Final Report Summary - CHEARTED (Gene- Environment interactions in heart development)
Executive summary:
March 2009 saw the launch of the seventh Framework Programme project "CHEARTED". The project was set up to study the biological and genetic effects of the environment on the developing child, focusing on cardiovascular malformation majorly as it is the commonest cause of childhood death in developed countries. In the EU there are 51,000 new cases each year and two million affected individuals. Yet, despite intensive research, the cause of 80% of CVM remains elusive. The mission of CHEARTED was to identify genetic and environmental pathways that can be modified with the goal of reducing preventable CVM incidence. Central to the proposal was to be the development of a bioinformatics tool complemented by an open-access Wiki-based database. The bioinformatics tool would combine sequence data and expression data generated by human and mouse studies with morphology and literature in order to prioritize genes and generate hypotheses.
Project Context and Objectives:
Research Line 1 Genetic Epidemiology
WP1a Collection of clinical samples
Tetralogy of Fallot samples. 979 samples were collected from affected individuals in addition to 262 parental samples. 1120 of these samples (the discovery cohort) were transferred to partner 5 (WP1b). Genotyping (WP1b) and data analysis (WP1c) revealed genetic loci associated with TOF. The standard procedure to assess such putative associations is to genotype single nucleotide polymorphisms at these loci in a second cohort of equivalent size.
WP1b Genome wide SNP analysis
About 1200 samples were received from all Partners involved in patient collection. QC was performed on all these samples. A comparison experiment between the Affymetrix 6.0 and Illumina 660Quad chips on 200 trio families was conducted. Following comparison of these data the remaining samples were genotyped on the Illumina 660Quad chips because of improved resolution of CNV detection. Genotypes underwent in house QC at CNG and data were released to the analysis groups (WP1c). Statistical analysis of the genotype data (WP1c) detected regions worthy of replication. 834 additional samples were received from partners for the replication studies. Following QC steps, 26 SNPs were analysed in all of these samples and results released for statistical analysis (WP1c).
WP1c Statistical Analysis of genotype data
We completed statistical analysis of whole-genome SNP and CNV data in 1733 individuals genotyped as part of WP1b (comprising 913 Tetralogy of Fallot (TOF) cases plus a number of unaffected relatives). We attempted to replicate all SNP association signals that passed a nominal significance level P less than=1x10-5 using an additional cohort of 798 UK, Dutch and French-Canadian TOF cases. In the whole-genome SNP analysis, a region on chromosome 12q24 was associated (P=1.4x10-7) and replicated convincingly (P=3.9x10-5) in 798 cases and 2931 controls (per allele OR=1.27 in replication cohort, P=7.7x10-11 in combined populations). SNPs in the glypican 5 gene (GPC5) on chromosome 13q32 were also associated (P=1.7x10-7) and replicated convincingly (P=1.2x10-5) in 789 cases and 2927 controls (per allele OR=1.31 in replication cohort, P=3.03x10-11 in combined populations). Four additional regions on chromosomes 10, 15 and 16 showed suggestive association accompanied by nominal replication. A paper describing this work has now been published (Cordell et al. (2013) Hum Molec Genet doi: 10.1093/hmg/dds552).
Research line 2 Animal Models
WP2a Mouse model of diabetic pregnancy
The first goal of our work package is the implementation of a mouse model for diabetic pregnancy that produces 20-30% malformed embryos. Initially the mouse models of diabetic pregnancy (streptozotocin-induced maternal diabetes and glucose-supplementation induced fetal hyperglycemia) yielded fewer malformations than anticipated. But changing to the low antioxidants, high fat and low microminerals mouse diet 9F now yields hyperglycemic embryos of which approximately50% are growth-retarded on embryonic day (ED) 8 and approximately10% on ED9 without differences between male and female embryos (identified by PCR for the male-specific Sry sequence). Using this protocol we have now harvested approximately 200 control and approximately 200 hyperglycemic embryos. 0.5-3 μg of total RNA was isolated from 52 individual male control and experimental embryos and processed for deep SAGE sequencing. The remaining RNA is used to determine miRNA composition. Maternal hyperglycemia caused differential gene expression in almost 1,500 genes in the anterior half of the embryo on ED8.5 and in greater than 5,000 genes in the cardiac neural crest and heart on ED9.5. The main difference in hyperglycemia-induced gene expression resided in genes regulating the cytoskeleton on ED8.5 (P less than 1.0E-6) and in oxidative phosphorylation- and cytoskeleton-associated pathways on ED9.5 (P less than 1.0E-10). The differences in expression revealed by microRNA sequencing were small compared to those in the mRNA population, but 7 and 3 microRNAs were differentially expressed on ED8.5 and 9.5 respectively.
WP2b Mouse model of outflow tract defects
To identify new genes and pathways specifically relevant to outflow tract malformations and tetralogy of Fallot using molecular techniques and an Nkx2-5 hypomorphic mouse model of outflow tract malformation. Two types of screens will be performed:
1. ENU screening depends on having a strain sufficiently different from the strain mutated to enable the mapping of mutations in pedigrees established during matings, whilst not directly affecting the extent or penetrance of the phenotype in itself. C57Bl/10 was selected as mapping strain because other possible mapping strains ameliorated the Nkx2-5 hypomorphic phenotype. However, it turned out to be impossible to find the correct mutagenic burden of ENU and this approach was abandoned and interesting mutations will have to rely on sequencing approaches. We have, however, continued to analyse the phenotype of Nkx2-5 hypomorphic mice and have discovered a significant defect in the proepicardium and epicardium, and subsequently the coronary vasculature. Characterisation of this phenotype is in progress.
2. Discovery of targets of cardiac transcription factors in the cardiac HL1 cell line using the DamID technique has initially resulted in data sets for Isl1, Nkx2-5, Gata4, Tbx1, Tbx5, Tbx20 and was later expanded to include SRF, Elk1, Elk4 and a mutant of Elk1 that does not interact with SRF. Data from mutant versions were collected for Nkx2-5. These data allowed us to construct preliminary cardiac networks to seek the logic of cardiac development and discover new transcription factors and other regulators. Our conclusions are that the ubiquitous factors Elk1 and Elk4 are embedded within the cardiac gene regulatory network at an executive level interacting with Nkx2-5 and other cardiac transcription factors. We have characterized a novel protein interaction domain within Nkx2-5 that sustains these interactions and showed that Elk1 and Elk4 are essential for heart development in zebrafish. Morpholino phenotypes resemble those due to knockdown of other kernel cardiac transcription factors such as Nkx2-5 and Tbx5. Because mutations in Nkx2-5 bind to hundreds of target genes and because Elk factors, along with other ubiquitous factors, play a role in directing to mutant forms of Nkx2-5 to off-targets, we hypothesise that off-targets of Nkx2-5 mutants play a role in congenital heart disease pathophysiology and impact on outflow tract phenotypes.
Research line 3 Genetic Bioinformatics
WP3a Genetic Bioinformatics: environmental factors influence on heart development gene networks, and candidate gene prioritization
The main goal of the research of this work package consists of adapting and extending our gene prioritization tool Endeavour to the specific purposes of the CHEARTED project. The first step consisted in building training sets, i.e. sets of genes involved in processes and that serve as seed (models) for the prioritization. Our training sets consist now of 79 genes implied in more than 80 syndromic as well as non-syndromic congenital heart defects when altered or absent. For example, among those genes, 22 are found to be implied in the Tetralogy of Fallot. The hierarchical method which performs multiple prioritizations based on different training sets, set up to increase the precision of the prioritization results, did give results similar to those obtained with the classical prioritization approach.
WP3b Genetic Morphology: 3D gene expression atlas of cardiac development
A comprehensive review on gene expression atlases describing vertebrate developmental stages was carried out. Eleven atlases were analysed and recommendations were formulated on the properties of the ideal gene expression atlas. The latter served as a guideline in the development of the Cardiogenetic Morphology Database (CMD). The morphological framework of this database is based on the generated high resolution reference models. The episcopic imaging method for generating of these models was further developed to determine optimal fixation procedures and to overcome problems caused by the presence of blood. To fill this database a series of 3D reconstructions of 3D gene expression patterns of 13 genes involved in cardiac development at 3 developmental stages was created. To avoid the subjective manual segmentation of gene expression patterns a combination of image analysis, local measurement of staining intensity and statistical clustering was developed. A genetic annotation was performed which is the annotation of cardiac compartments based on the profile of gene expression of individual genes at each location. These compartments only partly follow boundaries indicated by anatomical landmarks. To provide detailed insights into cardiac growth and proliferation rate of myocytes a quantitative 3D atlas of the forming mouse heart and the adjacent splanchnic mesoderm, from the onset of heart formation till late gestation was constructed. The 3D-mapping of the patterns of genes that were identified by the other work packages has started.
Line 4 Management and Dissemination
WP4a Management
Prof. Dr A.F.M. Moorman, was responsible for the internal and external communication, including the (report and financial) obligations to the EU. He was in charge of the Project Management Team and chaired the Board and the Council.
The CHEARTED logo and website were made. The CHEARTED website (see http://www.chearted.eu online) serves as a platform for the communication within the consortium, as well as between the consortium and the scientific and medical community at large. The former function will be served with 'news and views' items as well as the up-to-date listing of personnel and publications. In the near future the CHEARTED website will also contain an area providing information about the aims of the project targeting the general public.
Project meetings, dates and venues
- 4-5 March 2009 Amsterdam Kick-off Meeting and Council Meeting
- 8-9 June 2010 Annual and Council Meetings, Paediatric Hospital of the University of Zurich.
- 18-19 May 2011: Annual Meeting and Council Meetings in Bergen, Norway.
- 22-23 November 2012: Annual Meeting and Council Meetings in Leuven, Belgium.
WP4b Dissemination
The dissemination of the results of the CHEARTED project will be through the CHDwiki (see http://homes.esat.kuleuven.be/~bioiuser/chdwiki/ online) and the Cardiogenetic Morphology Database (CMD: http://cmd.amc.nl). The basic information on the project will be distributed through the CHEARTED website which will also serve as a portal for the other websites.
The CHEARTED website (see http://www.chearted.eu online) contains the basic information on the project and the beneficiaries.
Project Results:
Research Line 1 Genetic Epidemiology
WP1a Collection of clinical samples
1. Prepare samples for Tetralogy of Fallot genome wide association study.
2. Collect and prepare samples for association study of abnormal heart development in offspring of diabetic mothers.
WP1b Genome wide SNP analysis
1. To provide high resolution genotyping with SNP markers and copy number variants (CNVs) for tetralogy of Fallot cases, parental samples and controls.
2. To provide high resolution genotyping with SNP markers and copy number variants (CNVs) for offspring (+/- CVM) of diabetic mothers
3. To provide replication genotyping of additional samples as required.
WP1c Statistical Analysis of genotype data
1. To perform statistical analysis of whole-genome SNP and CNV data in order to detect genetic associations with CVM.
2. To perform statistical analysis of whole-genome SNP and CNV data in order to detect genetic associations with CVM in the presence of maternal diabetes.
3. To test whether the genetic factors involved in CVM are the same or differ depending on the presence of maternal diabetes.
1) Tetralogy of Fallot Genome wide Association study
Newcastle, Leuven, Erlangen and Sydney had begun collecting samples from patients with tetralogy of Fallot prior to CHEARTED. The samples held by the partners were checked for suitability for inclusion in the study. Some were excluded because not all four grand-parents were of Caucasian ethnicity others because there was insufficient DNA or the DNA did not meet the quality criteria. To achieve the aim of analysing 1000 samples we asked collaborators in Nottingham and Oxford to contribute samples. All samples were transferred to partner 5. The quality control (QC) evaluation of DNA samples, fragmentation and amplification and quantification determining the samples to be carried forward to the SNP genotyping excluded only 10 probands leaving 969 probands for the association study.
200 probands were typed on the Affymetrix 6.0 chip and re-genotyped on Illumina 660 quad chip to ascertain the optional platform, particularly with regard to the CNV analysis. While both platforms gave equivalent SNP calls, the resolution of CNV calls using the Illumina platform is greater thus the remaining samples were typed on the Illumina 660 quad chip.
After QC, the genotype data was distributed for statistical analysis groups using the Operon database software developed specifically for this purpose at the CNG.
Stringent quality control (QC) checks to ensure that only high-quality data were included in the final statistical analysis. QC procedures were carried out in the computer program PLINK with visualisation performed in the program R. Multidimensional scaling of our samples together with 210 unrelated Phase II HapMap individuals from four populations (CEU, JPT, CHB, YRI) (genotyped at same set of 41692 autosomal SNPs) was performed and identifed 33 individuals in our study that did not cluster with the CEU samples, suggesting non-European ancestry, these individuals were excluded. We also excluded samples exhibiting the known chromosome 22q deletion associated with TOF.
2) CHD association study in the children of diabetic women
The incidence of CHD is higher in the offspring of diabetic women. The aim of this part of the Workpackage was to collate clinical information on diabetic pregnancies and collect samples from the children, both those with CHD and those without for an association study. Partners 3 and 7 (UNEW and UiB) were tasked with collecting the samples.
UNEW: We had anticipated that collection in the north of England would be straightforward because there were linked diabetic pregnancy and congenital abnormality databases in the north-west and the north-east. However, regulatory processes changed in the period between submission of the CHEARTED application and the project commencing. We were no longer allowed to contact mothers of the 76 children born between 1999 and 2005 with cardiac defects whose mothers had diabetes cases For the north-east patients could still be approached but, instead of contact being made by the database staff, regulatory approvals had to be sought at the hospital at which the woman had received her care and the initial approach made by her consultant. The average time for obtaining regulatory approvals at a hospital was approximately six months. As the number of potential recruits was reduced by losing access to patients identified in the north-west diabetes pregnancy audit we asked additional sites across the UK to join the study. This was facilitated by the study being adopted onto the UK Clinical Research Network portfolio. By the end of the study approvals were in place in 26 English Hospitals and UKCRN and Diabetic Research Network (DRN) funded research nurses were recruiting in these centres. 1 additional UK centre has approval to participate and three further centres are in the process of obtaining regulatory approvals and regulatory approvals are in place to continue UK recruitment to 31/12/2013. We have recruited 36 affected offspring and 270 unaffected offspring along with all relevant clinical information on both children and pregnancy management.
UiB: Applications were submitted to the Regional Ethics Committee, Data Inspectorate, Directorate of Health and Medical Birth Registry of Norway (MBRN) by July 2009. By September, after minor adjustments due to requirements from REC and DI all applications were approved. In October 2009, delivery of data on potential cases and controls was requested from MBRN. All maternity units in Norway have were given about CHEARTED in the periodical Newsletter of MBRN (FØDSELSNYTT) and packs comprising information for the maternity unit, a letter of information for mothers, consent form, a questionnaire to the mother and kits for saliva sampling for distribution to women identified by MBRN. An important difference between UK and Norwegian samples was that whilst medical records were reviewed in the UK for HbA1C measurements etc. in Norway the measure of diabetic control depends on what the mother recorded. Patient information was received from MBRN in May 2010 and letters of invitation sent to 88 diabetic mothers of a child with a CHD (all identified in 1999 -2008) and 251 diabetic mothers of a child without CHD (randomly selected). Reminders were sent to all non-responders. Saliva samples and corresponding questionnaires have been obtained from 23 affected cases along with 13 unaffected siblings and 55 controls.
UMC Utrecht: 3 affected offspring and 106 unaffected offspring samples were transferred to UNEW.
As outlined above we recruited 62 affected cases and 444 unaffected controls. This number does not give adequate power for a statistically sound GWAS study. Rather than continue with an experiment that had no prospect of a meaningful outcome we considered other hypotheses that could be tested with this number of samples. One strong hypothesis is that maternal diabetes alters the epigenetic regulation of developmental genes, which in turn might lead to the CHD observed in the off-spring. We therefore deviated from the original plan to test this hypothesis, using the samples collected to investigate the methylation status of cardiac genes in affected and unaffected offspring. A custom panel of 29 candidate genes was selected based on two complementary approaches; a literature search using combinations of appropriate search terms and analysis of a mouse cardiac gene expression data set GEO GSE1479. Two housekeeping genes ACTC1 and MYLK3 were also included in the panel as 'control' genes. The promoter regions of all these genes were then analysed using CpGIslandExplorer v2.0 (see http://bioinfo.hku.hk/cpgieintro online) to locate CpG islands. A total of 96 CpG sites in these 31 genes were selected for the custom panel. 31 cases and 60 controls were selected for the plate, along with the recommended number of duplicates and standards. DNA was bisulphite modified and then prepared following the Illumina GoldenGate 'Methylation Assay for VeraCode' protocol and scanned on the Illumina BeadXpress. Data analysis is ongoing.
Research line 2 Animal Models
WP2a Mouse model of diabetic pregnancy
1. Catalogue the adaptive changes in gene expression in hyperglycemic embryos and hearts, and define the metabolic and signaling pathways that are specifically affected.
2. Determine the expression pattern of controlling genes in eu- and hyperglycemic embryos.
3. Establish causality for a subset of these genes in existing mutants or in in vitro embryo cultures.
The collection and preparation of embryonic samples for microarray analysis
Animal models for maternal hyperglycemia. We have evaluated a maternal diabetes (streptozotocin treatment, followed by treatment with "Linbit" controlled release insulin implants, so that overt diabetes develops only during pregnancy; Phelan et al., Diabetes. 1997, 46: 1189-1197) and maternal hyperglycemia model (hourly 25% glucose injections on ED7.5; Fine et al., Diabetes 1999, 48: 2454-2462) of diabetic pregnancy to study changes in gene expression in hyperglycemic embryos. Although glucose-injection approach offers the advantage of a short intervention, the frequent injections cause discomfort and are inconvenient.
Streptozotocin (STZ) model. The STZ protocol induced a 3-4-fold increase in blood glucose levels. Glycemic control with Linbit implants allowed the establishment of pregnancy in nearly all animals. Blood glucose at sacrifice was 3-4-fold elevated. However, on the standard SDS CRM(E) diet, too few malformations were seen. Switching to the commonly used 9F diet (Phelan et al., Diabetes. 1997, 46: 1189-1197) solved this problem. The malformed embryos often showed signs of cardiac failure (fluid accumulation in the pericardial cavity and apoptosis of cardiomyocytes). Based on the 9F/STZ protocol, we have collected approximately400 embryos. Male embryos were identified by PCR for the presence of the Sry gene. More importantly, perhaps, no fewer than 50% of the hyperglycemic embryos suffered from a severe growth retardation (approximately 1 day) on ED8.5 which led to their death on ED9.5 and to termination of pregnancy of litters in which greater than20% of the embryos were delayed in development on ED10.5. As a result, developmental delays and a reduction in litter size were no longer observed on ED10.5. The growth-retarding effect of hyperglycemia was similar in female and male embryos.
dSAGE libraries. We analyzed single embryos because of the observed growth retardations in approximately50% of the hyperglycemic embryos. We managed to construct SAGE libraries with approximately5% of the recommended amount of RNA, containing 30-40% identifiable tags, which is standard for dSAGE. A total of 52 single-embryos 3'-end cDNA SAGE libraries from 3 control (ED7.5 8.5 and 9.5) and 4 diabetic (ED8.5 and -9.5 non-retarded and growth-retarded) groups of male embryos were used in next generation SOLiD5 sequencing. We also processed RNA from another 48 embryos from the same control and diabetic groups to determine their microRNA expression profiles. Reproducibility of this procedure was established by repeating the entire procedure on the same starting RNA in a limited number of embryos.
Metabolic and signaling pathway analysis of the microarray data
SOLiD5 reads were mapped against Genome Reference Consortium Mouse Build 38 ("mm10"; December 2011). Reads mapping within exon locations of transcripts and with a NlaIII recognition site were counted and summed per gene. Unsupervised hierarchical clustering of top 100 expressing genes identified the respective biological groups, except that some ED8.5 control embryos mapped with diabetic ED8.5 embryos (even though littermate embryos did map to the control group;). Pathway analysis was performed by MetaCore software. On ED8.5 hyperglycemia induced almost 1,500 differentially expressed genes and greater than100 pathways were significantly regulated, in particular those regulating the cytoskeleton (P less than 10exp-6). On ED9.5 hyperglycemia induced greater than 5,000 differentially expressed genes and greater than200 pathways were highly regulated, with oxidative phosphorylation and cytoskeleton remodeling pathways being affected (P less than 10exp-11). Unsupervised hierarchical clustering of top 100 expressed miRNAs also identified the biological groups except that some ED8.5 control embryos mapped again with diabetic ED8.5 embryos. No differentially expressed microRNAs were identified by Deseq analysis, but the "G-test" identified 7 and 3 miRNAs on ED8.5 and -9.5 respectively, that were differentially affected by hyperglycemia. The "G-test" for classical SAGE analysis, which uses the G-statistic for analysis of groups of frequency distributions (Schaaf, Ruijter, et al. FASEB J. 2005;19:404-406), was adapted to handle the much larger number of tags identified in dSAGE.
Visualization of the expression pattern of controlling genes in eu- and hyperglycemic embryos / Testing relevant mouse mutants for changes in susceptibility to hyperglycemia or testing relevant genes for knock-down analysis in embryo cultures.
We visualized the expression of a few genes thus far (Pax3, Glut4), which was as predicted from literature. In the remaining part of the project (Jan-Aug 2013), we will test the prediction that oxidative phosphorylation is inhibited in diabetic embryos by measuring mitochondrial respiration and glycolysis with the Seahorse XFe Analyzer. In addition, we will explore the distribution of affected cytoskeletal proteins in the diabetic embryos.
Deviations from the project work programme, and corrective actions taken/suggested: identify the nature and the reason for the problem, identify beneficiaries involved.
We have replaced microarray analysis with deep-SAGE sequencing, mainly because it requires less mRNA and because the data obtained are more convenient for quantitative analysis of differential gene expression. The bioinformatics aspect of the project has, nevertheless, taken more time than anticipated. Two independent factors were responsible. Initially, the SAGE tags were sequenced with a SOLiD4 sequencer, which yields only approximately35 nt. Secondly, it turned out that the NlaIII endonuclease had cut only partially. Because this did not compromise the uniqueness of the tags, we adapted the search strategy, which however, meant a recount and reanalysis of all data.
WP2b Mouse Model of outflow tract defects
The objectives were to identify new genes and pathways specifically relevant to outflow tract malformations and Tetralogy of Fallot using molecular techniques and an Nkx2-5 hypomorphic mouse model of outflow tract malformation. Two types of screens will be performed:
1. High specificity/low yield: ENU screening for enhancers and repressors of the Nkx2-5 hypomorphic phenotype.
2. Medium specificity/high yield: microarray and transcription factor target gene analysis to identify genes involved in cardiac progenitor cell identity and behavior, and altered in outflow tract malformation.
Objective 1: High specificity/low yield: ENU screening for enhancers and repressors of the Nkx2-5 hypomorphic phenotype. We previously described using mouse models how proliferative defects in cardiac progenitor cells of the anterior second heart field arise as a result of mutation of the cardiac homeodomain transcription factor NKX2-5 1. In an NKX2-5 hypomorphic model, the lack of proliferative drive in progenitors leads to a spectrum of congenital heart disease-like manifestations, including atrial, ventricular and atrioventricular septal defects, over-riding aorta, double outlet right ventricle and Ebstein's anomaly. These resemble the more severe end of the congenital heart disease spectrum seen in humans carrying NKX2-5 mutations and hypomorphic mice die in the first 2 days after birth. We determined that repression of progenitor cell proliferation was due in part to elevation in the levels of BMP2 at early stages of heart tube formation. Other progenitor genes were also upregulated, leading us to propose that one of the early functions of NKX2-5 as cardiomyocytes are specified and begin to differentiate, is to repress progenitor genes, including those like BMP2 responsible for cardiac induction, thus allowing differentiation to proceed. In NKX2-5 null embryos, cardiomyocytes that do form show a deranged gene regulatory network with co-expression of both progenitor genes and cardiomyocyte differentiation genes. We found in fact that deletion of one allele of Smad1, which acts downstream in the BMP2 pathway, was sufficient to rescue progenitor proliferation and to significantly rescue structural malformations. However, lethality was not rescued and parallel studies indicatred that Smad1-mutant hypomorphs still show abnormalities of chamber growth and gene expression, and defects in epicardial development leading to abnormalities in coronary artery formation.
We found that outcrossing the hypomorphic strain from C57Bl/6 to QSi5, or indeed other common strains, lead to rescue of chamber growth, ventricular septal defects and of lethality. Hypomorphs on a 50:50 C57Bl/6:QSi5 background lived to old age without challenge. Preliminary breeding suggested that only one or a few genes on the C57Bl/6 background interacted with the NKX2-5 pathways to generate the spectrum of hypomorphic phenotypes. We therefore proposed an unbiased forward mutagenesis screen using ENU that might detect activators or suppressors of the NKX2-5 hypomorphic C57Bl/6 phenotype. The screen was set up in collaboration with the Australian Phenome Centre in Canberra, Australia, with costs partially offset from the National Collaborative Research Infrastructure Strategy of the Australian Government. We felt that this would be a low yield but high specificity screen that might detect genes in the NKX2-5 pathway, and those that would buffer mutations in the cardiac network, perhaps associated with the fetal environment. The initiation of the ENU screen was delayed because of troubles encountered identifying a mapping strain. The project depended on having a strain sufficiently different from the genetically modified strain so as to be able to map discovered mutations in pedigrees established during breeding, whilst not directly affecting the extent or penetrance of the phenotype itself. In short, we screened C3H, 129, CBA, FVB/N, BALBC, C57Bl/10 and LED, and all strains (except for C57Bl/10) show modifications of the phenotype, in most cases a degree of rescue.
This in fact gave us reassurance that there were specific alleles in multiple inbred strains of mice that can rescue the phenotype. The optimal ENU burden for C57Bl/10 had not been defined and when using the same conditions as used for C5Bl/6 mice, all C57Bl/10 males were sterile. This approach was abandoned and we did not consider trying to find a mapping strain further. We decided to rely instead on deep sequencing approaches and since this screen was started the APC has developed a robust pipeline for ENU mutation detection by whole genome sequencing. We went ahead and mutagenised C57Bl/6 studs to initiate the screen formally. The strategy was to mate ENU treated wt males with Nkx2-5-GFP heterozygous females, then to dissect N2 litters at 2 days after birth and screen for GFP-positive pups. Any pedigrees with positive pups at the time of dissection would be regarded as putative mutants, and used to produce subsequent litters by mating to Nkx2-5IRESCre mice. Further delays were encountered since the screen was set to be either too stringent (assessment for survivors at 2 weeks after birth) or too relaxed (assessment at 2 days), that latter showing some putative mutants that could not be replicated with further breeding. After setting the assessment time at 4 days after birth, we screened 148 pedigrees over 2 years without seeing putative mutations. Unfortunately, while by no means saturating, we had to abandon the screen because of lack of leads and lack of on-going funding from our Australian grants. While the screen was negative, we did learn a lot about the ENU technology and may return to a less stringent/higher yield screen in the future. As we further characterise the NKX2-5 hypomorphic strain, we use outbreeding as a tool to look at possible causes of dysfunction and death.
Objective 2: Medium specificity/high yield: microarray and transcription factor target gene analysis to identify genes involved in cardiac progenitor cell identity and behavior, and altered in outflow tract malformation. While we have not made strong progress on the first part of this aim microarray analysis of our NKX2-5 genetic model of outflow tract malformation, we have made a number of key discoveries related to gene pathways involved in outflow tract development. Mesp1 is a key transcription factor that is expressed in early cardiac and head fields and is absolutely required for the development of the heart. In a collaboration with Cedric Blanpain (Belgium), we showed that MESP1-GFP cells derived from ES cell cultures are enriched in mulitpotent heart progenitors and that transcriptional profiling of these cells uncovered new markers that allow their prospective isolation form ES cells and embryos 2. In this study, we also explored the interaction between MESP1 and ISL1, encoding another important cardiac transcription factor and results suggest that they interact to promote both cardiomyocyte and endothelial cell differentiation. In the NKX2-5 null model, ISL1 expression is retained in the differentiating myocardium where it is normally turned off. In collaboration with Margaret Buckingham (France), we elucidated a mechanism for how NKX2-5 might normally repress ISL1-mediated gene expression. Using FGF10 as an NKX2-5 target and up-regulated in NKX2-5 null myocardium, we showed that ISL1 and NKX2-5 share common binding sites within the second heart field enhancer of the FGF10 gene and these factors compete for binding and activation versus repression 3.
Other insights into cardiac developmental pathways include description of a genetic interaction between NKX2-5 and a gene encoding a component of the BRG1 chromatin-remodelling complex 4; demonstration of a role for Neuregulin 1 in maintaining the cardiac gene regulatory network in both trabecular and non-trabecular myocardium 5; role for NKX2-5 in suppression of blood development via GATA1 6; and recognition that the endocardium identified as having expressed NKX2-5 in its lineage history transiently retains hemogenic activity 7.
A major project under this aim has been to document the target genes of cardiac transcription factors using the DamID technique in the HL1 cardiomyocyte cell line, and to probe the mechanisms of congenital heart disease. This has been highly successful. Overall, we have defined target sets of NKX2-5, GATA4, ISL1, TBX1, TBX2, TBX5, TBX20, HAND1 and MEF2c, as well as mutants of NKX2-5, and we have recently expanded this initial screen to include SRF, ELK1, ELK4 and a mutant of ELK1 that does not interact with SRF. It has been necessary in our study to consider targets of both NKX2-5-DAM (C-terminal fusion) and DAM-NKX2-5 (N-terminal fusion) as steric influences become evident for some transcription factors, particularly when mutated.
We are principally interested in how cardiac transcription factors influence network logic and have made a number of key discoveries:
1) We have documented 757 high confidence NKX2-5 target genes in HL-1 cells covering a wide range of biological functions;
2) Contrary to expectation, mutations of NKX2-5, including a full homeodomain deletion, also bind hundreds of targets, some overlapping NXK2-5 WT targets but some unique, representing "off-targets";
3) NKX2-5 mutants can be directed to genuine NKX2-5 targets as well as off-targets via altered DNA-binding specificity and by association with cofactors;
4) A novel cofactor interface within NKX2-5 has been defined that facilitates homodimerisation, heterodimerisation with NKX2-5 mutants, and interactions with a range of cardiac-restricted and ubiquitous cofactors;
5) ELK proteins (ELK1 and ELK4) are ubiquitously expressed transcription factors that interact with NKX2-5 and are recursively wired into the early cardiac gene regulatory network;
6) the role of ELK1 and ELK4 in the cardiac gene regulatory network has been confirmed by gene knockdown studies in zebrafish, and by defining the targets of ELK1 and ELK4, was well as SRF which can associate with ELK proteins in some contexts, using DamID.
Research line 3 Genetic Bioinformatics
Workpackage 3a: Genetic Bioinformatics: environmental factors influence on heart development gene networks, and candidate gene prioritization
1. Inference of gene networks for normal and abnormal heart development
2. Inference of models for the environmental factors and their influence on heart development
3. Development of a bioinformatics tool for congenital heart defect candidate genes prioritization
The main goal of the research effort of this work package consisted of adapting, extending and applying the gene prioritization tool named as Endeavour (Aerts et al, 2006, Tranchevent et al, 2008) to the specific purposes of the CHEARTED project (i.e. applying it to congenital heart defects). As a reminder, Endeavour is a software application for the computational prioritization of candidates' genes. Endeavour is able to rank genes according to their similarity (considering various biological criteria such as sequence homology, GO classification, physical interaction) to a set of training genes (in our case, genes implied in the same congenital heart disease or in the same developmental process). This strategy is highly valuable considering the large sets of candidate genes that will be produced by the other partners in the project.
Building heart specific training sets
The first step consisted in building training sets, i.e. sets of genes involved in processes and that serve as seed (models) for the prioritization. After an extensive literature search (mainly performed by Drs Jeroen Breckpot and Bernard Thienpont) and the two year contribution of CHDWiki users, this collaborative web portal hosts' information about 79 syndromic genes, 274 associations between these genes and congenital heart defects as well as other relative information.
We also integrated the data produced by WP2.a where deep SAGE sequencing experiments were performed on embryos collected during pregnancy in a diabetic mouse model. On this new dataset we applied our new hierarchical prioritization method. The derived results and created datasets are hosted by CHDWiki.
Building environmental factors training sets and Build models specific to heart development
First, we developed a novel text-mining approach (called Beegle) that takes simple "PubMed queries" as input and returns a list of genes linked to that query. This approach was of primary interest to identify a list of genes and factors linked to congenital heart defects. This tool is accessible at the following URL: http://homes.esat.kuleuven.be/~sbrohee/beegle-new/. The tool works for any query (e.g. ranging from "Tetralogy of Fallot" to "apoptosis") and is based on the similarity between the abstracts returned by the query in Pubmed to the abstracts of the paper describing the genes in the GeneRIF database. The purpose of this tool is to be as simple as a "Google for genes" and we deliberately mimicked the display of Google when we implemented the online version of Beegle. The genes delivered by the Beegle text-mining approach can then be used as seeds for the PINTA tool (Nitch et al. 2011, Nucleic Acid Res, 39: 334-338), another tool developed by our lab. In short, PINTA looks at the neighbors of some seed genes in a functional interaction network (such as in the STRING database) to predict new candidates genes.
Set up a hierarchical prioritization method
In the beginning of the CHEARTED project, we tried to improve the, already satisfactory, performances of our Endeavour gene prioritization tool, which is available at our CHDWiki pages, proposing built-in training genes for various heart developmental processes. The idea was to combine various data sources according to the studied disease or process. Indeed, some data sources might be very redundant (e.g. several protein-protein interaction or homology databases) and biased toward already known information (like Gene Ontology) while some are not. The output of each (sub-) prioritization step would then be used together for the next prioritization step, using both biased and unbiased sources without giving a higher weight. Using this strategy, we hoped to build a tree of data sources. The obtained results when applying this hierarchical prioritization approach were somewhat unsatisfactory. Firstly, the performances were comparable to those obtained when no hierarchical prioritization was applied. Secondly, the algorithm was much slower (as many more prioritization steps were performed).
Prioritize candidates and analyze results
For the purposes of this project, two different candidate gene lists were created by two different prioritization methods and they are available at:
(see http://homes.esat.kuleuven.be/~bioiuser/chdwiki/index.php/CHD:Genes online). The first list was produced by Endeavour tool, which was integrated into CHDWiki. The second list was produced by the new hierarchical gene prioritization method that uses 4 prioritization tools to predict novel CHD candidate genes and the gene list produced by WP2.a as a training set. The users can further analyse these results by viewing separately each gene page where other kind of information are also offered. Moreover, CHDwiki integrates information from the major protein interaction databases (STRING, Biogrid, Intact, Mint, etc.) such as the list of interactions for each syndromic, non-syndromic or candidate gene contained in the CHDWiki gene lists. Finally, a viewer allows the users to navigate through the network of physical or functional interactions of the gene of their interest.
Build networks of genes (best candidates) involved in heart development and interactions with environmental factors.
The CHDwiki offers a visualization module that associates the candidate genes with congenital heart defects, their frequency as well as the physical protein-protein interactions as it can be seen at: http://homes.esat.kuleuven.be/~bioiuser/chdwiki/index.php/Chd_gene_networks. After doing this we could observe that genes sharing multiple phenotypes when mutated tend to act in the same molecular pathways or even encode proteins that directly interact.
Deviations from the project work programme, and corrective actions taken/suggested: identify the nature and the reason for the problem, identify Beneficiaries involved.
Workpackage 3b: Genetic Morphology: 3D gene expression atlas of cardiac development
1. Generate a Cardiogenetic Morphology Database (CMD) based on a standard series of 3D reconstructions of heart development, the expression patterns of 12 key marker genes and genetic annotation of cardiac compartments.
2. Develop an application to help researchers to automatically match (series) of individual cardiac sections to the reference series, to add gene expression information from these sections to the CMD and to display the association of the gene expression patterns with patterns of genes contained in the CMD.
3. Map novel genes involved in cardiac development into the CMD, and test hypotheses on their role in the signaling pathways.
The goal of WP3b was to establish a three-dimensional (3D) gene expression atlas of cardiac development. The work was distributed over 4 tasks. The first 3 tasks cover the generation of a series of high resolution reference models of cardiac development, the development of an application to match individual section to the 3D space of a reference model of the appropriate age, and the creation of a series of 3D reconstructions displaying the gene expression patterns of key genes in cardiac development. The fourth task was to map the 3D expression patterns of novel genes, identified by the other workpackages as being involved in cardiac development and affected by diabetic pregnancy.
Generate reference series of developing mouse hearts at 9.5 11.5 and 13.5 days of development
The basis of the 3D atlas is formed by a series of high resolution models of the embryonic mouse heart. These models were obtained by high resolution episcopic microscopy (HREM) at the MRC. Successive issues had to be addressed to reach the required results.
1) Whole embryos versus dissected hearts
Whole embryos have the obvious merit that they retain the anatomical relationships between the heart, great vessels and other organs in the thoracic cavity. However, it is extremely difficult to remove blood from whole embryos. Dissected hearts necessarily lose the anatomical context and run the danger of morphological distortion as a result of the dissection. Both kind of data sets were compared.
2) Exsanguination procedures
Removal of blood from mouse embryos required a non-invasive procedure to avoid damaging the heart itself. Embryo perfusion is only realistic at late stages (embryonic day (E) 14.5 and older). For younger stages prolonged maintenance of embryos at 37C with repeated removal of blood clots from severed vessels proved to be effective.
3) Staging of embryos
Different embryos within a litter are variable in developmental stage. Since developmental stage is evidently important in a study of morphological development, alternative "accurate" staging methods have been compared. From these we concluded that the most effective approach is to image a large number of samples for each apparent developmental stage and to order these based entirely on the established sequence of changes in cardiac morphology.
4) Image quality
Fluorescence of HREM tends to fade out when tissue is close to the surface of the block which results in apparent thickening of the tissue in the sectioning direction. To reduce this effect, an image filter was developed to remove this out-of-plane information.
Develop computer application that TRaces the Anatomical Context of single Tissue Sections (TRACTS).
The rapid morphological change of the developing heart makes it difficult to interpret single sections through such a heart. We developed a program (TRACTS) which fits arbitrary histological sections into a high resolution (HREM) 3D heart model. We implemented a straightforward brute force pixel-based approach, where the (2D) input section is compared to a selection of virtual (2D) cross sections of the (3D) reference model. Model cross sections are selected on basis of four image features: size, overall density, regional density and center of mass. The performance of TRACTS is assessed using virtual cross-sections in arbitrary directions from a second HREM dataset and using paraffin embedded material. A test, involving a panel of morphological experts who manually fitted a sample of sections to the reference model test showed that the program fits just as well as the best fits found by five morphological experts. The initial version, based on a reference model of E11.5 was extended to handle a reference model series of E9.5 E11.5 and E13.5.
Generate a timed series of 3D reconstructions of key marker genes
To understand the relation between genes and morphogenesis one first needs to know which genes are expressed in which parts of the developing organ. The widespread application of high throughput gene expression analysis platforms has lead to the identification of thousands of mammalian genes and the estimation of their activity in dozens of organs and parts of organs. However, virtually all such expression analyses are carried out on organ homogenates in which spatial information is lost. Only 3-dimensional (3D) reconstruction of specifically stained sections can provide a complete spatial expression pattern. The detailed gene expression information in 2D images or 3D reconstructions can be made accessible by combining such 3D reconstructions into a common reference model.
Line 4 Management and Dissemination
Workpackage 4a: Management
1. The management activities should be instrumental for a proper and timely implementation of the activities.
2. The management activities will secure the coordination and integration between the other activities in order to reach synergistic effects.
3. The starting point is that a project needs a clear and distinct description of roles, responsibilities and tasks also in order to allow for a proper accountability along the (financial) guidelines of the European Commission, the partner organizations and the principles of accountancy.
Consortium management tasks and achievements
Prof. Dr A.F.M. Moorman, was responsible for the internal and external communication, including the (report and financial) obligations to the EU. He was in charge of the Project Management Team and chaired the Board and the Council.
The CHEARTED logo and website (see http://www.CHEARTED.eu online) were designed by Dr Alexandre Soufan. The CHEARTED website (see http://www.CHEARTED.eu online) serves as a platform for the communication within the consortium, as well as between the consortium and the scientific and medical community at large. The former function will be served with 'news and views' items as well as the up-to-date listing of personnel and publications. In the near future the CHEARTED website will also contain an area providing information about the aims of the project targeting the general public.
List of project meetings, dates and venues
- 4-5 March 2009 Amsterdam Kick-off Meeting and Council Meeting
- Workshops 2009: 24 April Meeting UNW and UiB in Bergen and 5 October Workshop WP3a. WP3b and WP4b in Leuven
- 8-9 June 2010 Annual and Council Meetings, Paediatric Hospital of the University of Zurich. Invited speakers: Dr Mary Loeken from the Joslin Diabetes Center, Harvard and ESAC member Dr Ed Lammer from Children's hospital and Research Center Oakland, Calilifornia. ESAC member Prof Dietrich Rebholz-Schuhmann was also present
- 18-19 May 2011: Annual Meeting and Council Meetings in Bergen, Norway. Invited speaker: Prof. Rolv Skjærven, from the Medical Birth Registry of Norway, Norwegian Institute of Public Health Bergen, Norway. ESAC member Dr Ed Lammer was present
- 22-23 November 2012: Annual Meeting and Council Meetings in Leuven, Belgium Invited speakers: Prof. Marc-Philip Hitz (Sanger Institute, Cambridge, UK) and Dr Phil Barnett (Dept Anatomy, Embryology and Physiology, AMC, Amsterdam, NL). Two ESAC members were present: Dr Ed Lammer (Berkley, CA, USA) and Prof Cornelia van Duijn, Rotterdam, NL.
Problems which have occurred and how they were solved or envisaged solutions
In Research Line 1 the number of children with heart defects born to diabetic mothers is less than anticipated and measures have been taken to analyse the result in a different way. An association study is not possible due to insufficient numbers, but an epigenetic study was performed instead
Workpackage 4b: Dissemination
1. Facilitate the exchange of the gene association, transcriptional network and genetic morphology results of CHEARTED both between the Beneficiaries of the project, with the scientific community and with the public.
2. Enable the contribution of third parties to the databases developed within CHEARTED.
3. New Bulletins
To integrate and share knowledge on heart development, CHEARTED wishes to construct an interactive database, which integrates knowledge on different domains such as genetics, environmental factors, morphogenesis, gene expression etc
Cardiogenetics Knowledge Database
The CHDwiki website (see http://homes.esat.kuleuven.be/~bioiuser/chdwiki online) has been implemented as a collaborative resource. CHDwiki is a user-friendly knowledge base portal aimed at mapping genes involved in CHDs and understanding their relations with corresponding human phenotypes. CHDwiki is a collaborative resource, meaning that every registered user can add, delete or modify the data in the database. The database is described in Barriot R, Breckpot J, Thienpont B, Brohée S, Van Vooren S, Coessens B, Tranchevent LC, Van Loo P, Gewillig M, Devriendt K, Moreau Y. Collaboratively charting the gene-to-phenotype network of human congenital heart defects. Genome
Med. 2010 Mar 1;2(3):16. There are currently 163 registered users. The results of the knowledge acquisition are shown below (as of December 2012):
Features entries Features CHD
Genes 79 Genes 274
CHD types 113 Mutations 367
linkage regions 15 linkage regions 19
balanced translocations 25 balanced translocations 27
imbalances (cases) 374 Van Karnebeek 1999 regions 85
imbalances (regions) 410 Studies (mutation screens) 428
Van Karnebeek 1999 regions 46
References 253
Results of knowledge acquisition. The right column displays the number of different features present in CHDWiki (e.g. genes, CHD types, imbalances). The left column displays the total number of relations between Congenital Heart Defects (CHD) and currently managed features.
CHDWiki portal provides a collaborative knowledge base where the current knowledge on the genetic basis of CHDs is stored. Relationships between genes and CHDs have been mapped, derived from published data or single clinical cases making this platform a useful tool for geneticists, molecular biologists and clinicians. The genomic variations that are recorded in CHDWiki create relationships between associated genes, congenital heart defects and clinical reports. Lists of syndromic or nonsyndromic genes are available to the users while two different lists of candidate genes can be found produced by Endeavour and a newly introduced hierarchical gene prioritization method.
Also in this specialized portal, various modules are hosted such as:
- An integrated version of the Endeavour tool in order to perform genome wide gene prioritization. This software application allows the bioinformatic prioritization of candidates genes, based on a set of training genes. Different lists of training genes concerning different parts of the heart are defined. Endeavour was used to identify two CHD candidate genes, (Breckpot J, et al. K. Congenital heart defects in a novel recurrent 22q11.2 deletion harboring the genes CRKL and MAPK1. Am J Med Genet A. 2012 Mar;158A(3):574-80. and Breckpot J, et al. BMPR1A is a candidate gene for congenital heart defects associated with the recurrent 10q22q23 deletion syndrome. Eur J Med Genet. 2012 Jan;55(1):12-6).
- Besides Endeavour, we integrated two other text mining tools that establish direct links between genes and CHD and vice versa: Huge Navigator and AgeneApart. For each protein coded by a CHD implied gene, CHDwiki returns the list of physical as well as functional interactions in a network visualizer.
- A module that enable users to visualize the physical or the functional interactions of the gene of their interest.
- A map module, which provides a karyogram with all regions and genes annotated in the CHDWiki.
Cardiogenetics Morphology Database
The Cardiogenetic Morphology Database (CMD) was developed to hold 2D and 3D gene expression patterns of 3 stages of mouse heart development. The CMD website (see http://cmd.amc.nl/ online) has been set up to enable end-users, researchers in the field of embryology, to submit images of single sections in which a gene expression pattern is visualized and to query the database for (non-)overlapping genes. To this end, a semi-automatic pipeline was created to upload, process, and store the submitted images.
CHEARTED website
The CHEARTED website (see http://www.CHEARTED.eu online) was based on the website created for the FP6 project HeartRepair (see http://www.heartrepair.eu online). The website contains functionalities like a private domain, public access level, factual information, publications and contact list. The content of the site is managed by the dissemination manager (beneficiary 11). The CHEARTED website contains a link to the CHDwiki maintained by the partner K.U.-Leuven and thus
provides access to this data source.
News bulletins, Brochures
After a delayed start up in the first year, 8 News Bulletins and 2 flyers have been produced since January 2010. A final, concluding News Bulletin will be published in March 2013.
News Bulletins and flyers were distributed in a printed version and as PDF document to
- all researchers related to the CHEARTED project
- scientist and medical specialists in the field of congenital heart development and congenital heart diseases.
- the European heart foundations (European Heart Network (EHN))
The News bulletins give a nice overview of the research activities of the Workpackages:
- In the 1st news bulletin professors Koenraad Devriendt and Yves Moreau from the University of Leuven gave an introduction to the CHDwiki (workpackage 4b).
- In the 2nd news bulletin Professor Judith Goodship from Newcastle explained GWAS to identify genes involved in the occurrence of Tetralogy of Fallot as well as in the mechanism by which maternal diabetes affects cardiac development (workpackage 1a).
- In the 3rd news bulletin dr. Jan Ruijter from the Amsterdam Medical Center presented the work on the development of methods to visualize genetic networks regulating cardiac development (workpackage 3B). Furthermore, a short report of the first annual meeting was given by dr. Dietrich Rebholtz-Schumann, member of the External Scientific Advisory Committee of CHEARTED.
- The 4th news bulletin was dedicated to the statistic analyses of genetic data by the group of professor Heather Cordell from Newcastle (workpackage 1c). Professor Wout Lamers from Amsterdam described experiments on a mouse model for cardiac malformations caused by diabetes during pregnancy (workpackage 2a).
- The 5th news bulletin presented the second CHEARTED annual meeting, which was held in Bergen, Norway on May 18-19, 2011. Also, there is an extended reference to the national medical birth registry of Norway and to the first CHEARTED dissertations, successfully defended by dr. Bouke de Boer of the AMC and dr. Jeroen Breckpot of the KU Leuven.
- The 6th news bulletin presented an overview over the work of Dr. Bouke de Boer (workpackage 3b) and Dr. Jeroen Breckpot (workpackage 4b).
- The 7th news bulletin is dedicated to Professor Richard Harvey's group which played an important role in Workpackage 2b of the CHEARTED project.
- The 8th news bulletin included an extensive article by Professor Gerard H.A. Visser (University Medical Centre Utrecht, The Netherlands) about the role of diabetes in heart development.
Potential Impact:
We expected that the knowledge gained from our research would lead to an improvement in the prevention and treatment of cardiovascular diseases, which are a major cause of ill health and death in Europe and worldwide. Our research focused on mechanisms of normal and abnormal heart development leading to congenital heart diseases. This included studies on genetic basis and the interaction between genes and environment in the development of heart defects and built on existing epidemiological data. As we stated in the Annex I, the data we generated would not have direct clinical applications but would make a significant contribution to the understanding of CVM and the role of environment during fetal development. Moreover, to carry out this research, and to make the results accessible, we had to develop a number of informatics tools and databases that themselves would be of potential scientific and public value.
Fundamentally, the ultimate goal of CHEARTED was to reduce the of incidence of what could be considered preventable cardiovascular malformations. Environmental and developmental issues are presently trapped in a bottleneck. A wealth of information regarding cardiovascular development, and the factors which may potentially affect its course, is already available and accessible. Despite this, the cause for more than 80% of cardiovascular malformations (CVM) remains elusive. Only now, due to the vast improvements in recent years in genome wide association and molecular analysis techniques, is it possible to unravel complex diseases such as cardiovascular malformation (CVM). Indeed, recently we have seen the identification by genome-wide association studies of susceptibility alleles in conditions (such as coronary heart disease) that are known to be far less heritable than CVM. Thus, it is was timely to give a new and devoted impulse to collect, identify, analyze and describe etiological risk factors for CVM. A major aim of research line 1 was to identify genotypes which affect the metabolism or bioavailability of specific nutrients and ultimately affect heart development and to provide a real indication of the total contribution of common genetic variation to cardiovascular malformation. At the end of the last reporting period, we completed the statistical analysis of whole-genome SNP and CNV data in 1733 genotyped individuals, comprising 913 Tetralogy of Fallot (ToF) cases plus a number of unaffected relatives. Two loci were identified on chromosomes 12 and 13 that were statistically strongly associated with the risk of ToF. This study was published in a high impact journal and is, worldwide, the first genome-wide association study of any congenital heart malformation phenotype. In addition, work on the CNV data in ToF cases identified an excess of rare genic deletions, moreover it was found that both duplication and deletion of 1q21.1 are more common in ToF cases than in controls.
Two papers describing this work have now been published. The potential impact of this combined knowledge is enormous, but at the same time daunting. In general we can state that CHEARTED had an enormous impact as it was world-wide the first to uncover common variation underlying Tetralogy of Fallot. If we look in detail at the results from the GWAS study, we identified two loci in which multiple genes reside. We now know that these genes somehow result in ToF through an (as yet) unknown mechanism. One potential insight is the fact that one of the genes in these loci is the PTPN11 gene, a protein-tyrosine phosphatase, known to be causative in Noonans syndrome, a Mendelian multisystem disorder in which malformation of the cardiac outflow tract is a typical feature. The precise mechanism how variations in these loci underlie the development of Tetralogy of Fallot is subject of future research and forms a large part of the impact CHEARTED has had on European science.
List of Websites:
http://www.CHEARTED.eu
http://homes.esat.kuleuven.be/~sbrohee/beegle-new/
http://homes.esat.kuleuven.be/~bioiuser/chdwiki/index.php/CHD:Genes
http://cmd.amc.nl ; http://downloads.hfrc.nl
March 2009 saw the launch of the seventh Framework Programme project "CHEARTED". The project was set up to study the biological and genetic effects of the environment on the developing child, focusing on cardiovascular malformation majorly as it is the commonest cause of childhood death in developed countries. In the EU there are 51,000 new cases each year and two million affected individuals. Yet, despite intensive research, the cause of 80% of CVM remains elusive. The mission of CHEARTED was to identify genetic and environmental pathways that can be modified with the goal of reducing preventable CVM incidence. Central to the proposal was to be the development of a bioinformatics tool complemented by an open-access Wiki-based database. The bioinformatics tool would combine sequence data and expression data generated by human and mouse studies with morphology and literature in order to prioritize genes and generate hypotheses.
Project Context and Objectives:
Research Line 1 Genetic Epidemiology
WP1a Collection of clinical samples
Tetralogy of Fallot samples. 979 samples were collected from affected individuals in addition to 262 parental samples. 1120 of these samples (the discovery cohort) were transferred to partner 5 (WP1b). Genotyping (WP1b) and data analysis (WP1c) revealed genetic loci associated with TOF. The standard procedure to assess such putative associations is to genotype single nucleotide polymorphisms at these loci in a second cohort of equivalent size.
WP1b Genome wide SNP analysis
About 1200 samples were received from all Partners involved in patient collection. QC was performed on all these samples. A comparison experiment between the Affymetrix 6.0 and Illumina 660Quad chips on 200 trio families was conducted. Following comparison of these data the remaining samples were genotyped on the Illumina 660Quad chips because of improved resolution of CNV detection. Genotypes underwent in house QC at CNG and data were released to the analysis groups (WP1c). Statistical analysis of the genotype data (WP1c) detected regions worthy of replication. 834 additional samples were received from partners for the replication studies. Following QC steps, 26 SNPs were analysed in all of these samples and results released for statistical analysis (WP1c).
WP1c Statistical Analysis of genotype data
We completed statistical analysis of whole-genome SNP and CNV data in 1733 individuals genotyped as part of WP1b (comprising 913 Tetralogy of Fallot (TOF) cases plus a number of unaffected relatives). We attempted to replicate all SNP association signals that passed a nominal significance level P less than=1x10-5 using an additional cohort of 798 UK, Dutch and French-Canadian TOF cases. In the whole-genome SNP analysis, a region on chromosome 12q24 was associated (P=1.4x10-7) and replicated convincingly (P=3.9x10-5) in 798 cases and 2931 controls (per allele OR=1.27 in replication cohort, P=7.7x10-11 in combined populations). SNPs in the glypican 5 gene (GPC5) on chromosome 13q32 were also associated (P=1.7x10-7) and replicated convincingly (P=1.2x10-5) in 789 cases and 2927 controls (per allele OR=1.31 in replication cohort, P=3.03x10-11 in combined populations). Four additional regions on chromosomes 10, 15 and 16 showed suggestive association accompanied by nominal replication. A paper describing this work has now been published (Cordell et al. (2013) Hum Molec Genet doi: 10.1093/hmg/dds552).
Research line 2 Animal Models
WP2a Mouse model of diabetic pregnancy
The first goal of our work package is the implementation of a mouse model for diabetic pregnancy that produces 20-30% malformed embryos. Initially the mouse models of diabetic pregnancy (streptozotocin-induced maternal diabetes and glucose-supplementation induced fetal hyperglycemia) yielded fewer malformations than anticipated. But changing to the low antioxidants, high fat and low microminerals mouse diet 9F now yields hyperglycemic embryos of which approximately50% are growth-retarded on embryonic day (ED) 8 and approximately10% on ED9 without differences between male and female embryos (identified by PCR for the male-specific Sry sequence). Using this protocol we have now harvested approximately 200 control and approximately 200 hyperglycemic embryos. 0.5-3 μg of total RNA was isolated from 52 individual male control and experimental embryos and processed for deep SAGE sequencing. The remaining RNA is used to determine miRNA composition. Maternal hyperglycemia caused differential gene expression in almost 1,500 genes in the anterior half of the embryo on ED8.5 and in greater than 5,000 genes in the cardiac neural crest and heart on ED9.5. The main difference in hyperglycemia-induced gene expression resided in genes regulating the cytoskeleton on ED8.5 (P less than 1.0E-6) and in oxidative phosphorylation- and cytoskeleton-associated pathways on ED9.5 (P less than 1.0E-10). The differences in expression revealed by microRNA sequencing were small compared to those in the mRNA population, but 7 and 3 microRNAs were differentially expressed on ED8.5 and 9.5 respectively.
WP2b Mouse model of outflow tract defects
To identify new genes and pathways specifically relevant to outflow tract malformations and tetralogy of Fallot using molecular techniques and an Nkx2-5 hypomorphic mouse model of outflow tract malformation. Two types of screens will be performed:
1. ENU screening depends on having a strain sufficiently different from the strain mutated to enable the mapping of mutations in pedigrees established during matings, whilst not directly affecting the extent or penetrance of the phenotype in itself. C57Bl/10 was selected as mapping strain because other possible mapping strains ameliorated the Nkx2-5 hypomorphic phenotype. However, it turned out to be impossible to find the correct mutagenic burden of ENU and this approach was abandoned and interesting mutations will have to rely on sequencing approaches. We have, however, continued to analyse the phenotype of Nkx2-5 hypomorphic mice and have discovered a significant defect in the proepicardium and epicardium, and subsequently the coronary vasculature. Characterisation of this phenotype is in progress.
2. Discovery of targets of cardiac transcription factors in the cardiac HL1 cell line using the DamID technique has initially resulted in data sets for Isl1, Nkx2-5, Gata4, Tbx1, Tbx5, Tbx20 and was later expanded to include SRF, Elk1, Elk4 and a mutant of Elk1 that does not interact with SRF. Data from mutant versions were collected for Nkx2-5. These data allowed us to construct preliminary cardiac networks to seek the logic of cardiac development and discover new transcription factors and other regulators. Our conclusions are that the ubiquitous factors Elk1 and Elk4 are embedded within the cardiac gene regulatory network at an executive level interacting with Nkx2-5 and other cardiac transcription factors. We have characterized a novel protein interaction domain within Nkx2-5 that sustains these interactions and showed that Elk1 and Elk4 are essential for heart development in zebrafish. Morpholino phenotypes resemble those due to knockdown of other kernel cardiac transcription factors such as Nkx2-5 and Tbx5. Because mutations in Nkx2-5 bind to hundreds of target genes and because Elk factors, along with other ubiquitous factors, play a role in directing to mutant forms of Nkx2-5 to off-targets, we hypothesise that off-targets of Nkx2-5 mutants play a role in congenital heart disease pathophysiology and impact on outflow tract phenotypes.
Research line 3 Genetic Bioinformatics
WP3a Genetic Bioinformatics: environmental factors influence on heart development gene networks, and candidate gene prioritization
The main goal of the research of this work package consists of adapting and extending our gene prioritization tool Endeavour to the specific purposes of the CHEARTED project. The first step consisted in building training sets, i.e. sets of genes involved in processes and that serve as seed (models) for the prioritization. Our training sets consist now of 79 genes implied in more than 80 syndromic as well as non-syndromic congenital heart defects when altered or absent. For example, among those genes, 22 are found to be implied in the Tetralogy of Fallot. The hierarchical method which performs multiple prioritizations based on different training sets, set up to increase the precision of the prioritization results, did give results similar to those obtained with the classical prioritization approach.
WP3b Genetic Morphology: 3D gene expression atlas of cardiac development
A comprehensive review on gene expression atlases describing vertebrate developmental stages was carried out. Eleven atlases were analysed and recommendations were formulated on the properties of the ideal gene expression atlas. The latter served as a guideline in the development of the Cardiogenetic Morphology Database (CMD). The morphological framework of this database is based on the generated high resolution reference models. The episcopic imaging method for generating of these models was further developed to determine optimal fixation procedures and to overcome problems caused by the presence of blood. To fill this database a series of 3D reconstructions of 3D gene expression patterns of 13 genes involved in cardiac development at 3 developmental stages was created. To avoid the subjective manual segmentation of gene expression patterns a combination of image analysis, local measurement of staining intensity and statistical clustering was developed. A genetic annotation was performed which is the annotation of cardiac compartments based on the profile of gene expression of individual genes at each location. These compartments only partly follow boundaries indicated by anatomical landmarks. To provide detailed insights into cardiac growth and proliferation rate of myocytes a quantitative 3D atlas of the forming mouse heart and the adjacent splanchnic mesoderm, from the onset of heart formation till late gestation was constructed. The 3D-mapping of the patterns of genes that were identified by the other work packages has started.
Line 4 Management and Dissemination
WP4a Management
Prof. Dr A.F.M. Moorman, was responsible for the internal and external communication, including the (report and financial) obligations to the EU. He was in charge of the Project Management Team and chaired the Board and the Council.
The CHEARTED logo and website were made. The CHEARTED website (see http://www.chearted.eu online) serves as a platform for the communication within the consortium, as well as between the consortium and the scientific and medical community at large. The former function will be served with 'news and views' items as well as the up-to-date listing of personnel and publications. In the near future the CHEARTED website will also contain an area providing information about the aims of the project targeting the general public.
Project meetings, dates and venues
- 4-5 March 2009 Amsterdam Kick-off Meeting and Council Meeting
- 8-9 June 2010 Annual and Council Meetings, Paediatric Hospital of the University of Zurich.
- 18-19 May 2011: Annual Meeting and Council Meetings in Bergen, Norway.
- 22-23 November 2012: Annual Meeting and Council Meetings in Leuven, Belgium.
WP4b Dissemination
The dissemination of the results of the CHEARTED project will be through the CHDwiki (see http://homes.esat.kuleuven.be/~bioiuser/chdwiki/ online) and the Cardiogenetic Morphology Database (CMD: http://cmd.amc.nl). The basic information on the project will be distributed through the CHEARTED website which will also serve as a portal for the other websites.
The CHEARTED website (see http://www.chearted.eu online) contains the basic information on the project and the beneficiaries.
Project Results:
Research Line 1 Genetic Epidemiology
WP1a Collection of clinical samples
1. Prepare samples for Tetralogy of Fallot genome wide association study.
2. Collect and prepare samples for association study of abnormal heart development in offspring of diabetic mothers.
WP1b Genome wide SNP analysis
1. To provide high resolution genotyping with SNP markers and copy number variants (CNVs) for tetralogy of Fallot cases, parental samples and controls.
2. To provide high resolution genotyping with SNP markers and copy number variants (CNVs) for offspring (+/- CVM) of diabetic mothers
3. To provide replication genotyping of additional samples as required.
WP1c Statistical Analysis of genotype data
1. To perform statistical analysis of whole-genome SNP and CNV data in order to detect genetic associations with CVM.
2. To perform statistical analysis of whole-genome SNP and CNV data in order to detect genetic associations with CVM in the presence of maternal diabetes.
3. To test whether the genetic factors involved in CVM are the same or differ depending on the presence of maternal diabetes.
1) Tetralogy of Fallot Genome wide Association study
Newcastle, Leuven, Erlangen and Sydney had begun collecting samples from patients with tetralogy of Fallot prior to CHEARTED. The samples held by the partners were checked for suitability for inclusion in the study. Some were excluded because not all four grand-parents were of Caucasian ethnicity others because there was insufficient DNA or the DNA did not meet the quality criteria. To achieve the aim of analysing 1000 samples we asked collaborators in Nottingham and Oxford to contribute samples. All samples were transferred to partner 5. The quality control (QC) evaluation of DNA samples, fragmentation and amplification and quantification determining the samples to be carried forward to the SNP genotyping excluded only 10 probands leaving 969 probands for the association study.
200 probands were typed on the Affymetrix 6.0 chip and re-genotyped on Illumina 660 quad chip to ascertain the optional platform, particularly with regard to the CNV analysis. While both platforms gave equivalent SNP calls, the resolution of CNV calls using the Illumina platform is greater thus the remaining samples were typed on the Illumina 660 quad chip.
After QC, the genotype data was distributed for statistical analysis groups using the Operon database software developed specifically for this purpose at the CNG.
Stringent quality control (QC) checks to ensure that only high-quality data were included in the final statistical analysis. QC procedures were carried out in the computer program PLINK with visualisation performed in the program R. Multidimensional scaling of our samples together with 210 unrelated Phase II HapMap individuals from four populations (CEU, JPT, CHB, YRI) (genotyped at same set of 41692 autosomal SNPs) was performed and identifed 33 individuals in our study that did not cluster with the CEU samples, suggesting non-European ancestry, these individuals were excluded. We also excluded samples exhibiting the known chromosome 22q deletion associated with TOF.
2) CHD association study in the children of diabetic women
The incidence of CHD is higher in the offspring of diabetic women. The aim of this part of the Workpackage was to collate clinical information on diabetic pregnancies and collect samples from the children, both those with CHD and those without for an association study. Partners 3 and 7 (UNEW and UiB) were tasked with collecting the samples.
UNEW: We had anticipated that collection in the north of England would be straightforward because there were linked diabetic pregnancy and congenital abnormality databases in the north-west and the north-east. However, regulatory processes changed in the period between submission of the CHEARTED application and the project commencing. We were no longer allowed to contact mothers of the 76 children born between 1999 and 2005 with cardiac defects whose mothers had diabetes cases For the north-east patients could still be approached but, instead of contact being made by the database staff, regulatory approvals had to be sought at the hospital at which the woman had received her care and the initial approach made by her consultant. The average time for obtaining regulatory approvals at a hospital was approximately six months. As the number of potential recruits was reduced by losing access to patients identified in the north-west diabetes pregnancy audit we asked additional sites across the UK to join the study. This was facilitated by the study being adopted onto the UK Clinical Research Network portfolio. By the end of the study approvals were in place in 26 English Hospitals and UKCRN and Diabetic Research Network (DRN) funded research nurses were recruiting in these centres. 1 additional UK centre has approval to participate and three further centres are in the process of obtaining regulatory approvals and regulatory approvals are in place to continue UK recruitment to 31/12/2013. We have recruited 36 affected offspring and 270 unaffected offspring along with all relevant clinical information on both children and pregnancy management.
UiB: Applications were submitted to the Regional Ethics Committee, Data Inspectorate, Directorate of Health and Medical Birth Registry of Norway (MBRN) by July 2009. By September, after minor adjustments due to requirements from REC and DI all applications were approved. In October 2009, delivery of data on potential cases and controls was requested from MBRN. All maternity units in Norway have were given about CHEARTED in the periodical Newsletter of MBRN (FØDSELSNYTT) and packs comprising information for the maternity unit, a letter of information for mothers, consent form, a questionnaire to the mother and kits for saliva sampling for distribution to women identified by MBRN. An important difference between UK and Norwegian samples was that whilst medical records were reviewed in the UK for HbA1C measurements etc. in Norway the measure of diabetic control depends on what the mother recorded. Patient information was received from MBRN in May 2010 and letters of invitation sent to 88 diabetic mothers of a child with a CHD (all identified in 1999 -2008) and 251 diabetic mothers of a child without CHD (randomly selected). Reminders were sent to all non-responders. Saliva samples and corresponding questionnaires have been obtained from 23 affected cases along with 13 unaffected siblings and 55 controls.
UMC Utrecht: 3 affected offspring and 106 unaffected offspring samples were transferred to UNEW.
As outlined above we recruited 62 affected cases and 444 unaffected controls. This number does not give adequate power for a statistically sound GWAS study. Rather than continue with an experiment that had no prospect of a meaningful outcome we considered other hypotheses that could be tested with this number of samples. One strong hypothesis is that maternal diabetes alters the epigenetic regulation of developmental genes, which in turn might lead to the CHD observed in the off-spring. We therefore deviated from the original plan to test this hypothesis, using the samples collected to investigate the methylation status of cardiac genes in affected and unaffected offspring. A custom panel of 29 candidate genes was selected based on two complementary approaches; a literature search using combinations of appropriate search terms and analysis of a mouse cardiac gene expression data set GEO GSE1479. Two housekeeping genes ACTC1 and MYLK3 were also included in the panel as 'control' genes. The promoter regions of all these genes were then analysed using CpGIslandExplorer v2.0 (see http://bioinfo.hku.hk/cpgieintro online) to locate CpG islands. A total of 96 CpG sites in these 31 genes were selected for the custom panel. 31 cases and 60 controls were selected for the plate, along with the recommended number of duplicates and standards. DNA was bisulphite modified and then prepared following the Illumina GoldenGate 'Methylation Assay for VeraCode' protocol and scanned on the Illumina BeadXpress. Data analysis is ongoing.
Research line 2 Animal Models
WP2a Mouse model of diabetic pregnancy
1. Catalogue the adaptive changes in gene expression in hyperglycemic embryos and hearts, and define the metabolic and signaling pathways that are specifically affected.
2. Determine the expression pattern of controlling genes in eu- and hyperglycemic embryos.
3. Establish causality for a subset of these genes in existing mutants or in in vitro embryo cultures.
The collection and preparation of embryonic samples for microarray analysis
Animal models for maternal hyperglycemia. We have evaluated a maternal diabetes (streptozotocin treatment, followed by treatment with "Linbit" controlled release insulin implants, so that overt diabetes develops only during pregnancy; Phelan et al., Diabetes. 1997, 46: 1189-1197) and maternal hyperglycemia model (hourly 25% glucose injections on ED7.5; Fine et al., Diabetes 1999, 48: 2454-2462) of diabetic pregnancy to study changes in gene expression in hyperglycemic embryos. Although glucose-injection approach offers the advantage of a short intervention, the frequent injections cause discomfort and are inconvenient.
Streptozotocin (STZ) model. The STZ protocol induced a 3-4-fold increase in blood glucose levels. Glycemic control with Linbit implants allowed the establishment of pregnancy in nearly all animals. Blood glucose at sacrifice was 3-4-fold elevated. However, on the standard SDS CRM(E) diet, too few malformations were seen. Switching to the commonly used 9F diet (Phelan et al., Diabetes. 1997, 46: 1189-1197) solved this problem. The malformed embryos often showed signs of cardiac failure (fluid accumulation in the pericardial cavity and apoptosis of cardiomyocytes). Based on the 9F/STZ protocol, we have collected approximately400 embryos. Male embryos were identified by PCR for the presence of the Sry gene. More importantly, perhaps, no fewer than 50% of the hyperglycemic embryos suffered from a severe growth retardation (approximately 1 day) on ED8.5 which led to their death on ED9.5 and to termination of pregnancy of litters in which greater than20% of the embryos were delayed in development on ED10.5. As a result, developmental delays and a reduction in litter size were no longer observed on ED10.5. The growth-retarding effect of hyperglycemia was similar in female and male embryos.
dSAGE libraries. We analyzed single embryos because of the observed growth retardations in approximately50% of the hyperglycemic embryos. We managed to construct SAGE libraries with approximately5% of the recommended amount of RNA, containing 30-40% identifiable tags, which is standard for dSAGE. A total of 52 single-embryos 3'-end cDNA SAGE libraries from 3 control (ED7.5 8.5 and 9.5) and 4 diabetic (ED8.5 and -9.5 non-retarded and growth-retarded) groups of male embryos were used in next generation SOLiD5 sequencing. We also processed RNA from another 48 embryos from the same control and diabetic groups to determine their microRNA expression profiles. Reproducibility of this procedure was established by repeating the entire procedure on the same starting RNA in a limited number of embryos.
Metabolic and signaling pathway analysis of the microarray data
SOLiD5 reads were mapped against Genome Reference Consortium Mouse Build 38 ("mm10"; December 2011). Reads mapping within exon locations of transcripts and with a NlaIII recognition site were counted and summed per gene. Unsupervised hierarchical clustering of top 100 expressing genes identified the respective biological groups, except that some ED8.5 control embryos mapped with diabetic ED8.5 embryos (even though littermate embryos did map to the control group;). Pathway analysis was performed by MetaCore software. On ED8.5 hyperglycemia induced almost 1,500 differentially expressed genes and greater than100 pathways were significantly regulated, in particular those regulating the cytoskeleton (P less than 10exp-6). On ED9.5 hyperglycemia induced greater than 5,000 differentially expressed genes and greater than200 pathways were highly regulated, with oxidative phosphorylation and cytoskeleton remodeling pathways being affected (P less than 10exp-11). Unsupervised hierarchical clustering of top 100 expressed miRNAs also identified the biological groups except that some ED8.5 control embryos mapped again with diabetic ED8.5 embryos. No differentially expressed microRNAs were identified by Deseq analysis, but the "G-test" identified 7 and 3 miRNAs on ED8.5 and -9.5 respectively, that were differentially affected by hyperglycemia. The "G-test" for classical SAGE analysis, which uses the G-statistic for analysis of groups of frequency distributions (Schaaf, Ruijter, et al. FASEB J. 2005;19:404-406), was adapted to handle the much larger number of tags identified in dSAGE.
Visualization of the expression pattern of controlling genes in eu- and hyperglycemic embryos / Testing relevant mouse mutants for changes in susceptibility to hyperglycemia or testing relevant genes for knock-down analysis in embryo cultures.
We visualized the expression of a few genes thus far (Pax3, Glut4), which was as predicted from literature. In the remaining part of the project (Jan-Aug 2013), we will test the prediction that oxidative phosphorylation is inhibited in diabetic embryos by measuring mitochondrial respiration and glycolysis with the Seahorse XFe Analyzer. In addition, we will explore the distribution of affected cytoskeletal proteins in the diabetic embryos.
Deviations from the project work programme, and corrective actions taken/suggested: identify the nature and the reason for the problem, identify beneficiaries involved.
We have replaced microarray analysis with deep-SAGE sequencing, mainly because it requires less mRNA and because the data obtained are more convenient for quantitative analysis of differential gene expression. The bioinformatics aspect of the project has, nevertheless, taken more time than anticipated. Two independent factors were responsible. Initially, the SAGE tags were sequenced with a SOLiD4 sequencer, which yields only approximately35 nt. Secondly, it turned out that the NlaIII endonuclease had cut only partially. Because this did not compromise the uniqueness of the tags, we adapted the search strategy, which however, meant a recount and reanalysis of all data.
WP2b Mouse Model of outflow tract defects
The objectives were to identify new genes and pathways specifically relevant to outflow tract malformations and Tetralogy of Fallot using molecular techniques and an Nkx2-5 hypomorphic mouse model of outflow tract malformation. Two types of screens will be performed:
1. High specificity/low yield: ENU screening for enhancers and repressors of the Nkx2-5 hypomorphic phenotype.
2. Medium specificity/high yield: microarray and transcription factor target gene analysis to identify genes involved in cardiac progenitor cell identity and behavior, and altered in outflow tract malformation.
Objective 1: High specificity/low yield: ENU screening for enhancers and repressors of the Nkx2-5 hypomorphic phenotype. We previously described using mouse models how proliferative defects in cardiac progenitor cells of the anterior second heart field arise as a result of mutation of the cardiac homeodomain transcription factor NKX2-5 1. In an NKX2-5 hypomorphic model, the lack of proliferative drive in progenitors leads to a spectrum of congenital heart disease-like manifestations, including atrial, ventricular and atrioventricular septal defects, over-riding aorta, double outlet right ventricle and Ebstein's anomaly. These resemble the more severe end of the congenital heart disease spectrum seen in humans carrying NKX2-5 mutations and hypomorphic mice die in the first 2 days after birth. We determined that repression of progenitor cell proliferation was due in part to elevation in the levels of BMP2 at early stages of heart tube formation. Other progenitor genes were also upregulated, leading us to propose that one of the early functions of NKX2-5 as cardiomyocytes are specified and begin to differentiate, is to repress progenitor genes, including those like BMP2 responsible for cardiac induction, thus allowing differentiation to proceed. In NKX2-5 null embryos, cardiomyocytes that do form show a deranged gene regulatory network with co-expression of both progenitor genes and cardiomyocyte differentiation genes. We found in fact that deletion of one allele of Smad1, which acts downstream in the BMP2 pathway, was sufficient to rescue progenitor proliferation and to significantly rescue structural malformations. However, lethality was not rescued and parallel studies indicatred that Smad1-mutant hypomorphs still show abnormalities of chamber growth and gene expression, and defects in epicardial development leading to abnormalities in coronary artery formation.
We found that outcrossing the hypomorphic strain from C57Bl/6 to QSi5, or indeed other common strains, lead to rescue of chamber growth, ventricular septal defects and of lethality. Hypomorphs on a 50:50 C57Bl/6:QSi5 background lived to old age without challenge. Preliminary breeding suggested that only one or a few genes on the C57Bl/6 background interacted with the NKX2-5 pathways to generate the spectrum of hypomorphic phenotypes. We therefore proposed an unbiased forward mutagenesis screen using ENU that might detect activators or suppressors of the NKX2-5 hypomorphic C57Bl/6 phenotype. The screen was set up in collaboration with the Australian Phenome Centre in Canberra, Australia, with costs partially offset from the National Collaborative Research Infrastructure Strategy of the Australian Government. We felt that this would be a low yield but high specificity screen that might detect genes in the NKX2-5 pathway, and those that would buffer mutations in the cardiac network, perhaps associated with the fetal environment. The initiation of the ENU screen was delayed because of troubles encountered identifying a mapping strain. The project depended on having a strain sufficiently different from the genetically modified strain so as to be able to map discovered mutations in pedigrees established during breeding, whilst not directly affecting the extent or penetrance of the phenotype itself. In short, we screened C3H, 129, CBA, FVB/N, BALBC, C57Bl/10 and LED, and all strains (except for C57Bl/10) show modifications of the phenotype, in most cases a degree of rescue.
This in fact gave us reassurance that there were specific alleles in multiple inbred strains of mice that can rescue the phenotype. The optimal ENU burden for C57Bl/10 had not been defined and when using the same conditions as used for C5Bl/6 mice, all C57Bl/10 males were sterile. This approach was abandoned and we did not consider trying to find a mapping strain further. We decided to rely instead on deep sequencing approaches and since this screen was started the APC has developed a robust pipeline for ENU mutation detection by whole genome sequencing. We went ahead and mutagenised C57Bl/6 studs to initiate the screen formally. The strategy was to mate ENU treated wt males with Nkx2-5-GFP heterozygous females, then to dissect N2 litters at 2 days after birth and screen for GFP-positive pups. Any pedigrees with positive pups at the time of dissection would be regarded as putative mutants, and used to produce subsequent litters by mating to Nkx2-5IRESCre mice. Further delays were encountered since the screen was set to be either too stringent (assessment for survivors at 2 weeks after birth) or too relaxed (assessment at 2 days), that latter showing some putative mutants that could not be replicated with further breeding. After setting the assessment time at 4 days after birth, we screened 148 pedigrees over 2 years without seeing putative mutations. Unfortunately, while by no means saturating, we had to abandon the screen because of lack of leads and lack of on-going funding from our Australian grants. While the screen was negative, we did learn a lot about the ENU technology and may return to a less stringent/higher yield screen in the future. As we further characterise the NKX2-5 hypomorphic strain, we use outbreeding as a tool to look at possible causes of dysfunction and death.
Objective 2: Medium specificity/high yield: microarray and transcription factor target gene analysis to identify genes involved in cardiac progenitor cell identity and behavior, and altered in outflow tract malformation. While we have not made strong progress on the first part of this aim microarray analysis of our NKX2-5 genetic model of outflow tract malformation, we have made a number of key discoveries related to gene pathways involved in outflow tract development. Mesp1 is a key transcription factor that is expressed in early cardiac and head fields and is absolutely required for the development of the heart. In a collaboration with Cedric Blanpain (Belgium), we showed that MESP1-GFP cells derived from ES cell cultures are enriched in mulitpotent heart progenitors and that transcriptional profiling of these cells uncovered new markers that allow their prospective isolation form ES cells and embryos 2. In this study, we also explored the interaction between MESP1 and ISL1, encoding another important cardiac transcription factor and results suggest that they interact to promote both cardiomyocyte and endothelial cell differentiation. In the NKX2-5 null model, ISL1 expression is retained in the differentiating myocardium where it is normally turned off. In collaboration with Margaret Buckingham (France), we elucidated a mechanism for how NKX2-5 might normally repress ISL1-mediated gene expression. Using FGF10 as an NKX2-5 target and up-regulated in NKX2-5 null myocardium, we showed that ISL1 and NKX2-5 share common binding sites within the second heart field enhancer of the FGF10 gene and these factors compete for binding and activation versus repression 3.
Other insights into cardiac developmental pathways include description of a genetic interaction between NKX2-5 and a gene encoding a component of the BRG1 chromatin-remodelling complex 4; demonstration of a role for Neuregulin 1 in maintaining the cardiac gene regulatory network in both trabecular and non-trabecular myocardium 5; role for NKX2-5 in suppression of blood development via GATA1 6; and recognition that the endocardium identified as having expressed NKX2-5 in its lineage history transiently retains hemogenic activity 7.
A major project under this aim has been to document the target genes of cardiac transcription factors using the DamID technique in the HL1 cardiomyocyte cell line, and to probe the mechanisms of congenital heart disease. This has been highly successful. Overall, we have defined target sets of NKX2-5, GATA4, ISL1, TBX1, TBX2, TBX5, TBX20, HAND1 and MEF2c, as well as mutants of NKX2-5, and we have recently expanded this initial screen to include SRF, ELK1, ELK4 and a mutant of ELK1 that does not interact with SRF. It has been necessary in our study to consider targets of both NKX2-5-DAM (C-terminal fusion) and DAM-NKX2-5 (N-terminal fusion) as steric influences become evident for some transcription factors, particularly when mutated.
We are principally interested in how cardiac transcription factors influence network logic and have made a number of key discoveries:
1) We have documented 757 high confidence NKX2-5 target genes in HL-1 cells covering a wide range of biological functions;
2) Contrary to expectation, mutations of NKX2-5, including a full homeodomain deletion, also bind hundreds of targets, some overlapping NXK2-5 WT targets but some unique, representing "off-targets";
3) NKX2-5 mutants can be directed to genuine NKX2-5 targets as well as off-targets via altered DNA-binding specificity and by association with cofactors;
4) A novel cofactor interface within NKX2-5 has been defined that facilitates homodimerisation, heterodimerisation with NKX2-5 mutants, and interactions with a range of cardiac-restricted and ubiquitous cofactors;
5) ELK proteins (ELK1 and ELK4) are ubiquitously expressed transcription factors that interact with NKX2-5 and are recursively wired into the early cardiac gene regulatory network;
6) the role of ELK1 and ELK4 in the cardiac gene regulatory network has been confirmed by gene knockdown studies in zebrafish, and by defining the targets of ELK1 and ELK4, was well as SRF which can associate with ELK proteins in some contexts, using DamID.
Research line 3 Genetic Bioinformatics
Workpackage 3a: Genetic Bioinformatics: environmental factors influence on heart development gene networks, and candidate gene prioritization
1. Inference of gene networks for normal and abnormal heart development
2. Inference of models for the environmental factors and their influence on heart development
3. Development of a bioinformatics tool for congenital heart defect candidate genes prioritization
The main goal of the research effort of this work package consisted of adapting, extending and applying the gene prioritization tool named as Endeavour (Aerts et al, 2006, Tranchevent et al, 2008) to the specific purposes of the CHEARTED project (i.e. applying it to congenital heart defects). As a reminder, Endeavour is a software application for the computational prioritization of candidates' genes. Endeavour is able to rank genes according to their similarity (considering various biological criteria such as sequence homology, GO classification, physical interaction) to a set of training genes (in our case, genes implied in the same congenital heart disease or in the same developmental process). This strategy is highly valuable considering the large sets of candidate genes that will be produced by the other partners in the project.
Building heart specific training sets
The first step consisted in building training sets, i.e. sets of genes involved in processes and that serve as seed (models) for the prioritization. After an extensive literature search (mainly performed by Drs Jeroen Breckpot and Bernard Thienpont) and the two year contribution of CHDWiki users, this collaborative web portal hosts' information about 79 syndromic genes, 274 associations between these genes and congenital heart defects as well as other relative information.
We also integrated the data produced by WP2.a where deep SAGE sequencing experiments were performed on embryos collected during pregnancy in a diabetic mouse model. On this new dataset we applied our new hierarchical prioritization method. The derived results and created datasets are hosted by CHDWiki.
Building environmental factors training sets and Build models specific to heart development
First, we developed a novel text-mining approach (called Beegle) that takes simple "PubMed queries" as input and returns a list of genes linked to that query. This approach was of primary interest to identify a list of genes and factors linked to congenital heart defects. This tool is accessible at the following URL: http://homes.esat.kuleuven.be/~sbrohee/beegle-new/. The tool works for any query (e.g. ranging from "Tetralogy of Fallot" to "apoptosis") and is based on the similarity between the abstracts returned by the query in Pubmed to the abstracts of the paper describing the genes in the GeneRIF database. The purpose of this tool is to be as simple as a "Google for genes" and we deliberately mimicked the display of Google when we implemented the online version of Beegle. The genes delivered by the Beegle text-mining approach can then be used as seeds for the PINTA tool (Nitch et al. 2011, Nucleic Acid Res, 39: 334-338), another tool developed by our lab. In short, PINTA looks at the neighbors of some seed genes in a functional interaction network (such as in the STRING database) to predict new candidates genes.
Set up a hierarchical prioritization method
In the beginning of the CHEARTED project, we tried to improve the, already satisfactory, performances of our Endeavour gene prioritization tool, which is available at our CHDWiki pages, proposing built-in training genes for various heart developmental processes. The idea was to combine various data sources according to the studied disease or process. Indeed, some data sources might be very redundant (e.g. several protein-protein interaction or homology databases) and biased toward already known information (like Gene Ontology) while some are not. The output of each (sub-) prioritization step would then be used together for the next prioritization step, using both biased and unbiased sources without giving a higher weight. Using this strategy, we hoped to build a tree of data sources. The obtained results when applying this hierarchical prioritization approach were somewhat unsatisfactory. Firstly, the performances were comparable to those obtained when no hierarchical prioritization was applied. Secondly, the algorithm was much slower (as many more prioritization steps were performed).
Prioritize candidates and analyze results
For the purposes of this project, two different candidate gene lists were created by two different prioritization methods and they are available at:
(see http://homes.esat.kuleuven.be/~bioiuser/chdwiki/index.php/CHD:Genes online). The first list was produced by Endeavour tool, which was integrated into CHDWiki. The second list was produced by the new hierarchical gene prioritization method that uses 4 prioritization tools to predict novel CHD candidate genes and the gene list produced by WP2.a as a training set. The users can further analyse these results by viewing separately each gene page where other kind of information are also offered. Moreover, CHDwiki integrates information from the major protein interaction databases (STRING, Biogrid, Intact, Mint, etc.) such as the list of interactions for each syndromic, non-syndromic or candidate gene contained in the CHDWiki gene lists. Finally, a viewer allows the users to navigate through the network of physical or functional interactions of the gene of their interest.
Build networks of genes (best candidates) involved in heart development and interactions with environmental factors.
The CHDwiki offers a visualization module that associates the candidate genes with congenital heart defects, their frequency as well as the physical protein-protein interactions as it can be seen at: http://homes.esat.kuleuven.be/~bioiuser/chdwiki/index.php/Chd_gene_networks. After doing this we could observe that genes sharing multiple phenotypes when mutated tend to act in the same molecular pathways or even encode proteins that directly interact.
Deviations from the project work programme, and corrective actions taken/suggested: identify the nature and the reason for the problem, identify Beneficiaries involved.
Workpackage 3b: Genetic Morphology: 3D gene expression atlas of cardiac development
1. Generate a Cardiogenetic Morphology Database (CMD) based on a standard series of 3D reconstructions of heart development, the expression patterns of 12 key marker genes and genetic annotation of cardiac compartments.
2. Develop an application to help researchers to automatically match (series) of individual cardiac sections to the reference series, to add gene expression information from these sections to the CMD and to display the association of the gene expression patterns with patterns of genes contained in the CMD.
3. Map novel genes involved in cardiac development into the CMD, and test hypotheses on their role in the signaling pathways.
The goal of WP3b was to establish a three-dimensional (3D) gene expression atlas of cardiac development. The work was distributed over 4 tasks. The first 3 tasks cover the generation of a series of high resolution reference models of cardiac development, the development of an application to match individual section to the 3D space of a reference model of the appropriate age, and the creation of a series of 3D reconstructions displaying the gene expression patterns of key genes in cardiac development. The fourth task was to map the 3D expression patterns of novel genes, identified by the other workpackages as being involved in cardiac development and affected by diabetic pregnancy.
Generate reference series of developing mouse hearts at 9.5 11.5 and 13.5 days of development
The basis of the 3D atlas is formed by a series of high resolution models of the embryonic mouse heart. These models were obtained by high resolution episcopic microscopy (HREM) at the MRC. Successive issues had to be addressed to reach the required results.
1) Whole embryos versus dissected hearts
Whole embryos have the obvious merit that they retain the anatomical relationships between the heart, great vessels and other organs in the thoracic cavity. However, it is extremely difficult to remove blood from whole embryos. Dissected hearts necessarily lose the anatomical context and run the danger of morphological distortion as a result of the dissection. Both kind of data sets were compared.
2) Exsanguination procedures
Removal of blood from mouse embryos required a non-invasive procedure to avoid damaging the heart itself. Embryo perfusion is only realistic at late stages (embryonic day (E) 14.5 and older). For younger stages prolonged maintenance of embryos at 37C with repeated removal of blood clots from severed vessels proved to be effective.
3) Staging of embryos
Different embryos within a litter are variable in developmental stage. Since developmental stage is evidently important in a study of morphological development, alternative "accurate" staging methods have been compared. From these we concluded that the most effective approach is to image a large number of samples for each apparent developmental stage and to order these based entirely on the established sequence of changes in cardiac morphology.
4) Image quality
Fluorescence of HREM tends to fade out when tissue is close to the surface of the block which results in apparent thickening of the tissue in the sectioning direction. To reduce this effect, an image filter was developed to remove this out-of-plane information.
Develop computer application that TRaces the Anatomical Context of single Tissue Sections (TRACTS).
The rapid morphological change of the developing heart makes it difficult to interpret single sections through such a heart. We developed a program (TRACTS) which fits arbitrary histological sections into a high resolution (HREM) 3D heart model. We implemented a straightforward brute force pixel-based approach, where the (2D) input section is compared to a selection of virtual (2D) cross sections of the (3D) reference model. Model cross sections are selected on basis of four image features: size, overall density, regional density and center of mass. The performance of TRACTS is assessed using virtual cross-sections in arbitrary directions from a second HREM dataset and using paraffin embedded material. A test, involving a panel of morphological experts who manually fitted a sample of sections to the reference model test showed that the program fits just as well as the best fits found by five morphological experts. The initial version, based on a reference model of E11.5 was extended to handle a reference model series of E9.5 E11.5 and E13.5.
Generate a timed series of 3D reconstructions of key marker genes
To understand the relation between genes and morphogenesis one first needs to know which genes are expressed in which parts of the developing organ. The widespread application of high throughput gene expression analysis platforms has lead to the identification of thousands of mammalian genes and the estimation of their activity in dozens of organs and parts of organs. However, virtually all such expression analyses are carried out on organ homogenates in which spatial information is lost. Only 3-dimensional (3D) reconstruction of specifically stained sections can provide a complete spatial expression pattern. The detailed gene expression information in 2D images or 3D reconstructions can be made accessible by combining such 3D reconstructions into a common reference model.
Line 4 Management and Dissemination
Workpackage 4a: Management
1. The management activities should be instrumental for a proper and timely implementation of the activities.
2. The management activities will secure the coordination and integration between the other activities in order to reach synergistic effects.
3. The starting point is that a project needs a clear and distinct description of roles, responsibilities and tasks also in order to allow for a proper accountability along the (financial) guidelines of the European Commission, the partner organizations and the principles of accountancy.
Consortium management tasks and achievements
Prof. Dr A.F.M. Moorman, was responsible for the internal and external communication, including the (report and financial) obligations to the EU. He was in charge of the Project Management Team and chaired the Board and the Council.
The CHEARTED logo and website (see http://www.CHEARTED.eu online) were designed by Dr Alexandre Soufan. The CHEARTED website (see http://www.CHEARTED.eu online) serves as a platform for the communication within the consortium, as well as between the consortium and the scientific and medical community at large. The former function will be served with 'news and views' items as well as the up-to-date listing of personnel and publications. In the near future the CHEARTED website will also contain an area providing information about the aims of the project targeting the general public.
List of project meetings, dates and venues
- 4-5 March 2009 Amsterdam Kick-off Meeting and Council Meeting
- Workshops 2009: 24 April Meeting UNW and UiB in Bergen and 5 October Workshop WP3a. WP3b and WP4b in Leuven
- 8-9 June 2010 Annual and Council Meetings, Paediatric Hospital of the University of Zurich. Invited speakers: Dr Mary Loeken from the Joslin Diabetes Center, Harvard and ESAC member Dr Ed Lammer from Children's hospital and Research Center Oakland, Calilifornia. ESAC member Prof Dietrich Rebholz-Schuhmann was also present
- 18-19 May 2011: Annual Meeting and Council Meetings in Bergen, Norway. Invited speaker: Prof. Rolv Skjærven, from the Medical Birth Registry of Norway, Norwegian Institute of Public Health Bergen, Norway. ESAC member Dr Ed Lammer was present
- 22-23 November 2012: Annual Meeting and Council Meetings in Leuven, Belgium Invited speakers: Prof. Marc-Philip Hitz (Sanger Institute, Cambridge, UK) and Dr Phil Barnett (Dept Anatomy, Embryology and Physiology, AMC, Amsterdam, NL). Two ESAC members were present: Dr Ed Lammer (Berkley, CA, USA) and Prof Cornelia van Duijn, Rotterdam, NL.
Problems which have occurred and how they were solved or envisaged solutions
In Research Line 1 the number of children with heart defects born to diabetic mothers is less than anticipated and measures have been taken to analyse the result in a different way. An association study is not possible due to insufficient numbers, but an epigenetic study was performed instead
Workpackage 4b: Dissemination
1. Facilitate the exchange of the gene association, transcriptional network and genetic morphology results of CHEARTED both between the Beneficiaries of the project, with the scientific community and with the public.
2. Enable the contribution of third parties to the databases developed within CHEARTED.
3. New Bulletins
To integrate and share knowledge on heart development, CHEARTED wishes to construct an interactive database, which integrates knowledge on different domains such as genetics, environmental factors, morphogenesis, gene expression etc
Cardiogenetics Knowledge Database
The CHDwiki website (see http://homes.esat.kuleuven.be/~bioiuser/chdwiki online) has been implemented as a collaborative resource. CHDwiki is a user-friendly knowledge base portal aimed at mapping genes involved in CHDs and understanding their relations with corresponding human phenotypes. CHDwiki is a collaborative resource, meaning that every registered user can add, delete or modify the data in the database. The database is described in Barriot R, Breckpot J, Thienpont B, Brohée S, Van Vooren S, Coessens B, Tranchevent LC, Van Loo P, Gewillig M, Devriendt K, Moreau Y. Collaboratively charting the gene-to-phenotype network of human congenital heart defects. Genome
Med. 2010 Mar 1;2(3):16. There are currently 163 registered users. The results of the knowledge acquisition are shown below (as of December 2012):
Features entries Features CHD
Genes 79 Genes 274
CHD types 113 Mutations 367
linkage regions 15 linkage regions 19
balanced translocations 25 balanced translocations 27
imbalances (cases) 374 Van Karnebeek 1999 regions 85
imbalances (regions) 410 Studies (mutation screens) 428
Van Karnebeek 1999 regions 46
References 253
Results of knowledge acquisition. The right column displays the number of different features present in CHDWiki (e.g. genes, CHD types, imbalances). The left column displays the total number of relations between Congenital Heart Defects (CHD) and currently managed features.
CHDWiki portal provides a collaborative knowledge base where the current knowledge on the genetic basis of CHDs is stored. Relationships between genes and CHDs have been mapped, derived from published data or single clinical cases making this platform a useful tool for geneticists, molecular biologists and clinicians. The genomic variations that are recorded in CHDWiki create relationships between associated genes, congenital heart defects and clinical reports. Lists of syndromic or nonsyndromic genes are available to the users while two different lists of candidate genes can be found produced by Endeavour and a newly introduced hierarchical gene prioritization method.
Also in this specialized portal, various modules are hosted such as:
- An integrated version of the Endeavour tool in order to perform genome wide gene prioritization. This software application allows the bioinformatic prioritization of candidates genes, based on a set of training genes. Different lists of training genes concerning different parts of the heart are defined. Endeavour was used to identify two CHD candidate genes, (Breckpot J, et al. K. Congenital heart defects in a novel recurrent 22q11.2 deletion harboring the genes CRKL and MAPK1. Am J Med Genet A. 2012 Mar;158A(3):574-80. and Breckpot J, et al. BMPR1A is a candidate gene for congenital heart defects associated with the recurrent 10q22q23 deletion syndrome. Eur J Med Genet. 2012 Jan;55(1):12-6).
- Besides Endeavour, we integrated two other text mining tools that establish direct links between genes and CHD and vice versa: Huge Navigator and AgeneApart. For each protein coded by a CHD implied gene, CHDwiki returns the list of physical as well as functional interactions in a network visualizer.
- A module that enable users to visualize the physical or the functional interactions of the gene of their interest.
- A map module, which provides a karyogram with all regions and genes annotated in the CHDWiki.
Cardiogenetics Morphology Database
The Cardiogenetic Morphology Database (CMD) was developed to hold 2D and 3D gene expression patterns of 3 stages of mouse heart development. The CMD website (see http://cmd.amc.nl/ online) has been set up to enable end-users, researchers in the field of embryology, to submit images of single sections in which a gene expression pattern is visualized and to query the database for (non-)overlapping genes. To this end, a semi-automatic pipeline was created to upload, process, and store the submitted images.
CHEARTED website
The CHEARTED website (see http://www.CHEARTED.eu online) was based on the website created for the FP6 project HeartRepair (see http://www.heartrepair.eu online). The website contains functionalities like a private domain, public access level, factual information, publications and contact list. The content of the site is managed by the dissemination manager (beneficiary 11). The CHEARTED website contains a link to the CHDwiki maintained by the partner K.U.-Leuven and thus
provides access to this data source.
News bulletins, Brochures
After a delayed start up in the first year, 8 News Bulletins and 2 flyers have been produced since January 2010. A final, concluding News Bulletin will be published in March 2013.
News Bulletins and flyers were distributed in a printed version and as PDF document to
- all researchers related to the CHEARTED project
- scientist and medical specialists in the field of congenital heart development and congenital heart diseases.
- the European heart foundations (European Heart Network (EHN))
The News bulletins give a nice overview of the research activities of the Workpackages:
- In the 1st news bulletin professors Koenraad Devriendt and Yves Moreau from the University of Leuven gave an introduction to the CHDwiki (workpackage 4b).
- In the 2nd news bulletin Professor Judith Goodship from Newcastle explained GWAS to identify genes involved in the occurrence of Tetralogy of Fallot as well as in the mechanism by which maternal diabetes affects cardiac development (workpackage 1a).
- In the 3rd news bulletin dr. Jan Ruijter from the Amsterdam Medical Center presented the work on the development of methods to visualize genetic networks regulating cardiac development (workpackage 3B). Furthermore, a short report of the first annual meeting was given by dr. Dietrich Rebholtz-Schumann, member of the External Scientific Advisory Committee of CHEARTED.
- The 4th news bulletin was dedicated to the statistic analyses of genetic data by the group of professor Heather Cordell from Newcastle (workpackage 1c). Professor Wout Lamers from Amsterdam described experiments on a mouse model for cardiac malformations caused by diabetes during pregnancy (workpackage 2a).
- The 5th news bulletin presented the second CHEARTED annual meeting, which was held in Bergen, Norway on May 18-19, 2011. Also, there is an extended reference to the national medical birth registry of Norway and to the first CHEARTED dissertations, successfully defended by dr. Bouke de Boer of the AMC and dr. Jeroen Breckpot of the KU Leuven.
- The 6th news bulletin presented an overview over the work of Dr. Bouke de Boer (workpackage 3b) and Dr. Jeroen Breckpot (workpackage 4b).
- The 7th news bulletin is dedicated to Professor Richard Harvey's group which played an important role in Workpackage 2b of the CHEARTED project.
- The 8th news bulletin included an extensive article by Professor Gerard H.A. Visser (University Medical Centre Utrecht, The Netherlands) about the role of diabetes in heart development.
Potential Impact:
We expected that the knowledge gained from our research would lead to an improvement in the prevention and treatment of cardiovascular diseases, which are a major cause of ill health and death in Europe and worldwide. Our research focused on mechanisms of normal and abnormal heart development leading to congenital heart diseases. This included studies on genetic basis and the interaction between genes and environment in the development of heart defects and built on existing epidemiological data. As we stated in the Annex I, the data we generated would not have direct clinical applications but would make a significant contribution to the understanding of CVM and the role of environment during fetal development. Moreover, to carry out this research, and to make the results accessible, we had to develop a number of informatics tools and databases that themselves would be of potential scientific and public value.
Fundamentally, the ultimate goal of CHEARTED was to reduce the of incidence of what could be considered preventable cardiovascular malformations. Environmental and developmental issues are presently trapped in a bottleneck. A wealth of information regarding cardiovascular development, and the factors which may potentially affect its course, is already available and accessible. Despite this, the cause for more than 80% of cardiovascular malformations (CVM) remains elusive. Only now, due to the vast improvements in recent years in genome wide association and molecular analysis techniques, is it possible to unravel complex diseases such as cardiovascular malformation (CVM). Indeed, recently we have seen the identification by genome-wide association studies of susceptibility alleles in conditions (such as coronary heart disease) that are known to be far less heritable than CVM. Thus, it is was timely to give a new and devoted impulse to collect, identify, analyze and describe etiological risk factors for CVM. A major aim of research line 1 was to identify genotypes which affect the metabolism or bioavailability of specific nutrients and ultimately affect heart development and to provide a real indication of the total contribution of common genetic variation to cardiovascular malformation. At the end of the last reporting period, we completed the statistical analysis of whole-genome SNP and CNV data in 1733 genotyped individuals, comprising 913 Tetralogy of Fallot (ToF) cases plus a number of unaffected relatives. Two loci were identified on chromosomes 12 and 13 that were statistically strongly associated with the risk of ToF. This study was published in a high impact journal and is, worldwide, the first genome-wide association study of any congenital heart malformation phenotype. In addition, work on the CNV data in ToF cases identified an excess of rare genic deletions, moreover it was found that both duplication and deletion of 1q21.1 are more common in ToF cases than in controls.
Two papers describing this work have now been published. The potential impact of this combined knowledge is enormous, but at the same time daunting. In general we can state that CHEARTED had an enormous impact as it was world-wide the first to uncover common variation underlying Tetralogy of Fallot. If we look in detail at the results from the GWAS study, we identified two loci in which multiple genes reside. We now know that these genes somehow result in ToF through an (as yet) unknown mechanism. One potential insight is the fact that one of the genes in these loci is the PTPN11 gene, a protein-tyrosine phosphatase, known to be causative in Noonans syndrome, a Mendelian multisystem disorder in which malformation of the cardiac outflow tract is a typical feature. The precise mechanism how variations in these loci underlie the development of Tetralogy of Fallot is subject of future research and forms a large part of the impact CHEARTED has had on European science.
List of Websites:
http://www.CHEARTED.eu
http://homes.esat.kuleuven.be/~sbrohee/beegle-new/
http://homes.esat.kuleuven.be/~bioiuser/chdwiki/index.php/CHD:Genes
http://cmd.amc.nl ; http://downloads.hfrc.nl