Michael Adams-Conroy’s adopted parents were shocked when the nine-year-old boy passed away due to a grand mal seizure after presenting with flu-like symptoms in 1995. They were even more shocked to learn that Michael’s autopsy revealed extremely high levels of Prozac in his system. His parents were charged with his murder, and their other two soon-to-be adopted foster children were removed from their home until they were cleared. But Michael’s parents weren’t responsible for his overdose. His genes were.
Michael had a mutation in a metabolic gene known as Cytochrome P450 2D6 (Cyp2d6). The variant form of the enzyme CYP2D6 that he carried was less effective at breaking Prozac down, causing it to build up in his system, which led to toxicity, flu-like symptoms, and seizure. Michael’s tragedy was a rallying cry for some researchers and physicians around the globe to check patients more regularly for variants and to identify more genetic variants that affect individual drug metabolism.
Twenty-five years later, many companies now offer screening panels for key drugs, often including cytochrome P450 (CYP) enzymes. However, genes not directly related to metabolism can affect drug toxicity and efficacy as well. Genetic testing in cancer patients has become increasingly common, but other fields like psychiatry have failed to catch up. Experts think that early consideration of potential gene-drug interactions in clinical trials can prevent tragedies like Michael’s.
“Historically in drug development, we've been like ‘oh here's this therapeutic range for the drug. Give this human adult 40 milligrams. Let's see what happens,’” said Kristine Ashcraft, director of pharmacogenetics at Invitae, a company focused on increasing the influence of genetics in medicine. “As we get more into more personalized medicine, we're going to see both less side effects and better therapeutic benefit if we make sure that they're on a dose that's more appropriate for them as an individual.”
Some gene variants affect drug responses
The most common and well characterized set of genes involved in individual variability in drug response are the CYP enzymes. CYP enzymes are highly expressed in the liver, where they primarily oxidize molecules such as steroids, fatty acids, and toxins. CYP enzymes break down approximately 80% of all prescribed drugs (1).
Nearly 90% of people have a variant form of at least one of five Cyp genes: CYP2D6, CYP2C9, CYPC19, CYP3A4, and CYP3A5. CYP2D6, the gene that contributed to Michael’s death, is involved in metabolizing 25% of all prescribed drugs, including cancer treatments such as tamoxifen and psychiatric drugs such as antidepressants (2).
“Most medications we take in an active form. Some, called prodrugs, we take in an inactive form. They hit these drug metabolizing enzymes in our liver to be converted to what actually gives treatment benefit, and then they usually hit liver enzymes again to be converted into something that our body can get rid of,” said Ashcraft.
Oftentimes, one can predict how someone will respond to a drug by determining if the variant they have makes them process the drug faster or slower.
“I think about these like a highway system. We're pretending everybody has two lanes of these metabolic highways for drug metabolism, when genetically, if someone's an intermediate metabolizer, we’re down to one lane; poor metabolizers, no lanes; ultra-rapid metabolizers, three or more lanes. But the intermediate metabolizers are the most common variant,” said Ashcraft.
Following genetic testing, individuals are often placed into one of four categories based on the number of gene copies they carry (3):
- Poor metabolizers carry no functional enzyme.
- Intermediate metabolizers have some enzyme function, either because they have two copies of a partially active enzyme, or one copy each of a fully functional and non-functional enzyme.
- Extensive (normal) metabolizers have two functional copies of an enzyme.
- Ultra-rapid metabolizers carry more than two functional copies of an enzyme.
Ultra-rapid metabolizers often need lower doses of a prodrug to limit side effects, while poor metabolizers generally need higher doses of the prodrug, or an alternative treatment. For example, tamoxifen, an anti-estrogen hormone therapy commonly used to treat breast cancer, is broken down by CYP2D6 to its active form endoxifen. People with a CYPD26 variant that renders them a poor metabolizer need almost two or three times the dose of that of extensive metabolizers to result in the same levels of endoxifen in their system (4).
Nearly 40% of the population carries at least one non-functional copy of CYP2D6.
“For years, millions of women across the world were treated without knowing their CYP2D6 variant, and if no one looked at their plasma levels, it meant that those who were poor metabolizers never built enough concentration of the active drug to really prevent a relapse of breast cancer. And that’s serious. We're talking about life and death at this point,” said Daniel Mueller, a pharmacogenetics researcher at the University of Toronto.
Categories aren’t enough
Seyed Yahya Anvar, a pharmacology researcher at Leiden University Medical Center, isn’t satisfied with the current system for ranking someone as a “poor” or “ultra-rapid” metabolizer. “There is a high degree of variability between individuals, even in the same class,” he said. “We are quite far from not having a clue what’s happening; we at least understand a mechanism, and we understand there is a range. What would it take to make this truly personalized?”
Anvar’s team is training a neural network to predict how individuals will metabolize a given drug. The first algorithm they developed determined how variants of CYP2D6 affected the effectiveness and toxicity of tamoxifen in individuals with breast cancer (5).
The researchers fed the network CYP2D6 gene sequences from 561 breast cancer patients treated with tamoxifen. Their network predicted with 79% accuracy what an individual’s response to tamoxifen would be. The classic categorical approach averages a 54% success rate. Anvar was surprised that their network identified rare variants and more accurately predicted the effects of combined variants. His team is working now to develop this method for other genes and other disorders.
“The full panel still hasn’t been decided, but it’s going to be larger. For example, CYP2D6 is also heavily involved in antidepressants, so that’s another class of drugs that can be good targets,” said Anvar.
Data-backed panels for testing drug responses
Proteins without metabolic functions such as human leukocyte antigens (HLAs), which are involved in immune responses, also have gene variants that affect the efficacy and toxicity of antidepressants and antipsychotics. However, some gene variants in non-metabolic genes aren’t fully vetted yet. And keeping up with the growing research on gene variants is a full-time job.
“We had really good data and evidence to support the clinical use of this data, but at the time, it really wasn’t being used because the biggest barrier was the fact that we didn’t have guidelines on how to utilize the information. How do you take a genotype result and translate it to a phenotype? And then when you have that phenotype, how do you act on it? How do you change the prescribing habits based on it?” said Kelly Caudle, co-principal investigator of the Clinical Pharmacogenetics Implementation Consortium (CPIC) at St. Jude’s Children’s Research Hospital.
The Clinical Pharmacogenetics Implementation Consortium (CPIC) is the authority on what gene variants are ready to be used as a diagnostic source in the clinic. Although Caudle is based at St. Jude’s Children’s Research Hospital, the CPIC is mostly composed of volunteers around the globe who scour the literature and help write and publish clinical practice guidelines. CPIC members vet one variant and its effect on processing one drug at a time. They categorically rank the level and quality of evidence and whether they recommend testing before prescribing.
While this is a useful resource for practicing clinicians, it is also a useful tool for researchers developing new drugs.
“I hope it informs other researchers whose expertise is in drug development and discovery and allows those individuals to come up with better drugs to prevent or treat these toxicities,” said Mary Eileen Dolan, a cancer pharmacogenetics researcher at the University of Chicago.
Invitae uses the CPIC guidelines to select genes for screening clinical trial participants for genetic variants that influence drug metabolism and to determine any variability in efficacy or toxicity during the drug development process. They offer two testing panels. One includes the five CYP genes that primarily affect drug metabolism, while the other expands on those genes to include twenty other metabolic and non-metabolic genes.
Ashcraft is surprised that pharmacogenetic testing isn’t more prevalent among clinicians both prescribing and developing drugs. She thought we’d be further along nearly twenty-five years after nine-year-old Michael’s death.
Tragedy in Hawaii
The importance of considering the potential impact of genetic variation when developing a drug became prominent news earlier this year when Bristol-Myers Squibb and Sanofi were ordered to pay 834 million dollars to the state of Hawaii for failing to disclose that a drug known as Plavix could be ineffective for some individuals with a variant form of CYPC19.
Plavix is a blood thinner that is often prescribed to prevent heart attacks or strokes. Plavix is inactive when taken, but its active metabolite is released after it is metabolized by CYPC19 in the liver. Patients with mutations in CYPC19 that render them poor metabolizers either cannot break down the drug, or do so minimally, rendering the drug virtually ineffective.
CYPC19 variants show up frequently in Pacific Islanders, with some reports indicating that as many as 60% of Pacific Islanders carry at least one nonfunctional allele (6). In 2010, the FDA labeled Plavix to warn that patients may carry a mutation that renders the drug ineffective.
“When the FDA put a black box warning on Plavix… a few doctors started ordering tests, but most of them were like, ‘well, the FDA didn't mandate testing. They just said patients who are poor metabolizers are three and a half times more likely to have a heart attack or stroke,’” said Ashcraft.
Regular genetic testing at the clinical level could improve this, at least a little. Overall, Ashcraft hopes that pharmacogenetics can change the way we develop and use drugs. And the earlier pharmacogenetics is considered, the better.
“As soon as you have an understanding of how a drug is metabolized, you should partner with someone so that you can simultaneously get drug approval and diagnostics for any pharacogenetic testing that is needed,” said Caudle.
- McDonnell, A.M. et al. Basic review of the cytochrome p450 system. J Adv Pract Oncol 4, 263-268 (2013).
- Zhou, S-F. Polymorphism of human cytochrome P450 2D6 and its clinical significance: Part I. Clin Pharmacokinet 48, 689-723 (2009).
- Laika, B. et al. Intermediate metabolizer: increased side effects in psychoactive drug therapy. The key to cost-effectiveness of pretreatment CYP2D6 screening? The Pharacogenomics Journal 9, 395-403 (2009).
- Martinez de Duenas, E. et al. Adjusting the dose of tamoxifen in patients with early breast cancer and CYP2D6 poor metabolizer phenotype. The Breast 23, 400-406 (2014).
- Van der Lee, M. et al. Toward predicting CYP2D6-mediated variable drug response from CYP2D6 gene sequencing data. Sci Transl Med 13 (2021).
- Kaneko, A. et al. High and variable frequencies of CYP2C19 mutations: medical consequences of poor drug metabolism in Vanuatu and other Pacific islands. Pharmacogenetics 9, 581-590 (1999).