Single-nuclei chromatin profiling of ventral midbrain reveals cell identity transcription factors and cell-type-specific gene regulatory variation

Abstract

Background: Cell types in ventral midbrain are involved in diseases with variable genetic susceptibility, such as Parkinson's disease and schizophrenia. Many genetic variants affect regulatory regions and alter gene expression in a cell-type-specific manner depending on the chromatin structure and accessibility.

Results: We report 20,658 single-nuclei chromatin accessibility profiles of ventral midbrain from two genetically and phenotypically distinct mouse strains. We distinguish ten cell types based on chromatin profiles and analysis of accessible regions controlling cell identity genes highlights cell-type-specific key transcription factors. Regulatory variation segregating the mouse strains manifests more on transcriptome than chromatin level. However, cell-type-level data reveals changes not captured at tissue level. To discover the scope and cell-type specificity of cis-acting variation in midbrain gene expression, we identify putative regulatory variants and show them to be enriched at differentially expressed loci. Finally, we find TCF7L2 to mediate trans-acting variation selectively in midbrain neurons.

Conclusions: Our data set provides an extensive resource to study gene regulation in mesencephalon and provides insights into control of cell identity in the midbrain and identifies cell-type-specific regulatory variation possibly underlying phenotypic and behavioural differences between mouse strains.

Keywords: Cell-type identity; Genetic variation; Midbrain; Mouse strains; Single-nuclei ATAC-seq; Wnt signalling.

Fig. 1

Midbrain snATAC-seq identifies cell-type-specific accessibility. A Clustering of snATAC-seq from C57BL/6J and A/J with corresponding cell-type proportions. Major cell types can be identified based on snATAC-seq profiles, with neurons having the biggest proportion on both strains. Cell types in C57BL/6J and A/J have comparable proportions with more than half of nuclei being identified as neurons. B Cell-type-specific accessibility is observed in marker genes. The genomic tracks are from C57BL/6J midbrain snATAC-seq. The expression profiles measured as transcript per 100,000 in cluster. Rpl13a is used as a house keeping gene to normalize the snATAC-seq signal. See also Additional file 1: Figures S1 and S2

Fig. 2

Regions controlling cell-type identity can be defined by combining snATAC-seq and scRNA-seq. A Schematic workflow to define cell-type-specific signatures. Digital gene expression is obtained from DropViz. For each gene, the 85th percentile of its expression across all cell types was calculated. To define a gene as a cell-type-identity gene, at least 60% of the cells of a cell type should have expression more than the 85th percentile, while at the same time no other cell type was permitted to have the same gene among its top expressed genes (above the 85th percentile) in more than 40% of the cells. Enrichment analysis with cell-type-identity genes found GO terms corresponding to cell-type characteristics. The cell-type-identity peaks are defined by peaks overlapping with the regulatory regions of cell-type-identity genes (basal region ± 100 kb until nearby genes). Subsequently, the enriched motifs in cell-type-identity peaks are detected. B Examples of identified cell-type-identity genes. The identified cell-type-identity genes for six major cell types show selective expression in the respective cell types when observing scRNA-seq data of the entire population of midbrain cells

Fig. 3

Identification of cell-type-specific TFs controlling cellular identity. A Heatmap showing the enriched signal of cell-type-identity peaks in eight cell types. The analysis was done on C57BL/6J midbrain snATAC-seq. The background is constructed by merging the sampling reads (366,278 reads/cell type) from each cell type. The raw signal is normalized to the background and library, following log₂-transformation. The normalized signal is plotted 3 kb up- and downstream of peaks. B Motif enrichment analysis on cell-type-identity peaks. The PWM logos, names of the associated TFs and p values are shown for each motif. The motifs are ranked according to p values

Fig. 4

Association of differentially accessible regions with altered gene expression between C57BL/6J and A/J. A Differential peaks are highly associated with differential genes. Top differential peaks (labelled as red) are defined by Wilcoxon rank-sum test with FDR < 0.05 within top 1% of logFC. The read counts in peaks of snATAC-seq bulk are log₁₀-transformed. Peaks with low read count (less than median—1.5 median absolute deviation) are filtered out. To associate differential peaks to DEGs, peaks are overlapped with the regulatory region of DEGs (basal region ± 100 kb until nearby genes). As a control, random peaks are selected by bootstrapping with 1000 repetitions (p < 0.0099). The RPKM from bulk RNA-seq of C57BL/6J and A/J (n = 12 per strain) is also log₁₀-transformed, and DEGs are defined as FDR < 0.05 and log₂-fold change > 1 (labelled as red). B Cell-type-specific differential peaks correlate with gene expression in bulk RNA-seq. The differential peaks are labelled as green. *FDR < 0.05

Fig. 5

Putative regulatory variants are associated with differentially expressed genes and show cell-type-selective accessibility. A Examples of putative regulatory variants of A/J found in the enhancer region upstream of Ddhd1, Zfp619 and Rn4.5s. The putative regulatory variants are defined as variants disrupting TF footprints located in active enhancers (defined by H3K27ac). B Each DEG (5082, FDR < 0.05) is associated with an average of 0.4 putative regulatory variants, while non-DEGs are associated with only 0.14 variants. Random: 5000 genes are randomly selected from all expressed genes. C Putative regulatory variants have differential accessibility across cell types. More than half of the variants are accessible in more than 6 cell types, while 7% in only 1 cell type. D An example showing how putative regulatory variants with differential accessibility affect cell-type-specific gene expression. The variants locating near the TSS of Tekt5 are associated with TSS signal in neurons of C57BL/6J but not A/J, potentially resulting in upregulation of Tekt5 as shown in bulk RNA-seq. *FDR < 0.05

Fig. 6

TCF/LEF family as a mediator for trans-acting variation in neurons. A Motif enrichment analysis for regions of differential accessibility between C57BL/6J and A/J in neurons. B, C Regions of differential accessibility with motif for TCF/LEF family show increased accessibility in neurons compared to other cell types. D–E Lef1 is highly expressed in endothelial stalk, Tcf7l1 shows low expression in all cell types, while Tcf7l2 is abundant in neurons and polydendrocytes. See also Additional file 1: Figure S3