Principal Component Analysis reveals the effects of time of day, genotype, and Zeitgeber on variance between RNA-seq samples. (a) Average variance in expression values (y-axis) along PCs 2–4, with respect to time of day (x-axis) and genotype (red and black bars). Error bars are the SEM for the four biological replicates. PC1 was determined to be a likely technical artifact (Supplemental Figs. S1–3). (b) Example genes whose expression variance was positively or negatively correlated with PCs 2–4. The Pearson correlation coefficient of the gene’s expression variance with the PC is shown in parentheses.

Loss of Klf9 function affects expression of some genes irrespective of time of day. (a) Venn diagrams (produced using Venny;⁵⁴) showing numbers of genes found by DESeq2 to be differentially expressed (log₂FC ≥ 0.5, FDR ≤ 0.1) between klf9^−/− and wildtype larvae at each time point. (b) klf9^−/− induced differential expression of thy1 and hmgcra with respect to DESeq2 distribution at each time point. In this and all following figures, the violin plot shows the distribution of values across all measured genes. (c) Nanostring analysis of thy1 and hmgcra expression with respect to time over an 8-h time course. Each sample consisted of total RNA pooled from 4–5 larvae collected at the indicated times on 5 dpf (see Gans et al., 2021, ref.¹⁷).

Loss of Klf9 function affects expression of per1a and crystallin genes only at a specific time of day. (a) Correlation of per1a expression variance with PCs 2–4. (b) Mean normalized expression of per1a at each time point in each genotype; error bars indicate the SEM. (c) klf9^−/− induced differential expression of per1a with respect to the DESeq2 distribution at each time point; *Adjusted p-value = 0.0013. (d) Nanostring analysis of per1a expression with respect to time over an 8-h time course. Each sample consisted of total RNA pooled from 4–5 larvae collected at the indicated times on 5 dpf (see Gans et al., 2021, ref.¹⁷). (e) klf9^−/−-induced differential expression of 25 cry genes with respect to the distribution of all genes that are differentially expressed between klf9^−/− and WT at each time point. (f) Temporal profiles of cry gene expression. Error bars are the SEM of the four biological replicates.

Effect of the klf9^−/− mutation mapped onto different cell types identified in the single cell atlas of zebrafish development. Differentially expressed genes were mapped to cell clusters identified in the atlas of Farnsworth et al. (ref.²⁶) and annotated as described in the text (see Supplemental Materials, Table S3). Each differentially expressed gene is represented as a dot within the cluster. The locations of hmgcra (downregulated in klf9^−/− mutants) and thy1 (upregulated in klf9^−/− mutants) are shown.

Klf9^−/− mutants have larger livers than wildtype larvae. (a) Image of Alexafluor568-Streptavadin-stained larva showing how the liver was located and segmented to create a 3-D surface for estimating its volume in Imaris. (b) Measurements of liver volumes, combined from separate analyses performed blind by two different individuals. The two sets of measurements differed slightly but systematically, so were brought into register by mean-centering all measurements taken by each experimenter before integrating their data (see Methods and Supplemental Fig. S8). This allowed each individual measurement to be plotted as a quantity relative to the average (“0” on the y-axis). The green diamonds show the 95% confidence interval of the mean, which is indicated by the central horizontal green line; the other two horizontal green lines delimit the region within which the two means are not significantly different. Significance was calculated by t-test.