Weak coupling between intracellular feedback loops explains dissociation of clock gene dynamics

doi:10.1371/journal.pcbi.1007330

Weak coupling between intracellular feedback loops explains dissociation of clock gene dynamics

Fig 1

Statistical hypothesis testing indicates dissociation of Bmal1-ELuc and Per1-luc rhythms at the single cell level.

A) Gaussian kernel density estimates in the bivariate graph of Bmal1 and Per1 oscillation periods, estimated by a Lomb Scargle analysis of surrogate time lapse movies, generated under hypothesis , i.e., dynamical dissociation at the single cell level. B) Same as panel (A) in case of hypothesis , i.e., randomly located cells with either Bmal1 or Per1 signal of different periods. In both panels, N = 150 cells have been randomly drawn. Signal intensities of 1, Bmal1 period of 23h, Per1 period of 24h, cell sizes σ_G = 0.0132 and noise strength of σ_n = 1 were used. See S1 Fig for an example. C) Top: Average values (bold line) and standard deviations (shaded areas) of Per1-luc (blue) and Bmal1-ELuc (orange) signals from a cultured SCN of double transgenic mice. Bottom: Times of oscillation peaks (acrophases) in the averaged Per1-luc (blue) and Bmal1-ELuc (orange) signals. Compared to Per1-luc signals, phase drift of Bmal1-ELuc in oscillation peak times can be observed, suggesting a shorter Bmal1-ELuc period. D) Histograms of pixel-wise oscillation periods in the Per1-luc (blue) and Bmal1-ELuc (orange) signals as determined by a Lomb Scargle periodogram analysis. Bold lines denote fits of a (non-central) Student’s t-distribution to the histogram data. The Student’s t-distribution has been preferred over normal distribution for its lower sensitivity to outliers [26]. Fitted parameters for the location (similar to the mean of a Gaussian) and scale parameter (similar to the standard deviation of a Gaussian) are ≈ 23.87 ± 0.06 h and ≈ 23.40 ± 0.13 h in case of Per1-luc and Bmal1-Eluc signals, respectively. E) Pixel-wise comparison of Per1-luc and Bmal1-Eluc periods as shown by a scatter plot (crosses) together with the corresponding kernel density estimates. The broader distribution of Bmal1-ELuc periods can be due to the lower SNR (signal-to-noise ratio) in comparison to the Per1-luc signal. Data analyzed in panels (C)-(E) correspond to the ones shown in Figure 5 of reference [21].

doi: https://doi.org/10.1371/journal.pcbi.1007330.g001