Circadian neurons in the paraventricular nucleus entrain and sustain daily rhythms in glucocorticoids

To evaluate whether these circadian rhythms were intrinsic to the PVN, we next recorded GCaMP6s calcium fluorescence every hour from individual CRH neurons in ex vivo PVN slices. Similar to in vivo recordings, calcium fluorescence of individual CRH neurons in vitro peaked in the mid-subjective afternoon (CT 8.5; Supplementary Fig. 2i–k). However, unlike in vivo PVN^CRH calcium rhythms, ex vivo calcium rhythms did not persist longer than ~36 h, suggesting that CRH neurons require extra-PVN input for sustained daily rhythms.

To first determine the physiological consequence of ablating BMAL1 in CRH neurons, we used in vivo fiber photometry to monitor calcium fluorescence from PVN^CRH neurons in KO mice injected with Cre-dependent GCaMP6s (Fig. 3b). BMAL1-ablated PVN^CRH neurons responded to a stressor with a large increase in calcium fluorescence, indicating that the stress response of these neurons remained intact after BMAL1 ablation (Supplementary Fig. 3d, e). In LD, BMAL1-ablated PVN^CRH neurons showed significant daily rhythms in calcium event frequency and integrated calcium levels in all recorded mice, peaking in the mid-afternoon, identical to wild-type PVN^CRH neurons (events ZT 7.3, integrated calcium ZT 8.0; Fig. 3d, e and Supplementary Fig. 3f, g). However, the waveform of BMAL1-ablated PVN^CRH neuron calcium event frequency rhythm changed such that they responded acutely to light but failed to continue firing into the early night (Supplementary Fig. 3h, p < 0.01). Remarkably, in DD, BMAL1-ablated PVN^CRH neuron calcium event frequency and integrated calcium levels were each arrhythmic. Together, these results are consistent with light-induced masking and a lack of photoentrainment of BMAL1-ablated PVN^CRH neuron activity.

Next, we customized cages to allow for the non-invasive measurement of fecal corticosterone metabolites from individual mice (Fig. 3b). Mice habituated to the cages in LD for at least 1 week and then in DD for 24 h before we automatically collected fecal pellets every 4 h for three days. Because BMAL1 was present in PVN^CRH neurons of all control genotypes (Crh^Cre/+; Bmal1^fl/fl, Crh^Cre/Cre, and Bmal1^fl/fl), and because there was no significant difference in corticosterone rhythmicity among controls, we analyzed all control genotypes together as “wild-type.” We found that fecal corticosterone peaked each day around 4 h after subjective dusk in WT mice, but, instead, peaked at unreliable times with low amplitude in KO mice (Fig. 3f and Supplementary Fig. 3i). Mice lacking BMAL1 in CRH neurons had roughly half the average, maximum, and peak-to-trough amplitude levels of fecal corticosterone, with no difference in minimum corticosterone levels compared to WT (Supplementary Fig. 3j). Similarly, the peak-trough amplitude—the ratio between mean corticosterone levels from CT 10-18 (when corticosterone normally peaks) and CT 22-6 (the trough of the corticosterone rhythm)—was significantly lower in KO than in WT mice on all days of collection (Fig. 3g). We found that daily corticosterone peaks were synchronous among individual WT mice, peaking near subjective dusk on each day of collection, but were scattered throughout the subjective day in individual KO mice (Fig. 3h). Accordingly, the standard deviation of peak times of daily corticosterone was more than doubled in KO compared to WT mice on all days of collection (day 1, KO 7.5 h, WT 2.2 h; day 2 KO 5.9 h, WT 2.6 h; day 3 KO 6.4 h, WT 1.1 h). Together, these results demonstrate that the ablation of BMAL1 in CRH neurons results in arrhythmic PVN^CRH calcium activity and reduces the amplitude and precision of daily corticosterone release. We conclude that BMAL1 regulates the cell-autonomous circadian regulation of excitability in PVN^CRH neurons and their ability to control circadian corticosterone release, but not their response to light or stress.

Next, we tested whether the effects of SCN^VIP neuronal firing on corticosterone levels depend on time of day. We housed a new cohort of Vip-Cre; ChR2 and Vip-Cre; EGFP mice for at least 1 week in LD and then collected feces in DD for 3 days of corticosterone measurement every 4 h. On the second day of collection, we optogenetically stimulated the animals around subjective dawn (CT 23-1), before the daily increase in SCN^VIP neuron firing and near the trough of the daily corticosterone rhythm in WT mice. We found that activating SCN^VIP neurons at subjective dawn had no effect on the peak or the peak-trough amplitude of the daily rhythm in fecal corticosterone during or after stimulation (Fig. 4e, f and Supplementary Fig. 4c), but significantly delayed the time of peak corticosterone by ~2 h on the day after stimulation (Supplementary Fig. 4d). Together, these results demonstrate that activating SCN^VIP neurons can acutely suppress corticosterone levels when activated during the evening corticosterone surge.

To determine if these effects of SCN^VIP neuron activity on corticosterone patterns depend on acute or circadian changes in PVN^CRH neurons, we first needed to determine whether SCN^VIP and PVN^CRH neurons are synaptically connected. We performed retrograde and anterograde monosynaptic tracing experiments (Supplementary Fig. 6a–d) but only found sparse synaptic connections between SCN^VIP and PVN^CRH neurons. However, using double-label immunohistochemistry, we found dense VPAC2R labeling throughout the SCN and PVN that overlapped with PVN^CRH neurons (Supplementary Fig. 6e), consistent with prior reports^60–63. These results, in combination with our identification of several SCN^VIP axons terminating near PVN^CRH neurons (Supplementary Fig. 6f), suggest that SCN^VIP neurons volumetrically release VIP to communicate with PVN^CRH neurons.

Finally, we made ex vivo explants from triple-transgenic Vip^Cre/+; Ai32^fl/+; PER2^LUC/+ mice to record PERIOD2 expression in SCN and PVN cells while stimulating SCN^VIP neurons. We found that PER2 circadian rhythms in isolated PVN cultures rapidly damped (Supplementary Fig. 6g), consistent with what we observed for PVN^CRH neuron calcium rhythms and what has previously been reported for PVN firing rate and clock gene expression rhythms^44,46. When we recorded PER2 rhythms from the PVN in slices containing the SCN, we found PVN PER2 rhythms persisted for several days. Daily optogenetic stimulation of SCN^VIP neurons for 1 h (470 nm, 8 Hz, 10 ms duration) shifted PER2 rhythms such that after three days of stimulation they were delayed in the SCN and PVN by about 4.5 h compared to controls (Fig. 6i, j and Supplementary Fig. 6h). These results indicate that firing of SCN^VIP neurons can entrain the intrinsic daily rhythms of the PVN.

After weaning and prior to recording, we housed male mice in a 12 h:12 h light/dark (LD, where lights on is defined as zeitgeber time (ZT) 0; light intensity ~2 × 10¹⁴ photons/cm²/s) cycle with food and water provided ad libitum and constant temperature (~22 °C) and humidity (~40%). In all experiments, mice were at least 6 weeks old at the time of virus injection and ~2–5 months old during recordings. For fluorescent and bioluminescent fiber photometry recordings, we used heterozygous Crh-IRES-Cre knock-in mice (Crh^Cre/+)¹¹⁵ on a C57BL/6JN background or homozygous Crh-IRES-Cre mice crossed to homozygous Bmal1^fl/fl mice⁵⁴ on a C57BL/6JN background to yield Crh^Cre/Cre; Bmal1^fl/fl mice. For BMAL1 ablation with corticosterone collection experiments, we used Crh^Cre/Cre; Bmal1^fl/fl mice. As controls, we used either homozygous Crh^Cre/Cre (n = 3), homozygous Bmal1^fl/fl (n = 4), or heterozygous Crh^Cre/+; Bmal1^fl/fl mice (n = 6). For experiments manipulating SCN^VIP neurons and measuring corticosterone, and for in vitro tract-tracing experiments, we used heterozygous Vip-IRES-Cre mice¹¹⁵ on a C57BL/6JN background. For a subset of tracing experiments, we used heterozygous Vip-IRES-Cre mice crossed to heterozygous Ai14 (floxed tdTomato) mice¹¹⁶ on a C57BL/6JN background. For ex vivo experiments, we used Crh^Cre/+ mice crossed to heterozygous Vip-IRES-Flp mice¹¹⁷ backcrossed to a C57BL/6JN background for at least 4 generations to yield Crh^Cre/+; Vip^Flp/+ mice and Vip-IRES-Cre mice crossed with floxed Ai32¹¹⁸ and PER2::LUC¹¹⁹ mice to yield triple-transgenic Vip^Cre/+; Ai32^fl/+; PER2^LUC/+ mice on a C57BL/6JN background as previously published³⁵. All primers used for genotyping are listed in Supplementary Table 2. All experiments were approved by and performed in accordance with the guidelines of Washington University’s Institutional Animal Care and Use Committee.

We used the following viruses in our experiments: AAV9.CAG.FLEX.GCaMP6s (Penn Vector Core; final concentration of ~1 × 10¹³ genome copies (GC)/ml); AAV9.CAG.FLEX.EGFP (Penn Vector Core; ~1 × 10¹³ GC/ml); AAV8.DIO.hChR2(H134R).EYFP (Penn Vector Core; ~6 × 10¹² GC/ml); AAV8.hSyn.DIO.hM4D(Gi).mCherry (Addgene, ~2 × 10¹² GC/ml); AAV9.sCAG.DIO.NES.jRCaMP1b (University of Zurich Viral Vector Facility, ~5 × 10¹² GC/ml); AAVDJ.Ef1a.fDIO.hChR2(H134R).EYFP (UNC Vector Core, ~6 × 10¹² GC/ml), AAVrg.Ef1a.DO.DIO.tdTom.EGFP (Addgene, ~8 × 10¹² GC/ml), and AAVDJ.hSyn.FLEX.Synaptophysin-EGFP (University of Zurich Viral Vector Facility, ~6 × 10¹² GC/ml). We also obtained pAAV.Per2.DIO.Venus and pAAV.Per2.DIO.LUC plasmids as generous gifts from Dr. Eric Erquan Zhang (NBS)⁵⁰ and packaged them into Cre-inducible AAV8 viruses (Washington University Viral Core, ~1 × 10¹³ GC/ml each). To inject viruses, we placed anesthetized mice (1.5% isoflurane) into a stereotaxic frame on a heating pad to maintain their body temperature throughout the procedure. We then infused virus to the PVN (−0.80 mm posterior, ±0.20 mm lateral, −4.65 mm ventral from bregma) or the SCN (−0.46 mm posterior, ±0.15 mm lateral, −5.65 mm ventral from bregma) at a rate of 0.05 µl/min through a 30-gauge needle attached to a 1 µl syringe (Neuros; Hamilton, Reno, NV). We injected a total volume of 0.3 µl unilaterally (GCaMP6s, Per2-Venus, Per2-LUC, EGFP) for fiber photometry experiments, 0.25 µl bilaterally (ChR2) for optogenetics experiments, 0.5 µl bilaterally (hM4Di) for chemogenetics experiments, 0.3 µl unilaterally (CreSwitch, Syp-EGFP) for tracing experiments, and 0.3 µl unilaterally (jRCaMP1b) and 0.25 µl bilaterally (ChR2) for ex vivo experiments. After infusion, we left the needle in place for at least 8 min before slowly withdrawing over the course of 1 min. For fiber photometry experiments, we subsequently implanted a fiber optic cannula (5 mm in length, 400 µm diameter core, 0.48 NA; Doric Lenses, Quebec, Canada) immediately dorsal to the virus injection site. We secured the cannulas with a layer of adhesive cement (Parkell, Edgewood, NY) and opaque black dental cement and covered the dental cement with a thin layer of black nail polish. We then transferred mice to their home cages for ~4 weeks to allow for recovery and virus expression.

After recovery and virus expression, we transferred mice to individual open-topped cages maintained in a temperature-, humidity-, and light-controlled chamber (light cycles: 12 h:12 h LD, constant darkness (DD), or constant light (LL); light intensity ~2 × 10¹⁴ photons/cm²/s) with food and water provided ad libitum. For bioluminescent fiber photometry experiments, we subcutaneously implanted an osmotic pump (Alzet 1004) containing D-luciferin K (100 mM; Goldbio, St. Louis, MO). We chronically tethered freely-moving mice to a fiber optic cable (400 µm core, 0.48 NA, 1.5 m length; Doric Lenses) connected to the implanted optic fiber by a ceramic mating sleeve. We began fiber photometry recording after mice had at least 7 d to acclimate to being tethered using established methods³⁴. Briefly, for fluorescent fiber photometry experiments, blue (490 nm) and violet (405 nm) LED lights (Thorlabs, Newton, NJ) were sinusoidally modulated at 211 and 531 Hz, respectively, using a custom Matlab program (Mathworks, Natick, MA) and a multifunction data acquisition device (National Instruments, Austin, TX). The excitation lights passed through a fluorescence cube containing two separate excitation filters (450–490 nm and 405 nm), reflected off a dichroic mirror, and coupled into the tethered fiber optic cable. We used an optical power meter (Thorlabs) to set the light intensity for each wavelength to ~30 µW at the end of the fiber optic cable. Sinusoidally modulated fluorescence was collected through the same fiber optic cable, passed through an emission filter (500–550 nm), and focused onto a photoreceiver (Newport, Irvine, CA). The photoreceiver signal was then sent to two lock-in amplifiers (Stanford Research Systems, Sunnyvale, CA) that were synchronized to 211 or 531 Hz for demodulation. We collected voltage signals from the amplifiers at 10 kHz using a custom Matlab program and a multifunction data acquisition device and saved them to disk as binary files. For calcium activity recordings, we automatically recorded GCaMP6s fluorescence for 10 min per h for multiple days. For clock gene expression recordings, we automatically recorded Venus fluorescence for 60 s per 15 min for multiple days. We excluded animals from analysis if the 490 and 405 nm signals simultaneously fluctuated by ~100 mV (typically indicating fiber placement inside the third ventricle), if they did not exhibit circadian locomotor activity, or if the fiber or virus was off-target as determined by histology. For bioluminescent fiber photometry experiments, tethered animals housed in constant darkness were connected to a custom-built photomultiplier tube-photon counting head assembly (Hamamatsu H9319). Bioluminescence counts were measured over 10 min and recorded to disk using custom software.

Data were analyzed using established methods³⁴. For calcium activity analysis, we fit the calcium-independent isosbestic signal (violet) to the calcium-dependent signal (blue using a linear least-squares fit). We then calculated change in fluorescence over baseline fluorescence (ΔF/F) as\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{{{{{{\mathrm{Blue}}}}}}}\,{{{{{{\mathrm{signal}}}}}}}-{{{{{{\mathrm{fitted}}}}}}}\,{{{{{{\mathrm{violet}}}}}}}\,{{{{{{\mathrm{signal}}}}}}}}{{{{{{{\mathrm{Fitted}}}}}}}\,{{{{{{\mathrm{violet}}}}}}}\,{{{{{{\mathrm{signal}}}}}}}}$$\end{document}Bluesignal−fittedvioletsignalFittedvioletsignaland adjusted the final values so that a ΔF of 0 corresponded to the 2nd percentile of the signal. We converted ΔF/F values to events by using a peak-finding algorithm (Matlab) to count instances longer than 1 s when ΔF/F exceeded the median plus two standard deviations of the trace or 1.5% ΔF/F, whichever was greater. We also calculated the integrated calcium value of each trace by integrating the ΔF/F values above 0% over 10 min. For clock gene expression analysis, because Per2-Venus does not have an isosbestic point, we excluded the violet signal from analysis (and used it as a negative control, Supplementary Fig. 1c). We averaged the voltage signal from the blue channel over 60 s for each recording to obtain a single voltage value at each timepoint, fit the data with an exponential filter, and subtracted the fitted signal from the raw data to obtain Venus fluorescence values over time in arbitrary units.

We obtained PVN slices using methods modified from ref. ⁴⁴. We decapitated anesthetized mice and cut their brains into 250 µm coronal slices using a vibrotome. We then dissected and cultured bilateral PVN individually on cell culture membranes (Millicell-CM, Millipore, Burlington, MA) in a recording medium (DMEM supplemented with 10 mM HEPES, 100 U/ml penicillin/streptomycin, 4.5 g/l d-glucose, and 2% B27) placed in sealed Petri dishes. We placed dishes on an inverted Tsscope (TE2000-S; Nikon, Tokyo, Japan) within an incubation chamber held at 35°C. We then automatically imaged GCaMP6s fluorescence in PVN^CRH neurons using a digital EMCCD camera (iXon DU-897; Andor Technology, Belfast, UK) once per hour for 48 h using a custom Micro-Manager script. We measured the fluorescence values of individual static, cell-sized ROIs using ImageJ and a custom Python script, detrended the traces by division, and presented the data as a raster plot.

We placed mice in custom-built collectors maintained in a temperature-, humidity-, and light-controlled chamber with food and water provided ad libitum to allow for non-invasive measurement of fecal corticosterone metabolites from individual mice, a validated method for measuring circadian rhythms in corticosterone release^88,120 These collectors consist of a 20 cm × 20 cm plexiglass cage with a wire bottom placed over a funnel coated in silicone lubricant (#300012, WD-40, San Diego, CA) set above a slotted dish that automatically rotated to allow fecal pellets to fall into a new slot positioned under the funnel every hour. We monitored locomotor activity in these cages using a beam-break sensor positioned in front of the food hopper connected to Clocklab software (v6.1, Actimetrics, Evanston, IL). We allowed mice to acclimate to the cages in LD for at least 7 d before beginning fecal pellet collection, after which they were placed into DD for 24 h. Every 24 h after the start of collection, we removed the collected fecal pellets from the dish, placed them into individual tubes, and baked them at 62 °C until completely dry, typically 2–3 days. We calculated the total weight per timepoint, and we estimate that our collector cages allowed us to recover ~85–100% of total feces produced per hour. To account for any timepoints with no fecal pellets, we binned timepoints into four-hour groups for further processing and ground the resultant samples into a fine powder using a mortar and pestle. We weighed out 20 mg of each sample, added 1 ml of 80% methanol, and agitated for 40 min to extract steroids. After centrifuging, we transferred the supernatant to new tubes and placed them in a fume hood until the methanol fully evaporated, typically 6–7 days. We resuspended the dried samples in 200 µl enzyme-linked immunosorbent assay (ELISA) buffer (Cayman Chemical, Ann Arbor, MI) and subsequently diluted the resuspended samples 1:50 (determined empirically based on the standard curve) in ELISA buffer. We processed the samples in duplicate for corticosterone concentration using the instructions in the ELISA kit (Corticosterone ELISA Kit, Cayman Chemical). We subsequently measured absorbance values using a microplate reader at 415 nm (iMark; BioRad, Hercules, CA) and determined the final corticosterone concentrations (in pg/ml) based on the standard curve and dilution factor. There were no measurable differences in fecal weight between groups across all experiments. For our Crh^Cre; Bmal1^fl/fl and optogenetics phase plots, peak times of corticosterone were defined as the time of maximum corticosterone on each day of collection. When two timepoints on a given day differed in their corticosterone levels by <15%, we calculated the corticosterone peak as the midpoint between those two timepoints. Because of the multiple peaks in our chemogenetics data, corticosterone peaks in Fig. 5 and Supplementary Fig. 5 were detected using a peak-finding algorithm in Matlab with identical parameters for each animal.

We transferred control and ChR2-expressing mice to our custom-built feces collectors and chronically tethered them to a fiber optic cable (400 µm core, 0.39 NA, 1.5 m length; Thorlabs) connected to a high-powered 470 nm LED (Thorlabs) under the control of an LED driver. We used an optical power meter to set the light intensity at the fiber tip to ~10.8 mW when driven at 1000 mA. We allowed mice to acclimate to the cages and tethering in LD for at least 7 d before beginning fecal pellet collection. We automatically collected fecal pellets over three days every hour. On the second day of collection, we optogenetically stimulated mice (470 nm, 8 Hz, 10 ms duration) from CT 11-13 or 23-0 using a stimulus generator (Grass Technologies, West Warwick, RI) attached to the LED driver under the control of a light timer.

We transferred control and hM4Di-expressing mice to our custom-built feces collectors and allowed them to acclimate to the cages in LD for at least 7 d before beginning fecal pellet collection. We replaced the water bottles on the cages with 1% sucrose water daily for 3 d prior to the start of collection to acclimate mice to the removal and replacement of their water bottles. We then automatically collected fecal pellets for three days. From CT 6 on the second day of collection until CT 6 the following day, we replaced the water bottles on all cages with clozapine-N-oxide (CNO, 1 mg/kg; Hello Bio, Princeton, NJ) in 1% sucrose water (to mask the bitter taste of CNO). This method of continual CNO administration in the drinking water has been previously shown to successfully alter neuronal activity for the duration of CNO administration^121,122. Virus expression was similar between mice. Mice consumed between 5 and 8 ml of CNO water, receiving at least 1 mg/kg CNO throughout the day; there was no genotype difference in the amount consumed.

We obtained and cultured PVN slices as described above, after which we kept them at 35 °C for at least 1 h before imaging at projected CTs 11-12 or CTs 6-8. We imaged jRCaMP1b or GCaMP6s fluorescence for up to 15 min total using a TRITC or EGFP filter cube and a digital CCD camera (QIClick, QImaging, Tuscon, AZ) at 5 frames/s with QCapture (QImaging) and CamStudio software. For SCN^VIP stimulation experiments, we turned off the imaging excitation light after 4 min and optogenetically stimulated SCN^VIP terminals for 2 min (470 nm, 8 Hz, 10 ms duration, ~10.8 mW at fiber tip) with a fiber optic cable (400 µm core, 0.39 NA, 1.5 m length; Thorlabs) connected to a high-powered LED under the control of an LED driver (Thorlabs) positioned directly above the slice. We then stopped stimulation, turned the imaging excitation light back on, and continued imaging for an additional 8 min. For SCN^VIP inhibition experiments, we imaged for 5 min, pipetted 1 µl CNO (10 µM) directly onto the PVN slice, and continued imaging for an additional 5 min. We verified presence of EYFP (ChR2) or mCherry (hM4Di) expression in SCN^VIP terminals in PVN slices and SCN^VIP cell bodies in adjacent SCN slices at the end of each experiment. We measured the fluorescence values of individual static, cell-sized ROIs using ImageJ, detrended the traces by division, and calculated change in fluorescence over baseline fluorescence (ΔF/F) as\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\frac{{F}_{i}-{{{{{{\mathrm{median}}}}}}}\left(F\right)}{{{{{{{\mathrm{median}}}}}}}(F)}$$\end{document}Fi−medianFmedian(F)We then converted ΔF/F values to events by using a peak-finding algorithm (Matlab) and calculated the integrated calcium value of each trace as described above. We excluded from analysis a subset of slices that had good jRCaMP1b expression (healthy cells, no nuclear filling) but exhibited no calcium dynamics before or after stimulation (ΔF/F_pre and ΔF/F_post = 0 ± 0.01).

We recorded bioluminescence from SCN-PVN cocultures using established methods¹²³. We obtained and cultured slices containing the SCN and PVN as described above and transferred them to a light-tight incubator at 36 °C (Onyx, Stanford Photonics, Palo Alto, CA). We collected images with an electron-multiplying CCD camera (iXon, Andor Technology) with an integration time of 1 h. After ~3 d of imaging, we briefly paused our recordings, optogenetically stimulated the cocultures using a custom-built LED array (470 nm, 10 ms, 8 Hz, 1 h), and resumed imaging. We repeated this stimulation paradigm every 24 h for 3 d and then continued imaging for several additional days. For analysis, we loaded the movies in ImageJ, discarded the first 24 h of recording and the frames immediately before and after the time of stimulation, and drew individual ROIs around each PVN and SCN. We applied adjacent frame minimization to each movie to filter out camera noise or cosmic rays. We detrended raw pixel intensity values using a 4 h moving average, and detected peak times independently using a peak-finding algorithm (Matlab) and crossover analysis¹²⁴. To identify phase shifts, peak times on each day were compared using circular statistics.

For retrograde monosynaptic tracing experiments, we unilaterally injected the PVN of Vip^Cre/+ with 0.3 µl AAVrg.Ef1a.DO.DIO.tdTom.EGFP (CreSwitch) virus¹²⁵ that encodes tdTomato in the absence of Cre and EGFP in the presence of Cre. After transfection, starter neurons in the PVN and Cre^– projection neurons in the SCN would express tdTom, and Vip-Cre⁺ projection neurons in the SCN would express EGFP. For anterograde tracing experiments, we unilaterally injected the SCN of Vip^Cre/+; Ai14 mice with 0.3 µl AAVDJ.hSyn.FLEX.Synaptophysin-EGFP¹²⁶ that expresses EGFP-fused synaptophysin in Cre⁺ neurons. After transfection, Vip-Cre⁺ axons expressing tdTomato and synaptic terminals expressing EGFP would label monosynaptic connections between the SCN and PVN. For each experiment, after 4 weeks for virus expression and recovery, we injected colchicine into the lateral ventricles (1 µl per side, 1 µg/µl) of anesthetized mice, allowed the mice to recover, and perfused them 24 h later at CT 6. Brains were processed, and PVN slices were immunostained for CRH, tdTomato, and EGFP and imaged as described below.

To verify successful optogenetic activation of SCN^VIP neurons, at CT 12, we optogenetically stimulated mice for 1 h, and at CT 14, we anesthetized and transcardially perfused the mice with phosphate-buffered saline (PBS) followed by 4% paraformaldehyde (PFA) as in³⁵. To verify chemogenetic inhibition of SCN^VIP neurons, we injected mice with 1 mg/kg CNO at CT 11.5, presented mice with a 15 min light pulse at CT 12, and perfused 1 h after the start of the light pulse as in³⁴. To verify BMAL1 ablation in PVN^CRH neurons, for tract-tracing experiments, and for VPAC2 immunohistochemistry, we injected colchicine into the lateral ventricles (1 µl per side, 1 µg/µl) of anesthetized mice, allowed the mice to recover, and perfused them 24 h later at CT 6. After perfusion, we extracted the brains, post-fixed them overnight in 4% PFA, and cryoprotected them in 30% sucrose. We then obtained 40 µm frozen coronal sections using a cryostat (Leica, Wetzlar, Germany) and performed immunostaining using established methods³⁴. We used the primary antibodies goat anti-c-FOS (sc52-G, 1:1000; Santa Cruz Biotechnology, Dallas, TX), guinea pig anti-BMAL1 (AB2204, 1:2000; Sigma Aldrich, St. Louis, MO), rabbit anti-CRH (ab8901, 1:1000; Abcam, Cambridge, UK), guinea pig anti-CRH (T-5007, 1:1000; Peninsula Laboratories, San Carlos, CA), chicken anti-GFP (GFP-1020, 1:1000; Aves Labs, Davis, CA), rabbit anti-mCherry (ab167453, 1:1000; Abcam), and rabbit anti-VPAC2 (ab28624, 1:1000, Abcam) and the secondary antibodies DyLight 488 donkey anti-goat (SA5-10086, 1:500; Thermo-Fisher Scientific, Waltham, MA), DyLight 594 donkey anti-goat (SA5-10088, 1:500; Thermo-Fisher Scientific), Alexa 594 donkey anti-guinea pig (706-585-148, 1:500; Jackson ImmunoResearch, West Grove, PA), Alexa 488 donkey anti-rabbit (711-545-152, 1:500, Jackson ImmunoResearch), Alexa 555 goat anti-rabbit (ab150078, 1:500; Abcam), Alexa 488 donkey anti-chicken (703-545-155, 1:500; Jackson ImmunoResearch), and Alexa 647 goat anti-guinea pig (ab150187, 1:500; Abcam). Brains were processed and stained identically (same antibody concentrations, durations, etc.). After immunostaining, we imaged sections on a confocal microscope (Nikon A1, Tokyo, Japan) as a single z-stack using identical laser and capture settings for each image that was to be quantitively compared. We adjusted brightness and contrast settings identically for sections depicted in representative images.

We chose sample sizes to be sufficient for statistical analysis based on similar techniques used in previous publications^34,35. We had no specific methods to blind investigators to genotype or condition during the experiments themselves, but investigators were blinded to the genotype or condition when processing data (e.g, corticosterone samples). We performed data analysis blind to genotype or condition. Mice were randomly assigned to experimental groups when possible (e.g., EGFP vs. ChR2). All statistical tests are presented in Supplementary Table 1. We performed the following statistical analyses in Prism 8.0 (GraphPad, San Diego, CA): Brown–Forsythe ANOVA, two-way repeated-measures ANOVA (or a restricted maximum likelihood mixed-effects model), nested one-way ANOVA, and unpaired two-tailed Welch’s t-test. We performed the following statistical analyses using the Circular Statistics Toolbox¹²⁷ in Matlab: Rayleigh test, two-way circular ANOVA, and Watson–Williams test. We determined circadian rhythmicity using JTK Cycle¹²⁸ in R. We used Shapiro–Wilk and Brown–Forsythe tests to test for normality and equal variances, defined α as 0.05, and presented all data as mean ± SEM.

Further information on research design is available in thelinked to this article. Nature Research Reporting Summary

Supplementary Information Peer Review File Reporting Summary