Network Dynamics Mediate Circadian Clock Plasticity

Abstract

A circadian clock governs most aspects of mammalian behavior. Although its properties are in part genetically determined, altered light-dark environment can change circadian period length through a mechanism requiring de novo DNA methylation. We show here that this mechanism is mediated not via cell-autonomous clock properties, but rather through altered networking within the suprachiasmatic nuclei (SCN), the circadian "master clock," which is DNA methylated in region-specific manner. DNA methylation is necessary to temporally reorganize circadian phasing among SCN neurons, which in turn changes the period length of the network as a whole. Interruption of neural communication by inhibiting neuronal firing or by physical cutting suppresses both SCN reorganization and period changes. Mathematical modeling suggests, and experiments confirm, that this SCN reorganization depends upon GABAergic signaling. Our results therefore show that basic circadian clock properties are governed by dynamic interactions among SCN neurons, with neuroadaptations in network function driven by the environment.

Keywords: automation; autopatcher; in vivo; patch clamp; subcortical; thalamus; whole-cell.

Figure 1. Altered light:dark cycle length changes circadian behavior and SCN topology

(See also Figs S1-3.) (a) Actograms of wheel running for representative PER2::LUC mice before, during, and after entrainment to light:dark cycle lengths of 22 and 26 hours (T22, T26). (b) Free-running period (mean+/-s.e.m., here and elsewhere) after entrainment. (n=4-10/group, 1w-ANOVA F(2,21) = 135.51, p<0.0001, * Dunnett’s post hoc test, p<0.05). (c) PER2::LUC bioluminescence intensity (detrended bioluminescence relative to maximum) in representative SCN slices from mice entrained to different light:dark cycles. (d) Bar graph of data in (c), n=5-7 slices/group. 1w-ANOVA F(2,12)=37.89, p<0.0001; * Tukey’s HSD test post hoc test p<0.01. (e) Average phase maps for SCN after entrainment to T22, T24, and T26, illustrating regional differences in the circadian time (CT) of peak PER2::LUC expression on the first day in vitro. n=9-10/ group. (f) Quantification of average CT peak time on the first cycle in vitro for dorsal SCN (dSCN) and ventral SCN (vSCN), as illustrated in (e). dSCN: 1w-ANOVA F(2,28) = 15.89, p<0.0001; vSCN: 1w-ANOVA F(2,28) = 37.58, p<0.0001; * Dunnett’s post hoc test, p<0.05. (g) Average phase maps illustrating changes in regional phase over the first four days in vitro for SCN from T22, T24 and T26 mice. Color scale shows the time of peak PER2::LUC relative to that for the field rhythm of the whole slice. n=9-10/cycle/group. (h) Average period of dSCN and vSCN regions, quantified from data in (g). dSCN: 1w-ANOVA F(2,28) = 4.36, p<0.05; vSCN: 1w-ANOVA F(2,28) = 4.56, p<0.05; * Tukey’s HSD test, p<0.05.

Figure 2. Different light:dark cycle lengths drive region-specific methylation changes

See also Fig S4-S5. (a) Volcano plot depicting differentially methylated regions (DMRs). DMRs between the SCN sub-regions from different T-cycles are shown in red. Only MEDIP-sets with a least 8 unique mapped reads used for the analysis were shown. (b) Bar graph showing the number of DMRs between the SCN sub-regions. (c) Correlation analysis of the methylation profiles between the SCN sub-regions from different T-cycles across refseq mouse genes (version mm9). Color and number indicate the correlation co-efficient (Pearson r), calculated via (Liu et al., 2011).

Figure 3. Pathways affected by light:dark cycle-dependent methylation

See also Fig S4-S5. Negative Log of P-value plot showing the top 10 terms associated with molecular function of differentially methylation regions using the ENRICHR tool (Chen et al., 2013; Kuleshov et al., 2016).

Figure 4. SCN interregional communication drives light:dark cycle length-dependent period changes

(a) PER2::LUC rhythms of separated vSCN and dSCN slices collected from T22, T24, and T26 mice. (b) Bar graph showing period of separated vSCN and dSCN slices from T22, T24, and T26 mice. (n=8-9 mice/group). dSCN: 1w-ANOVA F (2,22) = 0.543, p=0.58); vSCN: 1w-ANOVA F(2,22) = 2.826, p=0.08. (c) PER2::LUC rhythms of SCN slices from mice entrained to different light:dark cycle lengths with and without 2μM TTX. (d) Bar graph showing period of data from (c) (n=6-7 mice/group). Vehicle: 1w-ANOVA F(2,15) = 13,269 p<0.0001, * Dunnett’s post hoc test, p<0.05; TTX: 1w-ANOVA F (2,15) = 5.249 p<0.05; * Dunnett’s post hoc test, T22 vs T26 p<0.05. (e) Average phase maps of PER2::LUC bioluminescence of SCN slices from mice entrained to different light:dark cycle lengths, with or without 2μM TTX. n = 8 mice/group. Color scale as in Fig 1e. (f) Quantification of regional period differences for SCN slices cultured with or without 2μM TTX. Y-axis, period of the dSCN subtracted from the period of the vSCN. Statistics -- Vehicle: 1w-ANOVA F(2,28) = 11.28, p<0.0001; TTX: 1w-ANOVA F(2,28) = 5.22, p<0.05; * Tukey’s HSD test, p<0.05.

Figure 5. Modeling predicts that GABAergic signaling is necessary for light:dark cycle length-dependent period changes

See also Fig S6. (a) Our model consists of two coupled phase-only oscillators, representing the core (vSCN) and shell (dSCN) regions of the SCN. The core region has intrinsic period τ_C and phase ϕ_C, while the shell region has intrinsic period τ_S and phase ϕ_S. The coupling between the core and shell is composed of two complementary mechanisms that depend on the difference in phase: a synchronizing signal S(Δϕ) and a signal D(Δϕ) that in some situations tends to desynchronize phases. (b) Actograms of model simulations of whole SCN entrained to T22 (left), T24 (middle), and T26 (right) conditions, followed by aftereffects in constant darkness, which correspond to the preceding cycle lengths. Darkness is indicated in grey. Blue indicates subjective night (CT12-24) of the core oscillator (vSCN) and red shows subjective night of the shell oscillator (dSCN), with overlap appearing purple. (c) Simulations of SCN explant rhythms under different light:dark cycle lengths. Y-axis, relative amplitude of summed components. A. The simulated field rhythms are a weighted sum of core (30%) and shell (70%) and exhibit reversed aftereffects on period. Core, shell, and field rhythms of a simulated SCN slice are also shown following entrainment to B. T22, C. T24, and D. T26. See Fig S6 for period values.

Figure 6. GABAergic signaling and DNA methylation are necessary for light:dark cycle-dependent period changes

(a) PER2::LUC rhythms of SCN slices from mice entrained to different light:dark cycle lengths with or without 10uM gabazine (GABAz). (b) Bar graph showing period of data from (a), n=7-10 mice/group. Vehicle: 1w-ANOVA F(2,24) = 12.417, p<0.0001; GABAz: 1w-ANOVA F(2,25) = 0.083, p=0.92; * Dunnett’s post hoc test, p<0.05. (c) PER2::LUC rhythms of SCN slices from mice entrained to different light:dark cycle lengths with or without 50uM Zebularine. (d) Bar graph showing period of data from (c), n=6-16 mice/group. Vehicle: 1w-ANOVA F(2,23) = 17.13, p<0.0001; Zebularine: 1w-ANOVA F(2,25) = 1.07, p=0.35; * Dunnett’s post hoc test, p<0.05.