Fig. 2: Spatiotemporal tracing of single-cell transcriptome.

a, Pulse-chase experiment design on HeLa cells. For the first five timepoints, we used 1 h metabolic labeling (pulse) followed by 0, 1, 2, 4 and 6 h chase. At the last timepoint, we labeled the cells metabolically for 20 h. All cells were then processed by TEMPOmap workflow measuring 998 genes. b, RNA reads (cDNA amplicons) per cell for each pulse-chase timepoint. n = 1,425; 4,024; 4,421; 3,521; 3,727 and 2,303 cells from different pulse-chase timepoints (left to right). Boxplots are defined in terms of mean (center line), 25–75% percentile (bounds of box), lower and upper quartile (whiskers) and outlier values (dots). c, 3D fluorescent images of inprocess TEMPOmap with zoomed-in views of representative single cells in sequencing cycle 1 at each timepoint. Z-stack range, 10 µm. Scale bar, 10 µm. d, Top, stacked bar plot summarizing the fraction of reads in each subcellular region of all cells at each timepoint. Data are presented as mean values ± s.d. The statistics compare the fractions of nuclear reads (blue) across the first five timepoints. ***P < 0.001, Kruskal–Wallis test with post hoc Tukey’s honestly significant difference test. The number of cells (n) at each timepoint is shown. Bottom, subcellular region assignment (nuclear, middle and periphery) of one representative cell. e–f, TEMPOmap single-cell (e) or nucleocytoplasmic (f) RNA measurements rendered as a visualization by PHATE and colored by pulse-chase timepoints (I, III) or cell-cycle marker gene expression (II, IV). Black arrows inferred by RNA degradation vectors indicate the directions of chase time progression. Bottom row, representative raw images of G2/M phase cells separated on PHATE coordinates. All images show mRNAs (white) in HeLa cells (DAPI in blue). Scale bar, 15 µm.