Table 1.
Clinical characteristics of the study subjects.
Figure 1.
Representative cycling and non-cycling DNA methylation sites.
Hash marks indicate means and 95% confidence intervals of the mean. Data are double plotted in 4 hour bins. Red line indicates best-fit cosine curve. (A,B) representative rhythmic DNA methylation sites. (C,D) representative arrhythmic DNA methylation sites. Abbreviations: PVE proportion of variance explained.
Figure 2.
Comparison of observed and permuted null datasets.
(A) Mean and (B) Median proportion of methylation variance collectively explained by cosine curves in the 10000 permuted null (open bars) and the observed (red line) datasets, considering all 420,132 sites. In none of the permuted datasets did the mean or median percentage of variance explained exceed that in the observed data, indicating that the probability of the observed data having occurred by chance alone is p<0.0001. (C) Comparison of the timing of the nadirs of methylation at all 420,132 sites in the null (black) and the observed (red) datasets. The nadir times are uniformly distributed in the null datasets. In contrast, the nadir times are significantly non-randomly clustered in the observed data (H2 = 10449.1; p<2.2×10−16 by the Rao test for equality of dispersions).
Figure 3.
The timing and amplitude of DNA methylation at sites near transcription start sites is distinct from other DNA methylation sites.
(A) Heat map of the distribution of methylation nadir times as a function of distance from the transcription start site, considering all DNA methylation sites, in 500 bp bins. Each row depicts a one-hour bin of clock time. Red indicates a greater than expected density of methylation nadir times; blue indicates a less than expected density of methylation nadir times. (B) The same data depicted as histograms stratified into three groups: sites between −20 kb and −1 kb relative to the nearest transcription start site, sites between −1 kb and +1 kb of the nearest transcription start site, and sites between +1 kb and +20 kb of the nearest transcription start site. DNA methylation sites near transcription start sites have a tendency to reach their methylation nadir in the early morning, peaking at 5:30, while sites elsewhere are most apt to reach their methylation nadir in the evening, peaking at 20:30 (H1 = 7136.9; p<2.2×10−16 by the Rao test for equality of polar direction). (C) The proportion of all DNA methylation sites that have a peak to trough difference of more than 10% of the mean methylation level in 500 bp bins relative to the nearest transcription start site. High-amplitude rhythmic DNA methylation sites are enriched in the 1000 kb around transcription start sites (χ2 64712.6, p<2.2×10−16).
Figure 4.
Sites in gene regions proximate to the transcription start site have a phase and amplitude distinct from other gene regions.
(A) Heat map of the distribution of methylation nadir times for DNA methylation sties in various gene regions, considering all GENCODE v14 transcripts. (B) The same data expressed as histograms. Sites in the 2 kb upstream of the nearest transcription start site and in the 5′UTR, 1st exon, and 1st intron are most likely to reach their methylation nadir in the early morning, while those in other exons/introns, the 3′UTR, noncoding transcripts, and intergenic regions are most likely to peak in the evening (H1 = 14683.8; p<2.2×10-16 by the Rao Test for equality of polar direction). (C) The proportion of DNA methylation sites with high amplitude (peak to trough difference of>10% of the mean methylation level) in each gene region. The 2 kb upstream of the TSS, the 5′UTR, and the 1st exon are relatively enriched in sites with high amplitude oscillations (χ2 51250.8, p<2.2×10−16).
Figure 5.
The timing of methylation for DNA methylation sites with high amplitude oscillations is time locked to the timing of peak RNA abundance.
(A) Heat map depicting the time of the nadir of methylation relative to the timing of RNA abundance for the associated transcript, for high amplitude DNA methylation sites (i.e. amplitude of oscillation at least 10% of the mean methylation level) in or near GENCODE v14 annotated transcripts. Negative numbers mean that the nadir of methylation precedes peak RNA abundance and positive numbers indicate that the nadir of methylation follows peak RNA abundance. Red indicates a greater than expected density, while blue indicates a lesser than expected density. (B) The same data depicted as histograms in 1-hour bins. The dark line indicates the timing of peak RNA abundance. The red line indicates the angular mean time of the methylation nadir. The p-value is for Rayleigh's test for uniformity on the circle. The nadirs of methylation for sites in the 2 kb upstream of the transcription start site, the 5′UTR, the 1st exon, and to a lesser extent the 1st intron were significantly temporally clustered relative to the timing of peak RNA abundance (p = 8.1×10−21, p = 1.4×10−15, p = 1.2×10−15, and p = 3.1×10−7 respectively by Rayleigh's test) with the nadir of methylation preceding peak RNA abundance by 1–3 hours. There was no such temporal clustering in other exons, other introns, and the 3′UTR.
Figure 6.
The effects of sex, age, Alzheimer's disease, and 24-hour activity rhythms on the timing of rhythms of methylation.
Circular histograms of the effect of clinical variables on the timing of the nadir of methylation of 21,282 high amplitude DNA methylation sites, in 1 hour bins. P-values determined using Rao's Test for the equality of polar directions. Dark lines indicate angular means. (A) Female vs. Male Sex. (B) Age above vs. below median (88.4 years). (C) Present vs. absent Alzheimer's Disease based on NIA-Reagan Criteria (intermediate/high = present; low/no = absent). (D) High vs. low actigraphic rhythmicity based on 24-hour activity rhythms determined ante-mortem in a subset of 134 participants by actigraphy and quantified using the interdaily stability metric as described in the text. Female sex, higher age, and absence of dementia were associated with an earlier timing of methylation nadirs.
Figure 7.
The effects of sex, age, Alzheimer's disease, and 24-hour activity rhythms on the amplitude of rhythms of methylation.
Amplitudes expressed as percentage of mean methylation at each site for 21,282 high amplitude DNA methylation sites. Horizontal lines indicate medians. Boxes indicate interquartile ranges. P-values calculated using the Wilcoxon rank-sum test. (A) Female vs. Male Sex. (B) Age above vs. below median (88.4 years). (C) Present vs. absent Alzheimer's Disease based on NIA-Reagan Criteria (intermediate/high = present; low/no = absent). (D) High vs. low actigraphic rhythmicity based on 24-hour activity rhythms determined ante-mortem in a subset of 134 participants by actigraphy and quantified using the interdaily stability metric as described in the text. Greater age, presence of AD by NIA-Reagan criteria, female sex, and lower ante-mortem actigraphic rhythmicity were associated with a lower median amplitude of cycling.