Post-genome respiratory epidemiology: a multidisciplinary challenge

Will genetics overwhelm the field of epidemiology? Should epidemiology maintain its focus on environmental and public health issues in this changing world? Is there a need for very large population studies? Whereas the shared interests between geneticists and clinicians may be pathophysiology, i.e. interest in pathways, the shared interests of geneticists and epidemiologists may well be design and analytical issues, as developed by genetic epidemiologists. One major problem is the belief that one discipline is too complicated to be understood and used by those outside the field, and the subsequent lack of confidence in those from other disciplines. A major aim of the present review is to list relevant questions pertinent to the setting up of multidisciplinary research involving respiratory epidemiologists, geneticists, environmental epidemiologists and clinicians, a topic of increasing relevance with the current development of national and international platforms. After describing specific questions pertinent to the field of asthma and chronic obstructive pulmonary disease (COPD), two main methodological aspects regarding relevant technological and scientific advances are presented. The first one concerns biological aspects in the genome and post-genome phases, i.e. how to study the genome, the transcriptome and the proteome. The second area concerns genetic epidemiology, with aspects related to design (case control and family based) and statistical analytical issues. In addition, implications for research, ethical considerations, public health and organisational aspects in the field of genetic and environmental respiratory epidemiology will be presented. In conclusion, key points are summarised, and challenges and needs for the next steps in respiratory epidemiology for the genome and post-genome phases are discussed.

Current challenges in the genetics of asthma and COPD, and in respiratory epidemiology

A rigorous approach to the study of the genetics of asthma and COPD is needed in which all aspects of the aetiology of such complex multifactorial diseases are identified. Appropriate phenotypic characterisation of asthma and COPD is crucial. The heterogeneity of the effects of the environment (short-term effects, long-term effects, cohort effects, cumulative exposure, protective and deleterious effects, etc.) needs tobe considered. Questions arise regarding ethical issues in relation to the potential identification of individuals particularly susceptible to disease, mainly resistant to a specific hazard. Detailed reviews on the genetics of asthma and COPD may be found elsewhere 1–9.

Genetics are studied to understand underlying disease mechanisms, to define subgroups of disease, to identify individuals at risk, to prevent disease and to define new targets for treatment. Genetic epidemiology is studied because the current state of knowledge is inadequate, there are many polymorphisms and many inconclusive results from small selected samples. Large samples in population-based studies are required in order to study disease subtypes, gene-gene interactions and gene-environment interactions 1, 2. The nonrandom combination of alleles at different loci on a chromosome (linkage disequilibrium), which depends on the genetic distance of the loci, the age of the mutation at either locus and the history of the population, complicates the study of relevant genes in population samples. In general, this means that several single nucleotide polymorphisms (SNPs) need to be studied simultaneously for each gene under investigation.

It is likely that the genes that predispose to asthma have in the past been selected to be common in the population. The situation of Australian Aborigines serves as a good illustration of selection pressure. In this population, which is highly infected with parasites, the total serum immunoglobulin E concentration is negatively related to skin-prick test positivity. The nature of genetic polymorphisms differs from those of Caucasian Australians and has stronger associations to atopic phenotypes. In Europe as well, the population geography may contribute much of the overall observed variation in disease phenotype. Europe contains many environments and has a nonuniform genetic structure. Genetic polymorphisms in Europeans show signs of the spread of farming from the Middle East between 10,000 and 6,000 yrs ago, followed by the probable spread of Uralic people to the North East ofEurope and the spread of pastoral nomads (and their successors) who domesticated the horse in the steppes and appeared towards the end of the farming expansion 10. The genetic gradients remaining from these changing populations will be relevant for the allele frequency and presence of asthma genes, and will add to the complexity of gene identification. Examples of haplotypes that may vary markedly between populations include the β₂ adrenergic receptor and the genes of the major histocompatibility complex.

Most known genetic influences on asthma have been identified by the study of candidate genes, i.e. genes of known function. Several groups are now attempting to identify newgenes by positional cloning. This laborious process involves three steps: 1) the study of genetic linkage (i.e. the co-inheritance of disease and a chromosomal region); 2) fine mapping of a region of interest; and 3) gene identification by sequencing and the study of gene expression. Linkage studies correspond to a group of methods that analyse the co-inheritance of DNA markers within families, to determine if a particular region of the genome contains a gene related to the phenotype of interest. The lod score which is often used in such studies is the result of a statistical test used to determine if genetic loci are linked, expressed as log10 of the odds of the observed data under two hypotheses, linked versus nonlinked. Linkage studies by genome-wide screens have revealed that linkages to some chromosomal regions coincide for asthma and other immune disease, and that there are also organ-specific coincidences, e.g. for atopic dermatitis and psoriasis. Such observations suggest that genes affecting disease expression may be concentrated within a few chromosomal regions, that there are clusters of genes with general effects, e.g. influencing dermal inflammation and immunity, and that the atopic component of atopic dermatitis may have a secondary rather than primary influence. In asthma, positional cloning has been successful, in particular, on chromosome 13 11, and has allowed the identification of a disintegrin and metalloprotease 33 (ADAM33) on chromosome 20 12 and dipeptidyl peptidase 10 (DPP10) on chromosome 2 13.

Although cigarette smoking is the major environmental risk factor for the development of COPD, there is marked variability in the development of airflow obstruction in response to smoking. The frequent development of COPD in individuals with severe α-1-antitrypsin (AAT) deficiency (e.g. protease inhibitor (PI) phenotype Z), a proven genetic risk factor forCOPD, has provided a foundation for the protease-antiprotease hypothesis for the pathogenesis of emphysema. Although only a small percentage of COPD patients (estimated at 1–2%) inherit severe AAT deficiency, AAT deficiency can serve as a model of the manner in which genetic and environmental factors may interact to lead to COPD 14. Additional genetic determinants, which have not yet been identified, probably influence the variable development of airflow obstruction in PI Z individuals. Besides PI, a variety of association studies have compared the distribution of variants in candidate genes hypothesised to be involved in the development of progressive and irreversible airway obstruction in COPD patients and control subjects 15. For each of the candidate genes, at least one study refutes the association. Several factors could contribute to the inconsistent results of case-control genetic association studies in COPD, including genetic heterogeneity and population stratification. In case-control studies, observed associations of gene polymorphisms with disease may be due to: 1) a direct effect of a candidate gene; 2) linkage disequilibrium, that is an association of susceptibility gene and genetic marker (between families, at population level); and 3) population stratification, resulting from incomplete matching between cases and controls, including confounding by ethnic background. In the Boston early-onset COPD study 16, a variety of phenotypes that demonstrate smoking-related susceptibility in first-degree relatives of early-onset COPD probands were identified including forced expiratory volume in one second (FEV₁), FEV₁/forced vital capcity, bronchodilator responsiveness and chronic bronchitis. Linkage analysis results on both qualitative and quantitative spirometric phenotypes show that regions suggestive (i.e. p<7.4×10⁻⁴) 17 for linkage for COPD or spirometric values were located on chromosomes 1, 2 and 12 18, 19.

The role that genetics could play in the understanding of the aetiology of asthma is a matter of debate. The allergen hypothesis does not appear to explain the global increase in asthma prevalence. This hypothesises that the primary cause of asthma (and, hence, of the global increases) is that allergen exposure causes atopic sensitisation, which, in turn, induces bronchial hyperresponsiveness (BHR) and asthma. In fact, BHR is not a good surrogate measure of asthma prevalence, and is more relevant to specific mechanisms of asthma 20, 21. The proportion of cases attributable to atopy is <40% in both children and adults 22. There is not enough evidence toconclude that allergen exposure early in life is a major primary risk factor for developing asthma, as insufficient cohort studies have been conducted to demonstrate that link and, in fact, several studies have found no association. An alternative hypothesis is that changes in susceptibility could be occurring through better hygiene, producing reduced infant infections, and thus a reduced T-helper (Th)-1 cytokine response and an increased Th2 immune response. Key features of westernisation, which have been suggested to play a role, include small family size, reduction of infections in early life (particularly gastro-intestinal), large size at birth and changes in diet (with more fatty acids and fewer vegetables). However, replacing the allergen hypothesis with the hygiene hypothesis may be replacing one dogma with another. The hygiene hypothesis is conceptualised within the allergen Th1/Th2 paradigm, which appears to account for at most one out of two asthma cases 22–24. Even this may be an overestimate because the association may not be entirely causal and sensitisation may be a consequence of an asthmatic predisposition. If the package of changes involved in increasing affluence and westernisation can account for the increase in asthma prevalence, it does not necessarily invoke classic atopic (Th2) mechanisms 25.

Genetic epidemiological studies may help in understanding the origin of the disease. The greatest potential for population-based genetic interventions occurs when the disease is known to occur through a single well-defined mechanism, when only a small number of genes are relevant to this mechanism andwhen there is little temporal or geographical variation. Asthma satisfies none of these conditions. It occurs through multiple mechanisms that are not well understood; many genes appear to be related to different aspects of the main (allergic) mechanism that has been studied to date, and there is major temporal or geographical variation that cannot be explained by genetic factors. To date, asthma genetics has hadlimited success in explaining a significant proportion of asthma cases and even less success in explaining the population patterns. The lack of replication does not mean that the observed associations are invalid, but it does limit their usefulness in both scientific and public health terms. The expression of asthma as the increase of asthma and allergy in the last 30 yrs could be primarily ascribed to environmental changes. Genetics cannot account for the global increase in asthma prevalence. Some considered that the importance of genetic factors for health has been overestimated, and that the ethical and practical problems in applying genetic knowledge to interventions have been underestimated. Searching for environmental causes of asthma is likely to yield useful and useable results, whilst the study of gene-environment interactions could play an important secondary role. A simplistic approach would be to ignore genetics, although this would not serve the epidemiological community well. To get to the roots of the disease will be so complex that all of the skills ofthe molecular geneticist and the epidemiologist will berequired, together with work on animal models. It is important that new theories of asthma causation, including nonallergic mechanisms, are developed and tested. The overall goal should be to incorporate more complex paradigms in order to disentangle the aetiology of such a complex disease or group of diseases. Subtypes should be considered; genetics have shown that diabetes is no longer one disease. Intermediate phenotypes are now popular, but difficult to define, because the definition of intermediate pathways is debatable. From the point of view of finding genes, the most promising intermediate phenotype would be one that is the closest to gene, in order to find linkage.

Perspectives of the genome and post-genome phases

Post genome refers to the period after the establishment of the complete human genome sequence. The true post-genome phase is yet to come. The knowledge of the genome sequence is only a first step, the search for SNPs and haplotypes within a given gene now requires a phase of resequencing, as a result of errors in the currently available maps. Genes need to be characterised by their sequence, their function, their regulation and polymorphisms. In the post-genome phase, both genetic epidemiology and technological considerations will play a key role. Phenotyping may be improved in the future through the study of the transcriptome and proteome.

The sequencing of the human genome has shown that there are fewer genes than anticipated. The drafts already available, whilst rendering genetics more accessible are imposing other aspects of complexity, as a result of the smaller than anticipated functional genes. A current challenge is to use this knowledge for epidemiological applications. The requirements for genetic epidemiology include the availability of DNA from large populations of patients, family members andcontrols, high quality phenotypes, and clinical data and methods for identifying gene or whole genome variation by fast, low-cost and flexible genotyping. Technically, numerous possibilities already exist. SNPs show variations in the genome in every few hundred bases 26–28. A large proportion of putative SNPs in public databases have not been verified and much more information is required on allelic frequences in different populations. Assessing only two polymorphisms in the coding region of each gene will not provide sufficiently dense information. The frequency of the variants is heavily dependent on the population and, in this regard, information on the population distribution is critical. Haplotype pattern information may be defined with a small number of SNPs 29–32. The sites within a gene that are common between different haplotypes and may be related to a quantitative phenotype (such as serum angiotensin converting enzyme (ACE) for the ACE gene) may provide critical information on the causal variant.

Epidemiological applications in large populations raise technical challenges after SNP discovery. Mass spectrometry is based on primer extension assay products of different molecular weights run on a matrix and separated by their size. Its strengths include speed (2 s per analysis), the possibility of analysis of large numbers of individuals, the high quality due to absolute result, the miniaturisation of sample preparation (0.5–2 ng DNA per reaction) and the flexibility for the choice of SNPs. Future technological trends include electric field arrays, which are good for small number of individuals and SNPs, oligonucleotide ligation, which is good for large numbers of SNPs, and period-based technology and mass spectrometry, which are the methods of choice for a large number of individuals. The most appropriate and valid methods for both large numbers of SNPs and individuals are not yet established, and it is an intensive area of research 33, 34.

The study of the transcriptome, through cDNA arrays, is difficult but technically feasible and provides a new area of quantitative phenotypic characterisation. Genome-wide gene expression profiling using DNA arrays is the equivalent of performing tens of thousands of Northern blots 35, 36. Recent developments in this technology make it possible to investigate interactions of a large number of molecules with a large gene library, thus providing fundamental insights into biological processes ranging from gene function to development, cancer, ageing and pharmacology. These techniques offer the possibility of a shift from small-scale, low-throughput molecular biology to a functional genomics approach. It will help further understanding of biological processes, genetic networks and living systems by using large-scale systematic, high-throughput technology, either in a purely discovery mode or in combination with traditional hypothesis-driven approaches. Systems biology is an emerging field that aims at system-level understanding of biological systems, which extends beyond the static description oftheir isolated parts and basic mechanisms based on a reductionist approach. It enables a systematic description of the entire sets of DNA, RNA and functional molecules of complete biological systems (genome, transcriptome and proteome) and their metabolic interactions (the interactome and metabolome).

Proteomics is the study of whole patterns of changes in protein expression and their modifications. The situation regarding proteomics is promising but even more difficult from a technological viewpoint than genomics. Whereas there are ∼30,000 genes, there may be 400,000–1,000,000 proteins 37. Generating hypotheses approaches and candidate (a priori hypothesis) protein approaches are both of interest. Technological aspects are still major considerations and have not yet been solved. The human proteome project will not be deliverable without access to high-affinity ligands, produced at genomically relevant cost and time. There is a need for an enormous increase in throughput and reduction in the cost of antibody development and ensuring that high-throughput array avoids cross reactivity. It is anticipated that in the next 5 yrs, the cost will decrease by a factor of 100, thus making the technology more accessible. Specialists aim to develop a matrix capable of following the protein output of each and every gene within the human genome. An aspect familiar to epidemiologists is the dynamic nature of the human proteome, which varies according to environmental characteristics and the time of the day. Potential future applications include so-called patient “cohorting”, which aims to characterise patients in order to develop markers of drug rejection, prognosis, early diagnosis and disease management, etc. Understanding biological networks is a challenge. Tools are currently only available for <5% of proteins but practical applications may be expected in the next few years in this area as a direct application of knowledge gained from sequencing the human genome 38.

Unsolved questions relate to the validity and quantitative accuracy of the observed changes including the precision and reproducibility of massive amounts of data, and the inferences on levels of gene expression that can be drawn from the hybridisation signal intensities of cDNA probes to gene expression levels. A central issue is the choice of genes, which should be the priority for further study. Studies of single genes and clusters of genes, analysed in terms of common functionality, interaction, coregulation and, ultimately, the understanding of protein networks through their underlying genes, may be informative. Setting standards will be essential in order to facilitate the comparability of gene expression data from different sources, and ensure the compatibility of different gene expression databases and data analysis software.

The limiting factor in future studies is likely to remain the phenotype rather than the genetic technology. Other aspects to be considered are quality control issues, including background noise versus signal, the need of replicates in relation to inter- and intra-subject variability, the number of samples studied, and protection of results and the sharing of data (and methods) in the context of public or private partnerships. The amount of variability concerned at the proteomic level is enormous, hence the challenge in applying these techniques. Issues of validation at the population level are critical.

Genetic epidemiology: designs and analyses

The classical case-control design has entered into a new phase in genetic epidemiology as a consequence of the development of large-scale facilities for genotyping (table 1⇓). Furthermore, for diseases such as COPD, for which age of onset is generally late, family-based designs have limitations. New types of case-control designs in the context of the search for genetic factors have appeared. Although well known among genetic epidemiologists, such designs could be used byrespiratory epidemiologists without extensive training. Contrary to classical linkage analyses, association studies are intuitive to epidemiologists familiar with associations, selection, confounding and interactions in the context of environmental factors.

From genetic epidemiology to public health

Studying the genetics of complex diseases is a multidisciplinary enterprise. Studying families can allow gene identification by positional cloning for monogenic diseases, with the major role devoted to human geneticists, along with an important contributing role for genetic epidemiologists, although a lesser role for clinicians. Positional cloning is still a paradigm for complex diseases, but it needs a different structure. Debates between scientists interested in function (gene-driven research) and those interested in disease aetiology (disease-driven research) are constructive and these two approaches need to be combined. Geneticists are no longer in the position to set up a project on their own, and the competence of genetic epidemiologists and clinicians is necessary. From the genetic epidemiologist's perspective, environmental contributions fall into the remit of molecular epidemiology. For example, a National Genomics Research Network has been established in Germany with disease-oriented networks and several platforms, amongst which there is one for high throughput genotyping, and one for design, data management and analyses called the Centers for Genetic Epidemiologic Methodology. It is only by fulfilling the need for competence and expertise in phenotypes, genotypes and in methodology that there is any chance of solving some of the genetic components of the aetiology of complex diseases (table 2⇑).

The Human Genome Epidemiology (HuGE) initiative is a collaborative effort of individuals and organisations that are committed to the development and dissemination of population-based genome data worldwide 64 with a public health perspective. Potential applications include primary prevention (such as aiding at understanding the biology of diseases, potentially leading to the development of strategies based onmass intervention or target risk intervention approaches), secondary prevention (such as the development of primary tests for population-wide screening) and therapy (such as choosing among alternative interventions) 65–67. Relevant information includes: 1) prevalence of gene variants; 2) assessment of the contribution of gene variants to the burden of disease and disability (penetrance and attributable fraction); 3) the magnitude of disease risk associated with gene-gene and gene-environment interactions in different populations and relevant to targeting intervention; 4) the clinical validity and utility of genetic tests; and 5) the determinants and impact of genetic tests and services. Epidemiological data on their own are clearly not enough, and it is important to develop policies to integrate genetic testing intomedical or public health programmes for public health professionals and policy makers, researchers, healthcare providers (primary and specialists) and consumers. HuGE Net aims at developing information exchange, the knowledge base (with reviews on specific aspects), training and policy.

The term epidemiologist encompasses several dimensions, with the respiratory epidemiologist, the risk factor epidemiologist, the environmental epidemiologist and the genetic epidemiologist. From a geneticist's point of view, epidemiologists and clinicians are often viewed as collectors of good family cohorts (i.e. providing phenotypes and DNA) for groups with genotyping facilities. One of the challenges of the post-genome period is to understand and get the appropriate training that will ensure a true integration of the various facets of the work of an epidemiologist interested in one given disease (table 2⇑).

Ethics

Genetic epidemiology, and DNA sampling and have banking raised new ethical questions. Presently, four different kinds of banks exist according to their mode of constitution: 1) those built through research activities in universities and hospitals; 2) those established following legislative decisions, such as in Iceland and Estonia, where whole countries have decided that DNA is a national resource to be exploited in their country; 3) commercial banks established by pharma or by start-up companies with the intent of commercialising results of research; and 4) virtual banks corresponding to entities which are soliciting DNA donors via the worldwide web and sending the results to individuals or through physicians.

Furthermore, it is important to point out the increased interest in the DNA of heterogeneous populations as opposed to high-risk populations, or of isolates (in population terms). The target has changed with the mapping of the genome and the study of SNPs in large populations 68. This implies, in turn, that ethical aspects not only have to consider the issues raised by the typology of emerging banks, but also have to move away from the monogenic model of research in homogeneous populations. Most ethics codes have concentrated on individual rights (autonomy, consent, discrimination, confidentiality, etc.). The guidelines proposed no longer completely respond to the new issues raised by population-banking studies. The development of normative frameworks appears necessary. Some of the new relationships between research, populations and society have already been addressed by bodies such as the United Nations Educational Scientific and Cultural Organisation, World Health Organization, Human Genome Organisation, and Council for Internal Organizations of Medical Sciences 69–71.

At least six important issues need to be addressed as follows. 1) Recruitment. 2) The principle of mutuality i.e. involving healthy populations and communities with no direct immediate benefit and consent to participate. Finding a way for the community to express itself in such a context is needed 72. Anonymised samples generally have limited use, thus it may be possible that in the next 10 yrs coded DNA could be used in an ongoing fashion provided there is ethical review. 3) Confidentiality as individual anonymisation or coding cannot mask the issue of genetic susceptibility profiles, which can emerge on a group or a region. 4) Banking has to be considered prospectively due to its sociological, psychological and cultural implications. In particular, policy clarifying aspects such as the change in principal investigator, lack of funds, the structural aspects of the institution and responsibility issues should be transparent. 5) Communication of results needs to be clear upfront. 6) Commercialisation issues also need to be addressed 73.

Moreover, while people generally cannot own their DNA, they do exercise control, and the institution, custodianship. Transformation of the DNA adds intellectual know-how to the sample. Patenting can only be on what is novel, inventive and has economic utility. There is ongoing debate between countries on how to distinguish between discovery and invention. Intellectual property rights may have commercial value and that should be clear from the outset. The notion of shared benefits for the people as a group has also emerged in the last 5 yrs. Current discussions centre on whether populations get recognised for their contribution or receive something back as immediate investments in public health or for humanitarian purposes 72. It is in that context that some countries have requested to co-patent the DNA, whilst others request that DNA should always remain in the country of origin or only leave the country after approval by a central representative body. The post-genome phase has raised various new aspects regarding confidentiality, traceability, commercialisation and patenting, feedback to the subjects, and differences between countries, which all need to be addressed.

Requirements for the inclusion of genetic research in respiratory epidemiology

The relevance of incorporating genetics in respiratory epidemiology is not obvious to all, although it is generally agreed that new concepts and approaches need discussion before putting them into practice. More information, training (both of respiratory epidemiologists and geneticists) and discussion appear necessary (table 2⇑).

From the epidemiological perspective the importance of valid case-control designs for association studies should beemphasised, which is within the area of expertise of the epidemiologist, and hence reassuring. Geneticists could be reminded of the central importance of good phenotypic characterisation, which is strongly grounded in respiratory epidemiology. The need for large sample size and replication, and the concept of population stratification (one type of confounding) are now well understood. Future studies should systematically collect information on the origins of individuals participating in such studies. Building nested case-control studies within cohorts and getting ready to react quickly to new genetic developments, as they became available would be efficient. Defining standards of study, thereby allowing the full realisation of the opportunities within existing cohorts, has been suggested. Improvements in organisational aspects that respect the existing framework of various studies are needed to encourage networking and possible aggregation of data banks, including recognition of the work requested to set up biological collections 74. In the field of respiratory epidemiology, there is already an excellent foundation with regard to standardisation, as evidenced by initiatives such as the British Medical Research Council questionnaire for COPD, the European Community Respiratory Health Survey, and the International study of Asthma and Allergy in Childhood studies for asthma.

Whilst not forgetting the major importance of environmental factors, and in particular of smoking in the context of COPD, there is now a consensus among epidemiologists that genetic research merits serious consideration. To understand the complexity of environmental influences, it is necessary to take into account the pertinent window of exposure for a disease with a variable expression over time. A real strength of the respiratory epidemiological community resides in methodological aspects, as many of the complex issues raised in genetic research have a great deal in common with issues and challenges in the design and analysis of traditional environmental epidemiology. New possibilities of collaboration could emerge between environmentalists and geneticists through the high level in statistical modelling of epidemiologists, with a current focus on environment. Design and statistical aspects are probably the bridge between the respiratory and genetic communities, at their common interface with genetic epidemiology. Training epidemiologists in these fields will provide the knowledge for deciding whether it is the time to incorporate the “new genetics” into epidemiological research and public health, and, conversely, training of basic scientists and clinicians in epidemiological methods is a necessary step to develop true multidisciplinary research. International networking, in particular at the European level, is now emerging with the support of various initiatives of European scientific societies, in particular for asthma and allergic diseases, and should provide the framework for the challenges of post-genome respiratory epidemiology 75, 76.