Comment on "Circadian rhythms in the absence of the clock gene Bmal1"

Abstract

Ray et al (Reports, 14 February 2020, p. 800) report apparent transcriptional circadian rhythms in mouse tissues lacking the core clock component BMAL1. To better understand these surprising results, we reanalyzed the associated data. We were unable to reproduce the original findings, nor could we identify reliably cycling genes. We conclude that there is insufficient evidence to support circadian transcriptional rhythms in the absence of Bmal1.

Fig. 1.. Reanalysis of transcriptional cycling in Bmal1^+/+ and Bmal1^−/− mouse skin fibroblast cells across three experiments performed by Ray et al.

(A) Genes detected as cycling in mouse skin fibroblasts at various significance levels for experiment i. P values were adjusted for multiple waveform tests using the ABH method of RAIN, followed by FDR correction across the multiplicity of genes. Relative to the values reported in figure 1E of (1), we identify far fewer cycling transcripts in both the wild type and the Bmal1^−/− knockout. (B) Upset plot showing overlap in genes detected as cycling at FDR < 0.1 in all arms of experiments i to iii; blue and green bars indicate wild type and Bmal1^−/−, respectively. Darker bars indicate the common 37°C, 48 hours post-DEX control condition; lighter bars indicate the experimental conditions (27° and 32°C in experiment ii; PM in experiment iii). The minimum set overlap is truncated at 25 for readability. Note that most genes detected as cycling are unique to a particular dataset and not reproduced by others. Inset: Untruncated upset plot for the common 37°C, 48 hours post-DEX control condition. (Overlaps between control arms may include genes that also overlap in non—control arms, shown in the main plot.) Note again that few genes are detected reproducibly among all three datasets despite the common conditions. Upper right: Scatterplots (upper triangle), histograms (diagonal), and rank correlation coefficients of FDR-adjusted P values for the common genes in the 37°C condition of the three experiments. Red lines in the upper triangle indicate FDR = 0.1, with points displayed on a −log₁₀ scale; genes above the lines on this plot have FDR < 0.1. In the lower triangle, conditional probabilities of having FDR < 0.1 along the y axis given FDR < 0.1 along the x axis (and vice versa) are also shown. (C) Relationship among AM FDR, PM FDR, and AM-PM phase difference in experiment iii. Significant genes at FDR < 0.1 lie above 1 on the −log₁₀ scale; the most significant cyclers lie close to the upper right corner. Highly significant genes tend to have large phase differences in Bmal1^+/+; this is not observed in Bmal1^−/−.