A Guide to Single-Cell Transcriptomics in Adult Rodent Brain: The Medium Spiny Neuron Transcriptome Revisited

Abstract

Recent advances in single-cell technologies are paving the way to a comprehensive understanding of the cellular complexity in the brain. Protocols for single-cell transcriptomics combine a variety of sophisticated methods for the purpose of isolating the heavily interconnected and heterogeneous neuronal cell types in a relatively intact and healthy state. The emphasis of single-cell transcriptome studies has thus far been on comparing library generation and sequencing techniques that enable measurement of the minute amounts of starting material from a single cell. However, in order for data to be comparable, standardized cell isolation techniques are essential. Here, we analyzed and simplified methods for the different steps critically involved in single-cell isolation from brain. These include enzymatic digestion, tissue trituration, improved methods for efficient fluorescence-activated cell sorting in samples containing high degree of debris from the neuropil, and finally, highly region-specific cellular labeling compatible with use of stereotaxic coordinates. The methods are exemplified using medium spiny neurons (MSN) from dorsomedial striatum, a cell type that is clinically relevant for disorders of the basal ganglia, including psychiatric and neurodegenerative diseases. We present single-cell RNA sequencing (scRNA-Seq) data from D1 and D2 dopamine receptor expressing MSN subtypes. We illustrate the need for single-cell resolution by comparing to available population-based gene expression data of striatal MSN subtypes. Our findings contribute toward standardizing important steps of single-cell isolation from adult brain tissue to increase comparability of data. Furthermore, our data redefine the transcriptome of MSNs at unprecedented resolution by confirming established marker genes, resolving inconsistencies from previous gene expression studies, and identifying novel subtype-specific marker genes in this important cell type.

Keywords: Cholinergic Receptor Muscarinic 4; basal ganglia; dopamine receptors; fluorescence-activated cell sorting; single-cell RNA sequencing; striatal medium spiny neurons.

FIGURE 1

Single-cell RNA-Seq work flow. Vibratome sections are enzymatically digested and mechanically triturated resulting in a single cell suspension. Density gradient centrifugation removes cellular debris. Single cells are collected by fluorescence-activated cell sorting (FACS). After library generation and quality control, single-cell transcriptomes are sequenced using Illumina protocols.

FIGURE 2

Nuclear dyes facilitate isolation of adult brain cells by FACS. (A) Top: Single-cell suspensions from adult brain tissue display a broad, indistinctive smear in the standard FSC/SSC plot used to identify cells by FACS due to (1) the overwhelming presence of debris from neuropil fragments and (2) the morphological diversity of CNS cell types (left). Introduction of the nuclear stain DRAQ5 allows for efficient isolation of cell bodies based on the presence of a cell nucleus (right). Bottom: Single-cell suspensions from cultured HEK293 cells allow isolation of single cells due to higher morphological homogeneity and the absence of debris present in dissociated brain tissue (left). Furthermore, a nuclear stain can enhance isolation by FACS for cultured cells (right). Photomultiplier tube (PMT) voltage for cell culture samples were set at 50% compared to adult brain cells due to the larger size of the cultured HEK293 cells. (B) Comparison of two common nuclear stains, Hoechst 33258 (left) and DRAQ5 (right), for single-cell isolation of adult CNS cells. Increased background fluorescence of a group of particles in the UV spectrum (asterisk) results in a lower signal/noise ratio for the nuclear dye Hoechst 33258 compared to DRAQ5.

FIGURE 3

Precise measurements of critical parameters for single-cell isolation. (A) Schematic view of experimental design. Key variables included diameter of tissue slices and performing individual. Two independent experiments were performed. (B–D) Examples of flow cytometry data: identification of cell bodies (B, based on nuclear dye DRAQ5), singlets (C, within cells), and live single cells (D, based on propidium-iodide, within singlets). (E,F) Impact of tissue slice thickness, performing individual (Ind), and experiment (Exp) on yield of singlets (E), and live single cells (F). Nested ANOVA was performed with individuals and experiments treated as replicate groups. (G) Effect of density gradient centrifugation on cell enrichment versus debris. Paired t-test was performed. Inset shows the density gradient (g) and supernatant (s) containing cellular particles (white bands). (H) Schematic view of experimental design for comparison of enzyme treatment. Slices of forebrain (FB), ventral midbrain (MB), and parts of hindbrain (myencephalon, MY), were treated with papain (in HABG buffer) or pronase (in carbogen-bubbled ACSF). (I) Flow cytometry measurement of live single cells based on DAPI incorporation comparing treatment with papain and pronase. Two-way ANOVA.

FIGURE 4

In vivo application of the cell isolation protocol for region-specific scRNA-Seq based on stereotaxic coordinates. (A) Schematic workflow for labeling cells in vivo in a cell type- and region-specific manner. BAC-transgenic reporters label the two major striatal MSN subpopulations based on their expression of the D1 and D2 dopamine receptors. Stereotaxic injection of the nuclear dye Hoechst 33258 allows specific labeling of cells in the DMS. Cells are isolated by FACS and RNA-Seq libraries produced for downstream Illumina sequencing. (B) Fluorescence microscopy image of a coronal section showing the right hemisphere of a double transgenic animal injected with a mixture of India ink (for visual guidance of tissue dissection) and Hoechst 33258 in the dorsomedial striatum. (C) Magnified view of an injection site. Note that the small molecule Hoechst 33258 spreads slightly further than the larger carbon particles contained in India ink. The approximate dissection area is indicated. (D) FACS data (comparable particle numbers) showing nuclear stain by Hoechst 33258 from a non-injected C57BL/6 control animal (left) and an injected double-transgenic animal (right). (E) Left: cells based on Hoechst 33258 gating shown in (D) assayed for cell death based on propidium-iodide (PI). Due to partial overlap of TdTomato and propidium-iodide with available filters, the gating was done setting TdTomato against propidium-iodide channels. Right: live cells based on left panel show Drd1a-TdTomato and Drd2-EGFP labeled MSN subpopulations within the dorsomedial striatum.

FIGURE 5

scRNA-Seq analysis identifies cell types. (A) Random down-sampling of cell numbers reveals relationship between stochasticity of gene expression in single cells and comprehensiveness of transcriptome capture with increasing cell numbers. Random sampling of defined cell numbers was repeated five times. (B) BAC-transgene expression in D1 and D2 MSN. (C) Comparison of reporter transgene fluorescence and mRNA levels. (D) PCA analysis for cell type identification reveals two Drd2-expressing cholinergic interneurons (ChI). Inset: Chat expression confirms cholinergic interneuron identity. (E) Box plots for previously reported MSN subtype-specific marker genes confirm subtype identity of analyzed MSN.

FIGURE 6

Gene expression specificity for MSN subtypes as measured by scRNA-Seq and compared to population-based gene expression data. BacTRAP data for this comparison was taken from Supplementary Table S2 by Heiman et al. (2008). (A) Scatter plot shows correlation for major MSN subtype-specific genes with adjusted p < 0.05 either for scRNA-Seq or BacTRAP. Notable examples are labeled. (B) Differential expression of major D1 and D2 MSN-specific genes commonly found by scRNA-Seq and reported by BacTRAP (adjusted p < 0.05). Genes displayed were found as statistically significant and with same specificity based on at least one microarray probe in the Heiman et al. (2008) data set. Genes are listed by differential expression from D1 MSN- (top, red) to D2 MSN-specific (bottom, green). (C) scRNA-Seq data of additional D1 and D2 MSN-specific markers identified by Heiman et al. (2008) using BacTRAP. (D) D1 and D2 MSN-specific gene expression patterns as identified by BacTRAP experiments that are not confirmed by scRNA-Seq. Genes with adjusted p < 0.01 based on BacTRAP are shown. For genes represented by several probes in the BacTRAP data, the adjusted p-value of the most significant probe is displayed. (E) Examples of reportedly D2 MSN-specific genes that are identified by scRNA-Seq as specific to striatal cholinergic interneurons. (F) Examples of reportedly D1 MSN-specific genes identified by scRNA-Seq as non-subtype-specific or not expressed in MSN.

FIGURE 7

Medium spiny neurons (MSN) subtype-specific genes not found as differentially expressed by population-based gene expression studies. Shown are genes with a differential gene expression of >10-fold and a base mean of >10. (A) Genes listed by differential expression from D1 MSN- (top, red) to D2 MSN-specific (bottom, green). Asterisk (^∗) indicates genes, which showed a tendency for the same MSN subtype in BacTRAP data but were not statistically significant in their differential gene expression (0.05 < p < 0.01). (B) Examples of differential gene expression between D1 and D2 MSN as assessed by scRNA-Seq. (C) Chrm4 read alignments visualized and compared to two distinct gene set annotations, RefSeq and Ensembl. Reads from D1 (top) and D2 (bottom) MSN were separately aligned. No reads for Chrm4 were detected in D2 MSN. For D1 MSN, note the bias of read accumulation at the 3′ end of the Chrm4 gene. The boxed area denotes gene expression signal only observed with the RefSeq gene set annotation, but not with Ensembl. (D) Read alignments in the Chrm4 gene using scRNA-Seq data by Gokce et al. (2016). (E) Boxplot showing differential Chrm4 gene expression between D1 and D2 MSN based on data from Gokce et al. (2016).