As a first step towards our goal, I developed a protocol for isolating single ECs from the adult mouse spinal cord. I used tissue from transgenic mice in which all ECs produce enhanced green fluorescent protein (EGFP) and thus can be easily identified. ECs are tightly bound to each other and bringing them into a healthy single-cell suspension was not an easy task. I ended up using a combination of enzymatic digestion (to breakdown the protein links between cells) and mechanical dissociation, and then removed remnants of broken neurons and cell aggregates to generate a single-cell suspension. Taking advantage of EGFP fluorescence in ECs, we used FACS to sort single ECs into 384-well plates containing lysis buffer. After sorting, the plates were immediately frozen until we ship them to San Francisco.
To prepare libraries of the cells and carry out scRNA-seq, we teamed up with Dr. Andy May from the Chan Zuckerberg Biohub (San Francisco, USA). We followed the Smart-seq2 protocol to prepare the cDNA libraries and sequenced them on the NovaSeq 6000 Sequencing System (Illumina) using 2x100bp paired-end reads. This returned 280 GB of raw data, which were transferred back to Dundee for computational analysis.
One of the main objectives of my project was to identify EC subtypes based on their transcriptional profiles. Initially, this is a computational challenge. After preprocessing and filtering out low-quality cells, the resulting dataset contains 995 high-quality cell transcriptomes. In preliminary analyses, I explored a number of approaches for normalisation, dimensionality reduction, manifold learning, clustering and differential expression analysis, to get a deep understanding of the computational tools and the dataset. I set on the broadly-used R package Seurat. Clustering of the cells revealed transcriptionally distinct clusters sharing known EC genes and others not associated with ECs before, some of which I could validate in the Allen Spinal Cord Atlas. To identify marker genes specific of each cluster, I carried out differential gene expression analysis between them and Gene Ontology enrichment analysis to find what set them apart and infer the biology and likely functions of each cluster. This part of the analysis is still ongoing, but we expect to publish our findings in summer of 2020.
An important aim of my project was to validate the computationally-defined EC subtypes experimentally, in intact tissue samples. For that, I used a combination of RNAscope and antibody staining in the same sample to detect multiple RNA transcripts and proteins respectively. Using the set of marker genes identified by differential expression analysis, I confirmed that the proportions of ependymal subtypes are accurately represented in our single-cell dataset. More importantly, with these experiments we are placing EC subtypes in the tissue context and investigating a potential correlation between molecular features and cell shape. Additionally, I am using these spatial methods to investigate whether the proportions of EC subtypes change in the spinal cord of younger and older mice. This data is helping us to establish, for example, whether the EC subtypes that we have defined are stable subtypes or transitional states.
I presented preliminary results of the project in international scientific conferences throughout the action, and the whole dataset will be made freely available for the research community as soon as we submit the paper describing our findings for publication.
