Fig. 2

Characteristics of nucleocyclic TSSs in the testes. a. We clustered representative TSSs into three groups (denoted by 1, 2, and 3) according to nucleosome consistency (autocorrelation) downstream of the TSSs in testes. A running average over 21-bp window is shown for smoothing of lines. The upper left box shows the number of TSSs in each group. b. Venn diagram of nucleocyclic TSSs (Group 1) in blastulae, testes, and liver. c. Histogram of probability distribution of gene expression in log scale. The gene expression of the nucleocyclic TSSs in Group 1 is significantly higher than the distribution in Group 2 or 3 (p = 1.9 x 10⁻¹⁷ by Wilcoxon’s ranksum test). d. Average incidences of nucleobase A, T, C and G at positions within 500 bp from representative TSSs in Groups 1, 2 and 3. A running average over 41-bp window is shown. The small histogram in each graph shows the significance of asymmetry of each incidence in individual group (see Methods). A p-value is depicted by –log (p-value) so that a lower p-value is represented by a higher bar to emphasize the higher significance of asymmetry. The difference between the mononucleotide compositions upstream and downstream of TSSs is most pronounced in Group 1. The graphs also indicate that the base incidences around the first nucleosome dyads and linkers significantly differ from those incidences around nucleosomes in the entire genome. e. To verify this hypothesis, individual first nucleosome dyads [+50, +150] downstream of nucleocyclic TSSs are identified (see Methods). f. Incidences of four bases in the region within 200 bp from nucleosome dyads in the entire genome (black line) and first dyads downstream of TSSs (red line). Lines are smoothed using a running average over 41-bp window. Around nucleosome dyads in the entire genome, the A/T (C/G, respectively) incidences around dyads are smaller (greater) than those around linkers, while this tendency is less pronounced around first dyads downstream of TSSs. Indeed, the difference in the tendency is significant, with p < 10⁻⁸ for any of A, C, G, and T by one-tailed Wilcoxon’s ranksum test (Methods)