Stylianos E Antonarakis

University of Geneva, Genève, Geneva, Switzerland

Are you Stylianos E Antonarakis?

Claim your profile

Publications (279)2740.78 Total impact

  • [Show abstract] [Hide abstract]
    ABSTRACT: Gene expression levels can be subject to selection. We hypothesized that the age of gene origin is associated with expression constraints, given that it affects the level of gene integration into the functional cellular environment. By studying the genetic variation affecting gene expression levels (cis expression quantitative trait loci [cis-eQTLs]) and protein levels (cis protein QTLs [cis-pQTLs]), we determined that young, primate-specific genes are enriched in cis-eQTLs and cis-pQTLs. Compared to cis-eQTLs of old genes originating before the zebrafish divergence, cis-eQTLs of young genes have a higher effect size, are located closer to the transcription start site, are more significant, and tend to influence genes in multiple tissues and populations. These results suggest that the expression constraint of each gene increases throughout its lifespan. We also detected a positive correlation between expression constraints (approximated by cis-eQTL properties) and coding constraints (approximated by Ka/Ks) and observed that this correlation might be driven by gene age. To uncover factors associated with the increase in gene-age-related expression constraints, we demonstrated that gene connectivity, gene involvement in complex regulatory networks, gene haploinsufficiency, and the strength of posttranscriptional regulation increase with gene age. We also observed an increase in heritability of gene expression levels with age, implying a reduction of the environmental component. In summary, we show that gene age shapes key gene properties during evolution and is therefore an important component of genome function. Copyright © 2014 The American Society of Human Genetics. Published by Elsevier Inc. All rights reserved.
    The American Journal of Human Genetics 12/2014; 95(6):660-74. · 11.20 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Alteration of the number of copies of double minutes (DMs) with oncogenic EGFR mutations in response to tyrosine kinase inhibitors is a novel adaptive mechanism of glioblastoma. Here we provide evidence that such mutations in DMs, called here amplification-linked extrachromosomal mutations (ALEMs), originate extrachromosomally and could therefore be completely eliminated from the cancer cells. By exome sequencing of seven glioblastoma patients we reveal ALEMs in EGFR, PDGFRA and other genes. These mutations together with DMs are lost by cancer cells in culture. We confirm the extrachromosomal origin of such mutations by showing that wild-type and mutated DMs may coexist in the same tumour. Analysis of 4,198 tumours suggests the presence of ALEMs across different tumour types with the highest prevalence in glioblastomas and low-grade gliomas. The extrachromosomal nature of ALEMs explains the observed drastic changes in the amounts of mutated oncogenes (like EGFR or PDGFRA) in glioblastoma in response to environmental changes.
    Nature Communications 12/2014; · 10.74 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Schizophrenia (SCZ) is a severe, debilitating mental illness which has a significant genetic component. The identification of genetic factors related to SCZ has been challenging and these factors remain largely unknown. To evaluate the contribution of de novo variants (DNVs) to SCZ, we sequenced the exomes of 53 individuals with sporadic SCZ and of their non-affected parents. We identified 49 DNVs, 18 of which were predicted to alter gene function, including 13 damaging missense mutations, 2 conserved splice site mutations, 2 nonsense mutations, and 1 frameshift deletion. The average number of exonic DNV per proband was 0.88, which corresponds to an exonic point mutation rate of 1.7×10−8 per nucleotide per generation. The non-synonymous-to-synonymous mutation ratio of 2.06 did not differ from neutral expectations. Overall, this study provides a list of 18 putative candidate genes for sporadic SCZ, and when combined with the results of similar reports, identifies a second proband carrying a non-synonymous DNV in the RGS12 gene.
    PLoS ONE 11/2014; · 3.53 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Introduction Hypophosphatemic rickets (HR) comprises a rare group of inherited diseases. Very recently, mutations in the dentin matrix protein 1 (DMP1) gene were identified in patients with an extremely rare autosomal recessive form of HR (ARHR). To date, very few cases of these mutations were reported. Materials and methods A Lebanese consanguineous family with 2 affected sisters was studied. Patients aged 45 and 47 years old presented with short stature, severe genu varum, cranial hyperostosis and a very high bone density that led to a diagnosis of a familial sclerosing bone dysplasia. Molecular analysis of known genes involved in osteopetrosis showed normal results. A combination of genotyping and exome sequencing was performed in order to elucidate the genetic basis of this pathology. Results Biochemical analysis was consistent with normal serum calcium and 1-25(OH)2D levels, low to normal serum phosphorus and elevated PTH values. Serum c-terminal FGF-23 was elevated in one of the two patients. A homozygous mutation disrupting the initiation codon of the DMP1 gene (OMIM 600980), NM_001079911.2: c.1A > G, p.Met1Val, was identified by exome sequencing and confirmed by Sanger sequencing. Conclusion We report here a family of ARHR secondary to a DMP1 mutation located in the first coding exon of the gene. Our cases show that some ARHR cases may develop with age an unaccountable increase in bone density and bone overgrowth.
    Bone 08/2014; · 4.46 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Hypophosphatemic rickets (HR) comprises a rare group of inherited diseases. Very recently, mutations in the dentin matrix protein 1 (DMP1) gene were identified in patients with an extremely rare autosomal recessive form of HR (ARHR). To date, very few cases of these mutations were reported.
    Bone 08/2014; · 4.46 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Most reported mutations in the FGFR3 gene are dominant activating mutations that cause a variety of short-limbed bone dysplasias including achondroplasia and syndromic craniosynostosis. We report the phenotype and underlying molecular abnormality in two brothers, born to first cousin parents. The clinical picture is characterised by tall stature and severe skeletal abnormalities leading to inability to walk, with camptodactyly, arachnodactyly and scoliosis. Whole exome sequencing revealed a homozygous novel missense mutation in the FGFR3 gene in exon 12 (NM_000142.4:c.1637C>A: p.(Thr546Lys)). The variant is found in the kinase domain of the protein and is predicted to be pathogenic. It is located near a known hotspot for hypochondroplasia. This is the first report of a homozygous loss-of-function mutation in FGFR3 in human that results in a skeletal overgrowth syndrome. This article is protected by copyright. All rights reserved.
    Human Mutation 05/2014; · 5.21 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: To study the prevalence of DEPDC5 mutations in a series of 30 small European families with a phenotype compatible with autosomal dominant nocturnal frontal lobe epilepsy (ADNFLE). Thirty unrelated families referred with ADNFLE were recruited in France, Italy, Germany, Belgium, and Norway. Whole-exome sequencing was performed in 10 probands and direct sequencing of the DEPDC5 coding sequence in 20 probands. Testing for nonsense-mediated messenger RNA decay (NMD) was performed in lymphoblastic cells. Exome sequencing revealed a splice acceptor mutation (c.2355-2A>G) in DEPDC5 in the proband of a German family. In addition, 3 nonsense DEPDC5 mutations (p.Arg487*, p.Arg1087*, and p.Trp1369*) were detected in the probands of 2 French and one Belgian family. The nonsense mutations p.Arg487* and p.Arg1087* were targeted by NMD, leading to the degradation of the mutated transcripts. At the clinical level, 78% of the patients with DEPDC5 mutations were drug resistant. DEPDC5 loss-of-function mutations were found in 13% of the families with a presentation of ADNFLE. The rate of drug resistance was high in patients with DEPDC5 mutations. Small ADNFLE pedigrees with DEPDC5 mutations might actually represent a part of the broader familial focal epilepsy with variable foci phenotype.
    Neurology 05/2014; · 8.30 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: The Hedgehog (Hh) family of secreted proteins act as morphogens to control embryonic patterning and development in a variety of organ systems. Post-translational covalent attachment of cholesterol and palmitate to Hh proteins are critical for multimerization and long range signaling potency. However, the biological impact of lipid modifications on Hh ligand distribution and signal reception in humans remains unclear. In the present study, we report a unique case of autosomal recessive syndromic 46,XY Disorder of Sex Development (DSD) with testicular dysgenesis and chondrodysplasia resulting from a homozygous G287V missense mutation in the hedgehog acyl-transferase (HHAT) gene. This mutation occurred in the conserved membrane bound O-acyltransferase (MBOAT) domain and experimentally disrupted the ability of HHAT to palmitoylate Hh proteins such as DHH and SHH. Consistent with the patient phenotype, HHAT was found to be expressed in the somatic cells of both XX and XY gonads at the time of sex determination, and Hhat loss of function in mice recapitulates most of the testicular, skeletal, neuronal and growth defects observed in humans. In the developing testis, HHAT is not required for Sertoli cell commitment but plays a role in proper testis cord formation and the differentiation of fetal Leydig cells. Altogether, these results shed new light on the mechanisms of action of Hh proteins. Furthermore, they provide the first clinical evidence of the essential role played by lipid modification of Hh proteins in human testicular organogenesis and embryonic development.
    PLoS Genetics 05/2014; 10(5):e1004340. · 8.52 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Down syndrome (DS) is due to increased copy number of human chromosome 21. The contribution of different genetic regions has been tested using mouse models. As shown previously, the Abcg1-U2af1 genetic region contributes to cognitive defects in working and short-term recognition memory in Down syndrome mouse models. Here we analyzed the impact of monosomy of the same genetic interval using a new mouse model, named Ms2Yah. We used several cognitive paradigms, and did not detect defects in the object recognition nor the Morris water maze tests. However, surprisingly, Ms2Yah mice displayed increased associative memory in a pure contextual fear conditioning test, and decreased social novelty interaction along with a larger long term potentiation recorded in the CA1 area following stimulation of Schaffer collaterals. Whole genome expression studies carried out on hippocampus showed that only the transcription of a small number of genes is affected, mainly from the genetic interval (Cbs, Rsph1, Wdr4), with a few additional ones, including the postsynaptic Gabrr2, Gabbr1, Grid2p, Park2 and Dlg1 and the components of the Ubiquitin mediated proteolysis (Anapc1, Rnf7, Huwe1, Park2). The Abcg1-U2af1 region is undeniably encompassing dosage sensitive genes or elements whose change in copy number directly impact learning and memory, synaptic function and autistic related behavior.
    Genetics 04/2014; · 4.39 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Trisomy 21 is the most frequent genetic cause of cognitive impairment. To assess the perturbations of gene expression in trisomy 21, and to eliminate the noise of genomic variability, we studied the transcriptome of fetal fibroblasts from a pair of monozygotic twins discordant for trisomy 21. Here we show that the differential expression between the twins is organized in domains along all chromosomes that are either upregulated or downregulated. These gene expression dysregulation domains (GEDDs) can be defined by the expression level of their gene content, and are well conserved in induced pluripotent stem cells derived from the twins' fibroblasts. Comparison of the transcriptome of the Ts65Dn mouse model of Down's syndrome and normal littermate mouse fibroblasts also showed GEDDs along the mouse chromosomes that were syntenic in human. The GEDDs correlate with the lamina-associated (LADs) and replication domains of mammalian cells. The overall position of LADs was not altered in trisomic cells; however, the H3K4me3 profile of the trisomic fibroblasts was modified and accurately followed the GEDD pattern. These results indicate that the nuclear compartments of trisomic cells undergo modifications of the chromatin environment influencing the overall transcriptome, and that GEDDs may therefore contribute to some trisomy 21 phenotypes.
    Nature 04/2014; 508(7496):345-50. · 38.60 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Cardiomyopathies are common, seemingly monogenic autosomal dominant cardiac disorders known as the primary cause of sudden cardiac death in young adults. These diseases are characterized by a remarkable genetic heterogeneity, which makes it difficult to unravel the causative mutation in a diagnostic laboratory that is very laborious and expensive by Sanger sequencing. To circumvent these limitations, we explored solutions of high throughput sequencing of targeted exomes with the aim to implement this approach in routine diagnostics. As a first test we designed a capture microarray with the total genomic length of 1 Mbp that includes all exons/splicing sites of 130 genes involved in cardiovascular mendelian disorders and analyzed simultaneously four samples by multiplexing patients with cardiomyopathies or Long-QT syndrome. Pathogenic mutations and variants of unknown significance were found thus resolving the genetic causes of the cardiopathy in three. In the fourth patient the mutation usually associated with hypertrophic cardiomyopathy was found with Long-QT. Further developments to next generation diagnostics are now in progress, and will be also discussed. In conclusion, high throughput sequencing holds considerable promises for molecular diagnosis of highly heterogeneous disorders in clinical practice and allows a better understanding of the complexity of mendelian disorders.
    Molecular Cytogenetics 01/2014; 7(Suppl 1)(7). · 2.66 Impact Factor
  • The Lancet 01/2014; 383(9912):123. · 39.21 Impact Factor
  • Source
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: There is increasing interest in clinical genetics pertaining to the utilization of high-throughput sequencing data for accurate diagnoses of monogenic diseases. Moreover, massive whole-exome sequencing of tumors has provided significant advances in the understanding of cancer development through the recognition of somatic driver variants. To improve the identification of the variants from HTS, we developed VariantMaster, an original program that accurately and efficiently extracts causative variants in familial and sporadic genetic diseases. The algorithm takes into account predicted variants (SNPs and indels) in affected individuals or tumor samples and utilizes the row (BAM) data to robustly estimate the conditional probability of segregation in a family, as well as the probability of it being de novo or somatic. In familial cases, various modes of inheritance are considered: X-linked, autosomal dominant, and recessive (homozygosity or compound heterozygosity). Moreover, VariantMaster integrates phenotypes and genotypes, and employs Annovar to produce additional information such as allelic frequencies in the general population and damaging scores to further reduce the number of putative variants. As a proof of concept, we successfully applied VariantMaster to identify (1) de novo mutations in a previously described data set, (2) causative variants in a rare Mendelian genetic disease, and (3) known and new "driver" mutations in previously reported cancer data sets. Our results demonstrate that VariantMaster is considerably more accurate in terms of precision and sensitivity compared with previously published algorithms.
    Genome Research 01/2014; · 14.40 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Children with Down syndrome (DS) and acute lymphoblastic leukaemia (ALL) have poorer survival and more relapses than non-DS children with ALL, highlighting an urgent need for deeper mechanistic understanding of DS-ALL. Here, using full-exome or cancer genes-targeted sequencing of 42 ALL samples from 39 DS patients, we uncover driver mutations in RAS, (KRAS and NRAS) recurring to a similar extent (15/42) as JAK2 (12/42) mutations or P2RY8-CRLF2 fusions (14/42). RAS mutations are almost completely mutually exclusive with JAK2 mutations (P=0.016), driving a combined total of two-thirds of analysed cases. Clonal architecture analysis reveals that both RAS and JAK2 drove sub-clonal expansions primarily initiated by CRLF2 rearrangements, and/or mutations in chromatin remodellers and lymphocyte differentiation factors. Remarkably, in 2/3 relapsed cases, there is a switch from a primary JAK2- or PTPN11-mutated sub-clone to a RAS-mutated sub-clone in relapse. These results provide important new insights informing the patient stratification strategies for targeted therapeutic approaches for DS-ALL.
    Nature Communications 01/2014; 5:4654. · 10.74 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Down syndrome (trisomy 21) is the most common viable chromosomal disorder with intellectual impairment and several other developmental abnormalities. Here, we report the generation and characterization of induced pluripotent stem cells (iPSCs) derived from monozygotic twins discordant for trisomy 21 in order to eliminate the effects of the variability of genomic background. The alterations observed by genetic analysis at the iPSC level and at first approximation in early development illustrate the developmental disease transcriptional signature of Down syndrome. Moreover, we observed an abnormal neural differentiation of Down syndrome iPSCs in vivo when formed teratoma in NOD-SCID mice, and in vitro when differentiated into neuroprogenitors and neurons. These defects were associated with changes in the architecture and density of neurons, astroglial and oligodendroglial cells together with misexpression of genes involved in neurogenesis, lineage specification and differentiation. Furthermore, we provide novel evidence that dual-specificity tyrosine-(Y)-phosphorylation regulated kinase 1A (DYRK1A) on chromosome 21 likely contributes to these defects. Importantly, we found that targeting DYRK1A pharmacologically or by shRNA results in a considerable correction of these defects.
    EMBO Molecular Medicine 12/2013; · 7.80 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Understanding the molecular basis of cardiomyocyte development is critical for understanding the pathogenesis of pre- and post-natal cardiac disease. MicroRNAs (miRNAs) are post-transcriptional modulators of gene expression that play an important role in many developmental processes. Here, we show that the miR-99a/let-7c cluster, mapping on human chromosome 21, is involved in the control of cardiomyogenesis by altering epigenetic factors. By perturbing miRNA expression in mouse embryonic stem cells, we find that let-7c promotes cardiomyogenesis by upregulating genes involved in mesoderm specification (T/Bra and Nodal) and cardiac differentiation (Mesp1, Nkx2.5 and Tbx5). The action of let-7c is restricted to the early phase of mesoderm formation at the expense of endoderm and its late activation redirects cells toward other mesodermal derivatives. The Polycomb complex group protein Ezh2 is a direct target of let-7c, which promotes cardiac differentiation by modifying the H3K27me3 marks from the promoters of crucial cardiac transcription factors (Nkx2.5, Mef2c, Tbx5). In contrast, miR-99a represses cardiac differentiation via the nucleosome-remodeling factor Smarca5, attenuating the Nodal/Smad2 signaling. We demonstrated that the identified targets are underexpressed in human Down syndrome fetal heart specimens. By perturbing the expression levels of these miRNAs in embryonic stem cells, we were able to demonstrate that these miRNAs control lineage- and stage-specific transcription factors, working in concert with chromatin modifiers to direct cardiomyogenesis.
    Stem Cell Research 11/2013; 12(2):323-337. · 4.47 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Large intergenic noncoding RNAs (lincRNAs) are still poorly functionally characterized. We analyzed the genetic and epigenetic regulation of human lincRNA expression in the GenCord collection by using three cell types from 195 unrelated European individuals. We detected a considerable number of cis expression quantitative trait loci (cis-eQTLs) and demonstrated that the genetic regulation of lincRNA expression is independent of the regulation of neighboring protein-coding genes. lincRNAs have relatively more cis-eQTLs than do equally expressed protein-coding genes with the same exon number. lincRNA cis-eQTLs are located closer to transcription start sites (TSSs) and their effect sizes are higher than cis-eQTLs found for protein-coding genes, suggesting that lincRNA expression levels are less constrained than that of protein-coding genes. Additionally, lincRNA cis-eQTLs can influence the expression level of nearby protein-coding genes and thus could be considered as QTLs for enhancer activity. Enrichment of expressed lincRNA promoters in enhancer marks provides an additional argument for the involvement of lincRNAs in the regulation of transcription in cis. By investigating the epigenetic regulation of lincRNAs, we observed both positive and negative correlations between DNA methylation and gene expression (expression quantitative trait methylation [eQTMs]), as expected, and found that the landscapes of passive and active roles of DNA methylation in gene regulation are similar to protein-coding genes. However, lincRNA eQTMs are located closer to TSSs than are protein-coding gene eQTMs. These similarities and differences in genetic and epigenetic regulation between lincRNAs and protein-coding genes contribute to the elucidation of potential functions of lincRNAs.
    The American Journal of Human Genetics 11/2013; · 11.20 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: We evaluated 25 protocol variants of 14 independent computational methods for exon identification, transcript reconstruction and expression-level quantification from RNA-seq data. Our results show that most algorithms are able to identify discrete transcript components with high success rates but that assembly of complete isoform structures poses a major challenge even when all constituent elements are identified. Expression-level estimates also varied widely across methods, even when based on similar transcript models. Consequently, the complexity of higher eukaryotic genomes imposes severe limitations on transcript recall and splice product discrimination that are likely to remain limiting factors for the analysis of current-generation RNA-seq data.
    Nature Methods 11/2013; advance online publication. · 23.57 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Genome sequencing projects are discovering millions of genetic variants in humans, and interpretation of their functional effects is essential for understanding the genetic basis of variation in human traits. Here we report sequencing and deep analysis of messenger RNA and microRNA from lymphoblastoid cell lines of 462 individuals from the 1000 Genomes Project-the first uniformly processed high-throughput RNA-sequencing data from multiple human populations with high-quality genome sequences. We discover extremely widespread genetic variation affecting the regulation of most genes, with transcript structure and expression level variation being equally common but genetically largely independent. Our characterization of causal regulatory variation sheds light on the cellular mechanisms of regulatory and loss-of-function variation, and allows us to infer putative causal variants for dozens of disease-associated loci. Altogether, this study provides a deep understanding of the cellular mechanisms of transcriptome variation and of the landscape of functional variants in the human genome.
    Nature 09/2013; · 38.60 Impact Factor

Publication Stats

18k Citations
2,740.78 Total Impact Points

Institutions

  • 1998–2014
    • University of Geneva
      • • Department of Genetic Medicine and Development (GEDEV)
      • • Department of Psychiatry
      • • Department of Cell Biology
      Genève, Geneva, Switzerland
    • Università degli Studi di Genova
      Genova, Liguria, Italy
  • 2013
    • Swiss Institute of Bioinformatics
      Lausanne, Vaud, Switzerland
    • Russian Academy of Sciences
      Moskva, Moscow, Russia
  • 2012
    • Institute of Genetics & Hospital for Genetic Diseases
      Bhaganagar, Andhra Pradesh, India
    • The Nebraska Medical Center
      Omaha, Nebraska, United States
  • 2011
    • National Research Council
      • Institute of Biophysics IBF
      Roma, Latium, Italy
    • Creighton University
      • Department of Pharmacology
      Omaha, NE, United States
  • 2009
    • The Walter and Eliza Hall Institute of Medical Research
      • Division of Immunology
      Melbourne, Victoria, Australia
  • 2007–2009
    • Duke University Medical Center
      • • Center for Human Genome Variation
      • • Institute for Genome Sciences and Policy
      Durham, NC, United States
    • University of Minnesota Duluth
      Duluth, Minnesota, United States
    • University of Pennsylvania
      Philadelphia, Pennsylvania, United States
    • Inselspital, Universitätsspital Bern
      Berna, Bern, Switzerland
    • Tel Aviv University
      • Department of Psychiatry
      Tel Aviv, Tel Aviv, Israel
  • 2005–2009
    • Wellcome Trust Sanger Institute
      Cambridge, England, United Kingdom
    • Walter And Eliza Hall Institute For Medical Research
      Melbourne, Victoria, Australia
    • Policlinique Médicale Universitaire Lausanne
      Lausanne, Vaud, Switzerland
    • Stanford Medicine
      • Center for Interdisciplinary Brain Sciences Research
      Stanford, California, United States
  • 2008
    • Academy of Athens
      Athínai, Attica, Greece
    • Telethon Institute of Genetics and Medicine
      Napoli, Campania, Italy
  • 2007–2008
    • University of Oslo
      Kristiania (historical), Oslo County, Norway
  • 2006–2008
    • University Pompeu Fabra
      • Center for Genomic Regulation (CRG)
      Barcelona, Catalonia, Spain
  • 2005–2008
    • University Hospital of Lausanne
      Lausanne, Vaud, Switzerland
    • University of Lausanne
      • • Department of Fundamental Microbiology (DMF)
      • • Center for Integrative Genomics (CIG)
      Lausanne, VD, Switzerland
  • 2006–2007
    • United Arab Emirates University
      Al Ain, Abu Dhabi, United Arab Emirates
  • 1992–2006
    • Johns Hopkins University
      • Department of Pediatrics
      Baltimore, MD, United States
  • 2004
    • University of Padova
      Padua, Veneto, Italy
  • 2003
    • Leiden University Medical Centre
      Leyden, South Holland, Netherlands
  • 2002
    • Stanford University
      • Department of Psychiatry and Behavioral Sciences
      Stanford, CA, United States
  • 1999
    • University of California, Irvine
      • Department of Pediatrics
      Irvine, CA, United States
  • 1996
    • Cantonal Hospital of Schwyz
      Schwyz, Schwyz, Switzerland