Cell population-specific expression analysis of human cerebellum

Deconvolution from bulk gene expression by leveraging sample-wise and gene-wise similarities and single-cell RNA-Seq data

BACKGROUND

The widely adopted bulk RNA-seq measures the gene expression average of cells, masking cell type heterogeneity, which confounds downstream analyses. Therefore, identifying the cellular composition and cell type-specific gene expression profiles (GEPs) facilitates the study of the underlying mechanisms of various biological processes. Although single-cell RNA-seq focuses on cell type heterogeneity in gene expression, it requires specialized and expensive resources and currently is not practical for a large number of samples or a routine clinical setting. Recently, computational deconvolution methodologies have been developed, while many of them only estimate cell type composition or cell type-specific GEPs by requiring the other as input. The development of more accurate deconvolution methods to infer cell type abundance and cell type-specific GEPs is still essential.

RESULTS

We propose a new deconvolution algorithm, DSSC, which infers cell type-specific gene expression and cell type proportions of heterogeneous samples simultaneously by leveraging gene-gene and sample-sample similarities in bulk expression and single-cell RNA-seq data. Through comparisons with the other existing methods, we demonstrate that DSSC is effective in inferring both cell type proportions and cell type-specific GEPs across simulated pseudo-bulk data (including intra-dataset and inter-dataset simulations) and experimental bulk data (including mixture data and real experimental data). DSSC shows robustness to the change of marker gene number and sample size and also has cost and time efficiencies.

CONCLUSIONS

DSSC provides a practical and promising alternative to the experimental techniques to characterize cellular composition and heterogeneity in the gene expression of heterogeneous samples.

From whole bodies to single cells: A guide to transcriptomic approaches for ecology and evolutionary biology

RNA sequencing (RNAseq) methodology has experienced a burst of technological developments in the last decade, which has opened up opportunities for studying the mechanisms of adaptation to environmental factors at both the organismal and cellular level. Selecting the most suitable experimental approach for specific research questions and model systems can, however, be a challenge and researchers in ecology and evolution are commonly faced with the choice of whether to study gene expression variation in whole bodies, specific tissues, and/or single cells. A wide range of sometimes polarised opinions exists over which approach is best. Here, we highlight the advantages and disadvantages of each of these approaches to provide a guide to help researchers make informed decisions and maximise the power of their study. Using illustrative examples of various ecological and evolutionary research questions, we guide the readers through the different RNAseq approaches and help them identify the most suitable design for their own projects.

Identifying tumor type and cell type-specific gene expression alterations in pediatric central nervous system tumors

Central nervous system (CNS) tumors are the leading cause of pediatric cancer death, and these patients have an increased risk for developing secondary neoplasms. Due to the low prevalence of pediatric CNS tumors, major advances in targeted therapies have been lagging compared to other adult tumors. We collect single nuclei RNA-seq data from 84,700 nuclei of 35 pediatric CNS tumors and three non-tumoral pediatric brain tissues and characterize tumor heterogeneity and transcriptomic alterations. We distinguish cell subpopulations associated with specific tumor types including radial glial cells in ependymomas and oligodendrocyte precursor cells in astrocytomas. In tumors, we observe pathways important in neural stem cell-like populations, a cell type previously associated with therapy resistance. Lastly, we identify transcriptomic alterations among pediatric CNS tumor types compared to non-tumor tissues, while accounting for cell type effects on gene expression. Our results suggest potential tumor type and cell type-specific targets for pediatric CNS tumor treatment. Here we address current gaps in understanding single nuclei gene expression profiles of previously under-investigated tumor types and enhance current knowledge of gene expression profiles of single cells of various pediatric CNS tumors.

imply: improving cell-type deconvolution accuracy using personalized reference profiles

Using computational tools, bulk transcriptomics can be deconvoluted to estimate the abundance of constituent cell types. However, existing deconvolution methods are conditioned on the assumption that the whole study population is served by a single reference panel, ignoring person-to-person heterogeneity. Here, we present imply, a novel algorithm to deconvolute cell type proportions using personalized reference panels. Simulation studies demonstrate reduced bias compared with existing methods. Real data analyses on longitudinal consortia show disparities in cell type proportions are associated with several disease phenotypes in Type 1 diabetes and Parkinson's disease. imply is available through the R/Bioconductor package ISLET at https://bioconductor.org/packages/ISLET/ .

Tumor type and cell type-specific gene expression alterations in diverse pediatric central nervous system tumors identified using single nuclei RNA-seq

Central nervous system (CNS) tumors are the leading cause of pediatric cancer death, and these patients have an increased risk for developing secondary neoplasms. Due to the low prevalence of pediatric CNS tumors, major advances in targeted therapies have been lagging compared to other adult tumors. We collected single nuclei RNA-seq data from 35 pediatric CNS tumors and three non-tumoral pediatric brain tissues (84,700 nuclei) and characterized tumor heterogeneity and transcriptomic alterations. We distinguished cell subpopulations associated with specific tumor types including radial glial cells in ependymomas and oligodendrocyte precursor cells in astrocytomas. In tumors, we observed pathways important in neural stem cell-like populations, a cell type previously associated with therapy resistance. Lastly, we identified transcriptomic alterations among pediatric CNS tumor types compared to non-tumor tissues, while accounting for cell type effects on gene expression. Our results suggest potential tumor type and cell type-specific targets for pediatric CNS tumor treatment. In this study, we address current gaps in understanding single nuclei gene expression profiles of previously uninvestigated tumor types and enhance current knowledge of gene expression profiles of single cells of various pediatric CNS tumors.

Transcriptomic analysis of isolated and pooled human postmortem cerebellar Purkinje cells in autism spectrum disorders

At present, the neuronal mechanisms underlying the diagnosis of autism spectrum disorder (ASD) have not been established. However, studies from human postmortem ASD brains have consistently revealed disruptions in cerebellar circuitry, specifically reductions in Purkinje cell (PC) number and size. Alterations in cerebellar circuitry would have important implications for information processing within the cerebellum and affect a wide range of human motor and non-motor behaviors. Laser capture microdissection was performed to obtain pure PC populations from a cohort of postmortem control and ASD cases and transcriptional profiles were compared. The 427 differentially expressed genes were enriched for gene ontology biological processes related to developmental organization/connectivity, extracellular matrix organization, calcium ion response, immune function and PC signaling alterations. Given the complexity of PCs and their far-ranging roles in response to sensory stimuli and motor function regulation, understanding transcriptional differences in this subset of cerebellar cells in ASD may inform on convergent pathways that impact neuronal function.

Niche deconvolution of the glioblastoma proteome reveals a distinct infiltrative phenotype within the proneural transcriptomic subgroup

Glioblastoma is often subdivided into three transcriptional subtypes (classical, proneural, mesenchymal) based on bulk RNA signatures that correlate with distinct genetic and clinical features. Potential cellular-level differences of these subgroups, such as the relative proportions of glioblastoma’s hallmark histopathologic features (e.g. brain infiltration, microvascular proliferation), may provide insight into their distinct phenotypes but are, however, not well understood. Here we leverage machine learning and reference proteomic profiles derived from micro-dissected samples of these major histomorphologic glioblastoma features to deconvolute and estimate niche proportions in an independent proteogenomically-characterized cohort. This approach revealed a strong association of the proneural transcriptional subtype with a diffusely infiltrating phenotype. Similarly, enrichment of a microvascular proliferation proteomic signature was seen within the mesenchymal subtype. This study is the first to link differences in the cellular pathology signatures and transcriptional profiles of glioblastoma, providing potential new insights into the genetic drivers and poor treatment response of specific subsets of glioblastomas.

GEOMETRIC STRUCTURE GUIDED MODEL AND ALGORITHMS FOR COMPLETE DECONVOLUTION OF GENE EXPRESSION DATA

Complete deconvolution analysis for bulk RNA-seq data is important and helpful to distinguish whether the differences of disease-associated GEPs (gene expression profiles) in tissues of patients and normal controls are due to changes in cellular composition of tissue samples, or due to GEPs changes in specific cells. One of the major techniques to perform complete deconvolution is nonnegative matrix factorization (NMF), which also has a wide-range of applications in the machine learning community. However, the NMF is a well-known strongly ill-posed problem, so a direct application of NMF to RNA-seq data will suffer severe difficulties in the interpretability of solutions. In this paper, we develop an NMF-based mathematical model and corresponding computational algorithms to improve the solution identifiability of deconvoluting bulk RNA-seq data. In our approach, we combine the biological concept of marker genes with the solvability conditions of the NMF theories, and develop a geometric structures guided optimization model. In this strategy, the geometric structure of bulk tissue data is first explored by the spectral clustering technique. Then, the identified information of marker genes is integrated as solvability constraints, while the overall correlation graph is used as manifold regularization. Both synthetic and biological data are used to validate the proposed model and algorithms, from which solution interpretability and accuracy are significantly improved.

Comprehensive evaluation of deconvolution methods for human brain gene expression

Transcriptome deconvolution aims to estimate the cellular composition of an RNA sample from its gene expression data, which in turn can be used to correct for composition differences across samples. The human brain is unique in its transcriptomic diversity, and comprises a complex mixture of cell-types, including transcriptionally similar subtypes of neurons. Here, we carry out a comprehensive evaluation of deconvolution methods for human brain transcriptome data, and assess the tissue-specificity of our key observations by comparison with human pancreas and heart. We evaluate eight transcriptome deconvolution approaches and nine cell-type signatures, testing the accuracy of deconvolution using in silico mixtures of single-cell RNA-seq data, RNA mixtures, as well as nearly 2000 human brain samples. Our results identify the main factors that drive deconvolution accuracy for brain data, and highlight the importance of biological factors influencing cell-type signatures, such as brain region and in vitro cell culturing.

Transcriptome deconvolution aims to estimate cellular composition based on gene expression data. Here the authors evaluate deconvolution methods for human brain transcriptome and conclude that partial deconvolution algorithms work best, but that appropriate cell-type signatures are also important.

A2Sign: Agnostic Algorithms for Signatures-a universal method for identifying molecular signatures from transcriptomic datasets prior to cell-type deconvolution

MOTIVATION

Molecular signatures are critical for inferring the proportions of cell types from bulk transcriptomics data. However, the identification of these signatures is based on a methodology that relies on prior biological knowledge of the cell types being studied. When working with less known biological material, a data-driven approach is required to uncover the underlying classes and generate ad hoc signatures from healthy or pathogenic tissue.

RESULTS

We present a new approach, A2Sign: Agnostic Algorithms for Signatures, based on a non-negative tensor factorization (NTF) strategy that allows us to identify cell-type-specific molecular signatures, greatly reduce collinearities and also account for inter-individual variability. We propose a global framework that can be applied to uncover molecular signatures for cell-type deconvolution in arbitrary tissues using bulk transcriptome data. We also present two new molecular signatures for deconvolution of up to 16 immune cell types using microarray or RNA-seq data.

AVAILABILITY AND IMPLEMENTATION

All steps of our analysis were implemented in annotated Python notebooks (https://github.com/paulfogel/A2SIGN). To perform NTF, we used the NMTF package, which can be downloaded using Python pip install.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

A Novel SCA3 Knock-in Mouse Model Mimics the Human SCA3 Disease Phenotype Including Neuropathological, Behavioral, and Transcriptional Abnormalities Especially in Oligodendrocytes

Spinocerebellar ataxia type 3 is the most common autosomal dominant inherited ataxia worldwide, caused by a CAG repeat expansion in the Ataxin-3 gene resulting in a polyglutamine (polyQ)-expansion in the corresponding protein. The disease is characterized by neuropathological, phenotypical, and specific transcriptional changes in affected brain regions. So far, there is no mouse model available representing all the different aspects of the disease, yet highly needed for a better understanding of the disease pathomechanisms. Here, we characterized a novel Ataxin-3 knock-in mouse model, expressing a heterozygous or homozygous expansion of 304 CAACAGs in the murine Ataxin-3 locus using biochemical, behavioral, and transcriptomic approaches. We compared neuropathological, and behavioral features of the new knock-in model with the in SCA3 research mostly used YAC84Q mouse model. Further, we compared transcriptional changes found in cerebellar samples of the SCA3 knock-in mice and post-mortem human SCA3 patients. The novel knock-in mouse is characterized by the expression of a polyQ-expansion in the murine Ataxin-3 protein, leading to aggregate formation, especially in brain regions known to be vulnerable in SCA3 patients, and impairment of Purkinje cells. Along these neuropathological changes, the mice showed a reduction in body weight accompanied by gait and balance instability. Transcriptomic analysis of cerebellar tissue revealed age-dependent differential expression, enriched for genes attributed to myelinating oligodendrocytes. Comparing these changes with those found in cerebellar tissue of SCA3 patients, we discovered an overlap of differentially expressed genes pointing towards similar gene expression perturbances in several genes linked to myelin sheaths and myelinating oligodendrocytes.

Cell type-specific potential pathogenic genes and functional pathways in Alzheimer's Disease

BACKGROUND

Alzheimer's disease (AD) is a pervasive age-related and highly heritable neurodegenerative disorder but has no effective therapy. The complex cellular microenvironment in the AD brain impedes our understanding of pathogenesis. Thus, a comprehensive investigation of cell type-specific responses in AD is crucial to provide precise molecular and cellular targets for therapeutic development.

METHODS

Here, we integrated analyzed 4,441 differentially expressed genes (DEGs) that were identified from 263,370 single-cells in cortex samples by single-nucleus RNA sequencing (snRNA-seq) between 42 AD-pathology subjects and 39 normal controls within 3 studies. DEGs were analyzed in microglia, astrocytes, oligodendrocytes, excitatory neurons, inhibitory neurons, and endothelial cells, respectively. In each cell type, we identified both common DEGs which were observed in all 3 studies, and overlapping DEGs which have been seen in at least 2 studies. Firstly, we showed the common DEGs expression and explained the biological functions by comparing with existing literature or multil-omics signaling pathways knowledgebase. We then determined the significant modules and hub genes, and explored the biological processes using the overlapping DEGs. Finally, we identified the common and distinct dysregulated pathways using overall DEGs and overlapping DEGs in a cell type-specific manner.

RESULTS

Up-regulated LINGO1 has been seen in both oligodendrocytes and excitatory neurons across 3 studies. Interestingly, genes enriched in the mitochondrial module were up-regulated across all cell types, which indicates mitochondrial dysfunction in the AD brain. The estrogen signaling pathway seems to be the most common pathway that is disrupted in AD.

CONCLUSION

Together, these analyses provide detailed information of cell type-specific and overall transcriptional changes and pathways underlying the human AD-pathology. These findings may provide important insights for drug development to tackle this disease.

Tutorial: Guidelines for Single-Cell RT-qPCR

Reverse transcription quantitative PCR (RT-qPCR) has delivered significant insights in understanding the gene expression landscape. Thanks to its precision, sensitivity, flexibility, and cost effectiveness, RT-qPCR has also found utility in advanced single-cell analysis. Single-cell RT-qPCR now represents a well-established method, suitable for an efficient screening prior to single-cell RNA sequencing (scRNA-Seq) experiments, or, oppositely, for validation of hypotheses formulated from high-throughput approaches. Here, we aim to provide a comprehensive summary of the scRT-qPCR method by discussing the limitations of single-cell collection methods, describing the importance of reverse transcription, providing recommendations for the preamplification and primer design, and summarizing essential data processing steps. With the detailed protocol attached in the appendix, this tutorial provides a set of guidelines that allow any researcher to perform scRT-qPCR measurements of the highest standard.

AutoGeneS: Automatic gene selection using multi-objective optimization for RNA-seq deconvolution

Knowing cell-type proportions in a tissue is very important to identify which cells or cell types are targeted by a disease or perturbation. Hence, several deconvolution methods have been proposed to infer cell-type proportions from bulk RNA samples. Their performance with noisy reference profiles and closely correlated cell types highly depends on the set of genes undergoing deconvolution. In this work, we introduce AutoGeneS, a platform that automatically extracts discriminative genes and reveals the cellular heterogeneity of bulk RNA samples. AutoGeneS requires no prior knowledge about marker genes and selects genes by simultaneously optimizing multiple criteria: minimizing the correlation and maximizing the distance between cell types. AutoGeneS can be applied to reference profiles from various sources like single-cell experiments or sorted cell populations. Ground truth cell proportions analyzed by flow cytometry confirmed the accuracy of AutoGeneS in identifying cell-type proportions. AutoGeneS is available for use via a standalone Python package (https://github.com/theislab/AutoGeneS).

Data-driven detection of subtype-specific differentially expressed genes

Among multiple subtypes of tissue or cell, subtype-specific differentially-expressed genes (SDEGs) are defined as being most-upregulated in only one subtype but not in any other. Detecting SDEGs plays a critical role in the molecular characterization and deconvolution of multicellular complex tissues. Classic differential analysis assumes a null hypothesis whose test statistic is not subtype-specific, thus can produce a high false positive rate and/or lower detection power. Here we first introduce a One-Versus-Everyone Fold Change (OVE-FC) test for detecting SDEGs. We then propose a scaled test statistic (OVE-sFC) for assessing the statistical significance of SDEGs that applies a mixture null distribution model and a tailored permutation test. The OVE-FC/sFC test was validated on both type 1 error rate and detection power using extensive simulation data sets generated from real gene expression profiles of purified subtype samples. The OVE-FC/sFC test was then applied to two benchmark gene expression data sets of purified subtype samples and detected many known or previously unknown SDEGs. Subsequent supervised deconvolution results on synthesized bulk expression data, obtained using the SDEGs detected from the independent purified expression data by the OVE-FC/sFC test, showed superior performance in deconvolution accuracy when compared with popular peer methods.

Unraveling Targetable Systemic and Cell-Type-Specific Molecular Phenotypes of Alzheimer's and Parkinson's Brains With Digital Cytometry

Alzheimer's disease (AD) and Parkinson's disease (PD) are the two most common neurodegenerative disorders worldwide, with age being their major risk factor. The increasing worldwide life expectancy, together with the scarcity of available treatment choices, makes it thus pressing to find the molecular basis of AD and PD so that the causing mechanisms can be targeted. To study these mechanisms, gene expression profiles have been compared between diseased and control brain tissues. However, this approach is limited by mRNA expression profiles derived for brain tissues highly reflecting their degeneration in cellular composition but not necessarily disease-related molecular states. We therefore propose to account for cell type composition when comparing transcriptomes of healthy and diseased brain samples, so that the loss of neurons can be decoupled from pathology-associated molecular effects. This approach allowed us to identify genes and pathways putatively altered systemically and in a cell-type-dependent manner in AD and PD brains. Moreover, using chemical perturbagen data, we computationally identified candidate small molecules for specifically targeting the profiled AD/PD-associated molecular alterations. Our approach therefore not only brings new insights into the disease-specific and common molecular etiologies of AD and PD but also, in these realms, foster the discovery of more specific targets for functional and therapeutic exploration.

Deconvolution of cellular subsets in human tissue based on targeted DNA methylation analysis at individual CpG sites

Background

The complex composition of different cell types within a tissue can be estimated by deconvolution of bulk gene expression profiles or with various single-cell sequencing approaches. Alternatively, DNA methylation (DNAm) profiles have been used to establish an atlas for multiple human tissues and cell types. DNAm is particularly suitable for deconvolution of cell types because each CG dinucleotide (CpG site) has only two states per DNA strand—methylated or non-methylated—and these epigenetic modifications are very consistent during cellular differentiation. So far, deconvolution of DNAm profiles implies complex signatures of many CpGs that are often measured by genome-wide analysis with Illumina BeadChip microarrays. In this study, we investigated if the characterization of cell types in tissue is also feasible with individual cell type-specific CpG sites, which can be addressed by targeted analysis, such as pyrosequencing.

Results

We compiled and curated 579 Illumina 450k BeadChip DNAm profiles of 14 different non-malignant human cell types. A training and validation strategy was applied to identify and test for cell type-specific CpGs. We initially focused on estimating the relative amount of fibroblasts using two CpGs that were either hypermethylated or hypomethylated in fibroblasts. The combination of these two DNAm levels into a “FibroScore” correlated with the state of fibrosis and was associated with overall survival in various types of cancer. Furthermore, we identified hypomethylated CpGs for leukocytes, endothelial cells, epithelial cells, hepatocytes, glia, neurons, fibroblasts, and induced pluripotent stem cells. The accuracy of this eight CpG signature was tested in additional BeadChip datasets of defined cell mixtures and the results were comparable to previously published signatures based on several thousand CpGs. Finally, we established and validated pyrosequencing assays for the relevant CpGs that can be utilized for classification and deconvolution of cell types.

Conclusion

This proof of concept study demonstrates that DNAm analysis at individual CpGs reflects the cellular composition of cellular mixtures and different tissues. Targeted analysis of these genomic regions facilitates robust methods for application in basic research and clinical settings.

Deciphering cellular transcriptional alterations in Alzheimer's disease brains

Large-scale brain bulk-RNAseq studies identified molecular pathways implicated in Alzheimer's disease (AD), however these findings can be confounded by cellular composition changes in bulk-tissue. To identify cell intrinsic gene expression alterations of individual cell types, we designed a bioinformatics pipeline and analyzed three AD and control bulk-RNAseq datasets of temporal and dorsolateral prefrontal cortex from 685 brain samples. We detected cell-proportion changes in AD brains that are robustly replicable across the three independently assessed cohorts. We applied three different algorithms including our in-house algorithm to identify cell intrinsic differentially expressed genes in individual cell types (CI-DEGs). We assessed the performance of all algorithms by comparison to single nucleus RNAseq data. We identified consensus CI-DEGs that are common to multiple brain regions. Despite significant overlap between consensus CI-DEGs and bulk-DEGs, many CI-DEGs were absent from bulk-DEGs. Consensus CI-DEGs and their enriched GO terms include genes and pathways previously implicated in AD or neurodegeneration, as well as novel ones. We demonstrated that the detection of CI-DEGs through computational deconvolution methods is promising and highlight remaining challenges. These findings provide novel insights into cell-intrinsic transcriptional changes of individual cell types in AD and may refine discovery and modeling of molecular targets that drive this complex disease.

Losing Dnmt3a dependent methylation in inhibitory neurons impairs neural function by a mechanism impacting Rett syndrome

Methylated cytosine is an effector of epigenetic gene regulation. In the brain, Dnmt3a is the sole 'writer' of atypical non-CpG methylation (mCH), and MeCP2 is the only known 'reader' for mCH. We asked if MeCP2 is the sole reader for Dnmt3a dependent methylation by comparing mice lacking either protein in GABAergic inhibitory neurons. Loss of either protein causes overlapping and distinct features from the behavioral to molecular level. Loss of Dnmt3a causes global loss of mCH and a subset of mCG sites resulting in more widespread transcriptional alterations and severe neurological dysfunction than MeCP2 loss. These data suggest that MeCP2 is responsible for reading only part of the Dnmt3a dependent methylation in the brain. Importantly, the impact of MeCP2 on genes differentially expressed in both models shows a strong dependence on mCH, but not Dnmt3a dependent mCG, consistent with mCH playing a central role in the pathogenesis of Rett Syndrome.

Informing disease modelling with brain-relevant functional genomic annotations

The past decade has seen a surge in the number of disease/trait-associated variants, largely because of the union of studies to share genetic data and the availability of electronic health records from large cohorts for research use. Variant discovery for neurological and neuropsychiatric genome-wide association studies, including schizophrenia, Parkinson's disease and Alzheimer's disease, has greatly benefitted; however, the translation of these genetic association results to interpretable biological mechanisms and models is lagging. Interpreting disease-associated variants requires knowledge of gene regulatory mechanisms and computational tools that permit integration of this knowledge with genome-wide association study results. Here, we summarize key conceptual advances in the generation of brain-relevant functional genomic annotations and amongst tools that allow integration of these annotations with association summary statistics, which together provide a new and exciting opportunity to identify disease-relevant genes, pathways and cell types in silico. We discuss the opportunities and challenges associated with these developments and conclude with our perspective on future advances in annotation generation, tool development and the union of the two.

Transcriptome analysis of LRRK2 knock-out microglia cells reveals alterations of inflammatory- and oxidative stress-related pathways upon treatment with α-synuclein fibrils

Several previous studies have linked the Parkinson's disease (PD) gene LRRK2 to the biology of microglia cells. However, the precise ways in which LRRK2 affects microglial function have not been fully resolved. Here, we used the RNA-Sequencing to obtain transcriptomic profiles of LRRK2 wild-type (WT) and knock-out (KO) microglia cells treated with α-synuclein pre-formed fibrils (PFFs) or lipopolysaccharide (LPS) as a general inflammatory insult. We observed that, although α-synuclein PFFs and LPS mediate overlapping gene expression profiles in microglia, there are also distinct responses to each stimulus. α-Synuclein PFFs trigger alterations of oxidative stress-related pathways with the mitochondrial dismutase Sod2 as a strongly differentially regulated gene. We validated SOD2 at mRNA and protein levels. Furthermore, we found that LRRK2 KO microglia cells reported attenuated induction of mitochondrial SOD2 in response to α-synuclein PFFs, indicating a potential contribution of LRRK2 to oxidative stress-related pathways. We validate several genes in vivo using single-cell RNA-Seq from acutely isolated microglia after striatal injection of LPS into the mouse brain. Overall, these results suggest that microglial LRRK2 may contribute to the pathogenesis of PD via altered oxidative stress signaling.

Computational deconvolution of transcriptomics data from mixed cell populations

Summary

Gene expression analyses of bulk tissues often ignore cell type composition as an important confounding factor, resulting in a loss of signal from lowly abundant cell types. In this review, we highlight the importance and value of computational deconvolution methods to infer the abundance of different cell types and/or cell type-specific expression profiles in heterogeneous samples without performing physical cell sorting. We also explain the various deconvolution scenarios, the mathematical approaches used to solve them and the effect of data processing and different confounding factors on the accuracy of the deconvolution results.

Contact

katleen.depreter@ugent.be.

Supplementary information

Supplementary data are available at Bioinformatics online.

Microenvironment of ruptured cerebral aneurysms discovered using data driven analysis of gene expression

BACKGROUND

It is well known that ruptured intracranial aneurysms are associated with substantial morbidity and mortality, yet our understanding of the genetic mechanisms of rupture remains poor. We hypothesize that applying novel techniques to the genetic analysis of aneurysmal tissue will yield key rupture-associated mechanisms and novel drug candidates for the prevention of rupture.

METHODS

We applied weighted gene co-expression networks (WGCNA) and population-specific gene expression analysis (PSEA) to transcriptomic data from 33 ruptured and unruptured aneurysm domes. Mechanisms were annotated using Gene Ontology, and gene network/population-specific expression levels correlated with rupture state. We then used computational drug repurposing to identify plausible drug candidates for the prevention of aneurysm rupture.

RESULTS

Network analysis of bulk tissue identified multiple immune mechanisms to be associated with aneurysm rupture. Targeting these processes with computational drug repurposing revealed multiple candidates for preventing rupture including Btk inhibitors and modulators of hypoxia inducible factor. In the macrophage-specific analysis, we identify rupture-associated mechanisms MHCII antigen processing, cholesterol efflux, and keratan sulfate catabolism. These processes map well onto several of highly ranked drug candidates, providing further validation.

CONCLUSIONS

Our results are the first to demonstrate population-specific expression levels and intracranial aneurysm rupture, and propose novel drug candidates based on network-based transcriptomics.

α1ACT Is Essential for Survival and Early Cerebellar Programming in a Critical Neonatal Window

Postnatal cerebellar development is a precisely regulated process involving well-orchestrated expression of neural genes. Neurological phenotypes associated with CACNA1A gene defects have been increasingly recognized, yet the molecular principles underlying this association remain elusive. By characterizing a dose-dependent CACNA1A gene deficiency mouse model, we discovered that α1ACT, as a transcription factor and secondary protein of CACNA1A mRNA, drives dynamic gene expression networks within cerebellar Purkinje cells and is indispensable for neonatal survival. Perinatal loss of α1ACT leads to motor dysfunction through disruption of neurogenesis and synaptic regulatory networks. However, its elimination in adulthood has minimal effect on the cerebellum. These findings shed light on the critical role of α1ACT in facilitating neuronal development in both mice and humans and support a rationale for gene therapies for calcium-channel-associated cerebellar disorders. Finally, we show that bicistronic expression may be common to the voltage-gated calcium channel (VGCC) gene family and may help explain complex genetic syndromes.

Variation among intact tissue samples reveals the core transcriptional features of human CNS cell classes

It is widely assumed that cells must be physically isolated to study their molecular profiles. However, intact tissue samples naturally exhibit variation in cellular composition, which drives covariation of cell-class-specific molecular features. By analyzing transcriptional covariation in 7,221 intact CNS samples from 840 neurotypical individuals, representing billions of cells, we reveal the core transcriptional identities of major CNS cell classes in humans. By modeling intact CNS transcriptomes as a function of variation in cellular composition, we identify cell-class-specific transcriptional differences in Alzheimer's disease, among brain regions, and between species. Among these, we show that PMP2 is expressed by human but not mouse astrocytes and significantly increases mouse astrocyte size upon ectopic expression in vivo, causing them to more closely resemble their human counterparts. Our work is available as an online resource ( http://oldhamlab.ctec.ucsf.edu/ ) and provides a generalizable strategy for determining the core molecular features of cellular identity in intact biological systems.

Inference of cell type content from human brain transcriptomic datasets illuminates the effects of age, manner of death, dissection, and psychiatric diagnosis

Psychiatric illness is unlikely to arise from pathology occurring uniformly across all cell types in affected brain regions. Despite this, transcriptomic analyses of the human brain have typically been conducted using macro-dissected tissue due to the difficulty of performing single-cell type analyses with donated post-mortem brains. To address this issue statistically, we compiled a database of several thousand transcripts that were specifically-enriched in one of 10 primary cortical cell types in previous publications. Using this database, we predicted the relative cell type content for 833 human cortical samples using microarray or RNA-Seq data from the Pritzker Consortium (GSE92538) or publicly-available databases (GSE53987, GSE21935, GSE21138, CommonMind Consortium). These predictions were generated by averaging normalized expression levels across transcripts specific to each cell type using our R-package BrainInABlender (validated and publicly-released on github). Using this method, we found that the principal components of variation in the datasets strongly correlated with the predicted neuronal/glial content of the samples. This variability was not simply due to dissection–the relative balance of brain cell types appeared to be influenced by a variety of demographic, pre- and post-mortem variables. Prolonged hypoxia around the time of death predicted increased astrocytic and endothelial gene expression, illustrating vascular upregulation. Aging was associated with decreased neuronal gene expression. Red blood cell gene expression was reduced in individuals who died following systemic blood loss. Subjects with Major Depressive Disorder had decreased astrocytic gene expression, mirroring previous morphometric observations. Subjects with Schizophrenia had reduced red blood cell gene expression, resembling the hypofrontality detected in fMRI experiments. Finally, in datasets containing samples with especially variable cell content, we found that controlling for predicted sample cell content while evaluating differential expression improved the detection of previously-identified psychiatric effects. We conclude that accounting for cell type can greatly improve the interpretability of transcriptomic data.

Brain Cell Type Specific Gene Expression and Co-expression Network Architectures

Elucidating brain cell type specific gene expression patterns is critical towards a better understanding of how cell-cell communications may influence brain functions and dysfunctions. We set out to compare and contrast five human and murine cell type-specific transcriptome-wide RNA expression data sets that were generated within the past several years. We defined three measures of brain cell type-relative expression including specificity, enrichment, and absolute expression and identified corresponding consensus brain cell “signatures,” which were well conserved across data sets. We validated that the relative expression of top cell type markers are associated with proxies for cell type proportions in bulk RNA expression data from postmortem human brain samples. We further validated novel marker genes using an orthogonal ATAC-seq dataset. We performed multiscale coexpression network analysis of the single cell data sets and identified robust cell-specific gene modules. To facilitate the use of the cell type-specific genes for cell type proportion estimation and deconvolution from bulk brain gene expression data, we developed an R package, BRETIGEA. In summary, we identified a set of novel brain cell consensus signatures and robust networks from the integration of multiple datasets and therefore transcend limitations related to technical issues characteristic of each individual study.

Mathematical Modeling and Deconvolution of Molecular Heterogeneity Identifies Novel Subpopulations in Complex Tissues

Tissue heterogeneity is both a major confounding factor and an underexploited information source. While a handful of reports have demonstrated the potential of supervised methods to deconvolve tissue heterogeneity, these approaches require a priori information on the marker genes or composition of known subpopulations. To address the critical problem of the absence of validated marker genes for many (including novel) subpopulations, we develop a novel unsupervised deconvolution method, Convex Analysis of Mixtures (CAM), within a well-grounded mathematical framework, to dissect mixed gene expressions in heterogeneous tissue samples. To facilitate the utility of this method, we implement an R-Java CAM package that provides comprehensive analytic functions and graphic user interface (GUI).

Transcriptomic profiling of the human brain reveals that altered synaptic gene expression is associated with chronological aging

Aging is a biologically universal event, and yet the key events that drive aging are still poorly understood. One approach to generate new hypotheses about aging is to use unbiased methods to look at change across lifespan. Here, we have examined gene expression in the human dorsolateral frontal cortex using RNA- Seq to populate a whole gene co-expression network analysis. We show that modules of co-expressed genes enriched for those encoding synaptic proteins are liable to change with age. We extensively validate these age-dependent changes in gene expression across several datasets including the publically available GTEx resource which demonstrated that gene expression associations with aging vary between brain regions. We also estimated the extent to which changes in cellular composition account for age associations and find that there are independent signals for cellularity and aging. Overall, these results demonstrate that there are robust age-related alterations in gene expression in the human brain and that genes encoding for neuronal synaptic function may be particularly sensitive to the aging process.

Defining Trends in Global Gene Expression in Arabian Horses with Cerebellar Abiotrophy

Equine cerebellar abiotrophy (CA) is a hereditary neurodegenerative disease that affects the Purkinje neurons of the cerebellum and causes ataxia in Arabian foals. Signs of CA are typically first recognized either at birth to any time up to 6 months of age. CA is inherited as an autosomal recessive trait and is associated with a single nucleotide polymorphism (SNP) on equine chromosome 2 (13074277G>A), located in the fourth exon of TOE1 and in proximity to MUTYH on the antisense strand. We hypothesize that unraveling the functional consequences of the CA SNP using RNA-seq will elucidate the molecular pathways underlying the CA phenotype. RNA-seq (100 bp PE strand-specific) was performed in cerebellar tissue from four CA-affected and five age-matched unaffected horses. Three pipelines for differential gene expression (DE) analysis were used (Tophat2/Cuffdiff2, Kallisto/EdgeR, and Kallisto/Sleuth) with 151 significant DE genes identified by all three pipelines in CA-affected horses. TOE1 (Log2(foldchange) = 0.92, p = 0.66) and MUTYH (Log2(foldchange) = 1.13, p = 0.66) were not differentially expressed. Among the major pathways that were differentially expressed, genes associated with calcium homeostasis and specifically expressed in Purkinje neurons, CALB1 (Log2(foldchange) = -1.7, p < 0.01) and CA8 (Log2(foldchange) = -0.97, p < 0.01), were significantly down-regulated, confirming loss of Purkinje neurons. There was also a significant up-regulation of markers for microglial phagocytosis, TYROBP (Log2(foldchange) = 1.99, p < 0.01) and TREM2 (Log2(foldchange) = 2.02, p < 0.01). These findings reaffirm a loss of Purkinje neurons in CA-affected horses along with a potential secondary loss of granular neurons and activation of microglial cells.

Mathematical modelling of transcriptional heterogeneity identifies novel markers and subpopulations in complex tissues

Tissue heterogeneity is both a major confounding factor and an underexploited information source. While a handful of reports have demonstrated the potential of supervised computational methods to deconvolute tissue heterogeneity, these approaches require a priori information on the marker genes or composition of known subpopulations. To address the critical problem of the absence of validated marker genes for many (including novel) subpopulations, we describe convex analysis of mixtures (CAM), a fully unsupervised in silico method, for identifying subpopulation marker genes directly from the original mixed gene expressions in scatter space that can improve molecular analyses in many biological contexts. Validated with predesigned mixtures, CAM on the gene expression data from peripheral leukocytes, brain tissue, and yeast cell cycle, revealed novel marker genes that were otherwise undetectable using existing methods. Importantly, CAM requires no a priori information on the number, identity, or composition of the subpopulations present in mixed samples, and does not require the presence of pure subpopulations in sample space. This advantage is significant in that CAM can achieve all of its goals using only a small number of heterogeneous samples, and is more powerful to distinguish between phenotypically similar subpopulations.

Computational deconvolution of genome wide expression data from Parkinson's and Huntington's disease brain tissues using population-specific expression analysis

The characterization of molecular changes in diseased tissues gives insight into pathophysiological mechanisms and is important for therapeutic development. Genome-wide gene expression analysis has proven valuable for identifying biological processes in neurodegenerative diseases using post mortem human brain tissue and numerous datasets are publically available. However, many studies utilize heterogeneous tissue samples consisting of multiple cell types, all of which contribute to global gene expression values, confounding biological interpretation of the data. In particular, changes in numbers of neuronal and glial cells occurring in neurodegeneration confound transcriptomic analyses, particularly in human brain tissues where sample availability and controls are limited. To identify cell specific gene expression changes in neurodegenerative disease, we have applied our recently published computational deconvolution method, population specific expression analysis (PSEA). PSEA estimates cell-type-specific expression values using reference expression measures, which in the case of brain tissue comprises mRNAs with cell-type-specific expression in neurons, astrocytes, oligodendrocytes and microglia. As an exercise in PSEA implementation and hypothesis development regarding neurodegenerative diseases, we applied PSEA to Parkinson's and Huntington's disease (PD, HD) datasets. Genes identified as differentially expressed in substantia nigra pars compacta neurons by PSEA were validated using external laser capture microdissection data. Network analysis and Annotation Clustering (DAVID) identified molecular processes implicated by differential gene expression in specific cell types. The results of these analyses provided new insights into the implementation of PSEA in brain tissues and additional refinement of molecular signatures in human HD and PD.

Correspondence regarding Zhong et al., BMC Bioinformatics 2013 Mar 7;14:89

Computational expression deconvolution aims to estimate the contribution of individual cell populations to expression profiles measured in samples of heterogeneous composition. Zhong et al. recently proposed Digital Sorting Algorithm (BMC Bioinformatics 2013 Mar 7;14:89) and showed that they could accurately estimate population-specific expression levels and expression differences between two populations. They compared DSA with Population-Specific Expression Analysis (PSEA), a previous deconvolution method that we developed to detect expression changes occurring within the same population between two conditions (e.g. disease versus non-disease). However, Zhong et al. compared PSEA-derived specific expression levels across different cell populations. Specific expression levels obtained with PSEA cannot be directly compared across different populations as they are on a relative scale. They are accurate as we demonstrate by deconvolving the same dataset used by Zhong et al. and, importantly, allow for comparison of population-specific expression across conditions.

Insights from cerebellar transcriptomic analysis into the pathogenesis of ataxia

IMPORTANCE

The core clinical and neuropathological feature of the autosomal dominant spinocerebellar ataxias (SCAs) is cerebellar degeneration. Mutations in the known genes explain only 50% to 60% of SCA cases. To date, no effective treatments exist, and the knowledge of drug-treatable molecular pathways is limited. The examination of overlapping mechanisms and the interpretation of how ataxia genes interact will be important in the discovery of potential disease-modifying agents.

OBJECTIVES

To address the possible relationships among known SCA genes, predict their functions, identify overlapping pathways, and provide a framework for candidate gene discovery using whole-transcriptome expression data.

DESIGN, SETTING, AND PARTICIPANTS

We have used a systems biology approach based on whole-transcriptome gene expression analysis. As part of the United Kingdom Brain Expression Consortium, we analyzed the expression profile of 788 brain samples obtained from 101 neuropathologically healthy individuals (10 distinct brain regions each). Weighted gene coexpression network analysis was used to cluster 24 SCA genes into gene coexpression modules in an unsupervised manner. The overrepresentation of SCA transcripts in modules identified in the cerebellum was assessed. Enrichment analysis was performed to infer the functions and molecular pathways of genes in biologically relevant modules.

MAIN OUTCOMES AND MEASURES

Molecular functions and mechanisms implicating SCA genes, as well as lists of relevant coexpressed genes as potential candidates for novel SCA causative or modifier genes.

RESULTS

Two cerebellar gene coexpression modules were statistically enriched in SCA transcripts (P = .021 for the tan module and P = 2.87 × 10-5 for the light yellow module) and contained established granule and Purkinje cell markers, respectively. One module includes genes involved in the ubiquitin-proteasome system and contains SCA genes usually associated with a complex phenotype, while the other module encloses many genes important for calcium homeostasis and signaling and contains SCA genes associated mostly with pure ataxia.

CONCLUSIONS AND RELEVANCE

Using normal gene expression in the human brain, we identified significant cell types and pathways in SCA pathogenesis. The overrepresentation of genes involved in calcium homeostasis and signaling may indicate an important target for therapy in the future. Furthermore, the gene networks provide new candidate genes for ataxias or novel genes that may be critical for cerebellar function.

Genome-wide DNA methylation analysis of human brain tissue from schizophrenia patients

Recent studies suggest that genetic and environmental factors do not account for all the schizophrenia risk, and epigenetics also has a role in disease susceptibility. DNA methylation is a heritable epigenetic modification that can regulate gene expression. Genome-wide DNA methylation analysis was performed on post-mortem human brain tissue from 24 patients with schizophrenia and 24 unaffected controls. DNA methylation was assessed at over 485,000 CpG sites using the Illumina Infinium HumanMethylation450 Bead Chip. After adjusting for age and post-mortem interval, 4641 probes corresponding to 2929 unique genes were found to be differentially methylated. Of those genes, 1291 were located in a CpG island and 817 were in a promoter region. These include NOS1, AKT1, DTNBP1, DNMT1, PPP3CC and SOX10, which have previously been associated with schizophrenia. More than 100 of these genes overlap with a previous DNA methylation study of peripheral blood from schizophrenia patients in which 27,000 CpG sites were analysed. Unsupervised clustering analysis of the top 3000 most variable probes revealed two distinct groups with significantly more people with schizophrenia in cluster one compared with controls (P=1.74 × 10(-4)). The first cluster composed of 88% of patients with schizophrenia and only 12% controls, whereas the second cluster composed of 27% of patients with schizophrenia and 73% controls. These results strongly suggest that differential DNA methylation is important in schizophrenia etiology and add support for the use of DNA methylation profiles as a future prognostic indicator of schizophrenia.