A single cell gene expression atlas of 28 human livers

Graphical Abstarct

Chronic liver disease (CLD) affects over 1 billion people worldwide¹ and non-invasive therapeutic options remain an unmet medical need. Thus, advances in uncovering the underlying molecular mechanism(s) of liver disease pathogenesis are essential to facilitate the discovery of novel therapeutic targets.

Next-generation sequencing technologies have revolutionized our understanding of human health and disease through its application to DNA and RNA sequencing. The latter comprises single-cell transcriptomics, which enables sequencing of thousands of cells in a single sample and analysis of gene expression of all cells individually. This creates an unbiased approach to assess cell identity and heterogeneity and to uncover rare cell populations otherwise obscured in bulk RNA-sequencing studies. Taking advantage of fully commercialized workflows, high-throughput single-cell RNA-sequencing (scRNA-seq) is increasingly being used to investigate the cell complexity and heterogeneity of human organs, including the liver. Whereas access to good quality fresh human healthy liver tissue has represented a major limitation in the field, in the last 2 years, 5 independent studies have analyzed human healthy livers at single cell resolution.^2-6 Thus, integration of these 5 scRNA-seq datasets can provide further insight into the transcriptomic architecture and stability of the human liver in physiological conditions, representing highly valuable, and previously unattainable, information for the liver research community worldwide. Herein, we summarize, integrate and analyze available human liver scRNA-seq data and provide an interactive open-access online cell browser for easy access to gene expression data across a variety of annotated parenchymal and non-parenchymal cells derived from 28 human healthy livers.

ScRNA-seq data was accessed from NCBI GEO and processed using Seurat v3.⁷ All datasets were filtered to only include cells expressing 250-2,500 genes as well as mitochondrial expression below 30%. Once filtered and normalized, Seurat was used to batch correct between the independent sources. Detailed methods and R code can be found at http://liveratlas-vilarinholab.med.yale.edu/ and https://github.com/joeb-liver/Single_Cell_Liver_Atlas, respectively.

Liver samples were obtained from individuals of both sexes, a wide range of age groups (21–65-years-old) and with a variety of underlying medical conditions. These include (i) liver resections for colorectal cancer metastasis or cholangiocarcinoma without history of CLD³; (ii) liver tissue from neurologically deceased donors deemed acceptable for liver transplantation²; (iii) liver resection for solitary colorectal metastasis⁶ and (iv) non-diseased liver tissue from deceased individuals which was not suitable for transplantation.⁵ In aggregate, the merged dataset comprises 26 clusters of a total of 36,188 liver cells. Cell lineage was inferred from a combination of unbiased clustering and gene expression profiles. This human liver single cell atlas is composed by 6,895 hepatocytes, 2,357 cholangiocytes, 6,876 liver endothelial cells, 1,604 mesenchymal cells and 18,223 immune cells. Importantly, an R Shiny application was created for interactive visualization and can be accessed at http://liveratlas-vilarinholab.med.yale.edu. This new resource facilitates user-friendly access and interactive visualization of gene expression for each cell (sub)type, which encompasses a range of abundant to rare liver cell populations. Moreover, this new web tool also gives information on which and what proportion of cell (sub)types express a gene of interest.

Collectively, this snapshot compiles all available healthy human liver scRNA-seq data and provides a valuable online resource to compare gene expression across the diverse liver cell populations. This web portal was designed to deliver up-to-date single cell gene expression data accessible to any researcher worldwide working on liver-related biology independent of their bioinformatic training and skills. Hence, this new resource tool has the potential to accelerate basic, translational and clinical research in liver health and disease globally.