Extended Data Fig. 12: Cell-type relationships inferred from single-cell data II.

a, Feature gene expression profiles of C1 single cells. Normalized log-transformed FPKM values (y-axis) are used for hierarchical clustering using Spearman coefficients with complete linkage. Major cell types (x-axis) together with an Lmo2⁺ mesenchyme subtype are highlighted using colours corresponding to Fig. 3a. The overall picture showed different numbers of marker genes across cell types. b, c CIBERSORT deconvolution of bulk data. CIBERSORT was used to deduce proportions of major cell types (y-axis) present in staged samples (x-axis) of independently produced forelimbs (b) and ENCODE mixed limb materials (c). The colour codes match Fig. 3a. d, Monocle lineage inference for four skeletal muscle clusters including cluster 22. Pseudotime, developmental time and cell type are shown on the left, and marker gene expression is mapped on the right. e, 20-micrometre sections of mouse E13.5 forelimb double-immunostained for Osr1 (green) and Myog (red) (left), with a DAPI (blue) counterstain (right). All images taken with 63X oil immersion objective. Images in upper panels are enlarged from boxed areas in lower panels. Arrowheads: green: Osr1(+) Myog(−) nucleus. Red: Myog(+) Osr1(−) nucleus. White: double (+/+) cells. Immunocytochemistry was repeated three times independently. f, TFs enriched in skeletal muscle cell types and mesenchyme in either 10x or C1 data. Cells were down-sampled for display (similar to Supplementary Fig. 3); cell types are colour-coded for cell cluster identity. Outlines highlight genes (Myod1; Plagl1) with early stage low-level expression detected in C1 but not 10x data versus pan-lineage markers (Six1; Pitx2) detected in both.