Research into genetics

Following publication of the ‘working draft’ of the human genome in 2001, it was discovered that there are more than 1.4 million single nucleotide polymorphisms (SNPs) in the genome, at an average density of one SNP per 1900 bases. Since 60,000 of these SNPs are estimated to lie within the regions of genes that code for protein molecules, there is a vast range of possible variations in human genes that could be associated with alterations in response to drugs. Furthermore, it is estimated that 85% of the known coding regions of genes lie within 5000 bases of an SNP, which may therefore be close enough to influence the gene’s level of expression or function.¹

Given the advances in DNA sequencing technology and the knowledge gained from the human genome project, it is unsurprising that there has been a growth in the volume of research into pharmacogenetics. Conventional clinical trials test medical interventions in large numbers of patients in order to balance out the variations in responses between patients, due to genetic and non-genetic factors. However, being able to use a genetic test to predict whether a particular drug should be used or its dosage is an appealing prospect.

This optimism is enhanced by the fact that the genetic make-up of a patient is constant over life and need only be sequenced once. Coupled with improved methods (such as DNA microarray technology) of detecting large numbers of polymorphisms simply, quickly, and cheaply, some advocates have suggested that personalised treatment for a range of conditions, including HIV, is imminent. Furthermore, some experts are so confident of the potential of pharmacogenetics that they have called for all clinical trials to involve taking patient-consented blood samples for use in future genetic association studies.².

However, while a few genetic variants have been linked to differences in drug levels, side-effects, or therapeutic responses, a huge number of published associations remain confined to research studies and have failed to reach the clinic. Unfortunately, few genetic associations meet these criteria.

The major reason for the poor rate of discovery of meaningful genetic variants lies in the complexity of the biological systems involved in mediating a drug’s effects. Once a drug has been administered, it must be absorbed and distributed to its site of action, where it interacts with its targets before it is broken down and excreted from the body. Any of these processes can be affected by genetic variation, as well as other factors, and each of them can be mediated by more than one gene product.

For example, the breakdown of many drugs is mediated by the cytochrome P-450 enzymes in the liver. Many of these enzymes are polymorphic, with different genetic variants producing enzymes with different rates of activity and at different levels of expression. However, the link between the activity or expression of these enzymes and drug levels is complicated by the fact that drugs can often be broken down by more than one member of this class of enzymes. Consequently, the presence of a genetic variation in one gene product may be concealed by the activity of other enzymes.

As an illustration, many people have a genetic inability to express the CYP3A5 gene due to a single change in the gene’s sequence.³ However, many drugs are metabolised by both CYP3A5 and CYP3A4, so the absence of CYP3A5 is obscured, unless the person also has a dysfunctional form of this second gene.

The final effects of a drug in the body therefore depend on the interaction of a huge number of gene products, all of which may be subject to genetic variation. The results of genetic analyses must therefore be interpreted cautiously: the chances of investigators discovering that a particular polymorphism is associated with a particular effect of a drug, such as increased drug levels in the blood or the increased incidence of side-effects, runs a high risk of finding false-positive results.

This is a particular danger with smaller studies, and with those that look at a wide range of genetic polymorphisms as possible causes of the trait in question. As techniques for detecting polymorphisms improve, allowing researchers to sequence many genes or detect hundreds of polymorphisms in small tissue samples, the risks of this are set to escalate.

Furthermore, genetic variation rarely causes noticeable effects on the outcomes of drug therapy. In contrast, the small changes in protein structure and function brought about by single nucleotide or amino acid changes may make a very small contribution to the drug’s effects. Only when considered alongside other polymorphisms within the same gene, within other genes, and within stretches of DNA that control the expression rates of these genes, can a fuller picture of the link between genotype and phenotype be painted. As the number of elements to be taken into consideration increases, so does the complexity of predicting a clear outcome.

For example, a mutation in the gene for the enzyme CYP2B6, which is responsible for the metabolism of the non-nucleoside reverse transcriptase inhibitor efavirenz (Sustiva), is known to have a strong influence on blood drug levels, the rate at which it is cleared from the body, and the severity of its side-effects. However, a recent study has shown that knowing which versions of this polymorphism an individual patient has is not sufficient to predict how long it will take for the drug’s level to fall in that patient, as there is considerable variability within each genotype group and overlap between groups.⁴ It is likely that this within-genotype variability is caused by variation in other factors, including other genetic elements, as well as non-genetic or environmental influences.

A further difficulty with genetic association studies is the risk of identifying a polymorphism that appears to be linked to an outcome, but that does not actually cause it. Since the DNA that makes up a gene can be thought of as a long string of bases, it is possible that a polymorphism identified in a study is not involved directly in affecting the outcome, but it is close to another polymorphism further along the gene that is responsible. The closer the two polymorphisms are to one another along the gene, the greater the chance that they will be passed on as a pair. The two polymorphisms are said to be ‘linked’: it is therefore possible that a polymorphism that is silent, having no effect on protein structure or function, will be linked to another polymorphism that does have an influence on an outcome, but which has not been identified in any association studies.

Hence, polymorphisms are best thought of as markers of changes in a protein’s structure and function and not necessarily as causes of change. For example, the link between the HLA-B*5701 polymorphism and abacavir (Ziagen) hypersensitivity, which has been introduced as the sole genetic test in HIV medicine, may in fact be a marker of another linked polymorphism that is actually responsible for the biological effect.⁵ Nevertheless, the association between the HLA-B*5701 allele and the drug reaction is impressive, and robust enough to have predictive power in the clinic. Until an alternative with better predictive power is identified and confirmed in large studies, the present test is likely to continue to be introduced into the care of patients starting treatment with abacavir.

Genetic variations in HIV co-receptor usage, chemokines, and cytokines can also affect disease progression. Researchers recently documented their finding that CCl3L1 copy number and CCR5 polymorphisms can affect susceptibility to HIV and immune reconstitution after antiretroviral therapy. Their results indicate a benefit to using this prognostic information, along with traditional laboratory measures, to determine the timing of therapy.^{6 7} See Genetic susceptibility in the section The immune system and HIV.

Because of the increasing research volume and large number of genes that can be surveyed, experts have drawn up specific guidelines to standardise pharmacogenetic approach and analysis.⁸ These guidelines, which must be followed in order for publication by most biomedical journals, are designed to eliminate the volume of false positive associations that are published. These include using sample sizes that are large enough to reveal meaningful associations between genes and outcomes, justifying the genes analyses on biological grounds, and using stringent statistical analyses.

Some simple relationships between genetic polymorphisms and the effects of drugs have been discovered and translated to the clinic. However, despite optimism that the human genome project would lead to individualised medicine within a few years, it seems that the complexity of drug response will result in genetic testing being limited to a few isolated examples until more information is gathered on the effects of genes, interactions between different genes, and interactions between genes and the environment.