Single-cell transcriptomics of human iPSC differentiation dynamics reveal a core molecular network of Parkinson’s disease

Parkinson’s disease (PD) is the second-most prevalent neurodegenerative disorder, characterized by the loss of dopaminergic neurons (mDA) in the midbrain. The underlying mechanisms are only partly understood and there is no treatment to reverse PD progression. Here, we investigated the disease mechanism using mDA neurons differentiated from human induced pluripotent stem cells (hiPSCs) carrying the ILE368ASN mutation within the PINK1 gene, which is strongly associated with PD. Single-cell RNA sequencing (RNAseq) and gene expression analysis of a PINK1-ILE368ASN and a control cell line identified genes differentially expressed during mDA neuron differentiation. Network analysis revealed that these genes form a core network, members of which interact with all known 19 protein-coding Parkinson’s disease-associated genes. This core network encompasses key PD-associated pathways, including ubiquitination, mitochondrial function, protein processing, RNA metabolism, and vesicular transport. Proteomics analysis showed a consistent alteration in proteins of dopamine metabolism, indicating a defect of dopaminergic metabolism in PINK1-ILE368ASN neurons. Our findings suggest the existence of a network onto which pathways associated with PD pathology converge, and offers an inclusive interpretation of the phenotypic heterogeneity of PD.

Singleton exome sequencing of 90 fetuses with ultrasound anomalies revealing novel disease-causing variants and genotype–phenotype correlations

Exome sequencing has been increasingly implemented in prenatal genetic testing for fetuses with morphological abnormalities but normal rapid aneuploidy detection and microarray analysis. We present a retrospective study of 90 fetuses with different abnormal ultrasound findings, in which we employed the singleton exome sequencing (sES; 75 fetuses) or to a lesser extent (15 fetuses) a multigene panel analysis of 6713 genes as a primary tool for the detection of monogenic diseases. The detection rate of pathogenic or likely pathogenic variants in this study was 34.4%. The highest diagnostic rate of 56% was in fetuses with multiple anomalies, followed by cases with skeletal or renal abnormalities (diagnostic rate of 50%, respectively). We report 20 novel disease-causing variants in different known disease-associated genes and new genotype–phenotype associations for the genes KMT2D, MN1, CDK10, and EXOC3L2. Based on our data, we postulate that sES of fetal index cases with a concurrent sampling of parental probes for targeted testing of the origin of detected fetal variants could be a suitable tool to obtain reliable and rapid prenatal results, particularly in situations where a trio analysis is not possible.

How much do model organism phenotypes contribute to the computational identification of human disease genes?

Computing phenotypic similarity has been shown to be useful in identification of new disease genes and for rare disease diagnostic support. Genotype--phenotype data from orthologous genes in model organisms can compensate for lack of human data to greatly increase genome coverage. Work over the past decade has demonstrated the power of cross-species phenotype comparisons, and several cross-species phenotype ontologies have been developed for this purpose. The relative contribution of different model organisms to identifying disease-associated genes using computational approaches is not yet fully explored. We use methods based on phenotype ontologies to semantically relate phenotypes resulting from loss-of-function mutations in different model organisms to disease-associated phenotypes in humans. Semantic machine learning methods are used to measure how much different model organisms contribute to the identification of known human gene--disease associations. We find that only mouse phenotypes can accurately predict human gene--disease associations. Our work has implications for the future development of integrated phenotype ontologies, as well as for the use of model organism phenotypes in human genetic variant interpretation.

Neuronal ROS-induced glial lipid droplet formation is altered by loss of Alzheimer’s disease–associated genes

A growing list of Alzheimer’s disease (AD) genetic risk factors is being identified, but the contribution of each variant to disease mechanism remains largely unknown. We have previously shown that elevated levels of reactive oxygen species (ROS) induces lipid synthesis in neurons leading to the sequestration of peroxidated lipids in glial lipid droplets (LD), delaying neurotoxicity. This neuron-to-glia lipid transport is APOD/E-dependent. To identify proteins that modulate these neuroprotective effects, we tested the role of AD risk genes in ROS-induced LD formation and demonstrate that several genes impact neuroprotective LD formation, including homologs of human ABCA1, ABCA7, VLDLR, VPS26, VPS35, AP2A, PICALM, and CD2AP. Our data also show that ROS enhances Aβ42 phenotypes in flies and mice. Finally, a peptide agonist of ABCA1 restores glial LD formation in a humanized APOE4 fly model, highlighting a potentially therapeutic avenue to prevent ROS-induced neurotoxicity. This study places many AD genetic risk factors in a ROS-induced neuron-to-glia lipid transfer pathway with a critical role in protecting against neurotoxicity.

Cellular development and evolution of the mammalian cerebellum

The expansion of the neocortex, one of the hallmarks of mammalian evolution, was accompanied by an increase in the number of cerebellar neurons. However, little is known about the evolution of the cellular programs underlying cerebellum development in mammals. In this study, we generated single-nucleus RNA-sequencing data for ~400,000 cells to trace the development of the cerebellum from early neurogenesis to adulthood in human, mouse, and the marsupial opossum. Our cross-species analyses revealed that the cellular composition and differentiation dynamics throughout cerebellum development are largely conserved, except for human Purkinje cells. Global transcriptome profiles, conserved cell state markers, and gene expression trajectories across neuronal differentiation show that the cerebellar cell type-defining programs have been overall preserved for at least 160 million years. However, we also discovered differences. We identified 3,586 genes that either gained or lost expression in cerebellar cells in one of the species, and 541 genes that evolved new expression trajectories during neuronal differentiation. The potential functional relevance of these cross-species differences is highlighted by the diverged expression patterns of several human disease-associated genes. Altogether, our study reveals shared and lineage-specific programs governing the cellular development of the mammalian cerebellum, and expands our understanding of the evolution of mammalian organ development.

Selection shapes the landscape of functional variation in wild house mice

Background Through human-aided dispersal over the last ~ 10,000 years, house mice (Mus musculus) have recently colonized diverse habitats across the globe, promoting the emergence of new traits that confer adaptive advantages in distinct environments. Despite their status as the premier mammalian model system, the impact of this demographic and selective history on the global patterning of disease-relevant trait variation in wild mouse populations is poorly understood. Results Here, we leveraged 154 whole-genome sequences from diverse wild house mouse populations to survey the geographic organization of functional variation and systematically identify signals of positive selection. We show that a significant proportion of wild mouse variation is private to single populations, including numerous predicted functional alleles. In addition, we report strong signals of positive selection at many genes associated with both complex and Mendelian diseases in humans. Notably, we detect a significant excess of selection signals at disease-associated genes relative to null expectations, pointing to the important role of adaptation in shaping the landscape of functional variation in wild mouse populations. We also uncover strong signals of selection at multiple genes involved in starch digestion, including Mgam and Amy1. We speculate that the successful emergence of the human-mouse commensalism may have been facilitated, in part, by dietary adaptations at these loci. Finally, our work uncovers multiple cryptic structural variants that manifest as putative signals of positive selection, highlighting an important and under-appreciated source of false-positive signals in genome-wide selection scans. Conclusions Overall, our findings highlight the role of adaptation in shaping wild mouse genetic variation at human disease-associated genes. Our work also highlights the biomedical relevance of wild mouse genetic diversity and underscores the potential for targeted sampling of mice from specific populations as a strategy for developing effective new mouse models of both rare and common human diseases.

Molecular Classification and Interpretation of Amyotrophic Lateral Sclerosis Using Deep Convolution Neural Networks and Shapley Values

Amyotrophic lateral sclerosis (ALS) is a prototypical neurodegenerative disease characterized by progressive degeneration of motor neurons to severely effect the functionality to control voluntary muscle movement. Most of the non-additive genetic aberrations responsible for ALS make its molecular classification very challenging along with limited sample size, curse of dimensionality, class imbalance and noise in the data. Deep learning methods have been successful in many other related areas but have low minority class accuracy and suffer from the lack of explainability when used directly with RNA expression features for ALS molecular classification. In this paper, we propose a deep-learning-based molecular ALS classification and interpretation framework. Our framework is based on training a convolution neural network (CNN) on images obtained from converting RNA expression values into pixels based on DeepInsight similarity technique. Then, we employed Shapley additive explanations (SHAP) to extract pixels with higher relevance to ALS classifications. These pixels were mapped back to the genes which made them up. This enabled us to classify ALS samples with high accuracy for a minority class along with identifying genes that might be playing an important role in ALS molecular classifications. Taken together with RNA expression images classified with CNN, our preliminary analysis of the genes identified by SHAP interpretation demonstrate the value of utilizing Machine Learning to perform molecular classification of ALS and uncover disease-associated genes.

Inducible transcriptional condensates drive 3D genome reorganization in the heat shock response

Mammalian developmental and disease-associated genes concentrate large quantities of the transcriptional machinery by forming membrane-less compartments known as transcriptional condensates. However, it is unknown whether these structures are evolutionarily conserved, capable of stress-inducible gene activation or involved in 3D genome reorganization. Here, we identify inducible transcriptional condensates in the yeast heat shock response (HSR). HSR condensates are biophysically dynamic spatiotemporal clusters of the sequence-specific transcription factor Heat shock factor 1 (Hsf1) with Mediator and RNA Pol II. Uniquely, HSR condensates drive the coalescence of multiple Hsf1 target genes, even those located on different chromosomes. Binding of the chaperone Hsp70 to a site on Hsf1 represses clustering, while an intrinsically disordered region on Hsf1 promotes condensate formation and intergenic interactions. Mutation of both Hsf1 determinants reprograms HSR condensates to become mammalian-like: constitutively active without intergenic coalescence. These results suggest that transcriptional condensates are ancient and flexible compartments of eukaryotic gene control.