not the c.1236G>A variant because the deep intronic variant is the functional variant responsible for decreased activity.”
Dierks and coauthors encourage the same (Dierks S, et al. Eur J Cancer. 2024;204:114076). They reported on a case of a 5-FU “clinically highly relevant underdose” in a German patient with gastric cancer whose pretherapeutic HapB3 analysis revealed a heterozygous c.1236G>A variant without the c.1129–5923C>G variant. Because of the assumed perfect linkage disequilibrium, the patient was classified as an intermediate 5-FU metabolizer and received a 50 percent 5-FU dosage reduction. The authors say their study “shows for the first time a clinically relevant discrepancy between genotype and phenotype interaction resulting from the reported incomplete HapB3 variant.”
Said Dr. Ly: “Laboratories that perform exome sequencing may test for the c.1236G>A as a proxy variant for the deep intronic variant as intronic variants are not detected. And laboratories should state a limitation in their report acknowledging the incomplete linkage disequilibrium and information about the deep intronic variant as the underlying causal variant.”
In tumor diagnostic testing, DPYD testing can be done when performing genomic analysis for other actionable therapeutic markers, Dr. Ly said, though the tumor may have additional somatic variants that are not present in the blood and liver where 5-FU metabolism occurs.
“We support consideration of DPYD testing in tumor diagnostic testing. However, if only the tumor is sequenced, then germline confirmation may be required.”
Germline diagnostic testing for autosomal recessive DPD deficiency in clinical laboratories may be distinct from DPYD pharmacogenomic testing, he said. “Sequencing is more commonly used for diagnostic testing, and there are differences in the variants analyzed, test design, interpretation, and clinical utilization. However, some laboratories may offer a single DPYD genetic test for both diagnostic and pharmacogenomic indications.”
For copy number variation, DPYD has been found to have partial or whole gene deletions, most notably the exon 4 deletion. In one study, the authors observed the exon 4 deletion at a high prevalence in a Finnish population, at 2.4 percent of those prescreened for DPD deficiency, Dr. Ly said (Saarenheimo J, et al. Cancer Chemother Pharmacol. 2021;87[5]:657–663). In another study, the authors reported a lower frequency of 0.2 percent for the exon 4 deletion in a Canadian population and included a patient with severe 5-FU toxicity (Wigle TJ, et al. Curr Oncol. 2023;30[1]:663–672). “The frequency of this deletion is likely population-specific and can vary considerably among different patient populations,” Dr. Ly said.
Interstitial deletions of exons 6, 12, and 14–16 as well as partial and whole gene DPYD deletions have been observed with DPD deficiency with variable phenotypes. “These exonic deletions meet the frequency for inclusion in either tier one or two but are not well defined at this time. So we currently have no recommendations for routine clinical testing,” he said.
If a laboratory decides to do DPYD copy number variation testing, multiple technologies can be used to identify recurrent or rare CNVs at the exon level: next-generation sequencing, chromosomal microarray, multiplex ligation-dependent probe amplification, TaqMan copy number assay, or exon array.
Though there are no recommendations for DPYD copy number variation testing and most clinical pharmacogenomic assays don’t include CNV analysis for DPYD, Dr. Ly said, in cases of 5-FU toxicity or DPD deficiency, “consider using copy number variation testing when a single pathogenic variant cannot explain the phenotype.”
Amy Carpenter is CAP TODAY senior editor.