Discrete photoentrainment of mammalian central clock is regulated by bi-stable dynamic network in the suprachiasmatic nucleus

Abstract

The biological clock synchronizes with the environmental light-dark cycle through circadian photoentrainment. While intracellular pathways regulating clock gene expression after light exposure in the suprachiasmatic nucleus are well studied in mammals, the neuronal circuits driving phase shifts remain unclear. Here, using a mouse model, we show that chemogenetic activation of early-night light-responsive neurons induces phase delays at any circadian time, potentially breaking the photoentrainment dead zone. In contrast, activating late-night light-responsive neurons mimics light-induced phase shifts. Using in vivo two-photon microscopy, we found that most neurons in the suprachiasmatic nucleus exhibit stochastic light responses, while a small subset is consistently activated in the early subjective night and another is inhibited in the late subjective night. Our findings suggest a dynamic bi-stable network model for circadian photoentrainment, where phase shifts arise from a functional circuit integrating signals to groups of outcome neurons, rather than a labeled-line principle seen in sensory systems.

Fig. 1. Activation of CT 16-trapped light-responsive neurons produces phase delay and breaks the circadian photoentrainment dead zone.

a Experimental scheme for DREADDs bilateral SCN injection in the TRAP2 (Fos-iCreER) mice. b Time table for the experimental procedure. c. Representative actogram for GFP-expressing CT 16 or CT 22-trapped mice before and after CNO injection. d Representative actogram for DREADDs (rM3Ds)-expressing CT 16-trapped mice before and after CNO injection. e Representative behavioral data for DREADDs (rM3Ds)-expressing CT 22-trapped mice before and after CNO injection. The red dots represent the time points of CNO injection, while the red line depicts an extended linear regression based on activity onsets prior to the treatment. The blue lines reflect the activity onsets following treatment. f Statistics of phase shift analysis for DREADDs (rM3Ds)-expressing CT 16-trapped mice. Here, the phase shift form TRAP-CT16 mice injected with CNO were calculated using onset with best of fit line generated according to 3–5 days prior to injection. g Statistics of phase shift analysis for DREADDs (rM3Ds)-expressing CT 22-trapped mice. Numbers indicate adjust p value and n.s. indicates p > 0.05, calculated by Holm-Šídák multiple t-test. n = 7/7, 5/6, 5/6 as control/DREADDs at CT 16, CT 22, CT 2 respectively for f. n = 7/7, 5/8, 5/8 as control/DREADDs at CT 16, CT 22, CT 2 respectively for g. Error bars indicate mean with SEM. Outline of mouse brain in a was drawn according to Paxinos and Franklin’s the Mouse Brain.

Fig. 2. Observation of SCN neuronal light responses in awake mice.

a, b An illustration of recording setup and surgical procedure. c. Representative two-photon in vivo GCaMP images (averaged) of SCN taken at different ZT times through the GRIN endoscope. d Representative 3D-projection of SCN GCaMP signal using z-stack images through the GRIN endoscope. The XY projection shows the summation of horizontal planes, while the XZ projection shows the summation of sagittal planes of the SCN. The XZ virtual 45-μm-thick section demonstrates the elongated point spread function in the z-axis through GRIN endoscopes. e Schemes for time points experiments (upper panel) and calcium imaging recording (lower panel). f Representative images of GCaMP and tdTomato signals collected simultaneously with fast-scanning two-photon system. Arrows indicate the colocalization of both channels. g Averaged light-response of identified neurons at 3 ZT times. Red: ZT 22, green: ZT 8, blue: ZT 16. n = 3 animals. Scale bars are 100 μm, according to the imaging side of the GRIN endoscope. Outline of mouse brain in b was drawn according to Paxinos and Franklin’s the Mouse Brain.

Fig. 3. Diverse neuronal light responses in the SCN.

a Representative heat map with GCaMP traces (normalized Z-score) from 113 identified neurons at ZT 16 (left), ZT 22 (middle), and ZT 8 (right), respectively. b, c 2D t-SNE map (b) shows the diversity of neuronal light-responses. 2875 recording traces from 113 neurons are plotted, excluding those with excessive motion artifacts. The 2875 recordings are classified into 7 clusters, indicated by different colors, with k-means. The 7 light responses are plotted in c, including averaged traces of all clusters (upper-left panel) and raw traces (gray) with average from each cluster (the other 7 panels). Stars above or below the average traces indicate significantly higher or lower z-score compared to the -5 ~ 0 second baseline, respectively. One-way ANOVA and Tukey post hoc test, * indicates p < 0.05. d Raw traces from the representative neuron that show high diversity of light-evoked responses through 9 repeated trials from 3 time points. e Histogram of neurons shows multiple types of light responses. n = 3 animals.

Fig. 4. VIPergic neurons show similar light responses to non-VIP neurons.

a Averaged light response traces from VIP⁺ neurons (blue) and VIP^- neurons (pink). n.s. indicates p > 0.05 by two-way ANOVA Tukey post hoc tests. b Composition of 7 clusters of light response at different time points among VIP⁺ and VIP^- neurons. Color code follows Fig. 3c. No significantly difference is observed between VIP⁺ and VIP^- neurons (two-way ANOVA). c Correlation maps comparing VIP⁺ and VIP^- neurons. Each pixel on the map represents the Pearson correlation coefficient between two neurons. d Violin plots summarize the Pearson correlation coefficient of each comparison pair among cell types, including VIP^- to VIP^- (pink), VIP⁺ to VIP^- (yellow), and VIP⁺ to VIP⁺ (blue). The number indicates p value, n.s. indicates p > 0.05, by one-way ANOVA Tukey post hoc tests. n = 3 animals. For violin plots, circles indicate median, thick vertical lines indicate interquartile ranges and thin vertical lines indicate 1.5X interquartile range.

Fig. 5. Dynamic light responses from individual SCN neurons that form distinct temporal circuit.

a Percentage of inhibitory clusters (Cluster 7 and C6 - 7) is significantly decreased at ZT 8. Error bars indicate mean with SEM. b. Heat map of combined normalized traces from three-time points for 113 identified neurons ranked by mean Z-score at ZT16. c Representative heat maps of Pearson correlation coefficients comparing light responses between neurons. Left panel shows comparison between different ZTs, right panel shows comparison between different trials within the same ZT. Blue outlines indicate same neurons in different trials, and yellow outlines indicate different neurons in same trials. d Cross-time correlation analysis from each neuron between different trials (R-cross, blue outline in c and blue plots in d) is significantly lower than the correlation coefficients for different neurons at the same ZT time within a single trial (R-same, yellow outline in c and orange plots in d). e Representative diagram showing highly correlated neuron pairs (average r from 9 repeats > 0.5) from each time points. f. Scatter plots of distance between pairs of neurons and their Pearson correlation coefficient in respective time points. The r² value of linear regression lines is 0.007, 0, and 0.004 for ZT 16, ZT 22, and ZT 8, respectively. For (a) and (d), numbers indicate p-value with one-way ANOVA, Tukey post hoc tests. Scale bars are 100 μm, according to the imaging side of the GRIN endoscope. n = 3 animals. For violin plots, circle indicate median, thick vertical lines indicate interquartile ranges and thin vertical lines indicate 1.5X interquartile range.

Fig. 6. Three distinct groups of SCN neurons with different time-dependent light responses.

a Representative light response analysis from 3 neurons. Three possible results are demonstrated here, including that the Z-score of the specific time point is significantly higher (orange), lower (blue), or not significantly changed (grey) compared to the baseline (one-way ANOVA and Tukey test). b The results in (a) are plotted together. Neurons are classified into three groups according to their significance pattern across all time points, separated by dotted lines. Group 1 neurons (green) are specific activated at ZT 16. Group 2 neurons (red) are specifically inhibited at ZT 22. While group 3 neurons (grey) do not show consistent positive or negative light response. c Averaged light-response traces from three groups of neurons at different time points. d Group 1 neurons show significantly higher activation clusters (C1 - 4) at ZT 16 and group 2 neurons show significantly higher inhibition clusters (C5 - 6) at ZT 22. Numbers indicate p-values from one-way ANOVA, Tukey post hoc tests. Error bars indicate mean with SEM. e Haar feature analysis of all neurons for each ZT using 18 bins of 5 sec average fluorescence intensity. f Neuron trajectories in the Haar-ZT space show distinct characters from three groups. g Neuron groups are labeled on the raw images. Green dots indicate the cell body of group 1 neurons, red dots indicate group 2 neurons, while grey dots indicate group 3 neurons. Scale bars are 100 μm, according to the imaging side of the GRIN endoscope. n = 3 animals.

Fig. 7. Hypothetic model of SCN circuit for circadian photoentrainment.

Proposed bi-stable functional circuit within the SCN comprised with 3 distinct groups of neurons according to Fig. 6b (classification), 6c (average light response) and Supplementary Fig. 12 (dynamic response). Group 1 and 2 neurons are outcome units to drive phase shift at delay or advance time respectively. Group 3 neurons are the computational unit using populational dynamic coding. Right part: diagram of average group 1-3 neuronal response.