Pharmacogenetics and personalised medicine Flashcards
Pharmacogenetics and personalised medicine
Personalised medicine centers the idea that drug treatments and regimes can be selected or avoided based on identified individual characteristics.
Pharmacogenetics is the study of genetic factors influencing an individual’s response to a specific drug. It includes selective drug targeting of genetic variants contributing to disease.
This is not a new field. In the 1930s, it was detected that the inability to detect bitter taste (phenylthiocarbamide) is heritable. This is controlled by the taste 2 receptor (variant 38), encoded by the TAS2R38 gene. There are two haplotypes from co-inheritance of 3 coding region SNPs. There is some suggestion that “non-taster” also correlates with alcohol/tobacco consumption! People that do not taste the bitterness are more likely to abuse these.
In 1975, individual variation in response to the antihypertensive debrisoquine was observed.
Pharmacogenetics has been accelerated by the human genome project (first sequence 1999) and genetic sequencing technology.
Pharmacogenomics is the application of individualised drug therapy in practice, in conjunction with a genomic diagnosis test.
The potential advantages of personalised medicine are improvements in clinical efficacy and avoidance of adverse drug reactions.
Cost-effective prescribing is also facilitated: groups that have neither adverse nor therapeutic effect still need to be identified for health economics, but also so that time is not wasted on non-effective treatment in time-sensitive diseases such as cancer.
There is a need for alternative strategies in drug discovery.
The idea of blockbuster drugs has changed with the rise of personalised medicine, as there is a shift from screening large libraries of compounds towards structure-based drug discovery targeting different protein variants.
Small groups of patients likely to benefit from a specific drug are identified.
Genetic variation
Germline mutations are inherited. For example, there can be inherited variation in liver metabolising enzymes due to different CYP350 isoforms.
Somatic mutations are any change to DNA that happens in any cell of the body after conception, except for germ cells. These can occur in cancer development.
- e.g. BRCA1/2 gene expression is associated with increased risk of breast cancer, so some women chose o have a prophylactic mastectomy.
The genome is 3x10^9 base pairs long. <2% represents coding DNA for proteins.
Common types of human genome variation include:
- Single nucleotide polymorphisms (SNP) - insertion/deletion/substitution of a single base pair. 2/3 are T/C variation and there is ~1 SNP every 200 bps.
- Small insertions/deletions (indels): ~1 indel every 3000 bps
- Large structural differences such as duplications or deletions. There are ~20 000 sites in the genome and this is often seen with CYP enzymes.
Single nucleotide polymorphisms
Synonymous - a change to a gene’s DNA sequence that doesn’t change the sequence of the resulting protein, made possible by the degenerate genetic code – e.g. alanine coded by GCT or GCC)
Non-synonymous/Missense – a DNA mutation that changes the amino acid sequence of a protein, resulting in a point mutation in protein e.g. Val GTC versus Ala GCC. This can change how the protein behaves in terms of its folding, structure and interactions with drugs or other proteins.
Loss of function:
- generation of a premature STOP codon → truncated protein
- frameshift mutation – single bp insertion/deletion
- mutation of mRNA splicing site → affects the boundary between exons and introns
In comparison to the reference human genome sequence, any individual is predicted to have, in their coding regions:
- 11-12k synonymous SNPs
- 10-11k non-synonymous SNPs
- 250–300 genes significantly affected by loss of function SNPs
SNPs in non-coding regions, including promoter, enhancer elements and introns, can increase or decrease protein expression in different tissues.
Individuals can be homozygous or heterozygous for a gene affected by an SNP (allele) due to paired chromosomes.
Closely spaced SNPs (e.g. within the same gene) are often genetically linked and inherited together as a haplotype.
Individual SNPs are easy to find but they normally have a very small effect. Looking at haplotypes can be more clinically relevant.
Study of common SNPs (>1 % frequency) has been the most tractable type of variation for genetic association studies with disease or drug response.
Larger structural changes
Copy number variations (CNVs) arise from large duplications or deletions in the genome (1-1000s kbps).
This is common with CYPs, causing poor vs ultra-fast metabolizers.
~0.4% of the genome differs by these types of change in unrelated individuals
The context of a genetic change can be critical:
- Gene-gene interaction - haplotypes or tumorigenesis (multiple somatic mutations required for cancer.)
- Gene-environment interactions - epigenetics, e.g. patterns of DNA methylation, or variations in mitochondrial DNA.
Effect of genetics on drug function
Genetics could result in altered interaction with the target (pharmacodynamics). This could affect selectivity of the binding site, activity or function, receptor regulation (e.g. after chronic drug exposure), expression and tissue localisation.
We could have changes in the receptor itself, but also other proteins involved in downstream signalling, such as the G protein.
This could also apply to “off target” side-effects/toxic effects that derive from interaction with cellular proteins.
- e.g. Terfenadine and Astemizole are non-sedating anti-histaminergic drugs, both withdrawn from the market due to potent potassium channel (hERG) blocking effect leading to QT-prolongation and Torsades de Pointes. This only occurred in people with a genetic mutation in the hERG gene, but it could be fatal.
Genetics can also result in altered drug concentration at the target site due to changes in pharmacokinetics (ADME). This can be as a result of changes in liver metabolism (CYPs) or transport across membrane barriers (multidrug resistance proteins), which can affect absorption, distribution, elimination and access to intracellular targets.
Cancer and personalised medicine
Cancer has an inherently genetic basis and is a very heterogenous disease.
Unrestricted cell growth requires multiple gene mutations, including in pro-proliferative proteins (e.g. EGF receptor, Ras), tumour suppressors (e.g. BRCA1/2 for DNA repair), and proteins involved in invasion/metastasis (e.g. VEGFR).
There are ~375k cancer diagnoses/yr in the UK.
The 10-year survival rate for breast cancer is 76%, but for lung cancer it is only 10%.
The key features of cancer supporting a pharmacogenetic approach are:
- Variability in molecular mechanisms for tumorigenesis
- Risks with conventional chemotherapeutics - side effects/toxicity, e.g. from non-specific inhibition of cell division
- Rational basis for “individualised” drug targets, e.g. with a selective inhibition of a mutated or overexpressed oncogene versus the native protein (pharmacodynamic mechanisms)
- Unmet Clinical need
- Cost effective prescribing - the expense of biologics (e.g. monoclonal antibodies) is £25-30k for a trastuzumab (herceptin) course, so selection of individuals who will benefit is desirable.
Basic individualised therapy in breast cancer
We can take biopsies of breast cancer tumours and carry out diagnostic tests by staining for receptors with immunohistochemistry. Genome sequencing is therefore not necessarily required to be able to apply personalised medicine.
~70% of tumours are oestrogen receptor (ER) positive. These would be treated by inhibition of ER dependent proliferation, e.g. with selective oestrogen receptor modulators like tamoxifen.
~15-20% of tumours are HER2 positive. These would be treated by inhibition of EGF receptor dependent proliferation, e.g. anti-HER2 mAbs like trastuzumab.
Unresponsive cohort: 15-20% of breast tumours are “triple negative” for ER, progesterone receptors and HER2 – e.g. often associated with BRCA1/2 genotype. These are treated with surgery.
ER/HER2 based therapies depend on genetic differences related to the therapeutic targets in tumours. This is pharmacodynamic targeting.
The selectivity of tamoxifen or trastuzumab relies on over-expression of ER or HER2 genes in tumours. Over-expression occurs due to mutations in regulatory regions.
Pharmacodynamic targeting of a single mutation
The Ras-Raf-MAPK pathway drives cell proliferation.
Constitutive activation of this pathway by mutations in Ras or Raf can lead to unrestricted cell growth.
The B-Raf activating mutation is caused by a SNP. This is a GTG to GAG mutation, which leads to a Val600Glu (V600E) mutation in the B-Raf protein. This stabilises the active conformation.
This mutation is present in 50-70% of melanomas, as well as some other cancers.
Small molecules were developed to be selective for the active B-Raf conformation stabilised by the V600E mutation.
Vemurafenib is a type I B-Raf inhibitor, binding at the ATP binding site.
The amino acid Phe595 moves between the inactive and active Raf conformations (DFG-in for active state and DFG-out for inactive). The “active” pocket is stabilised by the V600E mutation, allowing vemurafenib to bind with a high affinity.
Vemurafenib can form an ionic interaction with the glutamate side chain, but not with valine. This means that the treatment is selected for the mutated protein present in melanoma cells, but it does not interact potently with B-Raf in regular healthy cells.
Vemurafenib had dramatic effects in clinical trials for advanced melanoma. This treatment has been approved by NICE since 2012.
Dabrafenib is a 2nd class drug approved in 2014. It is more selective and effective against brain metastases and it also targets the V600K mutation.
There is also an inhibitor of the k-Ras G12C mutation, Adagrasib, licensed in 2022 by the FDA for advanced or metastatic NSCLS (11-16% tumours have G12C).
Vemurafenib is the first example of licensing a drug that achieves increased selectivity and efficacy by targeting a mutant protein. This demonstrates proof of concept for therapeutic targeting of a SNP, showing that this approach is feasible.
Broad challenges in personalised cancer treatment
Cancer pharmacogenetics are actually very complex. It is difficult to know which polymorphisms/mutations are the most relevant.
We also have to ask if selective targeting/screening is even possible in all of these cases.
Cancer is a multifactorial disease as multiple contributing mutations are always involved. We may be able to tackle this by combination therapy.
There is a requirement for routine diagnostic tests.
Pharmacogenetics: off target effects
Pharmacogenetics can lead to ‘idiosyncratic’ ADRs. These are side effects which cannot be predicted from the interactions between the drug and its intended molecular target.
These effects can occur through many different types of mechanisms, such as:
- Immune reactions: hypersensitivity (e.g. abacavir, carbamazepine, phenytoin, flucloxacillin) or agranulocytosis, the loss of WBCs (e.g. the antipsychotic clozapine).
- On- and off-target binding site similarities - e.g. aminoglycoside antibiotics (mitochondrial vs bacterial ribosomes)
Abacavir and hypersensitivity
Abacavir is a HIV reverse transcriptase inhibitor introduced in 1998.
5% of patients develop immune hypersensitivity, which can be fatal. This is known as Stevens-Johnson syndrome and it affects the skin and other mucous membranes.
Hypersensitivity reactions have a genetic basis.
The MHC on antigen presenting cells is encoded by HLA genes. There are >1000 genetic variants/alleles.
The MHC presents the antigenic peptide to T cells. The T cell receptor and CD4 bind the antigen/MHC complex. This activates T cells, triggering an immune response.
Abacavir can initiate this immune reaction by binding the MHC encoded by the HLAB5701 gene variant, leading to hypersensitivity.
Routine diagnostic screening for the HLAB5701 genotype has been advised in the UK and US prior to abacavir prescription since 2008.
Another HLA link established is that between carbamazepine and HLAB1502. HLAB1502 allele frequency depends on ethnicity - the UK Patient Information Leaflet (PIL) recommends genotyping for Asian patients.
Cytochrome P450 isoforms
CYP450s catalyse drug oxidation, a phase I reaction.
Individual gene alleles (e.g. CYP2D6) vary substantially within the population.
SNP and SNP combinations (haplotypes) can reduce enzyme activity.
CNVs can duplicate or delete CYP isoform genes, leading to higher or lower enzyme expression and activity.
Tamoxifen is metabolised by liver CYP2D6, as well as CYP3A4 and CYP2C9, into active metabolites such as 4-OH tamoxifen.
Around 7% of the population has a less active CYP2D6 isoform. This results in reduced active metabolite production and therefore a reduced anti-cancer effect. An increased tamoxifen dose is recommended for these individuals. However, CYP2D6 genotyping is not yet recommended for tamoxifen treatment - patients are stated on a low dose and the efficacy of this is then evaluated for dose adjustment.
CYP2D6 is also involved in the metabolism of opioids, converting the codeine pro-drug into the active drug morphine. Poor metabolizers have non-functional CYP2D6 isoforms, resulting in a lack of codeine analgesia. Ultra fast metabolizers have CYP2D6 gene duplication, resulting in increased morphine production and an increased risk of toxicity.
Clopidogrel is an antiplatelet drug for post-MI secondary prevention and for atrial fibrillation, used as an alternative to aspirin in patients sensitive to GI effects. Its metabolite acts as an antagonist at P2Y12 receptors for ADP on platelets to prevent aggregation.
CYP2C19 haplotypes are associated with altered levels of the active metabolite and clinical risk:
- CYP2C192 haplotype shows reduced activity (↓ active drug)
- CYP2C1917 haplotype is a gain of function (↑ active drug)
Patients with one or two copies of *2 have an increased risk of thrombosis after post-MI stent insertion. Meta analysis shows a ~2.5x higher risk. There is also some evidence for an increased risk of a major coronary event.
There is some weaker evidence that 17/17 patients have increased bleeding risk and greater therapeutic effect.
In the US, FDA has approved a Genetic Box Label for Clopidogrel, and some hospitals test for haplotype pre-treatment.
PK is probably the most developed area of pharmacogenetics, as it is easy to test for CYP enzyme differences.
Opportunities and considerations for personalised medicine and genetic association studies
We need to think about:
- How we can pharmacologically exploit genetic variants
- Health economics
- Regulatory framework - how are trials affected by a smaller target population? - Society and ethics - should insurance companies have access to your genetic information?
All of these have to be underpinned by scientific evidence as to whether particular mutations cause different responses.
A single identified monogenic cause for a disease or drug response variation is relatively uncommon. This is however seen in cystic fibrosis, haemoglobinopathies (e.g. sickle cell) and Abacavir hypersensitivity.
Most individual variation in drug response or disease development is a complex interplay of genetic, environmental and developmental factors.
Genetic association study design and limitations of GWAS
Studies often take the form of the ‘candidate gene’ approach - e.g. antipsychotic drugs are dopamine D2 antagonists, so we look for association between D2 receptor polymorphisms and response in schizophrenic patients. The advantage of this is that we are only looking at one gene, but it relies on you having a good hypothesis.
The GWAS approach is more common.
A genome-wide association study (GWAS) is a research method that compares the DNA of many people to find genetic variants linked to a disease or trait.
The process is:
- genotype >500,000 SNPs, covering >75% of the genome, by using microchip arrays
- take DNA from disease group and a healthy control population
- look for association with a disease trait: are there SNPs that are more common in the disease population than in the healthy population?
This is unbiased, so it doesn’t rely on our (incomplete) knowledge of biology.
It can implicate entirely novel genes in the disease process. However, we may not know what the genes does so there is a lot of work involved in understanding how this protein works and how to target it.
The association between the SNP and the gene is indirect - could be a protein downstream of the actual causal gene.
For most common traits, our knowledge of the number of gene variants involved and the magnitude of their effects does not account for the observed variability. Each SNP only contributes a small amount, so only targeting one of these may not offer therapeutic benefit.
In GWAS and candidate gene studies, SNPs and other small sequence variations are screened routinely.
However, structural changes in the genome, such as CNVs, large deletions, inversions and transpositions, have been more difficult to identify but can have a large impact.
Limitations are placed by the sample size needed to detect a statistically significant effect, requiring matched controls for disease traits and ethnicity. Recent T2DM analysis used ~150,000 individuals. Typically, the magnitude of the effect from an identified association is small - most T2DM loci have an odds ratio of 1.05–1.15.
Rarer alleles with greater effect may be more difficult to detect. This distribution might be expected for disease traits which reduce reproductive capability, so they are less likely to be passed on.
Low-frequency variants with an intermediate effect size are therefore the ones with the most potential for the development of personalised medicines.
Genetic variant effects may be underestimated due to:
- Matching control group (age, sex, ethnicity, environment)
- Accurate phenotype classification
- Complex context: gene-environment and gene-gene interactions
Genetic variation and Warfarin
Warfarin is an oral anticoagulant.
The optimal plasma concentration of warfarin must be tightly controlled. A lack of efficacy (dose too low) and bleeding (too high) are both problems.
There are 2,000,000 prescriptions/year in the US and 8 bleeding events per 100 patients, so there is a clear case for improving management.
There are 2 well-established genetic variants for warfarin efficacy and safety:
- Pharmacokinetic: CYP2C9 variants
- Pharmacodynamic: Vitamin K epoxide reductase (VKOR) haplotype
This led to genomic tests for CYP2C9 and VKOR being implemented in some US hospitals.
The FDA includes dosing algorithms based on genotype for warfarin on the product labelling.
However, the FDA/NICE have still not yet recommended genotype-guided dosing.
There are concerns over whether CYP2C9/VKOR are the only predictive genetic variants, even after GWAS.
It is also unclear what the value of the added benefit this approach has over the existing clinical practice: low dose first, monitor efficacy after ~4 weeks by international normalised ratio (INR) for clotting time, and adjust dosing accordingly. It may take 3 months to identify the ideal dose, but there are not enough studies to demonstrate that reaching this faster prevents a significant number of bleeding or clotting events.
Cystic fibrosis personalised medicine
Cystic fibrosis is an inherited, single gene disorder.
It leads to impaired function of an epithelial chloride channel, the cystic fibrosis transmembrane conductance regulator (CFTR), which drives fluid secretion.
This affects the lungs and digestive system.
This is a recessive disorder, so two mutant CFTR genes are required to cause the disease. 1 in 25 people are carriers and there are >10k patients in the UK.
>2000 mutations in CFTR are associated with the disorder. These produce a wide range of impacts on channel function, varying in severity.
The most common CFTR mutation is ΔF508 (deletion of Phe 508): >90% of patients have at least 1 copy and >40% are homozygous.
This mutation primarily affects the cell surface expression of the channel. The CFTR ΔF508 mutation prevents proper folding of the channel protein in the ER, which leads to its degradation before it can reach the cell surface.
This could be targeted through the development of a pharmacological chaperone, lumacaftor, to aid correct channel protein folding.
Lumacaftor is a “first in class” example of a pharmacological chaperone, approved by the FDA as Orkambi.
Pharmacological chaperones might provide a pharmacogenetics approach to rescue defective receptors, ion channels, enzymes etc. in the future.
However, this approach requires a well-defined monogenic cause of disease.
Precise diagnostics are required as lumacaftor does not work for all CFTR mutations.
This therapeutic offers modest effect/benefit over existing therapies. It also costs £500 m (NHS) versus £2.5 billion (Vertex). It works well, but not much better than other available drugs, so it is not considered ‘worth it’ in health economics and is therefore not covered by the NHS.
Cost effectiveness
Cost effectiveness has to be assessed to determine if development of novel therapeutics and diagnostic tests will be worthwhile.
This is influenced by:
- Magnitude of clinical benefit (efficacy/safety) compared to current approaches
- Savings by avoiding ineffective medication and lower ADR incidence
- Size of the patient cohort who will benefit
- Complexity of associated diagnostic test and its implementation in clinical setting
Routine pharmacogenomic tests have to be predictive and clinically robust, even with “simpler” inherited disorders.
Reduced sequencing costs over the following decades may lead to everybody’s genome being sequenced and stored electronically. This is already happening with DNA testing kits like 23andme.
Regulation
Routine pharmacogenomic information may force reassessment of the drug licensing framework.
International cooperation may occur to carry out large cohort studies for evaluation of drug efficacy/safety in rare genetic variants.
Evaluating safety decisions – can drug withdrawals for rare, serious side effects be minimised through genotyping and avoiding susceptible cohorts?
Society and ethics
Some questions surrounding the implementation of pharmacogenetics include:
- Should relevant genomic tests be compulsory for receiving a particular drug, if this is necessary for clinical assessment of efficacy and safety?
- Should patients be refused treatment on the basis of genetic evidence which suggests they will not benefit?
We must also consider the impact of routine personal genome information on wellbeing and family planning.
It is important to understand the limits of health predictions from the DNA genome sequence alone.
There are also ethical issues surrounding confidentiality and the idea of sharing knowledge of identified risk factors to family members, health insurance, employers, etc.
