DPYD: the gene behind severe fluorouracil and capecitabine toxicity
DPYD is the clearest example in pharmacogenomics of a test that exists because people died without it. Reduced enzyme activity turns a standard chemotherapy dose into an overdose, and the variants that cause it are individually rare but collectively common enough that Europe now recommends testing before treatment.
Also known as: dihydropyrimidine dehydrogenase, DPD.
What DPYD does
The gene makes dihydropyrimidine dehydrogenase, usually shortened to DPD, which is the enzyme that breaks down the fluoropyrimidine chemotherapy drugs. Most of a dose of fluorouracil is disposed of this way rather than acting on the tumour. If DPD activity is low, that disposal route narrows and the drug lingers, so the same infusion delivers far more exposure than intended. The toxicity that follows hits the gut lining, the bone marrow and the mucous membranes, and it can be fatal.
The metabolizer groups
DPYD is scored rather than typed. Each known reduced-function variant carries an activity value, the values a person carries are summed, and the total falls into a band: normal metabolizer at the top, intermediate in the middle, poor metabolizer at the bottom. Unlike the star-allele genes there is no long catalogue of named haplotypes doing the work here. A small number of variants carry nearly all of the clinical signal, and everything else reads as normal function.
Medications where it matters
fluorouracil (5-FU)
Fluorouracil is given intravenously in colorectal, breast, gastric and head-and-neck cancer, often as part of a combination regimen. It is the drug the DPYD guidance was built around, and the dose reductions below are among the largest CPIC publishes for any gene.
| If your result is | Published CPIC guidance |
|---|---|
| Intermediate Metabolizer | Per CPIC (A): "Reduce starting dose by 50% followed by titration of dose based on toxicity or therapeutic drug monitoring (if available)." |
| Normal Metabolizer | Per CPIC (A): "Based on genotype, there is no indication to change dose or therapy. Use label-recommended dosage and administration." |
| Poor Metabolizer | Per CPIC (A): "Avoid use of 5-fluorouracil or 5-fluorouracil prodrug-based regimens. In the event, based on clinical advice, alternative agents are not considered a suitable therapeutic option, 5-fluorouracil should be administered at a strongly reduced dose with early therapeutic drug monitoring." |
capecitabine (Xeloda)
Capecitabine is a tablet that the body converts into fluorouracil, so it inherits the same problem. People sometimes assume an oral drug is the gentler option; for a DPYD intermediate or poor metabolizer it is the same exposure by a different route.
| If your result is | Published CPIC guidance |
|---|---|
| Intermediate Metabolizer | Per CPIC (A): "Reduce starting dose by 50% followed by titration of dose based on toxicity or therapeutic drug monitoring (if available)." |
| Normal Metabolizer | Per CPIC (A): "Based on genotype, there is no indication to change dose or therapy. Use label-recommended dosage and administration." |
| Poor Metabolizer | Per CPIC (A): "Avoid use of 5-fluorouracil or 5-fluorouracil prodrug-based regimens. In the event, based on clinical advice, alternative agents are not considered a suitable therapeutic option, 5-fluorouracil should be administered at a strongly reduced dose with early therapeutic drug monitoring." |
Is this in your raw DNA file?
DPYD needs a sequenced file. Give us a VCF or gVCF, or take Deep Read from your aligned reads, and the gene is read normally. A 23andMe or AncestryDNA export cannot answer it, and the reason is specific rather than a general disclaimer about chips. Both variants that carry most of the risk, rs3918290 and rs67376798, are on the common consumer arrays. The second one is the problem: it is a T-to-A change, so its two alleles are each other's mirror image, and a chip records the pair of bases it saw without recording which strand of the DNA it read them from. T and A on one strand are A and T on the other. There is no arithmetic that recovers the answer, on any file, for anyone. Since a DPYD result is the sum of what every tested position carries, one position we cannot orient leaves the sum unfinished, so we report the gene as not called and say why. We would rather hand you that than a normal result built on a position nobody read. The clinical DPYD test your oncology service orders is a different assay and does not have this problem.
A genotyping chip reads only a few hundred thousand pre-selected positions, about 0.02% of your genome, chosen for common variation. It does not sequence the rest, so it cannot find the rare or novel pathogenic variants the clinical and carrier modules look for: a “no finding” from chip data means “this chip never looked”, far more so than with whole-genome sequencing. Chip data suits common-variant traits, pharmacogenomic tag SNPs, and haplogroup ancestry, not clinical or carrier screening.
Common questions
Is DPYD in my 23andMe or AncestryDNA raw data?
The two most important variants are usually on the array, and we still cannot call the gene from it. One of them, rs67376798, changes a T to an A, and a chip does not record which strand it read, so those two bases fit either orientation and the position cannot be resolved. We report DPYD as not called on a chip file rather than reporting a normal result we did not earn. A sequenced file, VCF or gVCF, or Deep Read from aligned reads, resolves it.
I am starting chemotherapy next week. Can I use this instead of a clinical test?
No. Use the clinical test, and if your service has not mentioned one, ask. European guidance recommends DPYD testing before fluoropyrimidine treatment, and a clinical laboratory result is what an oncologist can act on. A re-analysis of a consumer file is not that.
What does an intermediate metabolizer result actually change?
In practice it changes the starting dose and the intensity of monitoring, which is what the table above sets out. It does not usually mean the drug is off the table. Poor metabolizer is the result that pushes toward a different regimen entirely.
Nobody in my family has had a problem with chemotherapy. Does that settle it?
Only if a close relative has actually been treated with a fluoropyrimidine. Most carriers never meet these drugs, so the variant passes through a family silently for generations and shows up the first time someone is treated.