|
|
|
|
|
|
|
|
|
|
|
 |
 |
International Conference on Environmental
Epidemiology & Exposure
2 > 6 September 2006, la Villette
Conference Centre, Paris
Sunday September 3
Tuesday September 5
Wednesday September 6
Sunday September 3, 2006
SPL-1 On the continuum of exposures affecting children' health - Defining a conceptual paradigm for the earliest critical windows
Dr. Germaine Buck Louis from the National Institute of Child Health and Human Development, Bethesda, USA
Childhood is marked by a period of rapid development and underscores the need for researchers to be cognizant of the highly interrelated and time-dependent processes underlying successful human development and homeostasis. Without doubt, children, broadly defined, are often more adversely affected by environmental exposures than adults, indicative or their inherent biologic susceptibility and vulnerability arising from exposure pathways that are atypical of adult behavior. For example, children’s physiology differs from adults in that they have faster inhalation rates, higher consumption of food and liquids per body weight, a larger body surface to weight ratio as well as a host of other differences. The World Health Organization estimates that children represent approximately one-third of the world’s population and that approximately 30% or more of the global burden of disease can be attributed to environmental factors. Despite children’s at risk status for many environmental exposures, risk assessment paradigms typically do not capture the entire life span, underscoring the need for efforts to ensure the safety of all.
To aid in understanding the relation between environmental agents and children’s health status, several authors have developed conceptual paradigms for grounding research and to aid in the interpretation of results. Critical windows have been defined in relation to varying operational definitions of childhood such as chronologic time denoting age (inclusive of gestational age), in relation to school placement or other legal statues and, more recently, with regard to developmental stages. As yet, no universally accepted critical windows paradigm exists, though most epidemiologists recognize the presence of three broadly defined domains – conception, pregnancy and following birth. The latter two domains often are referred to as the prenatal or postnatal periods. Each of these three critical windows has been operationalized somewhat differently by authors. While there is a large amount of epidemiologic evidence regarding critical windows for in utero exposures and structural birth defects, timed exposures in relation to functional endpoints at birth, during childhood or beyond have received limited systematic study, underscoring one of many important critical data gaps for population-based epidemiologic research.
Growing interest in the so-called fetal origins of disease hypothesis (or more generically, the developmental origin of health and disease) during the past few decades has facilitated recognition and appreciation of the importance of critical windows in grounding epidemiologic research to derive answers to questions about the impact of environmental factors on children’s health. Reproductive epidemiologic research has continued to generate a body of evidence, albeit limited in scope, supporting the importance of the peri-conception interval as the earliest critical window for human development, one that is often overlooked. For some time, this oversight was a reflection of limited tools for capturing exposures in the absence of clinically recognized pregnancies coupled with limited statistical models and software for the widely known clustering of pregnancy outcomes and the longitudinal measurement of exposures. Fortunately, such methodologic limitations no longer preclude researchers’ ability to capture the peri-conception critical window. The absence of a biomarker for human conception prior to implantation has challenged epidemiologists in the design and conduct of research. However, this limitation is being overcome by the commercially available technologies for timing data collection to the menstrual cycle and, hence, to the time of ovulation and estimated conception. Thus, epidemiologists are not impeded by the absence of methodologies for population based research capturing the peri-conception window. In fact, this is an era of extreme excitement for designing research responsive to critical data gaps so that lingering questions may be addressed aided by the continual development and refinement of research methodologies including the use of the home for collection of data and biospecimens.
So, how might epidemiology proceed in answering questions about the impact of the environment on child health? Research aimed at estimating health risks in relation to environmental exposures requires prospective epidemiologic designs with longitudinal capture of data and biospecimens for classification of exposure, disease and even susceptibility. This approach allows a spectrum of reproductive and developmental outcomes to be assessed minimizing the risk of missing an adverse effect on one or more endpoints including those that might reflect subtle changes in function. The ability of such research to obtain valid and reliable data on exposure and a spectrum of health outcomes is dependent upon a conceptual framework that recognizes and accommodates the many complexities and eccentricities of human development. In assessing the spectrum of human developmental outcomes, epidemiologists are increasingly able to address competing risks for outcomes conditional on birth, such as embryonic or fetal mortality. This strategy is consistent with the timed dependent nature of human development and provides a more complete understanding of study findings conditional on a pregnancy resulting in a birth.
This talk presents an overview of a critical windows conceptual paradigm that offers a framework for delineating parental exposures and time-varying exposures, including the use of novel technologies suitable for population based epidemiologic research. This talk will emphasize the peri-conception critical window using the Longitudinal Investigation of Fertility and the EnvironmentStudy for illustration.
SPL-2 Early-life exposures and chronic disease risk in later life
Pr. George Davey-Smith from the University of Bristol, UK
Exposures acting during the pre-natal period and during infancy and childhood are increasingly being identified as potential causes of chronic disease in late adulthood. This presentation will review evidence from a range of study designs that implicate nutritional factors, infections and other environmental exposures acting during early-life sensitive periods in generating increased risk of metabolic and cardiovascular disorders in later life. The problems of establishing causality when outcomes are related to exposures acting across many years, with the possibilities of confounding, bias and reverse causation being manifold, will be discussed, as will novel study designs for establishing causality. Examples will be drawn from studies of early-life origins of cardiovascular disease, such as those relating exposures during the intrauterine period or during the early post-natal period with blood pressure, insulin resistance and cardiovascular disease in later life. Situations in which similar exposures acting at different stages of development have different long term consequences will be addressed. The notion that predictive adaptive responses may underlie some of the associations between early-life exposures and later-life health outcomes will be discussed.
Tuesday September 5, 2006
TPL-1 Understanding the relationship between genes and environmental hazards: promises for environmental epidemiology
Dr. Danielle Fallin from Johns Hopkins Bloomberg School of Public Health, Baltimore, USA
There is notable inter-individual heterogeneity in the consequences of exposure to environmental agents. Genes likely contribute substantially to this heterogeneity, and the identification of particular genetic variation involved may improve our understanding of the pathogenic mechanism of such agents, our ability to identify those at greatest risk, and our ability to design primary and tertiary interventions. This has been recognized through several recent US and international initiatives, such as the Environmental Genome Project at NIEHS.
The classical genetic model proposed that sources of disease or phenotype variation could be parsed into mutually exclusive genetic and environmental components, and the main etiologic work for a disease was then focused on either heritable or environmental causes, depending on the evidence for one source versus the other. Scientists have since acknowledged that these are rarely mutually exclusive contributions, and that there are often complicated interactions between inherited genes and environments that lead to disease. Some simple examples include the variable response to primaquine due to genotypes of the g6pd gene, which controls degradation of oxidizing agents. Those with risk variants of this gene are susceptible to anemia upon treatment with the antimalarial drug due to a decreased ability to detoxify this oxidizing agent. Other examples include variable adverse outcomes due to lead exposure, depending on genotype. More recently, work by my Hopkins colleague Brian Schwartz and by others has shown evidence for modification of lead-related risk for decreased cognition due to genes such as APOE, ALAD and VDR (Stewart et al. 2002;Weaver et al. 2005) . Many other examples like these are now emerging.
However, these simple gene-by-exposure interaction scenarios are likely to be only the tip of the iceberg. Many other genes that modify exposure risk are yet to be identified due to lack of adequate sample sizes and low power. Creative designs such as matching or case-only approaches can help to alleviate this to some extent. However, higher order interactions between multiple genes and potentially multiple exposures are also likely and data mining/pattern matching tools are emerging to help discover such effects. Further, such work to date has mostly focused on DNA variation as a modifier of exposure risk effects. Yet, the term “genetics” encompasses the central dogma of DNA, RNA, and even protein variation. Both quality and quantity matter, meaning a qualitative difference such as a nucleotide sequence change may affect the ultimate function of a protein and contribute to disease risk, but so too would the amount of DNA, or the amount of transcribed RNA. This is the focus of copy number variation work at the DNA level, and of expression work, such as expression arrays, at the RNA level.
Another important source of quantitative genetic variation is through epigenetics, which refers to the study of information within a cell that is heritable during cell division, but is not contained in the DNA sequence itself. Typically this is in the form of methylation of DNA cytosines, or chemical modification of histones (such as acetylation or phosphorylation). This source of genetic variation is of particular interest as a potential mechanism directly rela ting environmental exposure to gene expression, and ultimately to disease risk since environmental exposures may impact epigenetic modifications, leading to aberrant gene expression, and increasing risk for disease. Measurement of these epigenetic changes can therefore help explain observed relationships between exposures and disease phenotypes. Further, because epigenetic changes can increase or decrease gene expression, changes over time can amplify or attenuate the effects of DNA sequence variation. Currently, such “noise” often masks DNA sequence effects. Thus, a promising avenue for environmental epidemiology is the partnering with genetic and epigenetic measurements to illuminate relationships between environment, DNA variation, gene expression, and disease risk.
My colleagues and I, through the Center for Excellence in Epigenetics at Hopkins, led by Andy Feinberg, have developed a working conceptual framework for the interplay of sequence variation (‘traditional genetics’), epigenetics, environment, and disease risk (Bjornsson et al. 2004;Bjornsson et al. 2004) . Our model implies several testable hypotheses that we and others are currently addressing. First, if exposures act on the epigenome to influence disease, then there should be measurable intra-individual changes in epigenetic marks over time. No study has yet shown epigenetic changes in the same humans over time, although evidence from animal and cell work support this (Cooney 1993;Issa 2002) , and a recent study of 3-year old and 50-year old MZ twins has shown epigenetic differences by age, in a cross-sectional analysis. In that report, the similarity in methylation patterns between young twins compared to the dissimilar patterns among older twins argues strongly for environmental influence on the epigenome (Fraga et al. 2005) . We are currently studying methylation in the same individuals over a 15 year period to examine this longitudinally.
With evidence that epigenetic changes do occur in the same person over time, a second testable hypothesis is that particular environments can create or accelerate these changes. There is growing evidence for this as well, as reviewed in (Sutherland and Costa 2003;Van den Veyver 2002) . These include evidence of DNA methylation dependence on dietary methyl donors and folate (Van den Veyver 2002) , smoking influence on methylation and expression levels (Toyooka et al. 2004;Enokida et al. 2005) and methylation and histone modifications through exposure to metallotoxins such as cadmium, nickel, arsenic, and zinc (Beyersmann 2002;Sutherland and Costa 2003) . Since these particular exposures have also been shown to be important risk factors for several common human diseases, this is a promising avenue consistent with our conceptual framework.
Finally, a third hypothesis implied by our conceptual framework is that epigenetic mechanisms are directly involved in risk for disease. The strongest evidence for this is in cancer, although t here is also growing evidence linking epigenetic mechanisms with cardiovascular disease (Newman 1999) . In fact, it has been suggested that epigenetics may be a central mechanism of atherosclerosis (Newman, 1999), and the link between epigenetics and atherosclerosis has been supported in several animal and human studies (Zaina et al. 2005) .
This new framework, which jointly considers genetics, epigenetics, and environmental toxins, provides promise for elucida ting the mechanisms of exposure-related disease risk, allowing better detection of those at greatest risk for adverse outcomes of particular exposures, and affording opportunity to develop prevention strategies for such individuals. However, this new paradigm is not without limitation. We are currently restricted by the available technology to measure epigenetic and genetic information in large samples at affordable prices. Also, epigenetic variation, like gene expression, is tissue-specific. Therefore, one must think carefully about whether available tissue samples, typically lymphocytes, will be informative for diseases of interest. T here is reason to expect that they will be in some situations, given evidence that lymphocyte and colon DNA methylation is concordant (Sakatani et al. 2005) , and that loss of imprin ting in Rett Syndrome was found in lymphoblastoid cells from patients. However, careful comparisons of epigenetic information between tissue types are needed. Finally, w e also need creative designs for sampling populations that will best inform these hypotheses, and improved statistical techniques for handling such measurements and incorpora ting them into epidemiologic studies. G enetic concepts and measurements, from DNA sequence, to RNA, to epigenetic marks, do provide promising avenues for environmental epidemiology by helping to explain inter-individual variability in susceptibility to toxins, and by potentially connec ting those toxins to the disease mechanism.
Reference List
TPL-2 Analytical challenges and new methods in gene-environment studies
Pr. Nan Laird from Harvard School of Public Health, Boston, USA
The Human Genome and related projects, and the recent developments of the new genotyping technology have had a profound impact on gene mapping strategies. Association studies of candidate genes are now used extensively in gene mapping, either singly, or in tandem with linkage studies. Genome wide association studies are now a reality. Several studies with 100,000 SNPs genotyped on cases and their controls have been reported, and many more with 350,000-500,000 SNPs are planned or underway.
Mapping Mendelian disorders using linkage has been highly successful. In this case, disorders are caused generally by single gene mutations, and gene-environment associations, such as PKU and diet, can sometimes be identified without the need for molecular analysis (Hunter, 2005). By contrast, linkage has been less successful for complex disorders, which are characterized by variable age at onset, some occurring only late in life, difficulty of diagnosis, and numerous contributing factors, including life style and/or environmental exposures as well as genes, and possibly gene-gene and gene-environment interactions. Here association analysis holds much promise in sorting out the contributions of genes and the environment.
Locating gene-environment interactions has long been an interest to environmental epidemiologists, as there are now several important examples where seemingly harmless genetic polymorphisms have serious adverse effects in the presence of certain environmental exposures. This talk will address how recent developments in statistical methods for genetic association analysis for complex disorders offers promise for identifying gene-environment interactions. We will first briefly review association designs for finding genes, then discuss extensions to gene-environment interactions.
There are two major classes of association studies: those which use affected individuals and suitably chosen, unrelated controls, and those which use affected cases and their families. The designs can be extended to population samples which are not ascertained on the basis of disease. Each design has clear strengths and limitations. Case or cohort designs are generally more powerful than family designs, and statistical models can straightforwardly be used to model the simultaneous effects of genes and environment and their interaction. A major limitation of these designs is potential confounding factors which arise if disease rates and gene frequencies vary across subgroups of the population.
The family based designs have a major advantage over studies involving unrelated individuals in that they permit tests of gene effects on disease which are completely robust to any confounding effects, including population substructure. Most tests for gene-disease relationships are conditional tests, based on conditioning on disease status, and on parental genotype information. Such tests generally tend to be less powerful that unconditional tests, but have the advantage that they are statistically independent of the parental genotypes and disease trait (Laird and Lange, 2006). This latter information can thus be used to develop models and screen multiple traits, or genetic markers for testing, completely independently of the actual test. This offers tremendous potential for overcoming the multiple comparisons problem which arises in almost any gene mapping study, but especially in whole genome association scans. We will discuss an application to a whole genome scan of Body Mass Index (Herbert et al, 2006).
An additional feature of family based designs is that they are completely robust to any assumed model for the disease trait. That is, one need not assume a specific model for how the gene affects the trait in order to construct a valid test for the effect of the gene on the trait, although correct modeling of the trait can certainly make the test more powerful. For example, in constructing the test statistic, one can assume an additive, dominant, recessive or multiplicative model for the effect of the gene, but the test will remain valid under the null hypothesis regardless of the assumed model. When the trait is quantitative, it is not necessary to make any distributional assumptions on the trait. This becomes an issue, however, for both population based and family based tests when gene-environment effects are important.
The general statistical approach to the analysis of gene-environment interactions is to define a model for how the genotype affects the expected trait as a function of gene effects, environmental effects and gene-environment interaction. It is important to note, that a distinction is often made between biological interactions, or synergism, and statistical interactions (Thomas, 2004; Clayton and McKeigue, 2001). Interaction in the statistical approach will depend upon the chosen model for the analysis, i.e. linear or multiplicative, and presence or absence of statistical interactions may or may not coincide with presence of absence of biological interactions.
In the context of case-control or case-cohort studies, the use of logistic models for describing additive main effects and their interactions are popular because of selection on disease status. For quantitative outcomes, linear models are often used. Standard statistical methods can be used for testing interaction effects. Of course it is worth remembering that the validity of the tests rely on the underlying model for the main effects.
If one is willing to assume that genetic and environmental factors are independent in the population, and to further assume a multiplicative model for disease risk, then one can show that it is possible to make inference about interactions as departures from the main effects relative risk model by using cases only (Piegorsch et al., 1994). Umbach and Weinberg (1997) have shown how these assumptions can increase efficiency and reduce data collection in case-control and case-cohort designs as well, in a rare disease setting. Of course, the assumption that genetic and environmental factors are independent in the population is strong and may not be true in the presence of population stratification.
In testing effects of genes, family based designs are robust both to population substructure and to model selection, since the distribution of a proband’s genotype is computed assuming Mendelian transmissions. Joint tests for the main effect of genes and gene-environment interaction are similarly robust, and do no require correct specification of the model. However, they have the disadvantage that they cannot distinguish between main and interaction effects. As in non-family based designs, a model is required to distinguish between these two effects, as the test requires a null hypothesis that assumes a gene effect, against an alternative that assumes both a main effect and an interaction. Umbach and Weinberg (2000) and Lake and Laird (2003) have developed such tests for trio designs (cases and their parents), which make similar, but less restrictive assumptions as the case-only design. The relative risk model is assumed for the main effects, but it is necessary to assume only that genes and environmental factors are independent conditional on parental genotypes. This effectively means that an individuals’ genotype does not influence their exposure; this is considerably weaker than the assumption that the two are independent in the population. Vansteelandt and Lange (2006) have extended these tests to more general situations, including quantitative phenotypes, more complex family types and haplotypes.
References:
Wednesday September 6, 2006
WPL-1 Meeting the challenges of environmental health: promoting science and responsiveness to societal concerns
Nathalie Kosciusko-Morizet, Member of Parliament, Chair of the task force on health and the environment, Paris, France
Meeting society’s expectations without avoiding new concerns and without falling into manipulation, pessimism or fantasy. To seek this subtle balance, one should first recognize that, with regards environmental health, any concern deserves at least an echo, if possible an answer, and failing this, at least an open door towards research.
The « Environmental Health » group from the French Parliament, which I have the honour to chair, has, during its various studies, brought forward the fact that questions related to environmental health are not longer insignificant. Public queries, and sometimes concerns, are on the increase. Chemical substances, air quality, food… the subject is multiform, has various facets.
We have therefore organised in 2003 a conference on noise, then on the impact of chemical substances on health, and in 2005 on electromagnetic fields. The same conclusions apply to each event: answers must be provided.
This is one of the great challenges which the Environment Bill, now included in the French Constitution, intends to address. The first article of the Environment Bill states that:
“Everyone has the right to live in a stable and health friendly environment”.
In particular, in article 5, the Environment Bill provides a definition of the precautionary principle:
“When it is recognised that some prejudice, though such recognition is uncertain owing to the current scientific knowledge, may have a serious and irreversible impact on the environment, public authorities ensure, through the enforcement of the precautionary principle and within their field of competence, the implementation of risk assessment procedures and provisional and proportionate measures in order to prevent such prejudice”.
These clauses therefore attempt to provide an answer to everyone – anxious citizens, motivated or doubtful researchers and undecided decision-makers – by proposing a procedure, a protocol and maybe simply “good practice” to face the challenges of scientific uncertainty in environmental issues. We have been looking for such answers for over fifty years. Ecology, particularly environmental health, issues open wide access to a new world. This is now the dawn of the era of responsibility. And we cannot backtrack.
What are the first lessons? My experience is twofold, both as the Reporter of the constitutional Environmental Bill and as the Chair of the « Environmental Health » group from the French Parliament; it first points towards better coordination, better communication between the various actors. Competences are at present too widely spread, hence diluted. Let us take an example. In a field such as the impact of chemical substances, roughly twenty organisations are acknowledged as competent in France. How can we cross data, interpretations, recommendations? This is one the great political challenges.
WPL-2 Environmental genomics: an opportunity for the NIEHS
Dr. David Schwartz, Director of NIEHS, Director of the National Institute of Environmental Health Sciences, Research Triangle Park, USA
As I continue to consider new research opportunities for the NIEHS, my desire to support research in environmental genomics grows. While the accomplishments and available tools in genetics and genomics certainly enhance my enthusiasm for this field of research, my attraction to environmental genomics stems from my belief that environmental exposures can be used to understand the role of transcriptional regulation and genetic variation in the development and progression of common, yet complex human diseases.
A growing body of research helps to illustrate the opportunities and challenges that lie before us. The influence of environmental exposures on transcriptional regulation of genes is clearly highlighted by the field of epigenetics. Michael Skinner at Washington State University and colleagues recently demonstrated the potential transgenerational adverse effects of intrauterine exposure to endocrine disrupting pesticides on male fertility (Anway et al. 2005). Findings from Randy Jirtle’s laboratory at Duke University indicate that exposure through maternal diet to common methylating agents found in vegetables and vitamin supplements can have profound effects on gene expression in offspring that continue to be inherited in subsequent generations (Waterland and Jirtle 2003) . Moreover, since monozygotic twins diverge in the concordance of methylation as a function of age (Fraga et al. 2005) , it is abundantly clear that methylation is a dynamic process. These findings underscore the role that intrauterine exposures could potentially have on common complex diseases that involve developmentally vulnerable organ systems. Such research also indicates that environmental exposures may serve as biological clues to understanding the regulation of gene expression and the role that transcriptional regulation may have on the risk of developing disease, as well as point to novel therapeutic interventions.
Environmental exposures can also be used to simplify complex biological processes to both discover unique biological mechanisms and to narrow the pathophysiologic phenotype of complex human diseases. For instance, the discovery of the aryl hydrocarbon receptor (AHR) occurred as a direct result of the known toxicity of dioxin and polycyclic aromatic hydrocarbons. Not only did this discovery demonstrate the biological role of the AHR in mediating the toxicity to these agents, but it revealed the role of AHR in homeostatic and basic pathophysiologic processes. Most importantly, however, the identification of the AHR led to the ultimate discovery of the PAS (PER-ARNT-SIMS) superfamily of receptors that mediate response to various forms of environmental stress such as hypoxemia and circadian rhythm, and control basic physiologic activities such as vascular development, learning, and neurogenesis (Kewley et al. 2004; Nebert et al. 2004) . Likewise, understanding of environmental exposures can simplify complex disease processes by narrowing the pathophysiologic phenotype to elucidate the genetics and biology that underlie a particular condition. For example, diseases such as asthma arise from dozens of etiologic agents. Since asthma caused, or exacerbated, by dust mites, endotoxin, or ozone involves different genes and different biological mechanisms, the disease can be better studied by focusing the investigation on a specific etiologic type of asthma.
Given that an extensive number of animal genomes have been sequenced and have demonstrated the evolutionary conservation of biology and genetic structure, comparative genomics will be an important tool for identifying the genes that control response to specific environmental agents, which in turn will accelerate our discoveries in environmental health science. For instance, the discovery of the importance of the toll-like receptors in innate immunity in mammals occurred as a direct result of the observation that a defective receptor in flies caused them to be much more susceptible to Aspergillus fumigatus (Lemaitre et al. 1996; Medzhitov et al. 1997) . The ease with which we can observe and apply knowledge across model systems must be exploited so that we can efficiently understand the biological and clinical importance of environmentally responsive genes.
To facilitate progress in environmental genomics, we need to train young investigators in the discipline, and support scientific programs that focus on biological and clinical problems that can most directly be solved by employing these novel conceptual and methodological approaches. However, to truly have an impact on human health, we need to extend these approaches to understanding chronic complex human diseases including cardiac disease, cancer, diabetes, chronic lung disease, and cerebrovascular disease. While these diseases are known to account for substantial morbidity and mortality worldwide , avoidable environmental exposures and reversible behaviors play a critical role in their development (Willett 2002) . A clear challenge to the field of environmental health sciences will be to make the best use of environmental genomics to inform our understanding of the interaction between environmental exposures and genes in the development and progression of human diseases, and ultimately to translate this knowledge into effective prevention, intervention, and treatment strategies.
© Afsset - 253, avenue Général Leclerc 94701 Maisons Alfort Cedex - France |
Web site : www.afsset.fr |
