In vivo validation of late-onset Alzheimer's disease genetic risk factors.

INTRODUCTION: Genome-wide association studies have identified over 70 genetic loci associated with late-onset Alzheimer's disease (LOAD), but few candidate polymorphisms have been functionally assessed for disease relevance and mechanism of action. METHODS: Candidate genetic risk variants were informatically prioritized and individually engineered into a LOAD-sensitized mouse model that carries the AD risk variants APOE ε4/ε4 and Trem2*R47H. The potential disease relevance of each model was assessed by comparing brain transcriptomes measured with the Nanostring Mouse AD Panel at 4 and 12 months of age with human study cohorts. RESULTS: We created new models for 11 coding and loss-of-function risk variants. Transcriptomic effects from multiple genetic variants recapitulated a variety of human gene expression patterns observed in LOAD study cohorts. Specific models matched to emerging molecular LOAD subtypes. DISCUSSION: These results provide an initial functionalization of 11 candidate risk variants and identify potential preclinical models for testing targeted therapeutics. HIGHLIGHTS: A novel approach to validate genetic risk factors for late-onset AD (LOAD) is presented. LOAD risk variants were knocked in to conserved mouse loci. Variant effects were assayed by transcriptional analysis. Risk variants in Abca7, Mthfr, Plcg2, and Sorl1 loci modeled molecular signatures of clinical disease. This approach should generate more translationally relevant animal models.

Genetically diverse mouse models of SARS-CoV-2 infection reproduce clinical variation in type I interferon and cytokine responses in COVID-19.

Inflammation in response to severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) infection drives severity of coronavirus disease 2019 (COVID-19) and is influenced by host genetics. To understand mechanisms of inflammation, animal models that reflect genetic diversity and clinical outcomes observed in humans are needed. We report a mouse panel comprising the genetically diverse Collaborative Cross (CC) founder strains crossed to human ACE2 transgenic mice (K18-hACE2) that confers susceptibility to SARS-CoV-2. Infection of CC x K18-hACE2 resulted in a spectrum of survival, viral replication kinetics, and immune profiles. Importantly, in contrast to the K18-hACE2 model, early type I interferon (IFN-I) and regulated proinflammatory responses were required for control of SARS-CoV-2 replication in PWK x K18-hACE2 mice that were highly resistant to disease. Thus, virus dynamics and inflammation observed in COVID-19 can be modeled in diverse mouse strains that provide a genetically tractable platform for understanding anti-coronavirus immunity.

A novel human tau knock-in mouse model reveals interaction of Abeta and human tau under progressing cerebral amyloidosis in 5xFAD mice.

BACKGROUND: Hyperphosphorylation and intraneuronal aggregation of the microtubule-associated protein tau is a major pathological hallmark of Alzheimer's disease (AD) brain. Of special interest is the effect of cerebral amyloid beta deposition, the second main hallmark of AD, on human tau pathology. Therefore, studying the influence of cerebral amyloidosis on human tau in a novel human tau knock-in (htau-KI) mouse model could help to reveal new details on their interplay. METHODS: We studied the effects of a novel human htau-KI under fast-progressing amyloidosis in 5xFAD mice in terms of correlation of gene expression data with human brain regions, development of Alzheimer's-like pathology, synaptic transmission, and behavior. RESULTS: The main findings are an interaction of human beta-amyloid and human tau in crossbred 5xFADxhtau-KI observed at transcriptional level and corroborated by electrophysiology and histopathology. The comparison of gene expression data of the 5xFADxhtau-KI mouse model to 5xFAD, control mice and to human AD patients revealed conspicuous changes in pathways related to mitochondria biology, extracellular matrix, and immune function. These changes were accompanied by plaque-associated MC1-positive pathological tau that required the htau-KI background. LTP deficits were noted in 5xFAD and htau-KI mice in contrast to signs of rescue in 5xFADxhtau-KI mice. Increased frequencies of miniature EPSCs and miniature IPSCs indicated an upregulated presynaptic function in 5xFADxhtau-KI. CONCLUSION: In summary, the multiple interactions observed between knocked-in human tau and the 5xFAD-driven progressing amyloidosis have important implications for future model development in AD.

Genetically diverse mouse models of SARS-CoV-2 infection reproduce clinical variation in type I interferon and cytokine responses in COVID-19.

Inflammation in response to severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) infection drives severity of coronavirus disease 2019 (COVID-19) and is influenced by host genetics. To understand mechanisms of inflammation, animal models that reflect genetic diversity and clinical outcomes observed in humans are needed. We report a mouse panel comprising the genetically diverse Collaborative Cross (CC) founder strains crossed to human ACE2 transgenic mice (K18-hACE2) that confers susceptibility to SARS-CoV-2. Infection of CC x K18- hACE2 resulted in a spectrum of survival, viral replication kinetics, and immune profiles. Importantly, in contrast to the K18-hACE2 model, early type I interferon (IFN-I) and regulated proinflammatory responses were required for control of SARS-CoV-2 replication in PWK x K18-hACE2 mice that were highly resistant to disease. Thus, virus dynamics and inflammation observed in COVID-19 can be modeled in diverse mouse strains that provide a genetically tractable platform for understanding anti-coronavirus immunity.

In vivo validation of late-onset Alzheimer's disease genetic risk factors.

Introduction: Genome-wide association studies have identified over 70 genetic loci associated with late-onset Alzheimer's disease (LOAD), but few candidate polymorphisms have been functionally assessed for disease relevance and mechanism of action. Methods: Candidate genetic risk variants were informatically prioritized and individually engineered into a LOAD-sensitized mouse model that carries the AD risk variants APOE4 and Trem2*R47H. Potential disease relevance of each model was assessed by comparing brain transcriptomes measured with the Nanostring Mouse AD Panel at 4 and 12 months of age with human study cohorts. Results: We created new models for 11 coding and loss-of-function risk variants. Transcriptomic effects from multiple genetic variants recapitulated a variety of human gene expression patterns observed in LOAD study cohorts. Specific models matched to emerging molecular LOAD subtypes. Discussion: These results provide an initial functionalization of 11 candidate risk variants and identify potential preclinical models for testing targeted therapeutics.

Common Genetic Variants Contribute to Risk of Transposition of the Great Arteries.

RATIONALE: Dextro-transposition of the great arteries (D-TGA) is a severe congenital heart defect which affects approximately 1 in 4,000 live births. While there are several reports of D-TGA patients with rare variants in individual genes, the majority of D-TGA cases remain genetically elusive. Familial recurrence patterns and the observation that most cases with D-TGA are sporadic suggest a polygenic inheritance for the disorder, yet this remains unexplored. OBJECTIVE: We sought to study the role of common single nucleotide polymorphisms (SNPs) in risk for D-TGA. METHODS AND RESULTS: We conducted a genome-wide association study in an international set of 1,237 patients with D-TGA and identified a genome-wide significant susceptibility locus on chromosome 3p14.3, which was subsequently replicated in an independent case-control set (rs56219800, meta-analysis P=8.6x10-10, OR=0.69 per C allele). SNP-based heritability analysis showed that 25% of variance in susceptibility to D-TGA may be explained by common variants. A genome-wide polygenic risk score derived from the discovery set was significantly associated to D-TGA in the replication set (P=4x10-5). The genome-wide significant locus (3p14.3) co-localizes with a putative regulatory element that interacts with the promoter of WNT5A, which encodes the Wnt Family Member 5A protein known for its role in cardiac development in mice. We show that this element drives reporter gene activity in the developing heart of mice and zebrafish and is bound by the developmental transcription factor TBX20. We further demonstrate that TBX20 attenuates Wnt5a expression levels in the developing mouse heart. CONCLUSIONS: This work provides support for a polygenic architecture in D-TGA and identifies a susceptibility locus on chromosome 3p14.3 near WNT5A. Genomic and functional data support a causal role of WNT5A at the locus.

Transfer learning-trained convolutional neural networks identify novel MRI biomarkers of Alzheimer's disease progression.

INTRODUCTION: Genome-wide association studies (GWAS) for late onset Alzheimer's disease (AD) may miss genetic variants relevant for delineating disease stages when using clinically defined case/control as a phenotype due to its loose definition and heterogeneity. METHODS: We use a transfer learning technique to train three-dimensional convolutional neural network (CNN) models based on structural magnetic resonance imaging (MRI) from the screening stage in the Alzheimer's Disease Neuroimaging Initiative consortium to derive image features that reflect AD progression. RESULTS: CNN-derived image phenotypes are significantly associated with fasting metabolites related to early lipid metabolic changes as well as insulin resistance and with genetic variants mapped to candidate genes enriched for amyloid beta degradation, tau phosphorylation, calcium ion binding-dependent synaptic loss, APP-regulated inflammation response, and insulin resistance. DISCUSSION: This is the first attempt to show that non-invasive MRI biomarkers are linked to AD progression characteristics, reinforcing their use in early AD diagnosis and monitoring.

Author Correction: Molecular estimation of neurodegeneration pseudotime in older brains.

A Correction to this paper has been published: https://doi.org/10.1038/s41467-020-20261-6.

Molecular estimation of neurodegeneration pseudotime in older brains.

The temporal molecular changes that lead to disease onset and progression in Alzheimer's disease (AD) are still unknown. Here we develop a temporal model for these unobserved molecular changes with a manifold learning method applied to RNA-Seq data collected from human postmortem brain samples collected within the ROS/MAP and Mayo Clinic RNA-Seq studies. We define an ordering across samples based on their similarity in gene expression and use this ordering to estimate the molecular disease stage-or disease pseudotime-for each sample. Disease pseudotime is strongly concordant with the burden of tau (Braak score, P = 1.0 × 10-5), Aß (CERAD score, P = 1.8 × 10-5), and cognitive diagnosis (P = 3.5 × 10-7) of late-onset (LO) AD. Early stage disease pseudotime samples are enriched for controls and show changes in basic cellular functions. Late stage disease pseudotime samples are enriched for late stage AD cases and show changes in neuroinflammation and amyloid pathologic processes. We also identify a set of late stage pseudotime samples that are controls and show changes in genes enriched for protein trafficking, splicing, regulation of apoptosis, and prevention of amyloid cleavage pathways. In summary, we present a method for ordering patients along a trajectory of LOAD disease progression from brain transcriptomic data.

A novel systems biology approach to evaluate mouse models of late-onset Alzheimer's disease.

BACKGROUND: Late-onset Alzheimer's disease (LOAD) is the most common form of dementia worldwide. To date, animal models of Alzheimer's have focused on rare familial mutations, due to a lack of frank neuropathology from models based on common disease genes. Recent multi-cohort studies of postmortem human brain transcriptomes have identified a set of 30 gene co-expression modules associated with LOAD, providing a molecular catalog of relevant endophenotypes. RESULTS: This resource enables precise gene-based alignment between new animal models and human molecular signatures of disease. Here, we describe a new resource to efficiently screen mouse models for LOAD relevance. A new NanoString nCounter® Mouse AD panel was designed to correlate key human disease processes and pathways with mRNA from mouse brains. Analysis of the 5xFAD mouse, a widely used amyloid pathology model, and three mouse models based on LOAD genetics carrying APOE4 and TREM2*R47H alleles demonstrated overlaps with distinct human AD modules that, in turn, were functionally enriched in key disease-associated pathways. Comprehensive comparison with full transcriptome data from same-sample RNA-Seq showed strong correlation between gene expression changes independent of experimental platform. CONCLUSIONS: Taken together, we show that the nCounter Mouse AD panel offers a rapid, cost-effective and highly reproducible approach to assess disease relevance of potential LOAD mouse models.

Meta-Analysis of the Alzheimer's Disease Human Brain Transcriptome and Functional Dissection in Mouse Models.

We present a consensus atlas of the human brain transcriptome in Alzheimer's disease (AD), based on meta-analysis of differential gene expression in 2,114 postmortem samples. We discover 30 brain coexpression modules from seven regions as the major source of AD transcriptional perturbations. We next examine overlap with 251 brain differentially expressed gene sets from mouse models of AD and other neurodegenerative disorders. Human-mouse overlaps highlight responses to amyloid versus tau pathology and reveal age- and sex-dependent expression signatures for disease progression. Human coexpression modules enriched for neuronal and/or microglial genes broadly overlap with mouse models of AD, Huntington's disease, amyotrophic lateral sclerosis, and aging. Other human coexpression modules, including those implicated in proteostasis, are not activated in AD models but rather following other, unexpected genetic manipulations. Our results comprise a cross-species resource, highlighting transcriptional networks altered by human brain pathophysiology and identifying correspondences with mouse models for AD preclinical studies.

Transcriptomic stratification of late-onset Alzheimer's cases reveals novel genetic modifiers of disease pathology.

Late-Onset Alzheimer's disease (LOAD) is a common, complex genetic disorder well-known for its heterogeneous pathology. The genetic heterogeneity underlying common, complex diseases poses a major challenge for targeted therapies and the identification of novel disease-associated variants. Case-control approaches are often limited to examining a specific outcome in a group of heterogenous patients with different clinical characteristics. Here, we developed a novel approach to define relevant transcriptomic endophenotypes and stratify decedents based on molecular profiles in three independent human LOAD cohorts. By integrating post-mortem brain gene co-expression data from 2114 human samples with LOAD, we developed a novel quantitative, composite phenotype that can better account for the heterogeneity in genetic architecture underlying the disease. We used iterative weighted gene co-expression network analysis (WGCNA) to reduce data dimensionality and to isolate gene sets that are highly co-expressed within disease subtypes and represent specific molecular pathways. We then performed single variant association testing using whole genome-sequencing data for the novel composite phenotype in order to identify genetic loci that contribute to disease heterogeneity. Distinct LOAD subtypes were identified for all three study cohorts (two in ROSMAP, three in Mayo Clinic, and two in Mount Sinai Brain Bank). Single variant association analysis identified a genome-wide significant variant in TMEM106B (p-value < 5×10-8, rs1990620G) in the ROSMAP cohort that confers protection from the inflammatory LOAD subtype. Taken together, our novel approach can be used to stratify LOAD into distinct molecular subtypes based on affected disease pathways.

Loss of ADAMTS19 causes progressive non-syndromic heart valve disease.

Valvular heart disease is observed in approximately 2% of the general population1. Although the initial observation is often localized (for example, to the aortic or mitral valve), disease manifestations are regularly observed in the other valves and patients frequently require surgery. Despite the high frequency of heart valve disease, only a handful of genes have so far been identified as the monogenic causes of disease2-7. Here we identify two consanguineous families, each with two affected family members presenting with progressive heart valve disease early in life. Whole-exome sequencing revealed homozygous, truncating nonsense alleles in ADAMTS19 in all four affected individuals. Homozygous knockout mice for Adamts19 show aortic valve dysfunction, recapitulating aspects of the human phenotype. Expression analysis using a lacZ reporter and single-cell RNA sequencing highlight Adamts19 as a novel marker for valvular interstitial cells; inference of gene regulatory networks in valvular interstitial cells positions Adamts19 in a highly discriminatory network driven by the transcription factor lymphoid enhancer-binding factor 1 downstream of the Wnt signaling pathway. Upregulation of endocardial Krüppel-like factor 2 in Adamts19 knockout mice precedes hemodynamic perturbation, showing that a tight balance in the Wnt-Adamts19-Klf2 axis is required for proper valve maturation and maintenance.

Genetic perturbations of disease risk genes in mice capture transcriptomic signatures of late-onset Alzheimer's disease.

BACKGROUND: New genetic and genomic resources have identified multiple genetic risk factors for late-onset Alzheimer's disease (LOAD) and characterized this common dementia at the molecular level. Experimental studies in model organisms can validate these associations and elucidate the links between specific genetic factors and transcriptomic signatures. Animal models based on LOAD-associated genes can potentially connect common genetic variation with LOAD transcriptomes, thereby providing novel insights into basic biological mechanisms underlying the disease. METHODS: We performed RNA-Seq on whole brain samples from a panel of six-month-old female mice, each carrying one of the following mutations: homozygous deletions of Apoe and Clu; hemizygous deletions of Bin1 and Cd2ap; and a transgenic APOEε4. Similar data from a transgenic APP/PS1 model was included for comparison to early-onset variant effects. Weighted gene co-expression network analysis (WGCNA) was used to identify modules of correlated genes and each module was tested for differential expression by strain. We then compared mouse modules with human postmortem brain modules from the Accelerating Medicine's Partnership for AD (AMP-AD) to determine the LOAD-related processes affected by each genetic risk factor. RESULTS: Mouse modules were significantly enriched in multiple AD-related processes, including immune response, inflammation, lipid processing, endocytosis, and synaptic cell function. WGCNA modules were significantly associated with Apoe-/-, APOEε4, Clu-/-, and APP/PS1 mouse models. Apoe-/-, GFAP-driven APOEε4, and APP/PS1 driven modules overlapped with AMP-AD inflammation and microglial modules; Clu-/- driven modules overlapped with synaptic modules; and APP/PS1 modules separately overlapped with lipid-processing and metabolism modules. CONCLUSIONS: This study of genetic mouse models provides a basis to dissect the role of AD risk genes in relevant AD pathologies. We determined that different genetic perturbations affect different molecular mechanisms comprising AD, and mapped specific effects to each risk gene. Our approach provides a platform for further exploration into the causes and progression of AD by assessing animal models at different ages and/or with different combinations of LOAD risk variants.

Identifying and ranking potential driver genes of Alzheimer's disease using multiview evidence aggregation.

MOTIVATION: Late onset Alzheimer's disease is currently a disease with no known effective treatment options. To better understand disease, new multi-omic data-sets have recently been generated with the goal of identifying molecular causes of disease. However, most analytic studies using these datasets focus on uni-modal analysis of the data. Here, we propose a data driven approach to integrate multiple data types and analytic outcomes to aggregate evidences to support the hypothesis that a gene is a genetic driver of the disease. The main algorithmic contributions of our article are: (i) a general machine learning framework to learn the key characteristics of a few known driver genes from multiple feature sets and identifying other potential driver genes which have similar feature representations, and (ii) A flexible ranking scheme with the ability to integrate external validation in the form of Genome Wide Association Study summary statistics. While we currently focus on demonstrating the effectiveness of the approach using different analytic outcomes from RNA-Seq studies, this method is easily generalizable to other data modalities and analysis types. RESULTS: We demonstrate the utility of our machine learning algorithm on two benchmark multiview datasets by significantly outperforming the baseline approaches in predicting missing labels. We then use the algorithm to predict and rank potential drivers of Alzheimer's. We show that our ranked genes show a significant enrichment for single nucleotide polymorphisms associated with Alzheimer's and are enriched in pathways that have been previously associated with the disease. AVAILABILITY AND IMPLEMENTATION: Source code and link to all feature sets is available at https://github.com/Sage-Bionetworks/EvidenceAggregatedDriverRanking.

Meox2 Haploinsufficiency Accelerates Axonal Degeneration in DBA/2J Glaucoma.

Purpose: Glaucoma is a complex disease with major risk factors including advancing age and increased intraocular pressure (IOP). Dissecting these earliest events will likely identify new avenues for therapeutics. Previously, we performed transcriptional profiling in DBA/2J (D2) mice, a widely used mouse model relevant to glaucoma. Here, we use these data to identify and test regulators of early gene expression changes in DBA/2J glaucoma. Methods: Upstream regulator analysis (URA) in Ingenuity Pathway Analysis was performed to identify potential master regulators of differentially expressed genes. The function of one putative regulator, mesenchyme homeobox 2 (Meox2), was tested using a combination of genetic, biochemical, and immunofluorescence approaches. Results: URA identified Meox2 as a potential regulator of early gene expression changes in the optic nerve head (ONH) of DBA/2J mice. Meox2 haploinsufficiency did not affect the characteristic diseases of the iris or IOP elevation seen in DBA/2J mice but did cause a significant increase in the numbers of eyes with axon damage compared to controls. While young mice appeared normal, aged Meox2 haploinsufficient DBA/2J mice showed a 44% reduction in MEOX2 protein levels. This correlated with modulation of age- and disease-specific vascular and myeloid alterations. Conclusions: Our data support a model whereby Meox2 controls IOP-dependent vascular remodeling and neuroinflammation to promote axon survival. Promoting these earliest responses prior to IOP elevation may be a viable neuroprotective strategy to delay or prevent human glaucoma.

Enhancing face validity of mouse models of Alzheimer's disease with natural genetic variation.

Classical laboratory strains show limited genetic diversity and do not harness natural genetic variation. Mouse models relevant to Alzheimer's disease (AD) have largely been developed using these classical laboratory strains, such as C57BL/6J (B6), and this has likely contributed to the failure of translation of findings from mice to the clinic. Therefore, here we test the potential for natural genetic variation to enhance the translatability of AD mouse models. Two widely used AD-relevant transgenes, APPswe and PS1de9 (APP/PS1), were backcrossed from B6 to three wild-derived strains CAST/EiJ, WSB/EiJ, PWK/PhJ, representative of three Mus musculus subspecies. These new AD strains were characterized using metabolic, functional, neuropathological and transcriptional assays. Strain-, sex- and genotype-specific differences were observed in cognitive ability, neurodegeneration, plaque load, cerebrovascular health and cerebral amyloid angiopathy. Analyses of brain transcriptional data showed strain was the greatest driver of variation. We identified significant variation in myeloid cell numbers in wild type mice of different strains as well as significant differences in plaque-associated myeloid responses in APP/PS1 mice between the strains. Collectively, these data support the use of wild-derived strains to better model the complexity of human AD.

Copy number variation analysis in bicuspid aortic valve-related aortopathy identifies TBX20 as a contributing gene.

Bicuspid aortic valve (BAV) is the most common congenital heart defect (CHD), affecting 1-2% of the population. BAV is associated with thoracic aortic aneurysms (TAAs). Deleterious copy number variations (CNVs) were found previously in up to 10% of CHD cases. This study aimed at unravelling the contribution of deleterious deletions or duplications in 95 unrelated BAV/TAA patients. Seven unique or rare CNVs were validated, harbouring protein-coding genes with a role in the cardiovascular system. Based on the presence of overlapping CNVs in patients with cardiovascular phenotypes in the DECIPHER database, the identification of similar CNVs in whole-exome sequencing data of 67 BAV/TAA patients and suggested topological domain involvement from Hi-C data, supportive evidence was obtained for two genes (DGCR6 and TBX20) of the seven initially validated CNVs. A rare variant burden analysis using next-generation sequencing data from 637 BAV/TAA patients was performed for these two candidate genes. This revealed a suggestive genetic role for TBX20 in BAV/TAA aetiology, further reinforced by segregation of a rare TBX20 variant with the phenotype within a BAV/TAA family. To conclude, our results do not confirm a significant contribution for deleterious CNVs in BAV/TAA as only one potentially pathogenic CNV (1.05%) was identified. We cannot exclude the possibility that BAV/TAA is occasionally attributed to causal CNVs though, or that certain CNVs act as genetic risk factors by creating a sensitised background for BAV/TAA. Finally, accumulative evidence for TBX20 involvement in BAV/TAA aetiology underlines the importance of this transcription factor in cardiovascular disease.

Variable outcomes of human heart attack recapitulated in genetically diverse mice.

Clinical variation in patient responses to myocardial infarction (MI) has been difficult to model in laboratory animals. To assess the genetic basis of variation in outcomes after heart attack, we characterized responses to acute MI in the Collaborative Cross (CC), a multi-parental panel of genetically diverse mouse strains. Striking differences in post-MI functional, morphological, and myocardial scar features were detected across 32 CC founder and recombinant inbred strains. Transcriptomic analyses revealed a plausible link between increased intrinsic cardiac oxidative phosphorylation levels and MI-induced heart failure. The emergence of significant quantitative trait loci for several post-MI traits indicates that utilizing CC strains is a valid approach for gene network discovery in cardiovascular disease, enabling more accurate clinical risk assessment and prediction.

ROBO4 variants predispose individuals to bicuspid aortic valve and thoracic aortic aneurysm.

Bicuspid aortic valve (BAV) is a common congenital heart defect (population incidence, 1-2%)1-3 that frequently presents with ascending aortic aneurysm (AscAA)4. BAV/AscAA shows autosomal dominant inheritance with incomplete penetrance and male predominance. Causative gene mutations (for example, NOTCH1, SMAD6) are known for ≤1% of nonsyndromic BAV cases with and without AscAA5-8, impeding mechanistic insight and development of therapeutic strategies. Here, we report the identification of variants in ROBO4 (which encodes a factor known to contribute to endothelial performance) that segregate with disease in two families. Targeted sequencing of ROBO4 showed enrichment for rare variants in BAV/AscAA probands compared with controls. Targeted silencing of ROBO4 or mutant ROBO4 expression in endothelial cell lines results in impaired barrier function and a synthetic repertoire suggestive of endothelial-to-mesenchymal transition. This is consistent with BAV/AscAA-associated findings in patients and in animal models deficient for ROBO4. These data identify a novel endothelial etiology for this common human disease phenotype.