Impact of genomics on drug discovery and clinical medicine

Abstract

Genomics, particularly high‐throughput sequencing and characterization of expressed human genes, has created new opportunities for drug discovery. Knowledge of all the human genes and their functions may allow effective preventive measures, and change drug research strategy and drug discovery development processes. Pharmacogenomics is the application of genomic technologies such as gene sequencing, statistical genetics, and gene expression analysis to drugs in clinical development and on the market. It applies the large‐scale systematic approaches of genomics to speed the discovery of drug response markers, whether they act at the level of the drug target, drug metabolism, or disease pathways. The potential implication of genomics and pharmacogenomics in clinical research and clinical medicine is that disease could be treated according to genetic and specific individual markers, selecting medications and dosages that are optimized for individual patients. The possibility of defining patient populations genetically may improve outcomes by predicting individual responses to drugs, and could improve safety and efficacy in therapeutic areas such as neuropsychiatry, cardiovascular medicine, endocrinology (diabetes and obesity) and oncology. Ethical questions need to be addressed and guidelines established for the use of genomics in clinical research and clinical medicine. Significant achievements are possible with an interdisciplinary approach that includes genetic, technological and therapeutic measures.

Introduction

Genome research centers worldwide are engaged in the Human Genome Project (HGP) with the ultimate goal of elucidating and characterizing the complete sequence of the 3×10⁹ base pairs (bp) arranged in about 85 000 genes of the human genome. An even greater task is to determine their function and interplay. The genomic approach to mapping and sequencing the genome project has accelerated the rate of gene discovery. In 1990, 1772 human genes were identified and mapped to a specific chromosome or region of the genome. In September of 1996, this number was 3868 genes—a more than two‐fold increase. As of June 1996, 62 human genes linked to human diseases had been isolated by genomic technologies and, of these, 51 (82%) were available in the public domain as clones or as DNA sequences. Moreover, biomedical research is rapidly defining the molecular mechanisms of pharmacological effects, genetic determinants of disease pathogenesis, and functionally important polymorphisms in genes that govern drug metabolism and disposition. A radical new, but complementary, approach to drug development is now emerging which promises dramatic improvements in the efficiency and speed of drug development. This approach uses the emerging technological expertise from pharmacogenetics, pharmacogenomics and functional genomics to dissect, predict and monitor the nature of the individual response to medications. Ultimately, this may lead to smaller and faster clinical studies and to individually tailored pharmacological treatments, in which patients are screened to identify which therapeutic option most suits their genetic and physiological makeup and accurately monitored for their response. This approach is likely to have radical consequences in the planning, conduct of clinical trials and medical treatment of diseases.

One important outgrowth of molecular medicine is the development of technologies for the transfer of therapeutic genes to cells in culture and tissues in vivo, with potential applications both to medical research and the practice of clinical medicine. The use of genomic databases to find new targets for drug discovery and the rapid accumulation of human gene sequences is promising for clinical medicine, if the molecular level can be translated into improved interventions. If it can, therapeutic agents with specific molecular functions can be produced, be they gene products which are deficient or abnormal in the patients, or drugs with direct transcriptional or molecular effects. Individual genetic testing, with knowledge of disease genes, will help early diagnosis and early treatment. For example, recent advances in the genetics of complex traits (for example, diabetes, coronary heart disease and Alzheimer's disease) have to some extent reshaped disease phenotypic descriptions. The techniques developed for automated sequencing and analysis of DNA may eventually allow inexpensive screening of multiple loci for polymorphisms.

Molecular genetic techniques may also translate into gene therapy. The ability to clone and manipulate genes responsible for human disease and to re‐introduce functional copies of normal genes into living cells and tissues is one such therapeutic objective. The potential clinical applications of gene therapy are numerous, and a number of specific human genetic or environmentally‐induced diseases that result from a lesion in a single gene have been proposed as candidates for gene therapy (Table 1). For some of these diseases, the introduction of a functional homologue of the defective gene and the production of even small amounts of the missing gene product would have a beneficial effect; for example 10–20% production of the normal levels of factor IX can alleviate severe haemophilia B.¹ At the same time, overexpression of the gene product would not be expected to have deleterious effects. Thus, these genetic disorders are candidates for gene therapy because the expression of the transduced gene need not be strictly regulated. In contrast, it is not always necessary to correct the genetic lesion in the cell type that shows the defect. In such cases, a therapeutic gene may be introduced into another cell type so that the genetically modified cells functionally replace the defective cell type.

This paper discusses the impact of genomic science on drug discovery and clinical medicine and provides examples of treatment interventions in neuropsychiatry (migraine, neurological channelopathies and neurodegenerative disorders), cardiovascular medicine, endocrinology (diabetes and obesity) and oncology.

Table 1

Some examples of genetic diseases

Genetic disease	Deficient gene product	Cell type	References
Lesch‐Nyhan syndrome	Hypoxanthine phosphoribosyl transferase	Basal ganglia	239, 240
Duchenne muscular dystrophy	Dystrophin	Muscle	241
Parkinson's disease	Dopamine	Substantia nigra	242
Neurological channelopathies
Voltage‐gated
Paramyotonia congenita	α1 subunit of the sodium channel (SCN4A)	Muscle
Thomsen's disease (Autosomal dominant myotonia congenita)	Allelic disorders associated with mutations in a gene	Muscle
Becker's disease (Autosomal recessive myotonia congenita)	coding for skeletal muscle chloride channel (CLCN1)	Muscle
Familial hyperkalaemic periodic paralysis	α1 subunit of the sodium channel (SCN4A)	Muscle
Familial hypokalaemic periodic paralysis	α1 subunit of a skeletal muscle calcium channel	Muscle
	(CACNL1A3)
Episodic ataxia type 1	Potassium channel genes	Cerebellum and peripheral nerve
Episodic ataxia type 2	α1 subunit of a brain specific calcium channel	Brain	243
Ligand‐gated
Familial startle disease	α1 subunit of the glycine receptor		82
Nocturnal frontal lobe epilepsy	α4 subunit of the nicotinic acetylcholine receptor		83
Familial paroxysmal choreoathetosis	Chromosome 1p where a cluster of potassium channel		84
	genes is located
Familial hypercholesterolemia	Low‐density lipoprotein receptor	Liver cells	244

Genomics

Molecular genetics reached human genetics about 1976, when the first human genes were cloned.² Transgenic methods, ‘knock‐outs’ and ‘knock‐ins’ began in about 1986, and in about 1996, database searching became a fruitful way to do genomic research.³ The term ‘genome’ refers to an organism's complete set of genes and chromosomes. The term ‘genomics’ describes the scientific discipline of mapping, sequencing, and analysing genomes.⁴

Genome analysis may be divided into structural and functional genomics. Structural genomics is an initial phase of genome analysis, and has a clear endpoint which is the construction of high‐resolution genetic, physical, and transcript maps of an organism (its complete DNA sequence). This genotypic approach focuses on understanding how genotypic variation gives rise to phenotypic variation, relying on physical and genetic maps and easily‐typed DNA sequence polymorphisms. The expression approach (functional genomics) relies on the large collection of partially‐sequenced cDNA clones. The benefits of the information arising from the accumulation of human gene sequences includes developing systematic ways of finding genes of interest, and their functions; hence ‘functional genomics’. The genes cloned and their corresponding DNA sequences provide the tools for comprehensive characterization of the expression patterns of this entire set of genes, and for systematic experimental investigations of the functional properties of their products. Thus, functional genomics, which represents a new phase of genome analysis, makes use of the structural genomics information. The investigation is primarily a systematic approach to elucidate the genome and its functions.

The fact that most diseases do not follow a simple inheritance patterns has led to a significant challenge in the genetic dissection of the complex traits of diseases such hypertension, Alzheimer's disease, schizophrenia and diabetes.⁵ Four major approaches have been developed: linkage analysis, allele‐sharing methods, association studies in human populations, and polygenic analysis of experimental crosses in model organisms such as mouse and rat. The gold‐standard tests for human genes should include association studies demonstrating a clear correlation between functionally relevant allelic variations and the risk of disease in humans, and transgenic studies demonstrating that gene addition or gene knockout in animals produces a phenotypic effect. If these genetic approaches are successful, they may have significant relevance on drug research and clinical medicine.

Pharmacogenomics

Pharmacogenomics has its roots in pharmacogenetics. Whereas pharmacogenetics is the study of the linkage between an individual's genotype and that individual's ability to metabolize a foreign compound, pharmacogenomics is quite broad in scope, and is similar to molecular medicine, aiming to detect, monitor and treat the molecular causes of disease. Pharmacogenomics involves the application of genomics technologies such as gene sequencing, statistical genetics and gene expression analysis to drugs in clinical development and trials. Since many diseases develop as a result of a network of genes failing to perform correctly, pharmacogenomics can identify the genes or loci which are involved in determining the responsiveness to a given drug. In this way, genetic characterization of patient populations is becoming an integral part of the drug discovery and development process. Pharmacogenomics may aim to capitalize on these new molecular insights to discover new therapeutic targets and interventions and to elucidate the constellation of genes that determine the efficacy and toxicity of specific medications.

Impact on drug discovery and clinical research

Applying pharmacogenomics in the preclinical setting, one may start screening compounds with the least variation across individuals. If the target gene is selected, the compound that works best overall against all its subtypes may be chosen. Thus, drug selection is substituted for patient selection, decreasing the uncertainties that patient stratification introduces at the FDA and in marketing, as well as the need for a genetic screening. Genomics may also be used to select out adverse effects before drugs enter the clinic. For example, the gene‐expression pattern for the liver of an animal administered a drug can indicate whether gene pathways related to toxicity have been turned on. Variations in gene expression levels may prove just as useful as genetic variation in predicting drug response at any stage in the clinic and as a diagnostic. Pharmacogenetic data are vital during the development of a compound with a narrow therapeutic index or which is metabolized from a prodrug, as such information may influence decision of whether to discontinue development or design trials to study clinical responses in individual polymorphic for the relevant enzyme.

Significant issues at the preclinical level usually need to be addressed. Problems of medicinal chemistry, developing drugs with the appropriate absorption, metabolism, distribution, and elimination profiles still have an empirical basis. Nonetheless, small molecule drugs directed toward targets discovered by genomics may soon account for a great majority of drugs introduced into the marketplace.

Pharmacogenomics may benefit many stages of clinical drug development. It will significantly affect trial design, primarily through improved inclusion/exclusion criteria and more effective assessment of patient responses. Genes linked with drug metabolism in preclinical studies could be genotyped in patients recruited for phase I trials. Any genotype that correlates with adverse effects could then be used to screen out relevant patients in subsequent trials. Furthermore, if efficacy data are collected during phase I trials, polymorphisms in the drug target gene could be typed in phase I participants to assess whether they are linked with side‐effects or with variations in drug response. That analysis could obviously be further refined in phase II trials, enabling companies to undertake phase III trials in a subgroup of patients that responds well and exhibits fewer side‐effects. The resultant drugs would be expected not only to have better efficacy, but also a better safety profile.

At the clinical level, while the disease symptoms might appear to be uniform, individual‐to‐individual variations in these polygenic networks may make drugs healing for certain individuals while toxic for others. Pharmacogenomics can sometimes correlate gene variations with differential responses to the same drug leads, thereby hoping to accelerate novel drug discovery dramatically, by defining specific populations that will benefit from a drug. While this approach may maximize the medical utility of existing pharmaceuticals, it could also rescue dead drugs. Several products that have failed in recent years in late stage clinical trials may on retrospective analysis be effective in subsets of patients, although at the time, there was no clear way of recognizing such subsets clinically. A study of the genetic differences between the individuals could provide an answer. Consequently, traditional approaches that focus on broad groups of patients with a diagnosis (e.g. Alzheimer's disease) may need to be much more precisely divided into subsets of patients who may have a traditionally defined disease amenable to treatment based on a particular molecular target. These pharmacogenomic developments should lead to smaller, more rapid and cost‐effective trials, and ultimately to more individually focused and effective therapeutics.

Relevance to clinical medicine

In summary, patient segmentation and individual profiling will become increasingly important and pharmacogenomics analysis may serve to customize the use of pharmaceuticals for specific subgroups of patients. Over the past 20 years, genetic heterogeneity has been increasingly recognized as a significant source of variation in drug response. Some drugs work better in some patients than in others, and some drugs may even be highly toxic to certain patients. Pharmacogenomics can be used not only to predict drug efficacy in a particular patient, but also the likelihood of side‐effects, for example those due to differences in drug metabolism. This information could be used to rescue failed or failing drugs, by repositioning them for a defined patient population. Pharmacogenomics is about spotting correlations between such responses to drugs and the genetic profiles of patients.⁶ It generates data that are relevant to a drug's clinical performance. Ultimately, knowledge of the genetic basis for the drug disposition and response should make it possible to select many medications and their dosages on the basis of each patient's inherited ability to metabolize, eliminate and respond to specific drugs. Therefore, for clinical management in general, genetic subtyping of the patient response, whether due to different disease subtypes or differential drug effects, should make it possible gradually to replace the current trial‐and‐error‐based selection of the appropriate pharmacological intervention with a more informed and rational strategy. This will represent an important advance, particularly in diseases where it currently may require months or years of treatment to observe whether a positive response has occurred.

Genetic variations in pharmacokinetic and pharmacodynamic effects

Numerous factors, including genetics, affect drug metabolism and thus alter the bioavailability of therapeutic drugs. The best studied metabolizing enzymes are the cytochrome P450 (CYP450) isoenzymes, the N‐acetyl transferase (NAT) isoenzymes, the UDP‐glucuronosyl transferases, and the methyl transferases. Of these enzymes, the CYP450s are very important because they metabolize drugs into products that are readily excreted into the urine and faeces. In humans, six different forms of CYP450 (CYP1A2, CYP2C9, CYP2C19, CYP2D6, CYP2E1 and CYP3A4) are largely responsible for eliminating drugs. The rate of metabolism by several of the cytochrome CYP450 enzyme subfamilies varies, due to genetically‐determined polymorphisms in all populations studied. Recent research using phenotyping and genotyping techniques has reflected the interest and importance of these pharmacogenetic factors in determining drug responses. Some of the metabolizing enzymes such as CYP1A1, 1A2, 2A6, 2C9, 2C19, 2D6, 2E1, NAT1, NAT2 and NQO1 exhibit genetic polymorphism and alter responses to drugs (see Table 2). These metabolic polymorphisms are determined by gender (e.g. CYP1A2) and racial/ethnic origin.^7–8 Increased CYP1A activity (an enzyme catalysing a phase I oxidation reaction), coupled with slow acetylation (a phase II conjugation reaction), resulted in less myelosuppression from the active metabolites of the drug amonafide.⁹ Because every individual represents a combination of drug‐metabolizer phenotypes, given the large number of enzymes involved in drug metabolism, it is apparent that some individuals are likely to have unusual reactions to drugs, or to combination of drugs, due to the coincident occurrence of multiple genetic defects in drug‐metabolizing enzymes. Such an alignment of genotypes, particularly when coupled with polymorphisms in drug receptors, is likely to constitute part of the mechanism for the so‐called ‘idiosyncratic’ drug reactions.

Although no evidence to date suggests the CYP3A4 isoenzyme exhibits genetic polymorphism, in recent years there has been much discussion about the 3A4 system because of life‐threatening arrhythmic side‐effects that can occur as result of enzyme inhibition and accumulation of the antihistamines terfenadine, astemizole and cisapride.^10–12 Terfenadine has been removed from the market because of its serious cardiovascular drug interactions. Concerning CYP2C9, recent data suggest that patients who require low doses of warfarin (≤1.5 mg/day) carry point mutations (alleles CYP2C9*2 and CYP2C9*3) at the gene coding for CYP2C9 (which could occur at a frequency of 21% in the general population). These patients metabolized warfarin poorly, and responded to small doses of the drug with greater lengthening of the prothrombin time and higher international normalized ratio (INR) values than did carriers of the wild‐type allele CYP2C9*1.¹³ Genetically determined high‐responders to warfarin had bleeding complications four times more commonly than did a control group stabilized on larger doses of the drug. Knowledge of carriage of the hyper‐responsiveness alleles of CYP2C9*2 and CYP2C9*3 might help the clinician to decide against the use of warfarin (in favour of other coumarin derivatives such as phenoprocoumon and acenocoumarol, the metabolism of which is less influenced by CYP2C9), particularly in high‐risk elderly patients.

In addition to variation in drug metabolism or pharmacokinetics, the genetic variations in receptor function (and thereby pharmacodynamic effects) are important. Subtle differences in the sequences of receptor subtypes for dopamine, serotonin and catecholamines may result in individual differences in behavior and drug responses.⁶ Overall, a highly complex picture emerges in which genetic variation in both pharmacodynamic and pharmacokinetic factors contributes to drug responses. Some patients do not respond to a given drug because it is not processed efficiently; other patients do not respond because the disease gene defects or its pathway is not targeted by the drug.

Great progress has been made in understanding the molecular genetics of acetylation as well as the clinical consequences of being a rapid or slow acetylator. Inborn errors (several different alleles) at the hepatic arylamine N‐acetyltransferase‐2 (NAT2) locus are responsible for the traditional acetylator polymorphism.¹⁴ Rapid and slow acetylators reflect the genetically determined variation in the elimination of xenobiotics, as well as in NAT2 activity in the liver and other tissues.¹⁵ The human NAT2 gene contains an 870 bp intronless protein‐coding region.¹⁶ To date, one allele with a code for fast acetylation (wild‐type) and several mutated alleles with codes for impaired acetylation activity have been discovered.^17–18 Of all the NAT2 allelic variants that had been identified, three (NAT*5, NAT*6 and NAT*7) account for majority of the slow NAT2 acetylator genotype in White subjects.¹⁸ N‐acetylation status seems to be associated with several kinds of diseases, such as colon cancer, rheumatoid arthritis, and systemic lupus erythematosus.^19–21 The independent genetic feature as a rate of acetylation was shown to be related to the immunological system dysfunction. It may be one of the factors that makes an individual susceptible to the development of an atopic disease, and one study showed that up to 80% of individuals with chronic allergic rhinitis had a slow acetylation phenotype.²² A recent study which assessed the influence of NAT2 polymorphism on the risk of development of atopic disease also suggests that the risk of development of atopic diseases was five‐fold greater for homozygous slow acetylators compared to healthy subjects, and that slow acetylation genotype may be an important factor of individual susceptibility to atopic diseases.²³ This group of patients may also be at increased risk of adverse reactions after using drugs which are mainly metabolized by acetylation reaction. Among them, the mechanism of hypersensitivity to sulfonamides typical for slow acetylators seems to be of particular importance.

Studies have revealed variant alleles at the NAT1 locus as well. The consequences of pharmacogenetic variation in these enzymes include altered kinetics of specific drug substrates, drug–drug interactions resulting from altered kinetics, and idiosyncratic adverse drug reactions. The latter have been extensively investigated for the arylamine‐containing sulfonamide antimicrobial drugs. Individual differences in multiple metabolic pathways can increase the likelihood of covalent binding of reactive metabolites of the drugs to cell macromolecules with resultant cytotoxicity and immune response to neoantigens. This can result clinically in an idiosyncratic hypersensitivity reaction manifested by fever, skin rash and variable toxicity to organs including liver, bone marrow, kidney, lung, heart and thyroid.

Consideration of the genetic characteristics leads to population segmentation into groups, the slow metabolizers (having a slow metabolism) and fast metabolizers (having a normal metabolism). For example, in some Asian populations the incidence of poor metabolizers of the gastrointestinal drug omeprazole (due to polymorphism in CYP2C19) is 15–23%, compared to 2.5–6% in Caucasians.²⁴ In individuals with a poor‐metabolizer genotype for CYP2C19, the therapeutic efficacy of omeprazole (a proton‐pump inhibitor widely used as acid inhibitory agent for the treatment of upper gastrointestinal diseases and metabolized by CYP2C19) may be increased.²⁵ In patients with a poor‐metabolizer phenotype or genotype of CYP2C19, the area under the plasma concentration‐time curve of omeprazole is markedly increased, and the clinical effect of omeprazole is greater. Acid secretion in patients with a poor metabolizer status of CYP2C19 who are undergoing an omeprazole therapy is therefore assumed to be more strongly inhibited than those with the extensive metabolizer status. Cure rates for Helicobacter pylori were noted to be 28.6%, 60% and 100% in the rapid‐, intermediate‐, and poor‐metabolizer groups, respectively. The results of the genotyping test for CYP2C19 seem to predict the cure of Helicobacter pylori infection and peptic ulcer in patients who receive dual therapy with omeprazole and amoxicillin. A recent study designed to determine whether the effects of omeprazole on intragastric pH depends on CYP2C19 genotype status confirmed that after omeprazole administration, significant differences in mean intragastric pH values and plasma levels of gastrin, omeprazole and its metabolites were observed among the three groups of volunteers (homozygous extensive metabolizers, heterozygous extensive metabolizers and poor metabolizers), whereas no significant differences in these parameters were observed with the placebo administration.²⁶ Both the individual omeprazole AUC and mean intragastric pH values were greater in the poor metabolizer group compared with those in the homozygous extensive metabolizer and heterozygous extensive metabolizer groups. The results confirmed that the effects of omeprazole on intragastric pH significantly depends on CYP2C19 genotype status, and also suggest that the genotyping test of CYP2C19 may be useful for an optimal prescription of omeprazole.

Low metabolic activity of the CYP2D6 enzymes is inherited as an autosomal recessive gene and although CYP2D6 represents only about 1.5% of the total liver enzymes, it is involved in the metabolism of a number of commonly used drugs.⁷ There are now more than 20 identified variant CYP2D6 alleles which contribute to the variation in CYP2D6 metabolism. The most common allelic variations associated with poor‐metabolizers in Caucasians are CYP2D6*4 (75%), *3 (5%) and the gene deletion *5 (15%).²⁷ For drugs in which CYP2D6 plays a predominant role in metabolism, poor‐metabolizers will have high plasma concentrations and report the most severe adverse reactions.²⁸ Studies in Caucasian extensive‐metabolizers and poor‐metabolizers have uniformly demonstrated a 2‐ to 5‐fold difference in the capacity to metabolize CYP2D6 substrates, such as antidepressants and neuroleptics.²⁹ On the other hand, non‐Westerners (Asians and Indians) may require lower doses of several classes of psychotropics that are metabolized by CYP2D6 (e.g. conventional neuroleptics and tricyclic antidepressants) than do Westerners.³⁰ The poor‐metabolizers lack this enzyme as a result of an autosomal recessively transmitted defect in its expression. When drugs are converted to an active metabolite by 2D6 (e.g. conversion of codeine to morphine), the drug may be ineffective in poor‐metabolizers. Although significant interactions between 2D6‐metabolized drugs with the well‐known inducers rifampin and antiepileptics have been described, specific inducers of 2D6 have yet to be clearly identified. Administration of dextromethorphan followed by measurement of O‐demethylated metabolite excretion in urine is an accurate and non‐invasive way of phenotyping individuals as either extensive‐metabolizers or poor‐metabolizers for 2D6 activity.

Many opioid analgesics are activated by CYP2D6, rendering the 2–10% of the population who are homozygous for non‐functional CYP2D6 mutant alleles relatively resistant to opioid analgesic effects.³¹ It is thus not surprising that there is remarkable interindividual variability in the adequacy of pain relief when uniform doses of codeine are used.

Nothing is known about any particular advantage or disadvantage of any CYP2D6 variant. Because CYP2D6 occurs not only in the liver but also in the brain, it might affect some personality traits: this might affect fitness and thereby frequency of the variants.^32–33

Thiopurine methyltransferase (TPMT) is a cytosolic enzyme that catalyses the S‐methylation of aromatic and heterocyclic sulfhydryl compounds, including the thiopurine drugs 6‐mercaptopurine (6‐MP) and 6‐thioguanine.³⁴ Thiopurines are used to treat patients with neoplasia and autoimmune disease as well as recipients of transplanted organs. The TPMT genetic polymorphism may represent a striking example of the potential clinical importance of pharmacogenetic variation in expression of a drug‐metabolizing enzyme.³⁵ Individuals with genetically very low levels of TPMT activity are at a greatly increased risk for potentially life‐threatening toxicity when exposed to standard doses of thiopurines, while those with very high levels of this enzyme activity may be undertreated with the same dosages of these drugs.^36,37 Recent genetic data suggest that the active gene for the TPMT enzyme is ∼34 kb in length, consists of 10 exons and has been localized to chromosome band 6p22.3.³⁸ The wild‐type allele for high TPMT activity has been designated TPMT*1, and to date eight variants for very low TPMT activity have been reported.^38,39 The most common of these in Caucasians, TPMT*3A, represents 55–70% of all variant alleles for very low activity.³⁹ TPMT*3A contains two point mutations, G460→A and A719→G, resulting in Ala154→Thr and Tyr240→Cys amino acid substitutions, respectively.³⁸ However, because of the clinical significance of inherited variation in levels of TPMT activity, characterization of as many variant alleles responsible for very low TPMT activity as possible will be necessary so that DNA‐based diagnostic tests can be compared with the phenotypic test presently used to individualize therapy with thiopurine drugs. The ultimate aim is to minimize toxicity and improve the therapeutic efficacy of this important class of pharmacotherapeutic treatments.

Besides helping delineate such biological differences, genetic markers could also be used in selecting patients for clinical trials either for screening out individuals with genotypes that react adversely or for selecting patients who are more likely to respond well. The advent of DNA chip technology presents the opportunity not only to rapidly genotype individuals to provide information on polymorphic drug metabolism genes, but also to identify genes differentially expressed in response to a drug. Affymetrix is already marketing a CYP2C6/CYP2C19 genechip for identifying potential poor drug metabolizers. The recent development of a simple mouthwash method for obtaining genomic DNA clinical studies appears promising.⁴⁰ Using this cheap and simple‐to‐perform approach, subjects were successfully genotyped by PCR‐based assays for polymorphisms in the CYP1A1, CYP2E1 and NQO1 genes, confirming that the quality of DNA isolated from mouthwash samples was sufficient to support PCR amplification reliably. This mouthwash procedure may be suitable for large community‐based studies of genetic susceptibility to disease in which samples can be collected by the participants themselves. However, there should be agreed standards for the reproducibility and robustness of such systems.

The pharmacodynamics of drug action may be subject to genetic variation with respect to the sensitivity of the drug's target to its action (e.g. due to subtle conformational variations at the drug's binding site). Most drugs interact with specific target proteins to exert their pharmacological effects, such as receptors, enzymes, or proteins involved in signal transduction, cell cycle control, or many other cellular events. Molecular studies have shown that many of the genes encoding these drug targets exhibit genetic polymorphism, which in many cases alters their sensitivity to specific pharmacological treatments. Such examples include polymorphisms in β‐adrenergic receptors and their sensitivity to β‐agonists in asthmatics, sulfonylurea receptor and responsiveness to sulfonylurea hypoglycaemic agents, and 5‐hydroxytryptamine receptor and response to neuroleptics such as clozapine.^41,42 In addition, genetic polymorphisms that underlie disease pathogenesis can also be major determinants of drug efficacy, such as mutations in the apolipoprotein E (ApoE) gene and responsiveness of patients with AD to tacrine therapy.⁴³ Finally, the risk of adverse drug effects has been linked to genetic polymorphism that predispose to toxicity, such as dopamine D3 receptor polymorphism and the risk of drug‐induced tardive dyskinesia, potassium channel mutations and drug‐induced dysrhythmias and polymorphism in the ryanodine receptor and anesthesia‐induced malignant hyperthermia.⁴⁴

Bronchodilator responsiveness to β₂‐adrenergic receptor agonists in patients with asthma varies considerably and several missense mutations in the coding region of the β₂‐adrenergic receptor gene have been identified.^45,46 Among the general population (including patients with asthma), β₂‐adrenergic receptor alleles are distributed in the following approximate proportions: homozygous Arg (Arg16/Arg16), 15%; heterozygous (Arg16/Gly16) 38%; homozygous Gly 16 (Gly 16/Gly 16), 45%; homozygous Gln27 (Gln27/Gln27), 26%; heterozygous (Gln27/Glu27), 49%; and homozygous Glu27 (Glu27/Glu27), 22%.^45–48 The Gly6 allele has been associated with enhanced agonist‐promoted β₂‐receptor down‐regulation, whereas the Glu27 allele showed minimal down‐regulation compared with the Arg16 and Gln27 alleles.^49,50 Although asthma is primarily an inflammatory disease of the airways, mutations in the β₂‐adrenergic receptor may be risk factors in certain asthma phenotypes.⁴⁵ They may also be factors in determining responsiveness to β₂‐agonists.⁴⁶ In a study of 269 children with asthma, a glycine/arginine polymorphism at amino acid 16 was noted to be associated with a difference in responsiveness to albuterol.⁵¹ Individuals homozygous for the arginine variant were over five times more likely to respond to albuterol than individuals homozygous for the glycine variant. In another recent study determined to assess whether genetic polymorphisms of the β₂‐adrenergic receptor gene affect the relationship between albuterol, plasma concentrations and the forced expiratory volume in 1 s (FEV₁) in patients with moderate asthma, it was reported that the albuterol‐evoked FEV₁ was higher and the response was more rapid in Arg16 homozygotes compared with the cohort of carriers of the Gly16 variant (maximal percentage increase in FEV₁ was 18% vs. 4.9%, p<0.03).⁴¹ The results of this study confirm that β₂‐adrenergic receptor gene polymorphism is a major determinant of bronchodilator response to albuterol and that future pharmacodynamic studies of β₂‐agonists should include determination of β₂‐adrenergic receptor genotype.

The variation in cytochrome drug‐metabolizing genes that correlates with patients' adverse response or non‐response in clinical trials need to be considered. This information could be used to stratify clinical trials, leading to higher efficacy and limiting adverse reactions. Ultimately, detailed information about each patient's genetic variants relevant to drug treatments might eliminate the use of ineffective or even dangerous treatments. Prognosis of patients will be more informed, because more precise information on the aetiology of the illness, its pathophysiology and the effectiveness of therapeutic interventions will be available. Thus, the incorporation of pharmacogenetic information into trials as early as possible is recommended and appears very useful for effective drug development.

Table 2

Frequency of polymorphism in some important drug metabolism genes in different ethnic groups

Gene	Mutation/allele	Phenotypes	Frequency in	Frequency in	Frequency in other	References
			Caucasians	Orientals	populations
N‐acetyl transferase (NAT‐2)	T341C	Slow acetylator	>28%	7%		14
	C282T/G857A	Rapid acetylator	>5%	10–18%
Cytochrome P450 2D6 (CYP2D6)	C188T (P34S)	Poor‐metabolizer (mild)	3%	50%		245
	4	Poor‐metabolizer	20%	<1%	0–19% (Black)
Cytochrome P450 2C19 (CYP2C19)	m1 and m2	Poor‐metabolizer	2.5–6%	15–23%	19.1% CYP2C19*2 &	7, 246
					0% CYP2C19*3
					(Canadian Native Indians)

Gene therapy

Gene therapy in the broadest sense is the introduction of foreign genetic material into a cell with therapeutic intent. For deficiency states such as haemophilia, the aim is to add a normal gene to that cell to complement an abnormal counterpart. It has been estimated that 1 : 100 children have a serious genetic defect, and the possibility now exists of introducing engineered genes, with the protein product supplementing the defective gene. Ideally, the treatment should include replacing the defective gene with a function gene for a complete elimination of the disease‐provoking gene.

The primary problem in gene therapy is the method of gene delivery. Successful gene therapy depends on the availability of reliable methods for delivering a gene into the nuclei of selected target cells and subsequently ensuring the regulation of gene expression. Genes can be delivered (transfected) into cells by various systems.⁵² These fall into two main categories: gene delivery using recombinant viral vectors, and physical gene delivery. Somatic transfer of single genes is being attempted at the present time for terminal diseases such as cystic fibrosis and Duchenne muscular dystrophy, for which there is no effective treatment.^53,54 Understanding polygenic diseases represents a more difficult challenge, and another problem is presented by diseases such as diabetes mellitus in which feedback control of gene expression is important. The problem with some of the delivery systems is the size of the DNA that can be introduced. In the course of a typical infection, viruses insert their genetic material into cells of the victim, where this added genetic code directs the synthesis of various molecules needed to make new viral particles. Although natural viruses can be destructive, it is possible to tame and convert some of them so that they can carry a therapeutic gene and quietly deposit it inside a cell without causing damage. All ‘gene therapy’ that is reported in the literature is gene addition rather than replacement. Gene replacement is straightforward in yeast and is feasible in mammalian cell lines, but remains a long way from effective clinical applications.

Two general strategies have been developed for gene therapy: the in vivo approach and the ex vivo approach. These two approaches each have potential advantages and disadvantages which render them appropriate under different conditions. The in vivo approach is conceptually and technically more direct, involving the introduction of a gene directly into the tissues of an affected individual. In principle, it does not depend on the success of cell culture or subsequent survival of transplanted cells. The ex vivo approach, on the other hand, is technically more demanding. First, a suitable cell type is harvested from a donor and grown in tissue culture. Since mature neurons and glia are notoriously difficult to grow and genetically manipulate in culture, alternative cell types such as fibroblasts and myoblasts have been used.⁵⁵ Next the gene is introduced into the cells in vivo and cells expressing the transgene are amplified. The genetically altered cells are then harvested and reimplanted into an affected host. This approach is labor‐intensive and time‐consuming, and it requires the growth of suitable cells in vivo and their subsequent survival after implantation. However, one advantage of the ex vivo approach is that it does not require a highly efficient method for gene transfer, because genetically altered cells may be amplified in vivo prior to implantation.

Gene therapy is today a robust scientific discipline with several new reagents which are being released for specific clinical applications. Transgenic technology/transgenic animal models are continuously helping to set the stage for somatic gene therapy in humans. Encouraging results have been reported from long‐term animal studies of gene transfer and expression by direct injection of vector into brain, muscle and liver.^56,57 These data have led to an increased interest in adeno‐associated virus and expanded its use in human gene therapy trials. To date, clinical gene therapy has been attempted in only two central nervous system disorders, namely brain tumours and Canavan disease (an autosomal recessive leukodystrophy associated with spongiform degeneration of the brain and is characterized by mutations in the aspartoacylase gene, resulting in loss of enzyme activity).^58,59 Most early phase I clinical studies that have been performed were trials not designed to demonstrate efficacy at all, but instead to assess the safety of transferring cloned genes into humans. The results suggest that these approaches were safe; remarkably little morbidity and no deaths have been noted. Therefore with the advance of the engineering of new vectors including the adenoviruses, adeno‐associated viruses and lentiviruses which promise to greatly enhance the efficiency of in vivo gene delivery and to simultaneously reduce the immunogenicity of both vectors and transgenes, prospects for the clinical application of gene therapy appear good. Obviously, the cloning and sequencing of large numbers of new human genes and a better understanding of the genetic bases of human diseases have greatly increased the scope of diseases that may be amenable to treatment by gene therapy.

First genomic products in clinical trials

The number of clinical trials involving human gene therapy has dramatically increased since the initiation of the first approved trial in the US to treat adenosine deaminase deficiency (ADA, a hereditary deficiency of an enzyme essential to the immune system) in 1990.⁶⁰ Since then, >2100 patients have been enrolled in trials worldwide, with >1700 in the US.⁶¹ The majority of active trials involve gene therapy for malignancy (68%), AIDS (18%) and cystic fibrosis (8%).⁶¹ The key vectors used remain retroviruses (56%), but this high percentage is decreasing.

On 18 December 1997, Human Genome Sciences (HGS) announced the filing of its first investigational new drug (IND) application and was preparing to begin clinical testing of a chemokine called myeloid progenitor inhibitor factor‐1 (MPIF‐1) for the treatment of cancer patients to allow more potent doses of chemotherapy. According to HGS, MPIF‐1 is the first genomics‐derived therapeutic product to enter clinical trials. Among more than 50 candidate proteins tested in an extensive range of high‐throughput cellular assays, MPIF‐1 showed the greatest selectivity, the least inflammatory properties, and the best protective activity against a wide range of chemotherapeutics in over 100 primary human cell lines. MPIF‐1 moved into animal testing with excellent results. In mouse models, it reduced the severity of neutropenia, prolonged stem‐cell survival, and rapidly reverted white blood cell counts to normal following successive rounds of chemotherapy. For the development of MPIF‐1, HGS will be the sole sponsor of the phase I/IIa trial, and on completion, first Schering‐Plough and then SmithKline will have an option to co‐develop the protein in later trials.

Clinical trials are currently addressing a very broad range of potential delivery systems and disease targets. Of the 313 clinical studies listed in the public database maintained by the US National Institute of Health, 70% are involved in the treatment of cancer. This preponderance of cancer‐related trials may be surprising if one considers gene therapy as a treatment for genetic diseases, but in the broader context, gene therapy could be considered as another form of drug delivery, and this accounts for the wide variety of applications of this approach. Therefore, clinical gene therapy applications include treatments aimed at a diverse list of disorders such as arthritis, HIV infection, several types of cancers and extremely rare genetic diseases. Often, the number of patients enrolled in these trials is small (fewer than 20 patients) and this is mainly because of the necessity for ex vivo manipulation of the individual's patient's cells.

One of the more encouraging results in recent reports comes from the use of injections of DNA encoding vascular endothelial growth factors to promote angiogenesis in tissues affected by vascular insufficiency.⁶²

One of the most exciting applications of the use of viral vectors is adeno‐associated vector injection into muscle. A clinical trial was recently initiated to inject factor IX expressing adeno‐associated vectors into the muscles of patients with haemophilia B, and similar approaches have been suggested for retinitis pigmentosa, familial hypercholesterolemia, and muscular dystrophy.

The impact of genetic advances on clinical medicine

These genetic advances have greatly enhanced understanding of disease mechanisms and have begun to explain why the clinical course of common disorders such as diabetes is so variable. In future, presymptomatic population‐based genetic testing for common late‐onset disorders such as Alzheimer's disease may become widespread and bring important health benefits.⁶³ Genotyping may become part of routine investigations to help clinicians tailor drug treatment effectively. Soon medical prescriptions may be personalized to our genotype, along with specific neutraceutical foods. Some vaccines will be delivered through foods such as raw potatoes or bananas.⁶⁴

However, there are issues which need to be better understood. Even in the simple Mendelian disorders, the relation between the DNA sequence of a gene and the corresponding phenotype is far from clear. In late‐onset conditions, such as coronary heart disease and diabetes, where genetic, social, biological and environmental factors interact over time, predicting the clinical importance in a given patient of several different mutations of low penetrance genes is very difficult.⁶⁵ Whether testing will inevitably become widespread as more tests become available is uncertain. Much depends on the severity of the disease and the scope for effective treatment or prevention. Readiness to undergo testing also depends on how testing is offered and on personal, social, and psychological factors. Rigorous assessment of the benefits and costs, both economic and psychosocial is essential, not least because information from genetic screening tests carry implications for families as well as individuals.

Neuropsychiatry

Psychiatrists must usually rely on complex clinical symptoms and diagnostic schemes that although highly reliable, have no obvious biological criteria. Thus, the question of whether our modern definitions of clinical syndromes (considered as phenotypes) accurately reflect underlying genetic substrates (genotypes) remains. Genetic analysis of some psychiatric disorders might be improved by the identification of basic phenotypes for which a more homogeneous aetiology might be expected. The identification of these phenotypes could be achieved by two complementary strategies: the description of the affected subjects and the identification of vulnerability traits in non‐affected relatives of affected individuals. By studying the core symptoms of an illness, the clinical phenotype will be more aetiologically homogeneous. The extent to which genetic mapping is simplified by restriction of the phenotype redefinition can be assessed by measuring the recurrence risk for a relative of an affected person, divided by the risk in the general population.⁵

Factors such as genetic polymorphisms, age at onset, disease severity and family history can be helpful in the identification of homogeneous subtypes. Early onset is associated with increased familial risk in schizophrenia, bipolar affective disorder, major depressive disorder and obsessive‐compulsive disorder.^66–68 The study of associated symptoms and co‐morbid conditions has also proved helpful in the identification of subgroups. Increased familial risk can also provide a key in the identification of subgroups that have a genetic basis. In schizophrenia, a study of genetic polymorphism for drug metabolism (CYP2D6) and tardive dyskinesia suggests that heterozygous carriers of 2D6 mutated alleles may show an increased susceptibility to developing dyskinesia.⁶⁹ In Alzheimer's disease, point mutations in the gene encoding the amyloid precursor protein (chromosome 21), the gene encoding presenilin 1 (chromosome 14) and the gene encoding presenilin 2 (chromosome 1) were identified only after early‐onset familial cases that showed an autosomal dominant pattern of inheritance were recognized. Furthermore, subdivision according to age at onset and mode of inheritance has been particularly useful in the clarification of genetic heterogeneity in dementias of the Alzheimer type.

Migraine

Molecular genetics offers a novel approach to the understanding and management of migraine, since the disorder is known to have a strong genetic component. In a recent study, a Nocardia corallina‐I (NcoI) polymorphism in the gene encoding dopamine D₂ receptor was evaluated in a group of 250 unrelated individuals.⁷⁰ The major findings of the study was that susceptibility to migraine with aura is modified by dopamine D₂NcoI alleles. However, it is also clear that since not all individuals with the dopamine D₂NcoI A1/A1 genotype suffer from migraine with aura, multiple additional genes are involved in the pathogenesis of migraine.

A gene for familial hemiplegic migraine (FHM, a rare subtype of migraine with aura inherited as an autosomal dominant trait), has been assigned to chromosome 19.⁷¹ This gene codes for a brain‐specific calcium channel, and is responsible for FHM in 55% of the FHM families. Other FHM families have been linked to two different locations on chromosome 1. The identification of the genes for FFM may be a key to the identification of the genes for migraine with and without aura.

Neurological channelopathies

Disorders of ion channels (channelopathies) are increasingly being identified, making this a significant expanding area of neurology. Ion channel function may be controlled by changes in voltage (voltage‐gated), chemical interaction (ligand‐gated), or by mechanical perturbation and it has become obvious that genetic defects of both ligand‐ and voltage‐gated ion channels can cause diverse neurological disease.^72–74 A channelopathy may cause an abnormal gain of function (e.g. myokymia, myotonia, and epilepsy) or an abnormal loss of function (e.g. weakness or numbness) depending on whether change of channel function leads to excessive membrane excitability or inexcitability. The impact of these recent discoveries is that it is now understood that genetic channelopathies could affect nerves as well as muscles, and that they may also represent rare forms of more common disorders such as migraine and epilepsy.⁷⁵

The first diseases recognized as channelopathies were the voltage‐gated channelopathies causing inherited muscle diseases (the non‐dystrophic myotonias and familial periodic paralyses). Paramyotonia congenita is due to mutations in the gene coding for the α1 subunit of the sodium channel (SCN4A), while Thomsen's disease (autosomal dominant myotonia congenita) and Becker's disease (autosomal recessive myotonia congenita) are allelic disorders associated with mutations in a gene coding for skeletal muscle chloride channel (CLCN1). Familial hyperkalaemic periodic paralysis is due to mutations in the same sodium channel gene (SCN4A) as that affected in paramyotonia congenita, while familial hypokalaemia periodic paralysis results from mutations in the gene coding for the α1 subunit of a skeletal muscle calcium channel (CACNL1A3). DNA‐based diagnosis of periodic paralyses is now decreasing the need for time‐consuming and hazardous provocative testing.⁷⁶ Malignant hyperthermia is now known to be a disorder of regulation of skeletal muscle calcium and that mutations in the ryanodine receptor gene (RYR1) may cause malignant hyperthermia in some families.^77,78 Ryanodine receptor gene mutation analysis can now be used to identify those at risk of malignant hyperthermia in families with a known mutation in this gene.

One of the most intriguing recent discoveries in genetic channelopathies is that mutations in the same gene (CACNL1A4) can cause three different autosomal dominant disorders such as familial hemiplegic migraine, episodic ataxia type 2 and spinocerebellar degeneration type 6.⁷⁹ It is unclear how different mutations of the same gene can give rise to such different phenotypes. A CACLN1A4 mutation in a case of ‘non‐hemiplegic’ migraine suggests also that such mutations may associate with commoner forms of migraine.⁸⁰ The abnormalities on EEG which are noted in patients with familial hemiplegic migraine and episodic ataxia type 2 together with the recent observations of mouse models have led to the question of whether the gene CACLN1A4 is relevant to epilepsy in humans.⁸¹ In the case of myotonia congenita and familial hyperrekplexia, point mutations in the same gene can result in either autosomal recessive or dominant inheritance.

Ligand‐gated channelopathies that have recently been described include familial startle disease, which is due to mutations of the α1 subunit of the glycine receptor, and dominant nocturnal frontal lobe epilepsy, which is due to mutations of the α4 subunit of the nicotinic acetylcholine receptor.^82,83 A gene for familial paroxysmal choreoathetosis has been mapped to a region of chromosome 1p where a cluster of potassium channel genes is located.⁸⁴

The benefits of recent genetic findings are an improved classification of these neurological disorders and the availability of DNA‐based diagnosis. Such findings will certainly bring novel treatment interventions issued from pharmacogenomics research with the likely possibility of linking treatments to specific genotypes and modes of ion channel impairment.

Huntington's disease

Huntington's disease (HD) is an autosomal dominant genetic disorder caused by a mutated gene found on chromosome 4.^85,86 Efforts continue to determine how this genetic mutation leads to the symptoms of HD and recent investigations have provided further understanding of its genetics. It has been shown that a triplet mutation in a gene (IT15) on chromosome 4 p16.3 results in a protein with an expanded polyglutamine tract caused HD.⁸⁶ The HD gene transcript has been denoted IT15 and the protein it encodes has been termed ‘huntingtin’.⁸⁷ The CAG repeat number in the normal IT15 gene varies, but typically ranges between 10 to 36. In contrast, in patients with HD, the number of CAG repeats range from 37 to 121 repeats. The number of diseases identified as caused by (CAG)₁ continues to grow, and a common mechanisms could underlie these diseases. To date, eight such inherited neurological disorders have been identified to be caused by CAG‐repeat expansion in their respective genes (Huntington's disease, dentatorubral pallidoluysian atrophy, spino‐bulbar muscular atrophy and spinocerebellar ataxia types 1, 2, 3, 6 and 7).^88–90 Most of the diseases caused by expanded CAG repeats share common features which include neurodegeneration, a dominant pattern of inheritance and genetic anticipation (earlier onset and increased severity in successive generations).^91–93 Of these diseases, HD is the most prevalent, occurring in 4–10 per 100 000 individuals of Caucasian origin, so that a knowledge of the neuropathological events that lead to HD will probably provide insights into the disease mechanisms of the other triplet‐repeat disorders.

Symptoms of HD typically appear in adult life between the ages of 30 and 50, and the disease is most often characterized by chronic progressive chorea (quick, jerking, uncontrollable movements of the limbs, trunk and face) and dementia without remissions. An individual with HD often has problems in control of most bodily movements, intellectual functioning (poor short‐term memory and judgment) and emotional control (depression, irritability and apathy). The intensity and number of the above symptoms, however, varies with each HD patient.

The neuropathology of HD is most pronounced in the caudate and putamen forebrain region. Its pathogenesis is unknown, although roles for glutamate‐receptor activation and deteriorations in mitochondrial energy metabolism have been proposed as co‐factors.^94,95

There is no cure for HD; medication can alleviate some of the involuntary movements and emotional disorders for some HD patients. However, novel therapies in the near future may include interference with RNA splicing, blockade of the deleterious protein's effect or physiological interventions by pharmacological tools or neural transplants. At the RNA level of intervention, it may be possible to integrate viral or non‐viral vectors carrying catalytic antisense RNAs or ribozymes that bind to, and irreversibly cleave, abnormal mRNA molecules.^96,97 Antisense RNA has been used to hamper transcription, processing or translation of mRNAs in a variety of cell types. However, the results from animal studies suggest that although the antisense oligodeoxynucleotide targeted against the initiation codon of IT15 gene had penetrated several striatal cells and that they were not toxic to the brain, no significant decrease in levels of huntingtin was detected by immunostaining or Western blot analysis.⁹⁷ Therefore, improved methods for molecular modifications of IT15 may be needed for therapeutic initiatives. Considering the feasibility of gene therapy for HD, a recent study suggests that encapsulated human ciliary neurotrophic factor producing fibroblasts may prevent behavioral deficits and striatal degeneration in models of HD.⁹⁸

Alzheimer's disease

Pathologically, Alzheimer's disease (AD) is associated with generalized degeneration of the cerebral cortical and hippocampal neurons. Cholinergic neurons in the basal forebrain which project to cortex and hippocampus appear to be particularly vulnerable, and to an extent, so are serotonergic and noradrenergic afferents to cortical regions. The extracellular deposition of peptide fragments (amyloid‐beta) from the larger membrane precursor protein (APP) is typical in affected brain tissue. Intracellular accumulations of tau‐proteins (tangles) are present in many cortical and cortico‐limbic regions.⁹⁹

Genetic studies have led to the identification of three genes in which mutations can cause AD: the β‐amyloid precursor protein gene located on chromosome 21, presenilin 1 (PS1) located on chromosome 14, and presenilin 2 (PS2) located on chromosome 1.^100–102 In addition, the E4 allele of the apolipoprotein E (ApoE) gene is a risk factor for AD. While mutations associated with APP are extremely rare, the 50 or so mutations associated with PS1 may explain up to half of all cases of early‐onset AD. A study which investigated the association of two candidate genes (PS1 and α₁‐antichymotrypsin (ACT)) with the risk of sporadic Alzheimer's disease on chromosome 14 reported that the frequency of the ACT*A allele was significantly higher in AD patients than in controls and the stratification of the ACT data by PS1 genotypes showed that the risk associated with the ACT*A allele was confined to PS1*1 carriers only.¹⁰³ The two‐site haplotype data for PS1 and ACT indicated that the A1 haplotype, carrying the ACT*A and PS1 alleles, was more frequent in Alzheimer's disease patients, and these results may also suggest that there is a possible synergistic effect of these two loci on the risk of AD. In contrast to early‐onset AD, there is to date only one genetic factor indisputably linked with late‐onset forms of this disorder; the E4 allele of apolipoprotein E.¹⁰⁴ Differences in ApoE genotyping appear to explain differences in patients' responses to drug therapy. With tacrine, a better response was seen in patients with the ApoE E2 or ApoE E3 allele than in those carrying the ApoE E4. The ApoE E4 allele has an inverse relationship with residual brain choline acetyltransferase (the acetyl‐choline synthesizing enzyme) activity, and it appears that patients with this genotype may not have sufficient acetylcholine to benefit from a drug which acts as an inhibitor of acetylcholinesterase. However, patients with the ApoE E4 genotype appear to have a better response than other AD patients to treatment with another drug, Servier's S12024 (morpholinyl‐2 methoxyl‐8 tetrahydro‐1,2,3,4 quinoline) which is currently in phase II clinical trials. In fact, this drug had no detectable effect in patients with the other ApoE genotypes. S12024 does not appear to affect the cholinergic system, but rather to facilitate brain noradrenergic and vasopressinergic activity, and increases vasopressin synthesis and release in a dose‐dependent manner. There may be a balance between cholinomimetic and vasopressinergic pathways, according to ApoE E4 allele presence or absence.¹⁰⁵ With relevance to design of clinical trials, the important observation may be that alleles that appear to be conclusively associated with a therapeutically relevant phenotype can be used to select a subgroup of patients for clinical trials. A polymorphic site need not be part of the target for the drug; it only needs to be associated with a response to the treatment. In responsive patients, the selective treatment could be more effective, and associated with fewer or less severe side effects. Furthermore, pre‐emptive genotyping aimed at drug‐associated genes could mean that fewer drug candidates would fail to reach the market place because of poor toxicity/efficacy profiles in the general population. For example, genotyping of early‐onset AD is likely to include the two PS1 and PS2 genes involved in this disease. Predictive and diagnostic tests for PS1 mutations and diagnostic tests for ApoE alleles are already commercially available and other tests are being developed. Thus, genetic testing for AD exists for clinical use, and is likely to be used more often to stratify patients in Alzheimer's disease research, both in trials of preventive products and in tests of new pharmacological treatments. Therefore, predictive and diagnostic genetic testing for these highly penetrant mutations such as PS1 or PS2 may be appropriate for adults from families with a clear autosomal dominant pattern or inheritance, particularly those with a family history of early onset of symptoms. Testing is an option that could be discussed and that could reasonably be accepted or declined by the patients. However, the application of the ApoE test raises concerns, because although the E4 allele is associated with an increased risk of AD, its predictive value for individuals is quite limited.¹⁰⁶ The small increase in diagnostic confidence provided by ApoE genotyping does not justify the burdens of testing; such testing may have value in AD research, but its widespread clinical use is premature until practical benefits outweigh its costs.

A successful treatment approach might be the delivery of neurotrophic factors which prevent neuronal degeneration. Gene therapy may provide an effective alternative to traditional methods for delivery of trophic factors. For example, genes encoding trophic factors can be introduced into cells in tissue culture. Subsequent intracerebral implantation of these cells effectively produces a biological mini‐pump for constant local delivery of a trophic factor. The effectiveness of this approach has already been demonstrated in several animal models of neurodegeneration. In the first model, surgical lesions of the fimbria‐fornix in rodents interrupts cholinergic neurons in the basal forebrain from their normal supply of nerve growth factor (NGF) in the cortex.^107,108 In the absence of a constant supply of NGF, the cholinergic neurons degenerate. Early studies had documented that the dying neurons could be rescued by the intracerebral delivery of purified NGF via artificial mini‐pumps.

The appropriateness of NGF administration as treatment for AD and other neurodegenerative disorders remains a matter of debate. However, the animal experiments demonstrate the principle that gene therapy may be applied to prevent neurodegeneration by directing constant local production of a growth factor either by grafts of growth factor‐producing cells, or direct introduction of growth factor genes.

Amyotrophic lateral sclerosis

Amyotrophic lateral sclerosis (ALS) is an adult‐onset, chronic neuromuscular disorder, characterized by selective degeneration of cortical and spinal/bulbar motor neurons. The discovery that 15–25% of familial ALS patients have mutated Cu²⁺/Zn²⁺ superoxide dismutase (SOD1) links ALS to toxicity by reactive oxygen species.¹⁰⁹ SOD1 is thought to protect against cellular damage induced by oxygen radicals but the mechanism(s) through which mutations in SOD1 lead to late‐onset motor neuron degeneration remains unidentified.¹¹⁰ Two mutations which have been successfully used to generate transgenic mice that develop an ALS‐like syndrome are glycine 85 to arginine (G85R) and glycine 93 to alanine (G93A) with the mutant SOD1 allele overexpressed in a normal mouse genetic background.¹¹¹ Recent data assessing the axonal transport in animal models suggest that an important aspect of toxicity may arise from damage to transport, and that reduced transport of selective cargoes of slow transport, especially tubulin, arises months before neurodenegeration.¹¹² Therefore, damage to the machinery of slow transport may be an early feature of toxicity mediated by mutant SOD1.

Ciliary neurotrophic factor (CNTF) and glial‐cell‐line‐derived neurotrophic factor (GDNF) have been shown to rescue motor neurons from axotomy‐induced cell death and to slow the degeneration of motor neurons in a mouse mutant with progressive motor neuropathy (animal models of ALS), so neurotrophic factors may be of therapeutic importance for the treatment of neurodegenerative disease such as ALS.^113,114 A phase I clinical trial designed as an open‐label safety study using intrathecal delivery of human CNTF has been performed in six patients suffering from ALS.¹¹⁵ Baby hamster kidney (BHK) cells transfected with the CNTF gene were loaded in 5‐cm long polyethersulfone fibres and surgically placed within the lumbar intrathecal space. According to the protocol, the implant was retrieved after 3–4 months. The results of this small study showed that significant doses of CNTF can be delivered directly into the central nervous system, and that no limiting adverse effects in any of the patients exposed to low intrathecal doses of CNTF were noted. Further long‐term studies may therefore be conducted to assess efficacy and safety of this gene therapy approach for the treatment of ALS.

Parkinson's disease

In Parkinson's disease (PD), patients display cogwheel rigidity, resting tremor, and an impairment in the initiation and speed of movements. Although familial PD has been reported, most cases are thought to be acquired. The cause of the disease is unknown, but environmental toxins, metabolic derangements, infectious processes, and normal aging have all been hypothesized to play a role.¹¹⁶ Pathologically, this disorder arises because a part of the brain known as the substantia nigra degenerates over time. This region helps to regulate motor control, and destruction makes it hard for a person to initiate movements or execute complex coordinated motion.

Epidemiological studies have shown that positive family history and exposure to environmental toxins are considered risk factors for PD. Recent evidence indicated that there is a significant association between the slow acetylator genotype of N‐acetyltransferase 2 and familial PD.¹¹⁷ The proportion of slow acetylators was significantly higher among patients with familial PD (69%) than among normal controls (37%, p<0.001) and this difference remained significant after correction for multiple comparison (p=0.002). It is hypothesized that this difference may be due to the impaired ability to metabolize neurotoxic substances in patients with slow acetylation, resulting in the increased metabolic predisposition to familial PD.

With the recent identification of mutations in the α‐synuclein gene or linkage to a region on chromosome 2 in families with PD, it appears that genetic factors are important when the disease begins at or before age 50 years.^116,118,119 Mutations in the alpha‐synuclein gene may cause early‐onset PD by accelerating the formation of Lewy bodies. This findings could lead to a novel treatment for PD. It was observed that a mutant form of alpha‐synuclein, known as A53T, formed fibrils characteristics of Lewy bodies more quickly when concentrated and encouraged to aggregate than either wild‐type alpha‐synuclein or another mutant, A30P. The making of a new drug which targets inhibition of fibril formation should be possible.

In a strategy for therapeutic replacement of degenerated or dead dopamine cells for PD, experiments using heterologous donor tissue have mainly focused on tissue derived from the embryonic mesencephalon harvested at a time point when the cells start differentiating into dopaminergic neurons. Embryonic cells survive in the striatum after grafting and display a TH‐immunoreactive phenotype.^120,121 One of the main advantages of the use of fetal tissue over genetically engineered cells may lie in its potential to form functional synapses with the host brain and release dopamine in a regulated fashion.^122,123 Unfortunately, large amounts of tissue have been necessary to achieve behavioural recovery, and even with optimized techniques, only a minimal fraction of the transplanted cells will survive as TH‐IR neurons. Several clinical trials using fetal donor tissue have had various degrees of success, possibly due to significant differences in patient history and technical details.^124,125 So far, fetal tissue shows the greatest potential, but it may be difficult to obtain the significant amount of tissue to obtain the necessary amount of tissue for a great number of patients and ethical considerations have to be balanced with safety and efficacy issues. Progenitor cell or xenogenetically‐derived dopamine cells may provide useful sources for clinical cell implantation in PD.^126,127

To genetically engineer cells, because of the limitations of established cell lines, retroviral‐mediated gene transfer has been used to transduce primary fibroblasts and astrocytes.^108,128 The application of gene transfer techniques for the treatment of PD has been explored in animal models as a potential therapeutic intervention. Since the loss of striatal dopamine is thought to underlie the motor disturbances in PD, gene therapy strategies have focused on replacement of dopamine levels by genetically engineering cells to produce tyrosine hydroxylase (TH), the enzyme necessary for the conversion of tyrosine to L‐DOPA. Local production of L‐DOPA in the affected area may diminish the unwanted side‐effects observed with systemic administration. It is clear from behavioural analysis that grafted, genetically‐engineered, non‐neuronal cells can have significant effects as a biological L‐dopa pump system. However, in contrast to fetal dopamine neurons, these cells are unable to make connections with host cells. The importance of functional cellular integration in the restoration of finer motor control tasks remains to be determined.

A second strategy for treatment of PD is to prevent dopaminergic neurons from dying. Brain‐derived neurotrophic factor (BDNF), has been used to prevent cell death in either cell culture models of dopaminergic toxicity or animal models of PD.^129–131 The neurotrophin BDNF has been shown to have effect on damaged dopaminergic neurons in vivo. Although infusions of recombinant BDNF into the parenchyma failed to protect dopaminergic neurons from axotomy, delivery of BDNF by genetic engineering of nonneuronal cells has shown promising effects.¹³² These data suggest that this strategy may be a viable approach for enhancing the survival and function of dopamine neurons grafted into parkinsonian patients.

Cardiology

In pre‐clinical studies of potential gene therapies for cardiovascular disease, methods are being examined for genetically modifying cellular constituents of the vessel wall and the heart. Ex vivo approaches have included attempts to resurface specific arterial vessel segments with genetically modified endothelial cells or smooth muscle cells. The resurfacing of prosthetic vascular grafts with genetically modified cells has also been examined.¹³³

Most of the work indicating that gene therapy can show real clinical benefits in patients with heart disease has been done with the gene for vascular endothelial growth factor (VEGF). This approach is based on the idea that injection of the VEGF gene into the area where atherosclerotic plaques have caused the most blockade to the arteries will allow new blood vessels to grow which can by‐pass the blockage. This approach is being tested in both peripheral artery disease (who have blockages in the blood vessels in their legs) and those with coronary heart disease (in whom the arteries supplying the heart have become blocked). Many different groups have reported initial clinical trials in which the gene for VEGF induced the formation of new blood vessels. While most of the studies to date have been conducted in patients with peripheral vascular disease, several groups have now reported preliminary evidence of success in patients with coronary heart disease. If such research continues to progress well, the use of gene therapy could become an alternative to angioplasty or bypass surgery in coronary heart disease patients, and may even do away with the need for some drug therapies.

Cardiovascular risk factors

Genome‐wide linkage scans have revealed multiple quantitative trait loci contributing to susceptibility for several major cardiovascular risk factors. The challenge that remains, however, is how to improve these studies in order to narrow the genetic interval and find susceptibility genes.

The clinical course of the congenital long QT syndrome (LQTS), a hereditary arrhythmic disorder, can be predicted through genotypic analysis and that LQTS should no longer be excluded on purely clinical grounds^134,135 (see Table 3). The finding of a link between variations in genotype and differential clinical disease expression may allow clinicians to tailor the treatment of patients who have the syndrome according to their genotype; LQTS is caused by defective ventricular repolarization leading to prolongation of QT interval, and predisposes to ventricular tachyarrhythmias, largely associated with stressful settings. The prolonged pauses can cause cardiac syncope, and additionally serve as arrhythmogenic substrates, allowing ectopic beats to occur. Patients with LQTS may be asymptomatic or experience recurrent fainting spells, palpitations, torsade de pointes, and ventricular arrhythmias leading to sudden death. The observation of an increased lethality of cardiac events in patients with SCN5A mutation may suggest the need for more aggressive treatment of these patients. It also appears that gene‐specific therapy with class Ib sodium‐channel blockers may be more adequate treatment for these patients.^136–138 This approach of linking a defective gene to clinical outcome and treatment is an important major direction in clinical medicine.

Many areas of research are opened up by the exciting new molecular genetic findings in LQTS. Examples include the identification of drug therapies that might correct the defects in these mutations and studies to determine the relationship between clinical phenotypes and underlying genetic defects.¹³⁷ Preliminary data suggest a role for sodium‐channel blockers such as mexiletine in LQT3, linked to late‐opening sodium channels.¹³⁷ In patients with this defect, blockers of the late‐opening sodium channels might entirely reverse all the manifestations of the syndrome. Use of sodium‐channel blockers in patients with other defects might theoretically also shorten the QT interval by reducing inward current during the plateau.¹³⁹ However, they would not be expected to reverse the ion channel defect underlying the syndromes. In patients with LQTS due to abnormal potassium‐channel function, therapies directed at altering the function of the abnormal channels to create outward current (e.g. potassium‐channel openers) might achieve the latter effect. While these are interesting suggestions for future therapy after evaluation in vivo and in controlled trials, β‐blockers remain the current mainstay of therapy.

Familial hypertropic cardiomyopathy affects about 1 in 500 of the general population.¹⁴⁰ Most cases are inherited in an autosomal dominant manner with variable clinical expression.¹⁴¹ To date, disease‐causing mutations have been described in genes for seven cardiac sarcomeric proteins; β‐myosin heavy chain (chromosome 14q11), cardiac troponin T (chromosome 1q3), cardiac troponin 1 (chromosome 19p13.2‐q13.2), α‐tropomyosin (chromosome 15q2), cardiac myosin‐binding protein C (MyBPC, chromosome 11p11.2), and the essential and regulatory myosin light chains (chromosomes 3 and 12, respectively).^142–143 Mutations in the genes encoding β‐myosin heavy chain, troponin T and cardiac MyBPC account for over 50% of all reported cases. However, most published genetic studies are rather small and in highly selected populations; therefore this prevalence of different mutations may be imprecise. The discovery of mutations in genes for sarcomeric proteins in patients with hypertrophic cardiomyopathy has raised the possibility of gene testing in clinical care. Unfortunately, the genetic heterogeneity of this disease has made routine genetic testing impracticable, and testing is performed in research centers and limited to particular circumstances (e.g. patients with incomplete expression and a strong family history of sudden death). If the association between late‐onset hypertrophy and MyBPC gene mutations is confirmed, genetic screening in children of affected individuals may become appropriate.

Table 3

Genotype of long QT syndrome (from data in reference 134)

Mutant	Chromosome locus	Abnormal characteristics	Clinical course
genes
KVLQT1	LQT1 locus on chromosome 11	Encodes an abnormal potassium‐	The frequency of cardiac events
	(11p15.5)	channel protein (α subunit)	was higher among subjects
		that, when expressed with a	with mutations at the LQT1
		protein from the min K gene,	locus (63%) or the LQT2 locus
		reduces the current of the	(46%) than among subjects
		slowly activating, delayed	with mutations at the LQT3
		inwardly rectifying potassium	locus (18%). However, the
		channel.	likelihood of dying during
HERG	LQT2 locus on chromosome 7	Encodes an abnormal potassium‐	a cardiac event was
	(7q35–36)	channel protein that reduces	significantly higher among
		the current of the rapidly	families with mutations at the
		activating, delayed inwardly	LQT3 locus (20%) than among
		rectifying potassium channel.	those with mutations at the
SCN5A	LQT3 locus on chromosome 3	Encodes an abnormal sodium‐	LQT1 locus (4%) or the LQT2
	(3p21–24)	channel protein that does not	locus (4%).
		allow the complete inactivation
		of sodium inflow, resulting in
		continued entry of sodium into
		the myocardial cell during
		repolarization.
KCNE1	LQT5 locus on chromosome 21	Encodes β subunits that assemble
		with KVLQT1
		α subunits to form slowly
		activating delayed inwardly
		rectifying potassium channels.

Essential hypertension

Molecular genetics applied to essential hypertension has allowed the identification of some quantitative trait loci or genes that influence blood pressure.¹⁴⁴ A polymorphism of the genes for the cytoskeletal protein, adducin, which is linked to both rat and human hypertension, sodium sensitivity and to the pressor responsiveness to diuretic therapy has been identified.¹⁴⁵ Hypertension due to single‐gene abnormalities is very rare, and it appears that high blood pressure is, in general, a complex final phenotype that involves many control systems operating at each level of biological organization. The blood pressure difference between a hypertensive individual and a normotensive control has been attributed to the influence/interaction of 2–6 genetic loci.¹⁴⁶ In particular, through the candidate gene approach or random genome scanning, some specific genes have been identified such as angiotensinogen (AGT), angiotensin‐1 converting enzyme (ACE) and adducin. In humans, the problem of the association between a given genetic trait and hypertension can be approached both with case‐control and linkage studies. Only two genes, adducin and AGT were found to be involved in human essential hypertension when both types of studies were considered.^147,148

Considering hypertension due to single gene abnormalities, glucocorticoid‐remediable aldosteronism (GRA) is an autosomal dominant form of moderate‐to‐severe hypertension associated with both an excess of cerebral haemorrhage and Celtic ancestry.¹⁴⁹ The hypertension is caused by excessive secretion of aldosterone, and possibly additional adrenal mineralocortocoid hormones, where aldosterone secretion is regulated by corticotrophin rather than by angiotensin II. Kindreds with GRA demonstrate a novel gene on chromosome 8 that represents duplication arising from unequal crossover between the aldosterone synthase and 11 β‐hydroxylase genes, such that the regulatory sequences of 11 β‐hydroxylase are fused with coding sequences of aldosterone synthase.¹⁵⁰ Aldosterone synthase gene expression and enzymatic activity are therefore brought under the control of ACTH, which results in ectopic production of aldosterone, with hypertension due to increased salt and water retention. This type of hypertension responds to administration of physiological doses of glucocorticoids which suppress ACTH secretion and thereby suppress expression of the mutant gene.¹⁵⁰ It is now possible to define these individuals using a simple genetic test. This is an example of an approach in clinical practice to identify hypertensive individuals who should be treated with glucocorticoids as antihypertensive therapy.

Patients with autosomal‐dominant monogenic hypertension who tended to metabolic alkalosis with hypokalaemia were described by Liddle (Liddle syndrome). The patients had low‐renin and low‐aldosterone values; however, they did not respond to spironolactone, while thiazides and triamterene decreased the blood pressure. The responsible gene in a family with Liddle syndrome was localized to chromosome 16 and that the gene encodes the β‐subunit of the epithelial sodium channel (ENaC).¹⁵¹ The channel is amiloride‐ and triamterene‐sensitive, explaining the efficacy of these drugs in treating the syndrome. The channel remains inappropriately permeable even in high salt intake, thereby leading to salt‐sensitive hypertension. Another mutation in the γ‐subunit of ENaC has also been reported which can result in Liddle syndrome.¹⁵²

Evidence for association and genetic linkage of the ACE gene with hypertension and blood pressure in men, but not in women, was reported when data for over 3000 patients in a study were analysed.¹⁵³ Animal studies suggest that mutations in the α‐adducin gene may account for 50% of the observed differences in blood pressure between hypertensive and normotensive strains.¹⁵⁴ A study in human hypertensive sibling pairs also revealed support for linkage of the chromosomal region containing the α‐adducin gene to high blood pressure.¹⁵⁴ A polymorphism which exchanges tryptophan for glycine at position 460 (G460T) in the gene product associates with hypertension, and evidence suggests that this polymorphism may influence the response of blood pressure to sodium loading or depletion. It also appears that the G460T polymorphism may influence response to long‐term therapy with a thiazide diuretic.¹⁵⁵ Patients who were heterozygous for the polymorphism exhibited a greater blood pressure response to sodium depletion and long‐term therapy with a thiazide than those who were homozygous for the glycine variant. Confirmation of these data would indicate a way of predicting a diuretic‐responsive form of hypertension.

Considering the AGT genotype, an association has been reported between the T235 allele and both plasma AGT concentration and elevated blood pressure variation.^147,156 AGT genotype was also strongly associated with blood pressure reduction after 5 weeks' ACE inhibition, and the best blood pressure response was associated with the T235 allele.¹⁵⁷ These data suggest that a variation in the regulatory sequence of AGT may be more important in evaluating the response to blood‐pressure‐lowering treatment. It also appears that genetic understanding of hypertension will be used to develop predictions of pharmacological response and the development of new novel pharmacological therapeutic strategies. It is important that in the assessment of drug treatment, study designs include tests as to whether new diagnostics can predict treatment responses. This might provide an opportunity to focus on specific hypertensive subtypes and provide the most effective antihypertensive drugs to such defined hypertensive individuals.

A study which analysed data on the efficacy of specific drugs in individual patients concluded that 10–59% of patients failed to respond to diuretics, 12–86% failed to respond to β‐blockers, some patients exhibited heterogeneous responses to ACE inhibitors and calcium antagonists, and a small percentage of patients even showed an increase in blood pressure.¹⁵⁸ The variation in the individual response to antihypertensive drugs may be due to the heterogeneity of the mechanisms underlying hypertension, to interindividual variations of the pharmacokinetics of the drugs, or both. Although the choice of a proper therapeutic strategy through pharmacogenomics approach is complex, and targeting the disease gene itself may not be sufficient for the selection of the optimal therapeutic target, the gene identification involved in essential hypertension should provide clarification on the regulatory pathways triggered by the primary genetic defect and therefore lead to the identification of the optimal point of therapeutic intervention.

Diabetes

Diabetes can arise from a number of mutations in either the nuclear or mitochondrial genomes, and the search for diabetes‐associated genes is an important area of intense scientific investigation. In almost all cases of type 1 diabetes (insulin‐dependent diabetes mellitus, IDDM), the disease has an autoimmune component, whereby the system of insulin production by pancreatic β cells is slowly and irreversibly eroded by the body's own misdirected immune attack. The major histocompatibility complex on chromosome 6 has been shown to contain one or more major genetic determinants of disease susceptibility.¹⁵⁹ At least 40% of the familial aggregation of type 1 diabetes is accounted for by the human leucocyte antigen (HLA) genes: in particular, the HLA class II genes DQ and DR. In Caucasians, high‐risk class II molecules include DQA1*0501‐DQB1*0201 (associated with DR3) and DQA1*0301‐DQB1*0302 (associated with DR4). Dominant protection is apparently conferred by the DQ molecule DQA1*0102‐DQB1*0602, which is carried on DR2 haplotypes.¹⁶⁰

As yet there is no way to halt the destruction of insulin‐producing β‐cells in pancreatic islets, and by the time type 1 diabetes is diagnosed in children and young adults, the damage has been done, and the only treatment is daily insulin injections. Current evidence suggests that the age at diagnosis of type 1 diabetes is genetically determined, and that pairs of siblings tend to develop clinical diabetes at a similar age because of a common effect, such as puberty.¹⁶¹ One gene at fault is on chromosome 6p, within the HLA complex, but other genes can make additional contributions to the susceptibility of the insulin system to such immune attack. The only one noted so far is on chromosome 11p, and codes for insulin itself: it has two alleles, one (dominant) that is protective against type I diabetes and another (recessive) that is susceptible to disease development. So far, up to 20 additional genetic loci have been mapped by investigators looking for other contributory mutations; in every family with type 1 diabetes, a different number or selection of genes is implicated in its development. It is believed that identifying the faulty genes at an early age would provide clinicians with the opportunity to target prophylactic therapies at those likely to be susceptible to the disease. Obviously, a knowledge of the genes' function and how they are disrupted will provide clues to new therapies or even cures. However, evidence is easier to gather in a single‐gene disorder with a Mendelian pattern of inheritance; in a multigenic disorder, the case for linkage of markers is weaker.

Type 2 diabetes (formerly known as non‐insulin‐dependent diabetes mellitus or NIDDM) is a polygenic disease characterized by insulin‐resistance in muscle, fat and liver followed by a failure of pancreatic β‐cells to adequately compensate for this resistance despite increased insulin secretion. It is a genetically and clinically heterogeneous disorder and can be divided into early‐and late‐onset forms. The early‐onset forms includes maturity‐onset diabetes of the young (MODY), a genetically heterogeneous monogenic form of type 1 diabetes characterized by early age of onset (usually <25 years) and autosomal dominant inheritance.¹⁶² Mutation in the gene for insulin, the genes for insulin processing enzymes, and the genes for the insulin receptor have also been detected and associated with the early‐onset forms of the disease.^163,164 Linkage studies have identified genes that are mutated in different MODY pedigrees on chromosomes 20 (MODY1 locus, hepatocyte nuclear factor‐4α [HNF‐4α] gene, 7 (MODY2 locus, glucokinase gene), and 12 (MODY3 locus, HNF‐1α gene) and clinical studies indicate that mutations in these genes are associated with abnormal patterns of glucose‐stimulated insulin secretion^165–168 (Table 4). Regarding the identification of genes responsible for the more common forms of late‐onset type 2 diabetes, recent data suggest that the region of chromosome 12q close to MODY3 harbors a novel susceptibility gene or genes for type 2 diabetes.¹⁶⁹

The maternal inheritance pattern of diabetes has led researchers to look for genetic defects in the mitochondrial genome (mtDNA; mtDNA is maternally inherited, because the mitochondria contributed to the embryo by the sperm is much less than that present in the egg at the time of fertilization; the ratio is around 1 : 1000. Therefore mtDNA mutations are transmitted to the progeny via maternal lineage). To date, more than 42 different mtDNA mutations (point mutations, deletions and duplications) have been found to associate with the type 2 diabetes phenotype, and many are also associated with other disturbances such as central nervous system as well as disturbed muscle and renal function.^170,171 Patients with type 2 diabetes associated with mtDNA mutations generally are not obese, and they exhibit hyperglycaemia that is due to a significantly reduced insulin secretory capacity that progresses with age.^172,173 These patients develop a fatty liver, which is not an uncommon feature of diabetes mellitus. Beta‐cell loss is seen, as well as defects in glucose‐induced signaling of insulin release.¹⁷⁴ Any mutation of the mtDNA affects the synthesis of ATP and therefore the supply of energy. The discovery that mutations in the mtATPase 6 and 8 genes associate with type 2 diabetes has led to better understanding of this disorder, and to date five mutations that associate with type 2 diabetes have been reported in the genes for ATPase 6 and 8.^175,176

Animal models such as the BHE‐Cdb rat mimic the human with type 2 diabetes.^177,178 It develops moderate hyperglycaemia and impaired glucose tolerance as it ages, as well as a number of diabetic complications. With this animal model, hypothesis about the roles of age, diet and drugs in the pathology of type 2 diabetes may be studied. The usefulness of the BHE/Cdb rat as a tool to develop appropriate pharmacotherapies and nutriceuticals for the treatment of human mt diabetes should be recognized.

Table 4

Nuclear DNA mutations associated with type 2 diabetes

Gene	Chromosome	Comment	Reference
	location
Glucokinase (MODY 2)	7	Uncommon; heterogeneous	247
Insulin cleavage enzyme	?	Results in excessive amounts of proinsulin in the blood	248
MODY 3	12q	Encodes transcription factor 1α	249
MODY 1	20	Encodes transcription factor 4α	250
Glycogen synthase	19q 13.3	Two forms (liver and muscle); Highly polymorphic	251
FAD‐glycerol 3 phosphate	3	Associates with type 2 diabetes in the GK rat	252
dehydrogenase
Glucagon receptor	?	Associates with type 2 in French & Sardinian patients but	253
		not Japanese patients
CCKBR	11p 15.4	Associates with type 2 diabetes in French patients	254
IRS 1	?	Plays a role in downstream insulin signaling. When	255
		mutated, the signal pathway is aberrant

CCKBR, cholecystokinin B receptor; IRS 1, insulin receptor substrate 1.

Obesity

Obesity, a common multifactorial disorder, affects approximately one‐third of the United States population and 20% of Europeans. It predisposes to type 2 diabetes, hypertension and coronary heart disease, and is a major cause of morbidity and mortality. The evidence for genetic factors in human obesity has become increasingly strong, with studies of twins and families with and without obesity. All mutant genes responsible for animal obesities have now been identified, and strong homology exists between the rodent genes and human genes. The investigation of the mutant genes as well as the peptides (e.g. neuropeptide Y) with which they interact to determine treatment of obesity is important. Leptin protein and its functions appear to play a significant role in the organization of energy control and obesity. Studies have also shown consistently that about 40–70% of the variation in obesity‐related phenotypes, such as body mass index (BMI), skinfold thickness, fat mass, and leptin levels, is heritable.^179,180

Abdominal obesity is a heterogeneous phenotype with metabolic disturbance associated more closely with altered regional adiposity than obesity per se. Studies using precise measures of regional obesity (dual energy X‐ray absorptiometry, computed tomography and magnetic resonance imaging) indicate that visceral adipose tissue, independent of obesity, is a major determinant of insulin resistance and dyslipidaemia (syndrome X).^181–183 The phenotypic expression of adult obesity can be identified in children from the age of 6–10 years and increases in abdominal and hepatic fat probably begin at the same time.^184–186 By the time of adolescence, visceral obesity and syndrome X may be fully expressed, and by early adulthood apparently healthy men with increased central fat may already have cardiac pathology.^187,188

Visceral obese male individuals heterozygous for the apolipoprotein B‐EcoRI polymorphism are more prone to develop the dense LDL phenotype, and increased risk of coronary disease also occurs with some lipoprotein lipase gene variants.^189,190 Genetic studies also suggest that there may be specific genes responsible for visceral obesity. Genetic variations in the glucocorticoid receptor (bclI restriction fragment length polymorphism) and the fatty acid binding protein 2 (Ala54Thr FABP2) gene were associated with visceral adiposity in lean men.^191,192 Defects in the ob gene or receptor appear to be unlikely candidates, particularly because leptin levels correlate with total body fat rather than visceral adiposity.^136,193

Genes which appear to be implicated in obesity include agouti‐related transcript (ART), neuropeptide Y (NPY) and its receptors (NPY Y5R and Y6R), pro‐opiomelanocortin (POMC), uncoupling protein 2 (UCP2), and the melanocortin‐4 receptor (MC4‐R). The agouti gene modulates melanocortin signaling in the hypothalamus and inhibits the anorexogenic effect of melanocyte‐stimulating factor (MSH) in this brain region.¹⁹⁴ The fat and tubby mutations also appear to alter the neural circuits that regulate weight. The fat gene encodes carboxypeptide (CPE), an enzyme involved in neuropeptide processing, and the tubby gene is expressed in the paraventricular nucleus of the hypothalamus, a brain region known to play a role in the regulation of body weight. NPY is one of the most potent appetite stimulators in animals and it also appears to be one of the mediators of ob gene in the brain.^195,196 Mice deficient in MC4‐R develop late‐onset obesity and alterations in their peripheral metabolism.¹⁹⁴ Brain administration of glucagon‐like peptide 1 (GLP‐1) to rats reduces food intake and GLP‐1 receptor agonists may be useful for reducing food intake.¹⁹⁷ In general, these data suggest that body fat is controlled by a lipostat mechanism in which leptin is the afferent signal; the hypothalamus serves as an integrator and activates an output loop that modulates feeding behavior, energy expenditure, and fat and glucose metabolism.¹⁹⁸

Genetic testing may play a role in assisting the clinician to predict, monitor and prevent this problem, which is reaching epidemic proportions in Western society. There are many new therapeutic targets, and combinations of drugs with different modes of action may be necessary. The next generation of medicines to treat obesity may target the ob/leptin pathway. A drug that activates the ob pathway may have multiple therapeutic benefits such as suppressing appetite, increasing metabolic rate and reducing the amount of body fat. Leptin is an obvious choice, and trials of recombinant human leptin are in phase II. The development of new, highly specific and effective centrally‐acting agents remains an attractive possibility. Therefore, this approach may include agonists of GLP‐1, MC4‐R and specific neuropeptide receptor (e.g. NPY Y1, Y5) antagonists. However, for Y5‐receptor selective antagonists, side‐effects related to epileptogenesis, drug withdrawal symptoms, gut motility, sodium retention and the metabolic system must be investigated very carefully, as recent observations suggest that the Y5 receptor may not be very specific for chronic reduction of food intake.¹⁹⁹

Oncology

Critical and specific genetic events involving oncogenes, tumour suppressor genes, DNA repair enzymes have been characterized in many tumours, and these molecular flags of malignancy can be exploited for screening and treating cancers. A number of genetic aberrations have been identified and the molecular prognostic factors include BRCA‐1, BRCA‐2, p53, erbB oncogenes, loss of heterozygosity (LOH), chromosomal aberrations, microsatellite instability, transforming growth factor α (TGFα), and the multiple resistance (MDR) gene. Genetic heterogeneity has been shown or is suspected in inherited forms of melanoma, as well as colorectal, breast and prostate cancer (Table 5).

Case‐control studies have shown that breast cancer aggregates in familial patterns along with ovarian cancers and sometimes other cancers (male breast cancer, prostate cancer, and pancreas cancer, for example). The inherited defects include the BRCA‐1 gene at 17q21, a tumour suppressor gene, which when mutated dramatically increases a woman's risk for developing breast or ovarian cancer.²⁰⁰ In addition to BRCA‐1, another breast cancer susceptibility gene has been cloned, BRCA‐2, which is located at 13q12.3.²⁰¹ With regard to BRCA‐1, each of the 1 in 200 women who carry a mutation in this gene will have an estimated 85% lifetime risk of developing breast cancer. The fact that 15% of women who carry such a defect never develop breast cancer is an intriguing point which needs resolution. It is becoming clear that BRCA‐1 mutation screening could have a useful role in familial and early‐onset cases, but is unlikely to have a major impact on breast and ovarian cancer in the general population. While the risk of ovarian cancer among women in general is less than 2%, women with BRCA‐1 have a risk of developing ovarian cancer in their lifetimes as high as 45%. The risk is 25% among women with BRCA‐2. Since there is not yet a preventive therapeutic drug, screening for breast cancer may be of uncertain value in young women. Extensive clinical trials will need to address these issues before the benefits of BRCA‐1 and BRCA‐2 mutation testing can be determined. It is likely that within the next few years, genetic testing and identification of cell populations that share specific genetic alterations from samples obtained non‐invasively (saliva, sputum, urine, stool, mucosal washings) will become a method of screening populations for common cancers.²⁰²

Several studies have also shown that alterations of the p53 gene may be involved in breast cancer aetiology and pathogenesis. The dysfunction of the p53 protein is associated with tumour progression, since an association between p53 abnormalities and accumulation of genetic lesions (measured as overall allelic imbalance, homogeneously staining regions, and strong ErbB2 overexpression) was observed.²⁰³ A recent meta‐analysis of studies involving over 2000 patients has confirmed that somatic mutations of the p53 gene in the primary tumour are associated with worse prognosis.²⁰⁴

The erbB oncogene family provides an important example of how we can translate advances from basic research into clinically relevant tumour markers that may also serve as new therapeutic targets.²⁰⁵ This oncogene family encodes four different membrane‐bound growth factor receptors: erbB‐1 (epidermal growth factor, EGFR), erbB‐2 (HER 2/neu), erbB‐3 and erbB‐4. The EGFR gene is localized on chromosome 7 and encodes a 170 kDa glycoprotein which possesses tyrosine kinase activity.²⁰⁶ EGFR is expressed at detectable levels in 20–58% of the breast cancer specimens studied.²⁰⁷ The human proto‐oncogene c‐erbB‐2, also known as neu or HER2/neu, is localized on chromosome 17 and the oncogenic effects of erbB‐2 are manifested when the protein is overexpressed rather than mutated.^208,209 Overexpression of c‐erbB‐2 may be associated with resistance to hormone therapy. Pharmacogenomic applications of array‐based transcript profiling include analysis of patient tissues in response to therapy during clinical trials. Expression‐based studies appear to be especially appropriate in cancers, because RNA can be obtained from biopsies and surgical specimens. This technology readily detects the somatic changes associated with the development of some tumours and their response to chemotherapy. Somatic changes linked to therapeutic outcomes include the amplification of the oncogene erb‐B2, which predicts good response to cyclophosphamide‐methotrexate‐5‐fluorouracil (CMF) adjuvant therapy of breast cancer.²¹⁰ One study has proposed that overexpression of HER‐2/neu in ER‐positive patients is associated with a relative resistance to tamoxifen, while ER‐positive HER‐2 negative patients were more likely to respond.^211,212 From a clinical perspective, the assessment of c‐erbB‐2 overexpression in breast cancer might become a useful tool in the future treatment of patients by chemotherapy, since patients whose tumours show overexpression may benefit from higher doses of chemotherapy.²¹³

Two large epidemiological studies have clearly demonstrated the increased risk of breast cancer among blood relatives of patients with the ataxia‐telangiectasia (ATM) gene.^214,215 The evidence supporting the association between mutations at the ATM locus and breast cancer is strong, and no study with rigorous methods has refuted this association. Regarding the use of drug to prevent breast cancer, the approval of tamoxifen for reducing the incidence of breast cancer in healthy women with an increased risk of developing the disease is still being debated.²¹⁶

In an effort to guide the transition from clinical research to clinical application, the American Society of Clinical Oncology has recommended that clinical predisposition testing be offered when the prior probability of a positive result is high because of a strong family history of cancer or very early age of onset of cancer, the test can be adequately interpreted and the results will influence medical management of the patient or a family member.²¹⁷ Knowledge of molecular characteristics of certain cancers has made it possible to identify patients who could benefit from therapies that target those features. For example, leukaemic cells in the majority of patients with chronic myelogenous leukaemia have a telltale chromosomal abnormality.²¹⁸ This abnormality, a swapping of genetic material between chromosome 9 and 22, results in the production of an abnormal enzyme thought to be related to the development and proliferation of leukaemic cells in patients with chronic myelogenous leukaemia. Therefore, effective pharmacological intervention may be designed to target and specifically to inhibit this faulty enzyme. Concerning breast cancers, only women whose breast cancers overexpress a gene called HER2 are offered the medication herceptin, a drug that binds to the product of that gene, a growth factor receptor.²¹⁹ Women with breast cancers are also not treated with tamoxifen, if their cancers do not express the gene for the oestrogen receptor.²²⁰ Therefore, genotyping may help for treatment strategies and in some drug treatment choice, genotyping may be an absolute necessity. The adenoviral agent, ONYX‐015 (Onyx Pharmaceutical), for instance, aims to combat tumours caused by mutations in the tumour suppressor gene p53. As part of the clinical trials, Onyx is providing genotyping of patients for p53 mutations. This drug will require the development of an accompanying diagnostic tool to stratify the populations.

A number of gene therapy approaches have been taken in clinical trials for cancer therapy, including the delivery of tumour‐suppressing genes, notably p53, cytokines and the herpes simplex virus‐thymidine kinase gene plus gancyclovir treatment.^221–223 A phase I clinical trial of wild‐type p53 (p53^wt) gene therapy of patients with hepatocellular carcinoma (HCC) in the UK has already produced objective tumour responses in 50% of the patients treated.²²⁴ The principle behind using p53 for gene therapy is well established. A number of groups have demonstrated that p53^wt inhibits the growth of HCC cells in vivo; while others have shown that if p53 is injected into nude mice pretreated with human cancer cells, their tumours get smaller. This study is the first to attempt gene therapy with p53 of HCC patients. However, with all the above, a major limitation is the inefficiency of delivery into the tumour cells and the relatively inefficient killing of neighbouring untransduced cells by bystander effects.

A recent preclinical study suggests that interferon‐β gene therapy inhibits tumour formation and causes regression of established tumours in immune‐deficient mice, and it was argued that local interferon‐β gene therapy with replication‐defective adenoviral vectors might be an effective treatment for some solid tumours.²²⁵ The type 1 interferons (IFNs), the IFN‐α family and IFN‐β execute diverse biological functions including growth inhibition and immune cell stimulation, and have been shown to be inhibitors of angiogenesis (sprouting of new vessels from pre‐existing blood vessels), and therefore could inhibit tumour growth by blocking tumour vascularization.^226,227 This study demonstrated that ex vivo IFN‐β gene transduction by a replication‐defective adenovirus in as few as 1% of implanted cells blocked tumour formation. Direct in vivo IFN‐β gene delivery into established tumours generated high local concentrations of IFN‐β and inhibited tumour growth, and the results show a remarkable ability of IFN‐β gene therapy to block the formation of tumours de novo and to cause regression of established tumours. The ex vivo transduction results confirmed the potential bystander effects of a potent secreted cytokine, with as few as 0.3–1% transduced cells blocking the establishment of MDA‐MB‐468 tumours. A variety of other tumour cell lines that were also tested could be blocked with 1–10% of IFN‐β transduced cells. This dramatic regression of tumours appeared to be primarily the result of the direct antiproliferative or cytotoxic activity of IFN‐β, as the IFN‐β gene used in this study was of human origin, and human IFN‐β gene does not cross‐react appreciably with the host mouse cells. Therefore, local IFN‐β gene therapy may provide a promising strategy for the treatment of some solid tumours in humans. Another preclinical study which showed that intramuscular injection of plasmid DNA encoding murine IFN‐α leads to potent antitumour effects in mice, suggests that a novel type of in vivo cancer gene therapy may be used to treat primary tumours as well as to prevent the development of metastases in advanced cancer patients.²²⁸

Several of the approved clinical gene therapy protocols that involve cancer patients concern therapies designed to treat brain tumours.²²⁹ Factors that promote the use of gene therapy for gliomas include the failure and toxicity of conventional treatments, and the identification of the genetic abnormalities that contribute to the malignancy of gliomas. During the malignant progression of astrocitic tumours, several tumour suppressor genes are inactivated, and many growth factors and oncogenes are increasingly overexpressed. Therefore, in principle brain tumours could be treated by targeting their fundamental molecular defects, if the gene drug can be delivered to a sufficient number of malignant cells. Further improvement in the gene delivery systems would probably help to make significant progress in the successful clinical application of gene therapy in the treatment of brain tumour. Results obtained to date in this field suggest that gene therapy strategies for brain tumour are quite promising, provided more clinical research is performed, particularly in the vector field.

Because of the discovery of large amount of new genes involved in cancer, the National Cancer Institute is launching a major effort aimed to revolutionize how tumours are classified, a project that may have huge impact on how doctors in the future will diagnose and treat cancer.²³⁰ Pathologists traditionally have relied on tumour pathology as the basis for classifying tumours, and the weakness of such approach is that it is not possible to distinguish between tumours that have similar histopathological features but vary in clinical course and response to treatment. The Cancer Genome Anatomy Project (CGAP) which has been compiling a comprehensive record of all genes involved in human cancer indicates a record of more than 40 500 genes that are directly or indirectly active in one or more cancers, and it is believed that recognizing the detailed ‘tumour signatures’ and fully understanding how they correlate with their key clinical criteria will immensely improve cancer diagnostics and drug development. A recent study has examined the possibility of developing a more systematic approach to cancer classification based on the simultaneous expression monitoring of thousands of genes using DNA microarrays.²³¹ The class discovery procedure automatically discovered the distinction between acute myeloid leukaemia and acute lymphoblastic leukaemia without previous knowledge of these classes, and an automatically‐derived class predictor was able to determine the class of new leukaemia cases.²³¹ These preliminary results tend to indicate the feasibility of cancer classification based solely on gene expression monitoring and suggest a general strategy for discovering and predicting cancer classes for other types of cancer, independent of previous biological knowledge.

Table 5

Some examples of tumours, their chromosome location and the genes cloned

Tumours	Chromosome location	Cloned gene and its proposed function
Retinoblastoma	13q14.3	RB1; cell cycle and transcriptional regulation
Sarcomas, breast cancer	17p13.1	p53 (TP53); transcriptional factor; response to
		DNA damage and stress
	17q21	BRCA‐1; interacts with Rad51 protein; repair of
		double‐strand breaks
	13q12	BRCA‐2; interacts with Rad51 protein
Breast cancer, thyroid cancer	10q23	PTEN (MMAC1); dual‐specificity phosphatase with
(follicular type)		similarity to tensin
Colorectal cancer	5q21	APC; regulation of β‐catenin; microtubule binding
	2p16, 3p21, 2q32, 7p22	MSH2, MLH1, PMS1, PMS2; DNA mismatch
		repair
Neurofibromas	17q11.2	NF1; GAP for p21 ras proteins
Acoustic neuromas, meningiomas	22q12.2	NF2
Renal cancer (papillary type)	7q31	MET; transmembrane receptor for HGF
Medullary thyroid cancer	10q11.2	RET; transmembrane receptor tyrosine kinase for
		GDNF
Lymphoma	11q22	ATM, DNA repair
Exostoses (cartilaginous)	8q24.1, 11p11‐13, 19p	EXT1, EXT2, EXT3
protuberances on bones

GAP, GTPase‐activating protein, a negative regulator of the p21 ras guanine nucleotide‐binding proteins; APC, adenomatous polyposis coli; NF1, neurofibromatosis type 1.

Ethical issues

Scientific and industrial enthusiasm for the promises of genetic engineering have generated widespread concern about the potential dangers of their careless application, which have led to the suggestion of issuing regulations covering what could be done, and under what conditions. It is clear that the ability to modify the human germline may open vast medical possibilities, with unprecedented potential for reducing human suffering. However, consideration should be also be paid to the unforeseen hazards that might emerge from tampering with the human genome, and that are likely to be avoided only by careful monitoring. There is also the difficult task of establishing a boundary between acceptable and unacceptable applications. The ability to know who carryies which disease genes will bring about ethical, legal and social implications. It appears that five basic principles (autonomy, privacy, justice, equity and quality) need to be considered in order to harmonize eventual national regulation.²³²

Considering the concept of autonomy, it is argued that genetic testing and the resulting information is highly personal. Since most genetic information is only predictive and probabilistic, this imprecise nature of genetic information may necessitate further protection against social pressures and a reaffirmation of informed consent procedures. An exception to this principle may be applied to newborn screening programs for immediately treatable disorders. There also seems to be a consensus to limit genetic testing (including prenatal testing) to tests that are medically therapeutic and that decisions as to what tests are considered to be therapeutic remains the decision of individual countries with respect to their cultural, social and political norms. Increasingly, however, doctors may feel obliged to offer tests to healthy patients, including tests for future diseases to which the patient may never have thought about.

As studies to correlate genetic predisposition with clinical disease increase, issues of confidentiality and informed consent warrant more attention.²³³ Informed consent to undergo a predictive genetic test requires that the patient be aware of the possibility of false‐positive and false‐negative findings, and that treatment options after a positive finding may lack proved safety and efficacy, or may not be known. Respect for the privacy of the individual and for the confidentiality of genetic information is crucial, and some guidelines would prohibit any communication to all third parties without consent of the person concerned.^234–236 In the area of justice, the international community is united in its concern for vulnerable populations such as incompetent adults or minors and for future generations. Some argue that the intention of much genetic research is eugenic by implication, and legislation in China which has made this explicit has provoked much controversy.²³⁷ A recent ‘official’survey on ethical issues in genetic testing and screening in China shows that the majority of respondents favor offering genetic testing in the workplace for predisposition of executives to heart disease, cancer, and diabetes (94%) and testing children for genes for late‐onset disorders such as hypercholesterolemia (84%), alcoholism (69%) and AD (61%).²³⁸ There are serious concerns about the purely moral dimensions of deliberate intervention in the human germ line intended primarily to enhance culturally desirable characteristics. Several countries (e.g. Austria, France, Germany, Norway and Switzerland) prohibit germline alteration by statute. Furthermore, in the absence of treatment or prevention, the presymptomatic testing of children for late‐onset disease has not been recommended.

The quality issue has also been underlined so that specific criteria for laboratory test sensitivity, specificity and effectiveness has been recommended. Before a genetic test becomes clinically available for predictive purposes, the test developer must collect information regarding the test's validity. Therefore, there is an important responsibility that such developments and application take place in an acceptable manner. Both scientific and ethical considerations should be carefully weighed. The codification of these principles in an international instrument is probably imminent.

Discussion and conclusions

New gene targets for therapeutic intervention only provide a starting point in the long and difficult process of drug discovery. However, genomics will have an important impact in the later stages of drug development, especially in providing an understanding of the molecular nature of diseases and of the responses, both desirable and adverse to drugs. Modern genetics will bring about significant improvements in the provision and practice of healthcare by redefining disease and targeting treatment. It will also lead to the discovery of novel targets and effective treatments and the provision of more effective preventative healthcare. It is estimated that within the relatively short time frame of 7–10 years, functional genomics will begin to have a real impact on clinical medicine. The major targets of industrial R&D are, and will continue to be, the most common disease such as osteoporosis, diabetes, schizophrenia, asthma, AD, arthritis, obesity, bipolar affective disorders and atherosclerosis. It should become possible to combine genetic information with more conventional risk factors, such as smoking or lack of exercise, to determine quite accurately an individual's chance of developing serious heart or neuropsychiatric disease. These developments will have important implications for the pharmaceutical industry. By identifying those patients most likely to respond to novel drugs, it will be easier to demonstrate efficacy and safety. This would lead to smaller, more effective clinical trials with corresponding cost savings, in addition to a focused approach in which a suitably efficacious drug will be administered to the right patient.

Pharmacogenomics has major implications both for drug development and clinical management. Pharmacogenomics aims to complement the current ‘one‐medicine‐fits‐all’scenario with drugs that are based on a deeper understanding of gene variations and the effect of such variations on drug responses. Drugs that are more specific, not only in terms of the particular molecule they target, but also in the populations they affect, will be much more widely accepted and used in the future. In the short term, pharmacogenomics will be strategically used for clinical development of particular compounds with potential efficacy or toxicity issues. Pharmacogenomics may also be applied to approved drugs with restricted market share because of limited efficacy or high toxicity. As our molecular understanding of both human populations and preclinical models of disease and toxicity develops, pharmacogenomic information will optimize predictions of drug effects in humans. Perhaps a more certain outcome from the pharmacogenomics approach is a redefinition of what a disease is at the molecular level. There is little doubt that many of our current disease definitions such as diabetes, and heart disease are actually very different diseases at the molecular level that manifest themselves phenotypically in similar ways. Pharmacogenomics will not only enable companies to target drugs more precisely, but will also lead to customization. This should reduce the number of patients needed in any trial, but it would also bring about an increase in the number of trials needed. The creation of populations of ‘software people’ and ‘virtual trials’should reduce the amount of clinical resources required.

The therapeutic industry will soon be entering a time when solutions to therapeutic problems can be targeted to the individual. Using knowledge of gene functions and commercially available genomics tools, a genomics consumer will be able to employ focused, high‐speed technologies that will produce an individualized treatment in a short period of time. This is a fundamental change in research and clinical medicine. Instead of blockbuster drugs, there will be blockbuster therapeutic approaches that will be widely applicable to different populations. One way in which this approach might function would be that a patient who is not feeling well might initially go to a diagnosis centre, where a series of diagnostic tests would be run. These tests would focus on measuring not crude approximations of gene function, but on the gene functions themselves. Massively parallel testing would rapidly provide a clinician with a map of the patient's current pattern of gene expression. Sophisticated bioinformatics tools would match the gene function map to a particular therapy regimen, substantially reducing or even eliminating side‐effects, and provide the patient's doctor with detailed suggestions about continued therapy.

In linkage and association of genetic polymorphisms and disease, it is important at first to convince ourselves that putative associations are real. Consideration of the ethnic backgrounds of patients and the use of multiple, independent populations can help avoid this problem. An association that is not confirmed should be regarded as provisional, pending alternate proof of causality.

However, some important clinical issues will require to be quickly and adequately resolved. The pharmacokinetic and pharmacodynamic parameters of gene delivery vectors are largely uncharacterized in humans. Estimation of risk‐to‐benefit ratio and actual healthcare costs for gene therapy is not well documented. Undoubtedly, gene therapy will not be cheap, and requires resources and vigilance as trials become approved for human use. Although the current ethical, legal and social issues include some educational projects, a distinct and expanded program should be launched that is aimed for students, the general public, clinicians and genetic counselors.

Failure to appreciate the complexity and limitations of genetic tests and the fact that testing may provoke rather than allay uncertainty must be tackled. It will be the responsibility of clinicians to deal with the public's hopes and fears. Therefore, it is important to define and clarify service needs and how best to establish collaboration between geneticists, public health specialists, and primary care teams. Genomic research and its clinical application will require extensive cross‐border discussion, cooperation, and resolutions, not only regarding technical affairs such as patent protection but in more complex areas such as welfare, morality, and ethics.

Dr M. Ponchon is a research collaborator at the National Fund for Scientific Research, FNRS, Belgium. The authors are grateful to the following individuals for reviewing and providing comments on an earlier version of the manuscript: Hong‐Guang Xie, MD, PhD (Division of Clinical Pharmacology, Vanderbilt University Medical Center, Room 552, MRB‐1, Nashville, TN 37232–6602, USA), Michael Hanna, MD (Neurogenetics Section, University Department of Clinical Neurology, Institute of Neurology, Queen Square, London WC1N 3BG, UK) and Silvia G. Priori, MD, PhD (Molecular Cardiology Laboratories, Fondazione Salvatore Mangeri, Pavia, Italy).