Stem cell differentiation trajectories in Hydra resolved at single-cell resolution

Stefan Siebert, Jeffrey A. Farrell, Jack F. Cazet, Yashodara L. Abeykoon, Abby S. Primack, Christine E. Schnitzler, Celina E. Juliano. Science 365, eaav9314. 26 July 2019. doi: 10.1126/science.aav9314

The work was initially presented in the following bioRxiv preprint: https://doi.org/10.1101/460154

Contacts: Stefan Siebert (ssiebert@ucdavis.edu) and Celina Juliano (cejuliano@ucdavis.edu)

Overview: The adult Hydra polyp continuously renews all of its cells using three separate stem cell populations, but the genetic pathways enabling this homeostatic tissue maintenance are not well understood. We used Drop-seq to sequence the transcriptomes of 24,985 single Hydra cells to uncover the molecular signatures of a broad spectrum of cell states, from stem cells to terminally differentiated cells. This portal presents these data in an interactive format, and allows for visualization of sequencing statistics, clustering results, and gene expression across every cell in our dataset.

How to use this portal:

Selecting and Exploring Different t-SNE Plots:

To interact with the data, first click the "Explore" tab. This will bring up the default t-SNE plot of the whole transcriptome-mapped dataset with cells colored by cluster identity. To modify the plot being displayed, click on the "View Options" link located in the upper right-hand portion of the screen. This will expand a side bar that can modify what data is displayed. There are multiple interactive t-SNE plots available on this portal that can be selected from the "Load Cluster" dropdown menu. Available are plots for our whole dataset mapped to either the transcriptome ("Whole Transcriptome Clustering") or the genome ("Whole Genome Clustering"). Also available are various plots from subcluster analyses, wherein subsets of cells are re-clustered to allow for a more refined view of transcriptional states. We include here subcluster plots for the three lineages in Hydra ("Ectodermal Clustering", "Endodermal Clustering", and "Interstitial Clustering") as well as a plot of terminally differentiated and progenitor neuronal cells ("Neuronal Clustering"). 

The default view for all plots displays the cluster identify assigned to each cell as part of a clustering analysis performed on the entire dataset; however, additional metadata can be visualized through the "Select annotation" dropdown menu. "nUMI" colors cells based on the number of unique barcoded transcripts assigned to each cell. "nGene" colors cells based on the number of genes in each cell for which there is expression data. "orig.ident" colors cells by their library of origin. Hovering the cursor over individual cells in a plot will display the actual metadata values for the cell being highlighted.

Some of the available metadata will only be applicable to one of the plots. In those cases, the metadata will be placed in the "Cluster-based" subsection of the "Select annotation" dropdown menu and will vary depending on which plot has been selected. Cluster-based metadata include cluster identities that were assigned as part of a subcluster analysis. Cluster-based metadata also include metagene expression scores generated using non-negative matrix factorization (NMF). These metadata plots will color cells based on how strongly they express a specific set of co-expressed genes (e.g. wt80, a set of genes co-expressed in the basal disc).

Plotting Genes:

The portal can also be used to visualize the expression of a gene of interest. To find a gene of interest in either the Hydra 2.0 genome or transcriptome references, a BLAST search can be performed here: https://research.nhgri.nih.gov/hydra/sequenceserver/. To BLAST against the transcriptome reference, select the "Juliano aepLRv2" checkbox. To BLAST against the genome reference, select either the "Augustus Gene Models" checkbox for nucleotide sequence or the "Augustus Protein Models" for protein sequence. 

The ID for the gene of interest can be entered into the search bar in the upper left-hand corner of the "Explore" tab in the portal to visualize gene expression. Please note that the portal requires the query to be an exact match for a gene entry in our dataset. It is therefore recommended to make use of the search bar's autocomplete feature (especially in cases where a swissprot annotation has been appended to the ID, e.g. "t15393aep|sox4_mouse") to ensure the query will work. Also note that it is important that transcriptome IDs (denoted by a "t" prefix) be used only with transcriptome-mapped plots and genome IDs (denoted by a "g" prefix) be used only with genome-mapped plots. After entering the full gene ID, clicking the blue magnifying glass button to the right of the search bar will then generate the expression plots. The portal generates two plots to visualize gene expression, which are located in two different tabs. The "Distribution" tab will display violin plots of a gene's expression for each cluster. The "Scatter" tab will display a t-SNE plot with cells colored based on expression levels for the gene of interest.

Cluster Label Abbreviation Key: bat: battery cell, bd: basal disk, db: doublet cluster, ec: ectoderm, ecEP: ectodermal epithelial cell, en: endoderm, enEP: endodermal epithelial cell, fmgl: female germ-line, gc: gland cell, gmgc: granular mucous gland cell, i: cell of the interstitial lineage, id: integration doublet, mgl: male germline, mp: multiplet, nb: nematoblast, n: neuronal cell, nem: nematocyte, pd: suspected phagocytosis doublet, prog: progenitor, SC: stem cell, smgc: spumous mucous gland cell, tent: tentacle, zmg: zymogen gland cell

Visualizing gene expression on trajectories:

This requires a download of URD objects. A tutorial - URD_Hydra_Plotting_Tutorial.html - written by Jeff Farrell is available at https://github.com/cejuliano/hydra_single_cell.

File download: File download requires a single cell portal account and the user to be logged in. Seurat and URD objects are also available from Dryad: https://doi.org/10.5061/dryad.v5r6077.

For use and information on how these objects were generated see analysis code at  https://github.com/cejuliano/hydra_single_cell.