A Transcriptional Lineage of the Early C. elegans Embryo

Abstract

During embryonic development, cells must establish fates, morphologies, and behaviors in coordination with one another to form a functional body. A prevalent hypothesis for how this coordination is achieved is that each cell's fate and behavior is determined by a defined mixture of RNAs. Only recently has it become possible to measure the full suite of transcripts in a single cell. Here we quantify genome-wide mRNA abundance in each cell of the Caenorhabditis elegans embryo up to the 16-cell stage. We describe spatially dynamic expression, quantify cell-specific differential activation of the zygotic genome, and identify genes that were previously unappreciated as being critical for development. We present an interactive data visualization tool that allows broad access to our dataset. This genome-wide single-cell map of mRNA abundance, alongside the well-studied life history and fate of each cell, describes at a cellular resolution the mRNA landscape that guides development.

Figure 1. Single-cell mRNA-seq libraries for complete sets of cells from C. elegans embryos of the 1-, 2-, 4-, 8- and 16-cell stages

(A) Terminal cell fates of descendants of each cell of the 16-cell embryo. Terminal fates were calculated from Sulston et al. 1983, and refer to cell fates at the time of the first larval hatching. (B) Schematic of samples that were hand-dissected and prepared for scRNA-seq. The 4-cell stage is diagrammed below for illustration. (C) The total mass of mRNA detected from each embryo (diamonds). Embryos whose total mass of mRNA differed from the average by more than one standard deviation (plotted outside of gray band) were excluded from subsequent analyses. (D) The number of genes whose transcripts were detected in each whole embryo (diamonds). (E) The number of genes whose transcripts were detected in each individual cell (circle). (F) Key of the names of each cell from the zygote to the 16-cell stage. See also Table S1

Figure 2. Replicates of each cell type were grouped by transcript signatures and identified by candidate gene expression

(A–C) Transcriptomes of cells from the 2-cell stage (A) were subjected to Principle Component Analysis (PCA) (B) using only data from reproducibly differentially enriched genes, as selected by our algorithm (details in Experimental Procedures). (C) Genome browser tracks of the last exon of erm-1 (AB-enriched) and cpg-2 (P₁-enriched). Colors correspond to embryo of origin. Heights of tracks indicate read count density. All y-axes of genome browser tracks are scaled consistently within each panel. (D–G) Transcriptomes of cells from the 4-cell stage (D) were subjected to PCA (E). (F) PCA of the 10 transcriptomes that were not resolved in (E). (G) Genome browser tracks of the last exon of erm-1 (AB-enriched), med-2 (EMS-enriched) and cpg-2 (P₂-enriched). (H–L) Transcriptomes of cells from the 8-cell stage (H) were subjected to PCA using iteratively generated sets of informative genes (I–K). (L) Genome browser tracks of the last exon of W02F12.3 (ABxx-enriched), tbx-35 (MS-enriched), end-3 (E-enriched), mex-5 (C- and P₃-enriched), and cpg-2 (P₃-enriched). (M–Q) Transcriptomes of cells from the 16-cell stage (M) were subjected to PCA using iteratively generated sets of informative genes (N–P). (Q) Genome browser tracks of the last exon of T09B4.1 (ABxxx-specific), ceh-51 (MSx-specific), end-1 (Ex-specific), pal-1 (Cx- and P₄-specific), and cpg-2 (P₄-specific). See Figures S1 and S2 for further identification of D and P₄ transcriptomes. See also Figure S1, S2

Figure 3. Differential transcript enrichment of notch target genes in cells that could not be distinguished by global transcript signatures

(A) AB descendants from five replicates of the 8-cell stage embryo. (B) Genome browser tracks of ABxx transcriptomes, sorted into groups based on expression of notch target genes hlh-27, ref-1 and tbx-38 (Extended Methods). Last exons only are shown. (C) Example of smFISH targeting hlh-27 (C´, yellow arrows) and ref-1 (C′, purple arrows) transcripts in intact 6- or 8-cell stage embryos (hlh-27 pattern seen in 100% of embryos, n=4. ref-1 pattern seen in 75% of embryos, n=4. Remaining embryo showed ubiquitous ref-1 staining). (D) Example of smFISH targeting tbx-38 (D´, yellow arrows) in intact 8-cell stage embryos (pattern seen in 33% of embryos, n=3. 67% of embryos showed equal tbx-38 expression in ABal and ABar). (D′) nos-2 (P₃-specific) marks the posterior of the embryo. (E) AB descendants from six replicates of the 16-cell stage embryo. (F) Genome browser tracks of ABxxx transcriptomes, sorted into four groups based on a PCA using only notch target gene expression (shown in Figure S2D). Last exons only are shown. (G) Example of smFISH targeting hlh-27 (G´, yellow arrows) and ref-1 (G′, purple arrows) transcripts in intact 15-cell stage embryos (both patterns seen in 100% of embryos, hlh-27 n=5, ref-1 n=2). (H) Example of smFISH targeting tbx-38 (H´, yellow arrows) in intact 15-cell stage embryos (pattern seen in 100% of embryos, n=14). nos-2 (P₄-specific) marks the posterior of the embryo. See Figures S1 and S2 for further identification of ABx, ABxx, and ABxxx transcriptomes. See also Figure S1, S2

Figure 4. Differential activation of the zygotic genome in each cell lineage

(A) Transcript abundances of six genes with previously known expression patterns, heat-mapped on to pictograms of the embryo (key in Figure 1F). Asterisks indicate the cells in which we expected expression, based on the literature; sdz-38 expected in E, Ex (Ea and Ep); tbx-37 expected in ABalx (ABala and ABalp), ABarx (ABara and ABarp); ceh-51 expected in MS, MSx (MSa and MSp); elt-7 expected in Ex (Ea and Ep); cwn-1 expected in Cx (Ca and Cp), D; cey-2 expected in P₀, P₁, P₂, P₃, P₄ (references in Main Text). (B) Heatmap of transcript abundances of all 8,575 detected genes (y-axis) in each cell throughout time and space (x-axis). Only transcriptomes that passed quality filtration were plotted (164 out of 219). The y-axis along the top third of the heatmap is scaled twice as large as the bottom two thirds, to show detail. See Figure S3 for comparisons to related previously published datasets. (C) Transcript abundance data for daz-1 (a maternally inherited gene required for meiosis; Karashima et al. 2000), an example of a transcript we detected in only the germ cells and their sister cells. (D) Transcript abundance data for skr-10 (a member of the ubiquitin ligase complex; Yamanaka et al. 2002), an example of a transcript we detected in only somatic cells. (E) The number of upregulated genes for each cell type. Genes were scored as upregulated in a cell if their transcripts were at least twice as abundant as in any ancestors of that cell. (F) The number of downregulated genes for each cell type. Genes were scores as downregulated in a cell if their transcript abundances were half or less that of an ancestor. (G) The number of cell-specific, or unique, genes. Genes were scored as unique to a cell type if their transcript abundance was at least 10 times higher than in any other cell type in the dataset. (H) Percentage of each cell type’s unique genes, as defined in (G), that are transcription factors. (I) Mass of mRNA per cell as calculated using concentrations of control mRNA spike-ins. (J) Number of genes detected above 25 RPKM in each cell. (K) Length of cell cycle for each cell. (L) Pearson correlation of E-K across all cell types (excluding germ cell precursors, which are transcriptionally distinct; Schaner & Kelly 2006). (M) Matrix of the correlation coefficients of all cell types’ transcriptomes. Six branches of highly correlated cell types are color coded in the cartoon to the right. See also Figure S3

Figure 5. Spatially dynamic gene expression is revealed by high resolution data

(A) A cell lineage map and a pictogram of the 1- through 16-cell stages. Color corresponds to transcript abundance data for tbx-32 in each cell type. (B) smFISH of tbx-32. 100% of 2-cell stage embryos (n=2), 83% of 4-cell stage embryos (n=6, one embryo showed ubiquitous staining), 100% of 6- to 8-cell stage embryos (n=6), and 100% of 12- to 15-cell stage embryos (n=3) showed this pattern. (C) Pictograms for 5 genes showing transcript enrichment patterns similar to that of tbx-32.

Figure 6. Previously unappreciated paralogous, synexpressed genes are critical for development

(A) Correlations of expression patterns for sets of 2–5 genes that are similar to each other in sequence. 295 sets of genes had a correlation coefficient greater than 0.25, and were considered paralogous and synexpressed. (B) Gene set correlations in a scrambled dataset. (C) Histogram of the number of synexpressed paralogous gene sets detected in our dataset (red bar) and in 100 datasets randomized by scrambling gene names without replacement (gray bars). (D) Lethality phenotype observed in embryos in which T24E12.1 and T24E12.13 were targeted by co-injection of dsRNA. Error bars represent 95% confidence interval. See Figure S4 for embryonic lethality in single injections and other pairs of genes co-injected. (E) Pictograms showing quantitative transcript abundance data for the genes highlighted in (D,E). (F) Percent of genes targeted by RNAi in Kamath et al. 2003 that show embryonic lethality. Genes are filtered by their transcript abundance as detected in present study. See also Figure S4; Table S3

Figure 7. An interactive data visualization tool for querying the transcriptional lineage

Still image of data visualization tool. Full version available in Chrome and Firefox browsers at http://tintori.bio.unc.edu. (A–C) Sample selection. The user clicks on the cells or whole embryos they wish to compare on the top (A) and bottom (C) of the plot. When a new sample is selected, the plot (B) is redrawn to reflect the selected comparison. Size of points in B scales to the number of genes represented by each dot. (D–H) Gene selection. (D) The user can filter genes by adjusted P-value of differential enrichment between samples. (E) Clicking on a point or selecting a swath of points on the plot adds genes and their data to the Selected Genes table (F). Known genes can be added directly, by typing their names into the search bar. (G) The Watched Genes table is curated by adding Selected Genes individually or in bulk. (H) The Watched Genes table can be exported, and lists of genes can be imported to the Watched Genes table in bulk. (I–J) Gene expression metrics. (I) The gene tables are sortable by name, average expression level, fold change, significance of differential enrichment, and expression levels in either sample being compared. (J) Clicking on a gene in the table reveals a cartoon of the embryo over all five stages. Each cell is colored corresponding to the transcript level of the highlighted gene.