VExD: A curated resource for human gene expression following viral infection

MotivationHuman viruses cause significant mortality, morbidity, and economic disruption worldwide. The human gene expression response to viral infection can yield important insights into the detrimental effects to the host. To date, hundreds of studies have performed genome-scale profiling of the effect of viral infection on human gene expression. However, no resource exists that aggregates human expression results across multiple studies, viruses, and tissue types.ResultsWe developed the Virus Expression Database (VExD), a comprehensive curated resource of transcriptomic studies of viral infection in human cells. We have processed all studies within VExD in a uniform manner, allowing users to easily compare human gene expression changes across conditions.Availability and ImplementationVExD is freely accessible at https://vexd.cchmc.org for all modern web browsers. An Application Programming Interface (API) for VExD is also available. The source code is available at https://github.com/pdexheimer/[email protected], [email protected]

Dynamically expressed single ELAV/Hu orthologue elavl2 of bees is required for learning and memory

Changes in gene expression are a hallmark of learning and memory consolidation. Little is known about how alternative mRNA processing, particularly abundant in neuron-specific genes, contributes to these processes. Prototype RNA binding proteins of the neuronally expressed ELAV/Hu family are candidates for roles in learning and memory, but their capacity to cross-regulate and take over each other’s functions complicate substantiation of such links. Honey bees Apis mellifera have only one elav/Hu family gene elavl2, that has functionally diversified by increasing alternative splicing including an evolutionary conserved microexon. RNAi knockdown demonstrates that ELAVL2 is required for learning and memory in bees. ELAVL2 is dynamically expressed with altered alternative splicing and subcellular localization in mushroom bodies, but not in other brain regions. Expression and alternative splicing of elavl2 change during memory consolidation illustrating an alternative mRNA processing program as part of a local gene expression response underlying memory consolidation.

Genetic control of the dynamic transcriptional response to immune stimuli and glucocorticoids at single cell resolution

Synthetic glucocorticoids are used to treat many immune conditions, such as asthma and severe COVID-19. Single cell data capture fine-grained details of transcriptional variability and dynamics to gain a better understanding of the molecular underpinnings of inter-individual variation in drug response. We used single cell RNA-seq to study the dynamics of the transcriptional response to glucocorticoids in activated PBMCs from African American donors. We employed novel statistical approaches to calculate a mean-independent measure of gene expression variability and a measure of transcriptional response pseudotime. We demonstrated that glucocorticoids reverse the effects of immune stimulation on both gene expression mean and variability. Our novel measure of gene expression response dynamics separated cells by response status and identified dynamic transcriptional patterns along the response pseudotime. We identified genetic variants regulating gene expression mean and variability, including treatment-specific effects, and demonstrated widespread genetic regulation of the transcriptional dynamics of the gene expression response.

Integrative genetic analysis of the amyotrophic lateral sclerosis spinal cord implicates glial activation and suggests new risk genes

Amyotrophic lateral sclerosis (ALS) is a progressively fatal neurodegenerative disease affecting motor neurons in the brain and spinal cord. We used 380 post-mortem tissue RNA-seq transcriptomes from 154 ALS cases and 49 control individuals from cervical, thoracic, and lumbar spinal cord segments to investigate the gene expression response to ALS. We observed an increase in microglia and astrocyte expression, accompanied by a decrease in oligodendrocytes. By creating a gene co-expression network in the ALS samples, we identify several activated microglia modules that negatively correlate with retrospective disease duration. We map molecular quantitative trait loci and find several potential ALS risk loci that may act through gene expression or splicing in the spinal cord and assign putative cell-types for FNBP1, ACSL5, SH3RF1 and NFASC. Finally, we outline how repeat expansions that alter splicing of C9orf72 are tagged by common variants, and use this to suggest ATXN3 as a putative risk gene.

Chromatin Dynamics and Gene Expression Response to Heat Exposure in Field-Conditioned versus Laboratory-Cultured Nematostella vectensis

Organisms’ survival is associated with the ability to respond to natural or anthropogenic environmental stressors. Frequently, these responses involve changes in gene regulation and expression, consequently altering physiology, development, or behavior. Here, we present modifications in response to heat exposure that mimics extreme summertime field conditions of lab-cultured and field-conditioned Nematostella vectensis. Using ATAC-seq and RNA-seq data, we found that field-conditioned animals had a more concentrated reaction to short-term thermal stress, expressed as enrichment of the DNA repair mechanism pathway. By contrast, lab animals had a more diffuse reaction that involved a larger number of differentially expressed genes and enriched pathways, including amino acid metabolism. Our results demonstrate that pre-conditioning affects the ability to respond efficiently to heat exposure in terms of both chromatin accessibility and gene expression and reinforces the importance of experimentally addressing ecological questions in the field.

MichiGAN: sampling from disentangled representations of single-cell data using generative adversarial networks

Deep generative models such as variational autoencoders (VAEs) and generative adversarial networks (GANs) generate and manipulate high-dimensional images. We systematically assess the complementary strengths and weaknesses of these models on single-cell gene expression data. We also develop MichiGAN, a novel neural network that combines the strengths of VAEs and GANs to sample from disentangled representations without sacrificing data generation quality. We learn disentangled representations of three large single-cell RNA-seq datasets and use MichiGAN to sample from these representations. MichiGAN allows us to manipulate semantically distinct aspects of cellular identity and predict single-cell gene expression response to drug treatment.