- However, there are exceptions where only one of the two alleles is expressed. - Arguably, many of the previous studies actually investigated allele-biased expression rather than monoallelic expression, including our published work [14], as pointed out by recent reports [15, 16].. - Full list of author information is available at the end of the article. - Finally, an RNA-seq study using human prenatal brain tissues echoed the importance of lncRNAs in human neo- cortex development, as it demonstrated that LOC646329, one of the most radial glia-enriched lncRNAs, regulated cell proliferation [30]. - We called SNPs for 34 embryos after merging scRNA-seq data from different cells of the same embryo (Additional file 1: Figure S1a).. - To be highly conservative while keeping a reasonable number of hetSNPs for ana- lysis, we decided to use the cutoff of 20 for each of the two alleles in pooled RNA-seq data, yielding a positive predictive value of 99.46% on average (Fig. - b Summary of the mouse scRNA-seq dataset from 34 embryos at different developmental stages. - Boxplots are used to summarize data from individual embryos of the same developmental stages (see color legend in b). - ~7% of hetSNPs identified above), of which 77% were present in the dbSNP. - Moreover, we compared the allelic ratios of the SNPs from the GATK pipeline and those from our method, and found the allelic ratios of most SNPs missed by GATK were deviated from 1:1 (Additional file 1: Figure S1d). - The original study provided a clear overview of cellular heterogen- eity and complexity of the adult and fetal human cor- tical regions at the single cell transcriptome level [28].. - In the end, we ana- lyzed 323 single cells from six adults (adult21, adult37, adult47, adult50, adult54, and adult63B) (Additional file 2: Table S1). - analysis of the mouse scRNA-seq datasets (Table 1). - Due to different genetic backgrounds and differences in the number of cells analyzed, we obtained variable numbers of hetSNPs for the six individuals (Additional file 1: Figure S4a. - ~80% of the called hetSNPs were located in the genic re- gions (Additional file 1: Figure S4b), with ~25% in exons.. - We pre- dicted the functional impacts of the alternative alleles at hetSNP sites using wANNOVAR [37] and examined if the deleterious alleles were expressed in fewer cells (Fig. - Cell-type MA genes. - Analysis of data from other brains re- vealed a similar pattern except in the adult47 brain, in which the “ deleterious ” and “ probably damaging ” alleles showed similar or lower expression of the reference al- leles when compared to synonymous SNVs (Additional. - For each of the hetSNPs, we evaluated its allele-biased expression in a single cell by performing a binomial test of the read counts for the two alleles and considering the allelic ratio (see Methods and Add- itional file 1: Figure S2). - To get an overview of the allelic bias status, we first summarized the allelic bias at 284,220 sites in all cells among the six individuals. - Data are for all cells and for all hetSNPs in the six adult brains. - For 136,422 monoallelic (MA) sites, the distribution of the reference allelic ratios was bimodal, with ratios near 0 or 1 (Fig. - We also noted that the number of MA sites was approximately 5× larger than that of the BA sites, while 42.90% (“Unknown”) of the tested sites were not well covered for statistical inference at the single cell level (Fig. - To see how allelic expression in single cells would be reflected at the cell population level, we pooled scRNA- seq data from the same cell types of the same individuals and then called allelic expression from the pooled scRNA- seq reads (to mimic bulk RNA-seq analysis). - 3c, d), which had the largest number of cells for a specific cell type in any of the six individuals (Additional file 2: Table S1), we found that at the single cell level for 83.87% of the 99,723 total hetSNPs, the reference allele ratios were near 0 or 1 (i.e., strongly MA) in the 50 cells. - However, when scRNA-seq reads from the 50 neurons were pooled and analyzed as bulk RNA-seq data, we found that of the 17,559 non-redundant hetSNPs showed reference allele ratios between 0.4 and 0.6 (Fig. - 158,899 out of the 198,690 genes (79.97%) had only one hetSNP detected by scRNA-seq from a particular cell, while harbored two hetSNPs and contained >2 hetSNPs (Fig. - We also compared the expression level of the three groups of genes and found that genes of BA expression were overall expressed at a higher level (BA- BA vs. - We began by identifying genes that exhibited MA (or BA) across multiple cells of the same cell type. - We randomly sampled a subset of the 50 cells (from 1 to 49) and then determined the number of allelically expressed genes by our method. - For example, in adult50 neurons, 2160 out of the 3488 MA genes (61.93%) were called MA in less than four cells. - Nevertheless, we reasoned that a gene needs to exhibit MA “consistently” if its MA expression would confer any functional affect to a brain cell type, so we analyzed the genes that were called MA expression in at least four cells of the same type, with no cell exhibiting a BA. - When we took the total hetSNPs called for an individual into consideration, on average 5.37% of the heterozygous genes in the six individuals showed strong biased allelic gene expression in one of the six brain cell types (Table 1). - We found that 688 of our 1515 MA genes were present with experimental evidence, among which 65 genes (9.45%) were biased in at least one of the eight human tissues recorded in the dbMAE. - This is a summary of all cells in the six individuals. - c Distribution of the numbers of genes with three patterns of biased gene expression across individual cells. - This comparison shows that there is a broad agreement between our calls and the allelic expression reported in the dbMAE.. - the top 30% in expression level in each cell of the same cell type, were combined and used as background genes. - To determine if the MA genes from astrocytes and neurons in different individ- uals showed similar functions, we used a network to il- lustrate the relationship between groups of cells of the same cell type in different individuals and enriched GO terms (see Methods) (Fig. - and immune response in the CNS [44]. - Although we found that 50 and 67 of the MA genes have been implicated in autism and schizophre- nia, the lists of MA genes in the brain cell types as a whole showed no significant enrichment for either autism or schizophrenia (Fisher’s exact test, p = 0.33 for autism, p = 0.83 for schizophrenia) (Additional file 2: Table S4).. - We first compared MA genes in neurons and. - 6b), indicating that most of the MA genes were cell type-specific. - We found some shared MA genes in the same cell types across individuals. - For example, 86 of the 1006 MA genes in neurons showed MA expression in at least. - 6 Cell type-specific expression of MA genes. - The mean values of the two groups are significantly different (t-test, p <. - PCDH9 is a member of the cadherin superfamily of calcium-dependent cell adhesion molecules and was previously reported to show monoal- lelic expression [4, 5]. - In Oligodendrocytes, 11 of the 196 MA genes showed MA expression in at least 2 indi- viduals. - The expression profiles of the eigengenes for the three modules also supported the idea that these mod- ules were highly expressed in only one particular type of cell (Fig. - The 52 genes in the salmon4 module were enriched for astro- cyte differentiation functions. - For example, neuron marker genes, such as TMEM130 [50], MAP2, MAP1B, SNAP25, PGM2L1 and SCG2 [51], were in the magenta module. - oligodendrocyte marker genes, such as the ma- ture oligodendrocyte marker MBP [51, 52], CLDN11 [53], OPALIN [54], ERMN [55], PLP1, HSPA2, MOG and PPP1R14A [51] were in the salmon2 module. - and astro- cyte marker genes, such as AQP4 [56], ATP1A2, ALDOC, SLC1A2, GLUL and AHCYL1 [51] were in the salmon4 module. - Interestingly, we found that MA genes in neurons of adult50 were enriched in the magenta mod- ule (Fisher exact test, p value = 2.9E-3) (Fig. - of the MA genes in adult50 neurons also showed MA expression in at least one of other samples, indicating that some MA genes may function at different develop- mental stages. - and serves as an interactor of the p75 neurotrophin re- ceptor, linking neurotrophin signaling to the cell cycle [62]. - While most of the previous studies were performed in cell lines or stem cells, in this study, we re-analyzed scRNA-seq data of adult brain cells and found that at the single cell level most of the genes show allele-biased expression, indicat- ing that monoallelic expression seems to be the norm rather than exception. - Our finding is consistent with re- cent in vitro studies [20, 22] and indicates that neurons and other cell types in the brain all display widespread monoallelic gene expression at the single cell level. - Although we did not find a big difference among brain cell types in terms of the extent of monoallelic gene expres- sion, we showed that some MA genes were expressed monoallelically in specific brain cell types and MA genes in oligodendrocytes and neurons were involved in cellular functions specific for them. - brain development by disrupting different cellular compo- nents of the brain.. - For example, a meta-analysis based allele-specific expression detection for ASE expression (MBASED) works quite well by ag- gregating information across multiple SNPs of the same. - 7 Co-expression of MA genes. - c Network shows co-expression of genes in the neuron module.. - Of the genes with hetSNPs, we found 5.37%. - on average in the six individuals showed MA expression at the cell-type level (Table 1). - Much of the difference could be due to the different definition of allelic expression (i.e., monoallelic vs allele-biased), but the small number of cells used in our scRNA-seq datasets may be a key reason behind the difference, as discussed above (Additional file 1: Figure S13). - We also observed that 50 and 67 genes exhibiting MA expression in multiple cells of the same cell types were implicated in autism and schizophrenia, but the overlap is statistically insignificant.. - Recently, investigators have begun to understand gene regulation from the perspective of the 3D genome, which refers to gene expression changes caused by inter- and intra- chromosomal interactions. - Both Hi-C and ChIA-PET data demonstrate that the 3D organization of the genome shows cell type-specific patterns [70 – 73]. - When comparing our cell type- specific MA genes with the cell-type WGCNA modules, we found that neuron MA genes were enriched in the magenta module, which is involved in various neural func- tions, such as synaptic transmission and neuron projec- tion. - On the other hand, monoallelic expression could also increase disease susceptibility, conceivably, if one copy of the MA gene possesses deleterious mutations such that its expression or lack of expression leads to abnormal function of a specific brain region derived from the clonal expansion of a precursor in which MA expression first oc- curred. - Our method of deriving hetSNPs from RNA- seq data will miss hetSNPs that express only one of the two parental alleles across all cells in an individual.. - We first masked these SNP sites with “N” in the human genome (hg19) or mouse genome (mm10) and then aligned pooled scRNA-seq reads to the modified genomes by STAR (ver. - We considered a site heterozygous if each of the two alleles was supported by a minimal num- ber of reads (read depth cutoff), which was tested from 1 to 30. - If neither of the two alleles was the reference allele, the SNP position was excluded from further analysis. - Only a few such posi- tions (min 5, max 92) were observed in the samples.. - After a list of hetSNPs was called for each individual from the pooled scRNA-seq data, to get an overview of the genomic distribution of the hetSNPs, we first anno- tated the hetSNPs based on Ensembl gene annotation (Release 74), which contains 63,677 genes including 22,810 protein-coding and 56,337 non-coding genes. - The allele expression pat- tern of the remaining hetSNPs (with at least one read) was classified as “Unknown.”. - After assessing the biased expression states of each gene in each cell, we compared a gene’s biased expres- sion across cells of the same cell type (from the same individual) to evaluate cell-type biased expression. - found that the allelic losses are a function of the expression level with low expressed genes showing a high rate of allelic losses [12]. - To minimize the effect of gene expression level on gene bias decision, especially for low expressed genes, we classified a gene as BA ex- pression in a cell type if it was called BA in any cell of the cell type. - For MA expression in a cell type, we set the following criteria: (1) all cells (at least four) support the MA expression in the cell type. - Function analysis of MA genes. - Enrichment of MA genes in WGCNA module. - Numbers of hetSNPs sites with different reference allele ratios, after scRNA-seq reads from cells of the same type in individual brains were pooled. - Cell numbers used for scRNA-seq of the brains. - This table is based on the cell classification in the original study (Darmanis et al., 2015). - The number of cells supporting the monoallelic gene expression was in column SupportingCellNum and the corresponding single-cell RNA-seq files (GEO accession IDs) were in the column scRNAseqFiles.. - The funding agency had no roles in the design of the study, collection, analysis and interpretation of data, or writing the manuscript.. - Monoallelic yet combinatorial expression of variable exons of the protocadherin-alpha gene cluster in single neurons.. - Epigenetics: Monoallelic expression in the immune system. - Autosomal monoallelic expression in the mouse. - Monoallelic expression of multiple genes in the CNS. - Single-cell analysis of long non-coding RNAs in the developing human neocortex. - CRISPR/Cas9-mediated heterozygous knockout of the autism gene CHD8 and characterization of its transcriptional networks in neurodevelopment. - Influence of a genetic variant of the neuronal growth associated protein Stathmin 1 on cognitive and affective control processes: an event-related potential study. - Bex1 is involved in the regeneration of axons after injury. - Bex1, a novel interactor of the p75 neurotrophin receptor, links neurotrophin signaling to the cell cycle. - Genome-wide map of regulatory interactions in the human genome
Xem thử không khả dụng, vui lòng xem tại trang nguồn hoặc xem
Tóm tắt