CSMD1 regulates brain complement activity and circuit development.

Complement proteins facilitate synaptic elimination during neurodevelopmental pruning, but neural complement regulation is not well understood. CUB and Sushi Multiple Domains 1 (CSMD1) can regulate complement activity in vitro, is expressed in the brain, and is associated with increased schizophrenia risk. Beyond this, little is known about CSMD1 including whether it regulates complement activity in the brain or otherwise plays a role in neurodevelopment. We used biochemical, immunohistochemical, and proteomic techniques to examine the regional, cellular, and subcellular distribution as well as protein interactions of CSMD1 in the brain. To evaluate whether CSMD1 is involved in complement-mediated synapse elimination, we examined Csmd1-knockout mice and CSMD1-knockout human stem cell-derived neurons. We interrogated synapse and circuit development of the mouse visual thalamus, a process that involves complement pathway activity. We also quantified complement deposition on synapses in mouse visual thalamus and on cultured human neurons. Finally, we assessed uptake of synaptosomes by cultured microglia. We found that CSMD1 is present at synapses and interacts with complement proteins in the brain. Mice lacking Csmd1 displayed increased levels of complement component C3, an increased colocalization of C3 with presynaptic terminals, fewer retinogeniculate synapses, and aberrant segregation of eye-specific retinal inputs to the visual thalamus during the critical period of complement-dependent refinement of this circuit. Loss of CSMD1 in vivo enhanced synaptosome engulfment by microglia in vitro, and this effect was dependent on activity of the microglial complement receptor, CR3. Finally, human stem cell-derived neurons lacking CSMD1 were more vulnerable to complement deposition. These data suggest that CSMD1 can function as a regulator of complement-mediated synapse elimination in the brain during development.

Protein interaction networks in the vasculature prioritize genes and pathways underlying coronary artery disease.

Population-based association studies have identified many genetic risk loci for coronary artery disease (CAD), but it is often unclear how genes within these loci are linked to CAD. Here, we perform interaction proteomics for 11 CAD-risk genes to map their protein-protein interactions (PPIs) in human vascular cells and elucidate their roles in CAD. The resulting PPI networks contain interactions that are outside of known biology in the vasculature and are enriched for genes involved in immunity-related and arterial-wall-specific mechanisms. Several PPI networks derived from smooth muscle cells are significantly enriched for genetic variants associated with CAD and related vascular phenotypes. Furthermore, the networks identify 61 genes that are found in genetic loci associated with risk of CAD, prioritizing them as the causal candidates within these loci. These findings indicate that the PPI networks we have generated are a rich resource for guiding future research into the molecular pathogenesis of CAD.

Leveraging polygenic enrichments of gene features to predict genes underlying complex traits and diseases.

Genome-wide association studies (GWASs) are a valuable tool for understanding the biology of complex human traits and diseases, but associated variants rarely point directly to causal genes. In the present study, we introduce a new method, polygenic priority score (PoPS), that learns trait-relevant gene features, such as cell-type-specific expression, to prioritize genes at GWAS loci. Using a large evaluation set of genes with fine-mapped coding variants, we show that PoPS and the closest gene individually outperform other gene prioritization methods, but observe the best overall performance by combining PoPS with orthogonal methods. Using this combined approach, we prioritize 10,642 unique gene-trait pairs across 113 complex traits and diseases with high precision, finding not only well-established gene-trait relationships but nominating new genes at unresolved loci, such as LGR4 for estimated glomerular filtration rate and CCR7 for deep vein thrombosis. Overall, we demonstrate that PoPS provides a powerful addition to the gene prioritization toolbox.

Using brain cell-type-specific protein interactomes to interpret neurodevelopmental genetic signals in schizophrenia.

Genetics have nominated many schizophrenia risk genes and identified convergent signals between schizophrenia and neurodevelopmental disorders. However, functional interpretation of the nominated genes in the relevant brain cell types is often lacking. We executed interaction proteomics for six schizophrenia risk genes that have also been implicated in neurodevelopment in human induced cortical neurons. The resulting protein network is enriched for common variant risk of schizophrenia in Europeans and East Asians, is down-regulated in layer 5/6 cortical neurons of individuals affected by schizophrenia, and can complement fine-mapping and eQTL data to prioritize additional genes in GWAS loci. A sub-network centered on HCN1 is enriched for common variant risk and contains proteins (HCN4 and AKAP11) enriched for rare protein-truncating mutations in individuals with schizophrenia and bipolar disorder. Our findings showcase brain cell-type-specific interactomes as an organizing framework to facilitate interpretation of genetic and transcriptomic data in schizophrenia and its related disorders.

Protein interaction studies in human induced neurons indicate convergent biology underlying autism spectrum disorders.

Autism spectrum disorders (ASDs) have been linked to genes with enriched expression in the brain, but it is unclear how these genes converge into cell-type-specific networks. We built a protein-protein interaction network for 13 ASD-associated genes in human excitatory neurons derived from induced pluripotent stem cells (iPSCs). The network contains newly reported interactions and is enriched for genetic and transcriptional perturbations observed in individuals with ASDs. We leveraged the network data to show that the ASD-linked brain-specific isoform of ANK2 is important for its interactions with synaptic proteins and to characterize a PTEN-AKAP8L interaction that influences neuronal growth. The IGF2BP1-3 complex emerged as a convergent point in the network that may regulate a transcriptional circuit of ASD-associated genes. Our findings showcase cell-type-specific interactomes as a framework to complement genetic and transcriptomic data and illustrate how both individual and convergent interactions can lead to biological insights into ASDs.

Current advancements of modelling schizophrenia using patient-derived induced pluripotent stem cells.

Schizophrenia (SZ) is a severe psychiatric disorder, with a prevalence of 1-2% world-wide and substantial health- and social care costs. The pathology is influenced by both genetic and environmental factors, however the underlying cause still remains elusive. SZ has symptoms including delusions, hallucinations, confused thoughts, diminished emotional responses, social withdrawal and anhedonia. The onset of psychosis is usually in late adolescence or early adulthood. Multiple genome-wide association and whole exome sequencing studies have provided extraordinary insights into the genetic variants underlying familial as well as polygenic forms of the disease. Nonetheless, a major limitation in schizophrenia research remains the lack of clinically relevant animal models, which in turn hampers the development of novel effective therapies for the patients. The emergence of human induced pluripotent stem cell (hiPSC) technology has allowed researchers to work with SZ patient-derived neuronal and glial cell types in vitro and to investigate the molecular basis of the disorder in a human neuronal context. In this review, we summarise findings from available studies using hiPSC-based neural models and discuss how these have provided new insights into molecular and cellular pathways of SZ. Further, we highlight different examples of how these models have shown alterations in neurogenesis, neuronal maturation, neuronal connectivity and synaptic impairment as well as mitochondrial dysfunction and dysregulation of miRNAs in SZ patient-derived cultures compared to controls. We discuss the pros and cons of these models and describe the potential of using such models for deciphering the contribution of specific human neural cell types to the development of the disease.

Autism genes converge on asynchronous development of shared neuron classes.

Genetic risk for autism spectrum disorder (ASD) is associated with hundreds of genes spanning a wide range of biological functions1-6. The alterations in the human brain resulting from mutations in these genes remain unclear. Furthermore, their phenotypic manifestation varies across individuals7,8. Here we used organoid models of the human cerebral cortex to identify cell-type-specific developmental abnormalities that result from haploinsufficiency in three ASD risk genes-SUV420H1 (also known as KMT5B), ARID1B and CHD8-in multiple cell lines from different donors, using single-cell RNA-sequencing (scRNA-seq) analysis of more than 745,000 cells and proteomic analysis of individual organoids, to identify phenotypic convergence. Each of the three mutations confers asynchronous development of two main cortical neuronal lineages-γ-aminobutyric-acid-releasing (GABAergic) neurons and deep-layer excitatory projection neurons-but acts through largely distinct molecular pathways. Although these phenotypes are consistent across cell lines, their expressivity is influenced by the individual genomic context, in a manner that is dependent on both the risk gene and the developmental defect. Calcium imaging in intact organoids shows that these early-stage developmental changes are followed by abnormal circuit activity. This research uncovers cell-type-specific neurodevelopmental abnormalities that are shared across ASD risk genes and are finely modulated by human genomic context, finding convergence in the neurobiological basis of how different risk genes contribute to ASD pathology.

Endothelial ARHGEF26 is an angiogenic factor promoting VEGF signalling.

AIMS: Genetic studies have implicated the ARHGEF26 locus in the risk of coronary artery disease (CAD). However, the causal pathways by which DNA variants at the ARHGEF26 locus confer risk for CAD are incompletely understood. We sought to elucidate the mechanism responsible for the enhanced risk of CAD associated with the ARHGEF26 locus. METHODS AND RESULTS: In a conditional analysis of the ARHGEF26 locus, we show that the sentinel CAD-risk signal is significantly associated with various non-lipid vascular phenotypes. In human endothelial cell (EC), ARHGEF26 promotes the angiogenic capacity, and interacts with known angiogenic factors and pathways. Quantitative mass spectrometry showed that one CAD-risk coding variant, rs12493885 (p.Val29Leu), resulted in a gain-of-function ARHGEF26 that enhances proangiogenic signalling and displays enhanced interactions with several proteins partially related to the angiogenic pathway. ARHGEF26 is required for endothelial angiogenesis by promoting macropinocytosis of Vascular Endothelial Growth Factor Receptor 2 (VEGFR2) on cell membrane and is crucial to Vascular Endothelial Growth Factor (VEGF)-dependent murine vessel sprouting ex vivo. In vivo, global or tissue-specific deletion of ARHGEF26 in EC, but not in vascular smooth muscle cells, significantly reduced atherosclerosis in mice, with enhanced plaque stability. CONCLUSIONS: Our results demonstrate that ARHGEF26 is involved in angiogenesis signaling, and that DNA variants within ARHGEF26 that are associated with CAD risk could affect angiogenic processes by potentiating VEGF-dependent angiogenesis.

Translating polygenic risk scores for clinical use by estimating the confidence bounds of risk prediction.

A promise of genomics in precision medicine is to provide individualized genetic risk predictions. Polygenic risk scores (PRS), computed by aggregating effects from many genomic variants, have been developed as a useful tool in complex disease research. However, the application of PRS as a tool for predicting an individual's disease susceptibility in a clinical setting is challenging because PRS typically provide a relative measure of risk evaluated at the level of a group of people but not at individual level. Here, we introduce a machine-learning technique, Mondrian Cross-Conformal Prediction (MCCP), to estimate the confidence bounds of PRS-to-disease-risk prediction. MCCP can report disease status conditional probability value for each individual and give a prediction at a desired error level. Moreover, with a user-defined prediction error rate, MCCP can estimate the proportion of sample (coverage) with a correct prediction.

ZBTB33 is mutated in clonal hematopoiesis and myelodysplastic syndromes and impacts RNA splicing.

Clonal hematopoiesis results from somatic mutations in cancer driver genes in hematopoietic stem cells. We sought to identify novel drivers of clonal expansion using an unbiased analysis of sequencing data from 84,683 persons and identified common mutations in the 5-methylcytosine reader, ZBTB33, as well as in YLPM1, SRCAP, and ZNF318. We also identified these mutations at low frequency in myelodysplastic syndrome patients. Zbtb33 edited mouse hematopoietic stem and progenitor cells exhibited a competitive advantage in vivo and increased genome-wide intron retention. ZBTB33 mutations potentially link DNA methylation and RNA splicing, the two most commonly mutated pathways in clonal hematopoiesis and MDS.

Coexpression network architecture reveals the brain-wide and multiregional basis of disease susceptibility.

Gene networks have yielded numerous neurobiological insights, yet an integrated view across brain regions is lacking. We leverage RNA sequencing in 864 samples representing 12 brain regions to robustly identify 12 brain-wide, 50 cross-regional and 114 region-specific coexpression modules. Nearly 40% of genes fall into brain-wide modules, while 25% comprise region-specific modules reflecting regional biology, such as oxytocin signaling in the hypothalamus, or addiction pathways in the nucleus accumbens. Schizophrenia and autism genetic risk are enriched in brain-wide and multiregional modules, indicative of broad impact; these modules implicate neuronal proliferation and activity-dependent processes, including endocytosis and splicing, in disease pathophysiology. We find that cell-type-specific long noncoding RNA and gene isoforms contribute substantially to regional synaptic diversity and that constrained, mutation-intolerant genes are primarily enriched in neurons. We leverage these data using an omnigenic-inspired network framework to characterize how coexpression and gene regulatory networks reflect neuropsychiatric disease risk, supporting polygenic models.

Genoppi is an open-source software for robust and standardized integration of proteomic and genetic data.

Combining genetic and cell-type-specific proteomic datasets can generate biological insights and therapeutic hypotheses, but a technical and statistical framework for such analyses is lacking. Here, we present an open-source computational tool called Genoppi (lagelab.org/genoppi) that enables robust, standardized, and intuitive integration of quantitative proteomic results with genetic data. We use Genoppi to analyze 16 cell-type-specific protein interaction datasets of four proteins (BCL2, TDP-43, MDM2, PTEN) involved in cancer and neurological disease. Through systematic quality control of the data and integration with published protein interactions, we show a general pattern of both cell-type-independent and cell-type-specific interactions across three cancer cell types and one human iPSC-derived neuronal cell type. Furthermore, through the integration of proteomic and genetic datasets in Genoppi, our results suggest that the neuron-specific interactions of these proteins are mediating their genetic involvement in neurodegenerative diseases. Importantly, our analyses suggest that human iPSC-derived neurons are a relevant model system for studying the involvement of BCL2 and TDP-43 in amyotrophic lateral sclerosis.

Targeting acute myeloid leukemia dependency on VCP-mediated DNA repair through a selective second-generation small-molecule inhibitor.

The development and survival of cancer cells require adaptive mechanisms to stress. Such adaptations can confer intrinsic vulnerabilities, enabling the selective targeting of cancer cells. Through a pooled in vivo short hairpin RNA (shRNA) screen, we identified the adenosine triphosphatase associated with diverse cellular activities (AAA-ATPase) valosin-containing protein (VCP) as a top stress-related vulnerability in acute myeloid leukemia (AML). We established that AML was the most responsive disease to chemical inhibition of VCP across a panel of 16 cancer types. The sensitivity to VCP inhibition of human AML cell lines, primary patient samples, and syngeneic and xenograft mouse models of AML was validated using VCP-directed shRNAs, overexpression of a dominant-negative VCP mutant, and chemical inhibition. By combining mass spectrometry-based analysis of the VCP interactome and phospho-signaling studies, we determined that VCP is important for ataxia telangiectasia mutated (ATM) kinase activation and subsequent DNA repair through homologous recombination in AML. A second-generation VCP inhibitor, CB-5339, was then developed and characterized. Efficacy and safety of CB-5339 were validated in multiple AML models, including syngeneic and patient-derived xenograft murine models. We further demonstrated that combining DNA-damaging agents, such as anthracyclines, with CB-5339 treatment synergizes to impair leukemic growth in an MLL-AF9-driven AML murine model. These studies support the clinical testing of CB-5339 as a single agent or in combination with standard-of-care DNA-damaging chemotherapy for the treatment of AML.

Systematic auditing is essential to debiasing machine learning in biology.

Biases in data used to train machine learning (ML) models can inflate their prediction performance and confound our understanding of how and what they learn. Although biases are common in biological data, systematic auditing of ML models to identify and eliminate these biases is not a common practice when applying ML in the life sciences. Here we devise a systematic, principled, and general approach to audit ML models in the life sciences. We use this auditing framework to examine biases in three ML applications of therapeutic interest and identify unrecognized biases that hinder the ML process and result in substantially reduced model performance on new datasets. Ultimately, we show that ML models tend to learn primarily from data biases when there is insufficient signal in the data to learn from. We provide detailed protocols, guidelines, and examples of code to enable tailoring of the auditing framework to other biomedical applications.

Leadership.

We asked group leaders how they foster mutually reinforcing research productivity and psychological safety in their teams.

Cohesin mutations alter DNA damage repair and chromatin structure and create therapeutic vulnerabilities in MDS/AML.

The cohesin complex plays an essential role in chromosome maintenance and transcriptional regulation. Recurrent somatic mutations in the cohesin complex are frequent genetic drivers in cancer, including myelodysplastic syndromes (MDS) and acute myeloid leukemia (AML). Here, using genetic dependency screens of stromal antigen 2-mutant (STAG2-mutant) AML, we identified DNA damage repair and replication as genetic dependencies in cohesin-mutant cells. We demonstrated increased levels of DNA damage and sensitivity of cohesin-mutant cells to poly(ADP-ribose) polymerase (PARP) inhibition. We developed a mouse model of MDS in which Stag2 mutations arose as clonal secondary lesions in the background of clonal hematopoiesis driven by tet methylcytosine dioxygenase 2 (Tet2) mutations and demonstrated selective depletion of cohesin-mutant cells with PARP inhibition in vivo. Finally, we demonstrated a shift from STAG2- to STAG1-containing cohesin complexes in cohesin-mutant cells, which was associated with longer DNA loop extrusion, more intermixing of chromatin compartments, and increased interaction with PARP and replication protein A complex. Our findings inform the biology and therapeutic opportunities for cohesin-mutant malignancies.

Comprehensive characterization of amino acid positions in protein structures reveals molecular effect of missense variants.

Interpretation of the colossal number of genetic variants identified from sequencing applications is one of the major bottlenecks in clinical genetics, with the inference of the effect of amino acid-substituting missense variations on protein structure and function being especially challenging. Here we characterize the three-dimensional (3D) amino acid positions affected in pathogenic and population variants from 1,330 disease-associated genes using over 14,000 experimentally solved human protein structures. By measuring the statistical burden of variations (i.e., point mutations) from all genes on 40 3D protein features, accounting for the structural, chemical, and functional context of the variations' positions, we identify features that are generally associated with pathogenic and population missense variants. We then perform the same amino acid-level analysis individually for 24 protein functional classes, which reveals unique characteristics of the positions of the altered amino acids: We observe up to 46% divergence of the class-specific features from the general characteristics obtained by the analysis on all genes, which is consistent with the structural diversity of essential regions across different protein classes. We demonstrate that the function-specific 3D features of the variants match the readouts of mutagenesis experiments for BRCA1 and PTEN, and positively correlate with an independent set of clinically interpreted pathogenic and benign missense variants. Finally, we make our results available through a web server to foster accessibility and downstream research. Our findings represent a crucial step toward translational genetics, from highlighting the impact of mutations on protein structure to rationalizing the variants' pathogenicity in terms of the perturbed molecular mechanisms.

Prediction of cancer driver genes through network-based moment propagation of mutation scores.

MOTIVATION: Gaining a comprehensive understanding of the genetics underlying cancer development and progression is a central goal of biomedical research. Its accomplishment promises key mechanistic, diagnostic and therapeutic insights. One major step in this direction is the identification of genes that drive the emergence of tumors upon mutation. Recent advances in the field of computational biology have shown the potential of combining genetic summary statistics that represent the mutational burden in genes with biological networks, such as protein-protein interaction networks, to identify cancer driver genes. Those approaches superimpose the summary statistics on the nodes in the network, followed by an unsupervised propagation of the node scores through the network. However, this unsupervised setting does not leverage any knowledge on well-established cancer genes, a potentially valuable resource to improve the identification of novel cancer drivers. RESULTS: We develop a novel node embedding that enables classification of cancer driver genes in a supervised setting. The embedding combines a representation of the mutation score distribution in a node's local neighborhood with network propagation. We leverage the knowledge of well-established cancer driver genes to define a positive class, resulting in a partially labeled dataset, and develop a cross-validation scheme to enable supervised prediction. The proposed node embedding followed by a supervised classification improves the predictive performance compared with baseline methods and yields a set of promising genes that constitute candidates for further biological validation. AVAILABILITY AND IMPLEMENTATION: Code available at https://github.com/BorgwardtLab/MoProEmbeddings. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

TCF12 haploinsufficiency causes autosomal dominant Kallmann syndrome and reveals network-level interactions between causal loci.

Dysfunction of the gonadotropin-releasing hormone (GnRH) axis causes a range of reproductive phenotypes resulting from defects in the specification, migration and/or function of GnRH neurons. To identify additional molecular components of this system, we initiated a systematic genetic interrogation of families with isolated GnRH deficiency (IGD). Here, we report 13 families (12 autosomal dominant and one autosomal recessive) with an anosmic form of IGD (Kallmann syndrome) with loss-of-function mutations in TCF12, a locus also known to cause syndromic and non-syndromic craniosynostosis. We show that loss of tcf12 in zebrafish larvae perturbs GnRH neuronal patterning with concomitant attenuation of the orthologous expression of tcf3a/b, encoding a binding partner of TCF12, and stub1, a gene that is both mutated in other syndromic forms of IGD and maps to a TCF12 affinity network. Finally, we report that restored STUB1 mRNA rescues loss of tcf12 in vivo. Our data extend the mutational landscape of IGD, highlight the genetic links between craniofacial patterning and GnRH dysfunction and begin to assemble the functional network that regulates the development of the GnRH axis.