September 2009 | Vol 6 | N.º 9 | CNIC-12 [ PDF (140 KB)]

The regulatory basis of common disease: how can genome-wide association studies help us understand cardiovascular disorders?

M. Eva Alonso and Miguel Manzanares
Correspondence:
Miguel Manzanares.
Department of Cardiovascular Developmental Biology
Melchor Fernandez Almagro 3, 28049 Madrid, Spain
Email mealonso@cnic.es, mmanzanares@cnic.es

Abstract
Understanding the genetic component of complex human diseases, such as type 2 diabetes or cardiovascular disorders, was a daunting task until the beginning of the 21^st century. Although classical genetic analysis has proved successful in identifying genes and mutations that underlie monogenic diseases, it was not until the development of whole-genome association scans of single-nucleotide polymorphisms that we began to obtain a glimpse of the genes and networks involved in the most prevalent human diseases. To date, more than 200 genetic loci have been robustly and reproducibly associated with various disorders. However, in the vast majority of these cases, we still lack any understanding of how the identified variants relate to the disease aetiology. In addition, many of the loci uncovered by genome-wide association studies fall in intergenic regions, which complicates identification of the culprit gene. We now know that the non-coding fraction of the genome harbours multiple cis-regulatory elements that control the fine-tuned temporal and spatial transcription of genes. These elements are prime candidates for being altered in the disease-associated variants that do not map to coding regions. Uncovering these elements and how they relate to regulatory networks and pathways underlying common human pathologies is still a major challenge.

Background
Human genetics has taken a giant leap in the past few years with the advent of genome-scale analyses of variation that associate specific genes with a given disease in large cohorts. This progress has been a direct result of the efforts of the Human Genome Project, followed shortly by the Human HapMap Project, which provided information on millions of polymorphic sites consisting of single-base changes (single-nucleotide polymorphisms, or SNPs)¹. New technologies are being developed to rapidly and comprehensively analyse the genetic variation defining patients’ individual genetic profile. Most importantly, and contrary to classical genetic analysis, efforts to identify the genetic contributions to disease susceptibility are developing even in the absence of an identified familial inheritance pattern. This approach to the screening of genetic variation in patients with and without a given disease has been termed genome-wide association studies (GWAS)². In these studies, common variants are analysed and can serve as guides to linked loci that are directly related to a disease via the linkage disequilibrium structure of the genome. Therefore, by analysing 500,000 SNPs, it is possible to identify common variants that associate with a small but statistically significant increase in disease risk³.

Whole-genome association studies and emerging sequencing technologies are being used to describe the genetic architecture of common diseases, revealing new susceptibility genes and offering new clues about the mechanisms behind a wide range of health disorders. The strategic focus of common disease genetics is progressing from the basic identification of susceptibility genes to understanding the extent to which rare and common variants explain inherited susceptibility to common diseases, discovering new disease mechanisms, and pioneering and evaluating the transfer of these advances to potential clinical application⁴,⁵.

Identification of variations for common diseases
GWAS have reproducibly identified hundreds of associations of common genetic variants with more than 80 diseases and traits. The rapid increase in the number of GWAS provides an unprecedented opportunity to examine the potential impact of common genetic variants on complex disease by systematically cataloguing and summarising key characteristics of the observed associations and the trait/disease-associated SNPs underlying them. Although some of these aspects have been examined on a smaller scale for individual diseases and syndromes, such as type 2 diabetes⁶, cancer⁷-¹⁰, inflammatory bowel disease¹¹, restless legs syndrome¹² and atrial fibrillation¹³, further insight will be revealed as certain genomic regions are repeatedly associated with different diseases. Some of these associations are not surprising, such as the recurrent association of the HLA/MHC region on chromosome 6p21 with several autoimmune diseases, or that of the gene desert in the vicinity of MYC on 8q24 with various cancers¹⁴,¹⁵. Others have been more unexpected; for example, the region of 9p21 mapping near the tumour suppressor INK4/ARF locus is associated with a number of seemingly unrelated conditions, such as type 2 diabetes, coronary heart disease and glioma. A possible explanation for this clustering of disease associations is that the variants affect birth weight, which has been described as a risk factor for all three diseases¹⁶. As more loci associated with different diseases are identified, we will surely see more examples in which a common aetiology can be identified for various pathologies.

From association to function
Despite the success of GWAS in identifying novel disease-related loci, determining the exact sequence variants that cause an increased risk and how they do so is far from being achieved. As mentioned above, GWAS search for common SNPs that statistically associate with a disease or trait but are not necessarily the causative variants. These could be rare SNPs, with very low occurrence in the population, or even small structural variants that are in linkage disequilibrium with the tested SNPs. Identification of the true causal SNPs might be difficult; it involves the re-sequencing of the region from multiple samples. Nevertheless, this task will be aided by the near completion of the 1,000 Genome Project, which will dramatically increase the coverage of common genetic variants across the genome¹⁵. An even more complex task, necessarily linked to the above, will be to understand the role of these variants in disease aetiology. In this task, the use of functional assays—aided, if possible, by the use of animal models—will be needed despite being demanding and time consuming.

How to address regulatory variation
When a sequence variant is located in the coding region of a gene and, furthermore, causes a change in the coded amino acid residue, it is easy to envision how this variation can affect gene function. In fact, disease-associated SNPs located in coding regions are enriched in non-synonymous sites¹⁴. However, the vast majority (nearly 90%) of reported trait/disease-associated SNPs fall into non-coding regions, either intronic or intergenic¹⁴, which suggests that the disease-associated variants have a regulatory effect.

Non-coding genomic DNA harbours multiple regulatory elements that are crucial for proper gene function by controlling the levels and temporal-spatial patterns of transcription. These can include enhancer and silencer elements, insulators, boundaries or nuclear matrix attachment regions¹⁷. However, when analysing regulatory variation, we face additional problems. While the causal variants would be in strong linkage disequilibrium with the disease-associated SNP, this does not mean that the gene whose expression is affected by the causal variant will also be linked, or even nearby¹⁵. Hence, the regulatory function of disease-associated genomic regions has been elucidated in only a few studies. One example is the finding that the likely colorectal-risk causal variant, identified by re-sequencing the region to be located near the SMAD7 gene, alters the activity of an evolutionarily conserved transcriptional enhancer that drives expression in the rectum in a Xenopus transgenic assay¹⁸. Another example is the 8q24 cancer-risk variant, which also possesses enhancer activity and physically interacts with the MYC gene, located more than 300 kilobases away. Furthermore, in this case, there is evidence that the causal SNP changes the ability of the TCF7L2 transcription factor to bind to a consensus site located on this enhancer¹⁹,²⁰.

An example of how we can start to address the regulatory capacities of non-coding disease–associated variants with a combination of evolutionary genomics and functional transgenic assays in model organisms is illustrated in Fig. 1. In a GWAS on type 2 diabetes⁶, an association was found between chromosome 10 in the IDE-KIF11-HHEX region and two SNPs located in a 50 kb interval downstream of HHEX (Fig. 1A). This gene encodes a homeodomain-containing transcription factor that is involved in pancreas development in mice²¹, making it an obvious potential target of this non-coding variant. As part of a collaborative effort with groups from Bergen, Padua, Edinburgh and Seville, we investigated whether this region contained enhancer elements that could be related to HHEX function22. We used a multi-species genomic comparison of the region to identify non-coding sequence elements that are evolutionarily conserved. This approach, termed phylogenetic footprinting²³, aims at identifying stretches of conserved sequence across different genomes. If these stretches are located in non-coding regions, the fact that they have been conserved during evolution argues strongly for them having a function that has been positively selected against the random mutational drift of the genome²⁴. One such function is acting as an enhancer of gene transcription. When we analysed the HHEX downstream region, we found that the type 2 diabetes–associated SNPs are located in DNA regions that are conserved between human, mouse, chick and frogs (Fig. 1B). In order to test the regulatory capacities of these sequences, we ligated them to a minimal promoter that has no activity and the lacZ reporter gene. These constructs were used to generate transgenic mouse embryos in which we could observe beta-galactosidase activity (the product of the lacZ gene) in the developing pancreas (Fig. 1C, upper panels). We confirmed that these beta-galactosidase positive cells were located in the endocrine pancreas by double labelling with an antibody directed against glucagon (Fig. 1C, lower panels). In conclusion, we have described how the risk-associated region from HHEX is capable of acting as a tissue-specific transcriptional enhancer in the developing pancreas. Further work will be needed to identify the causal variants in this element and how these variants affect HHEX expression. Only then will we be able to understand how subtle changes in the expression of a developmentally regulated transcription factor can result in adult late-onset diseases such as type 2 diabetes.

Genome-wide association studies and cardiovascular diseases
GWAS has only recently been used to analyse the genetic component of cardiovascular disease, as indicated by the fact that thus far only 22 such studies have been published on cardiovascular disorders, half of them this year (data from the National Human Genome Research Institute –NHGRI– catalogue of published genome-wide association studies, www.genome.gov/gwastudies). These studies include some not very unexpected findings, such as the strong association of a regulator of the nitric oxide synthase pathway (NOS1AP) with myocardial function²⁵. However, there have beensome surprises, such as the association of a region near the tumour suppressors INK4/ARF with myocardial infarction²⁶,²⁷; this region has recently been shown to contain an enhancer of the poorly characterised non-coding RNA gene ANRIL28. However, it is expected that many more GWAS on cardiovascular disease will be carried out in the near future. Crucial to this is the availability of large cohorts that have been followed over years, such as in the Framingham Heart Study; the data from these studies can be mined with the genome-wide association tools that are now available²⁹.

We have surveyed the literature for cases in which variants associated with cardiovascular disease are located in non-coding regions of the genome and examined their evolutionary conservation (Table 1). It is interesting that in a number of these cases, the nearest genes correspond to transcription factors and other genes involved in early development. Studying whether subtle defects that occur in the embryo can result in late-onset cardiac disease will be of great interest in the future.

Atrial fibrillation (AF), the most common sustained cardiac arrhythmia in humans, is characterised by chaotic electrical activity of the atria. Studies of familial AF have shown that it is caused by mutations in potassium-channel genes, but these mutations account for only a small fraction of all cases³⁰. A genome-wide association study followed by replication studies in three populations of European descent and a Chinese population from Hong Kong demonstrated a strong association between two sequence variants on chromosome 4q25 and AF13. Both variants are adjacent to the PITX2 gene, a developmental gene coding for a homeodomain transcription factor that has a critical function in left-right asymmetry and heart development³¹. The regions in which the variants are located are conserved in all amniote vertebrates examined (Table 1), making them excellent candidates for regulatory elements. Two recent studies identified an intronic region of the ZFHX3 gene as a second AF-associated variant³²,³³. It is interesting that in one of these studies this gene was also associated with ischaemic stroke, and in other reports with various cancers as well as with Kawasaki disease. ZFHX3 codes for a homeodomain and zinc finger transcription factor but has been poorly studied, so it remains to be ascertained whether it has a common underlying role in the aetiology of these diseases.

Interestingly, other studies have also found associations between developmental genes and cardiovascular disorders. As an example, an association was found between hypertension and a region in a gene desert located more than 500 kilobases away from the T-box transcription factors TBX3 and TBX5 ³⁴, both of which have roles during heart development³⁵. Gene deserts (large genomic regions devoid of known protein-coding genes ³⁶) are often associated with complexly regulated transcription factors and harbour multiple regulatory elements. The gene-desert region associated with hypertension is conserved up to avians (Table 1). Another example is the association of a region in the vicinity of the flamingo-related cadherin CELSR2 with myocardial infarction³⁷. This adhesion molecule has important roles in planar cell polarity during development and is strongly expressed in the mouse embryo³⁸. Although no role has been established for CELSR2 in cardiovascular development, it is worth exploring this possibility.

Finally, two large studies of genomic association to variation in the QT interval³⁹,⁴⁰ found variants in non-coding regions near a number of ion channel genes that had previously been found to be related to familiar cases of AF or cardiac rhythm syndromes such as Brugada syndrome³⁰,⁴¹. A number of these regions are conserved between human and mouse, and it is noteworthy that many of the genes encoding for subunits of these channels show specific expression patterns in the developing heart⁴². The study of how these genes are regulated during development and disease will be an exciting area of future research that will help us understand how embryonic patterning and tissue physiology are integrated.

Conclusions and future prospects
Recent GWAS have robustly demonstrated that common genetic variation contributes to the risk of developing diseases, and an increasing number of genomic regions have been shown to be associated with an increased risk of cardiovasculopathies. However, because the majority of GWAS are based on an analysis of tagging SNPs that enable common genetic variation to be interrogated, the genotyped SNPs associated with various diseases are not necessarily the functional variants¹⁴.

Although a formal proof will be required for any SNP associated with a disease to be considered functional, we emphasise that, for the case of non-coding variants, the expression patterns driven in transgenic animals by these genomic regions can yield useful clues, in addition to revealing their regulatory function through potential changes in transcription factor sites. In some cases, the affected genes have a complex spatiotemporal expression pattern and multiple roles in development and differentiation; thus, different SNPs within large regions can affect different roles of the same gene. Other functional assays, such as examination of the behaviour of non-coding regions in cell differentiation assays, will also need to be developed. Elucidation of the functional relationship between conserved non-coding elements and their targets genes may hold a key to understanding the disease mechanism uncovered by the analysis of common genomic variation.

The characterisation of new disease-associated loci may serve as the basis for future approaches to early detection of high-risk individuals. However, the power of genetic associations revealed by whole-genome scans is related as much to the development of new diagnostic tools as it is to the discovery of unforeseen genetic determinants. This knowledge will permit the exploration of novel targets for intervention, because such studies will uncover previously unknown pathways and networks underlying disease.