Fig 1.
A. Overview of the study and the various data sources (listed in the callouts). Data were compiled from published studies of mouse liver in animals held under rhythmic light-dark (LD) conditions (green), or of Drosophila (blue), or both (black). Rhythmic transcriptional activities were estimated from Nascent-Seq data as visualized for the Saa1 transcript. Rhythmic transcriptional activities thought to be generated by circadian transcription factors (CTFs) such as REV-ERB α and β, E4BP4, and the CLOCK/BMAL1 heterodimer in mouse, or CLK in Drosophila. The core promoter (beige box) mediates rhythms in CTF binding to rhythmic transcription initiation. This mediation may involve chromatin and nucleosomal remodeling, interacting with rhythms in general transcription factor and Pol II binding to core promoter elements such as the TATA box. B. Overview of promoter classification. Promoters were classified according to whether they drive circadian or constitutive transcription, according to core promoter sequence features, and according to detected CTF binding at the promoter.
Fig 2.
Basic properties of circadian promoters.
A. Fractions promoters with TATA box for different promoter classes, with 95% confidence intervals assuming binomial distributions. Legend: const. = constitutive; circ. = circadian; circ. HA & HT = circadian with high amplitude and high average transcriptional activity (upper quartiles, Methods). Bars from left to right: n = 5829, 1895, 4892, 116. B. Averaged nucleosome occupancies for different promoter classes. Legend as in panel A. Position 0 refers to the TSS. Pileup refers to normalized MNase-Seq reads aligned to a given position. Median pileups for each position and promoter class are visualized. C. Transcriptional amplitude increases with nucleosome occupancy. Mean pileups for all circadian promoters at positions 101 to 1 bp upstream of the TSS were ranked and divided into 5 quantiles. Amplitudes in each such quantile are visualized (medians with 95% confidence intervals). D. Distribution of nucleosome occupancies immediately upstream of the TSSs of circadian promoters. Especially circadian promoters with TATA box appear to have a low occupancy subpopulation coexisting with a subpopulation with high nucleosome occupancies (arrow).
Fig 3.
Identifying strong circadian promoters (SCPs).
A. Amplitudes and averages of transcriptional activities as measured by Nascent-Seq were quantified for the transcript corresponding to each promoter. Promoters were stratified as either TATA box-containing and/or LCpG, or TATA-less and HCpG, as well as exhibiting either no, or at least one CTF ChIP-Seq peaks (CTFs, circadian transcription factors BMAL1, REV-ERBα, REV-ERBβ, and E4BP4). TATA boxes and LCpG were strongly associated with high circadian transcriptional amplitudes, while CTFs rather were associated with high average transcriptional activities. Box plots show medians, 25/75% quantiles, and minimum/maximum values. Promoter groups from top to bottom: n = 665, 605, 330, 295. B. Nucleosome occupancies 101 to 1 bp upstream of the TSSs for SCPs and type I circadian promoters (with TATA box or LCpG but not CTF binding). Occupancies were computed from MNase-Seq data (Methods). Data are presented as kernel densities. C. Nucleosome occupancies around TSSs for different promoter classes. Pileups were computed from MNase-Seq data (Methods) and averaged over the promoter classes for each position relative to the TSS (excluding top and bottom 1% values, respectively, due to a few outlier promoters). Circ. = Circadian; Const. = Constitutive; Type I defined as in panel B, type II = CTF binding TATA-less and with HCpG. D. Fraction of TFIID-dependent promoters for different classes of circadian promoters, with 95% confidence intervals assuming binomial distributions. Bars from left to right: n = 721, 330, 605, 78, 160, 57. E. Mature mRNA expression levels and transcript half lives for different promoter classes. Mature mRNA expression levels were quantified from poly(A)+ RNA-Seq data for the transcript corresponding to each promoter; transcript half lives were compiled from two literature sources (Methods). Mean expression levels and half lives were median averaged over the promoter classes and are shown as points. Error bars represent 95% confidence intervals for the medians. Circ. = Circadian; Const. = Constitutive; TATA = with TATA box; type I and type II as in panels B and C.
Fig 4.
Pol II, nucleosome, and H3K4me3 dynamics at circadian promoters.
A. Pol II levels around TSSs for different promoter classes. Pileups were computed from ChIP-Seq data (Methods) and averaged over the promoter classes for each position relative to the TSS. Circ. = Circadian; Const. = Constitutive; NC = non-TATA box, CTF binding; NN = non-TATA box, non-CTF binding. B. Time-resolved pileups of nucleosome occupancies (MNase-Seq data, methods), and Pol II levels (Pol II ChIP-Seq data, Methods) around the TSSs, averaged over all SCP promoters, phase-shifted according to transcriptional phase. Dot plots represents averages over positions 41 to 121 bp downstream of the TSS (nucleosomes), and positions 51 to 151 bp downstream of the TSS (Pol II). The Pol II levels depicted in the dot plot were rhythmic (harmonic regression p = 0.01), but the nucleosome occupancies were not (harmonic regression p = 0.52). C. Summary of the kinetic model analyzed. The model includes recruitment, initiation, and pausing release steps with accompanying rate constants. Increased Pol II recruitment rate only leads to both increased paused Pol II levels and increased rhythm propagation, if combined with increased Pol II initiation rates and decreased Pol II pause release rate constants. D. Transcriptional amplitudes decrease with higher H2A.Z levels. Mean H2A.Z pileups for all circadian promoters at the +1 nucleosome normalized to the +1 nucleosome peak (Methods) were ranked and divided into 5 quantiles for visualization. Medians and 95% confidence intervals of amplitudes in each such quantile are shown. E. H4K3me3 levels around TSSs for different promoter classes. Pileups were computed from ChIP-Seq data (Methods) and averaged over the promoter classes for each position relative to the TSS (excluding top and bottom 1% values, respectively, due to a few outlier promoters). Circ. = circadian; Const. = constitutive. NC = non-TATA box, CTF binding; NN = non-TATA box, non-CTF binding. Arrow highlights the striking high H3K4me3 signal continuing into the gene body.
Fig 5.
Properties of Drosophila circadian promoters.
A. Fractions promoters with TATA box for different promoter classes. Legend: Const. = constitutive; Circ. = circadian; Circ. HA = circadian with high amplitude (upper 25% quantile). From left to right: n = 5134, 247, 1541, 62. B. Amplitudes and averages of transcriptional activities as measured by Nascent-Seq were quantified for the transcript corresponding to each promoter. Promoters were stratified as TATA box-containing and TATA-less (top), or as exhibiting either no, or at least one CLK ChIP-Seq peak (bottom). Medians and 95% confidence intervals are shown. From top to bottom, left to right: p = 0.0023, p = 0.16, p = 0.29, p = 0.066, rank sum tests. The last p value (mean for CLK vs. no CLK) became much smaller when considering all expressed transcripts: p = 2.5×10−17, median ratio 1.4 CLK vs. no CLK (S3 Table). C. Amplitude increases with nucleosome occupancy, and decreases with levels of nucleosome variant H2A.Z. Mean nucleosome pileups at positions 101 to 1 bp upstream of the TSS, and normalized H2A.Z levels at the +1 nucleosome peak, respectively, for all circadian promoters were ranked and divided into 4 quantiles. Amplitudes in each such quantile are visualized. D. Nucleosome occupancies around TSSs for different Drosophila promoter classes. Pileups were computed from MNase-Seq data (Methods) and averaged over the promoter classes for each position relative to the TSS (excluding top and bottom 1% values, respectively, due to a few outlier promoters). Circ. = circadian; Const. = constitutive.