Output from specific brain cells controls daily body rhythms

While SCN VIP neurons are now known to exhibit robust circadian timekeeping properties^30–33, the daily electrophysiological profiles of individual SCN neurons in the intact adult SCN and their relationship to whole-animal physiological rhythms remain unknown. To address this, we employed a mouse line^30,34 where VIP cells express the light-gated cation channel, channelrhodopsin2 (Vip^+/cre; Ai32^+/−, termed here VIP-ChR2; Fig. 1a) and performed perforated multielectrode array (pMEA) recordings from the SCN of acutely prepared adult mouse brain slices (Fig. 1b).

Initially, we evaluated the impact of brief (0.3–3 ms) blue light flashes via an optical fibre positioned over the SCN during pMEA recordings across a full circadian cycle ex vivo. Under these conditions, the number of pMEA recording sites where we observed excitatory responses increased as a function of flash duration, with sites located further from the flash centre (and therefore receiving less light) requiring longer durations to reach firing threshold (Supplementary Fig. 1a–c). This effect plateaued for ≥2 ms flashes, where the distribution of electrode sites showing optogenetic responses matched that expected for SCN VIP cells^{30–32,34,35}. Accordingly, the observed response latencies (mean ± SEM: 1.9 ± 0.1 ms; n = 98 responding sites) were similar those reported previously for direct activation in patch recordings³⁴. As a result of this near synchronous activation of VIP neurons, single unit isolation from these population recordings was challenging. As such, subsequent experiments employed a ‘ramped’ stimulus (Supplementary Fig. 1d) comprising a train of multiple brief flashes to de-correlate the activation of individual VIP neurons. As predicted, these allowed us to readily isolate optogenetically activated neurons across appropriately located SCN recording sites (Supplementary Figs. 1d, e and 2a–c). Owing to the use of these ramped pulses, response latencies were, however, longer than for simple flashes (mean ± SEM: 6.9 ± 0.7 ms; n = 31 activated neurons from 4 slices).

To confirm that the observed responses reflected a direct ChR2-mediated excitation, we then applied antagonists of the main signalling pathways utilised by VIP neurons. Excitatory optogenetic responses persisted unchanged during application of a VIP receptor antagonist³⁶ and the GABA_A receptor locker (+)-bicuculline (Supplementary Fig. 1e, g), ruling out possible synaptic mechanisms. By contrast, among rarer cells displaying optogenetically-driven inhibition, responses persisted following application of VIP antagonist but were subsequently abolished under GABA_A-receptor blockade (Supplementary Fig. 1f, h), confirming a synaptically-mediated origin. Given the lack of VIP antagonist effect in these experiments, we also validated the sufficiency of this manipulation for preventing exogenous VIP-driven responses. As expected³⁷, bath application of 100 nM VIP evoked excitatory or inhibitory responses in subsets of SCN neurons which, in both cases, were reliably prevented by antagonist pre-treatment (Supplementary Fig. 2d–g). As reported previously then³⁴, the optical stimuli employed here do not drive any noticeable VIP-receptor mediated responses in VIP-ChR2 mice.

Having established an approach for reliable identification of VIP neurons, we employed ramped stimuli during long-term pMEA recordings from VIP-ChR2 SCN slices (1pulse/min over ≥26 h). From 16 recordings (slices prepared early or late day, n = 8/time-point), 59 neurons were clearly identified as VIP cells based on optogenetic response (Supplementary Fig. 3a). The majority of these VIP cells (n = 52/59) displayed robust circadian variation in spontaneous activity (see methods), with peak firing strongly clustered across mid-late projected day (Fig. 1c, e, f). These peak firing times were independent of the time of slice preparation (median ZT 6.0 vs. 6.6 h for early vs. late day slices, P = 0.5 from modified-linear multilevel mixed-effects model; n = 35 and 17 respectively), consistent with previous data indicating that acute electrophysiological recordings of this type reliably reflect SCN rhythms in the intact animal^{14,19,38–40}.

We also identified many SCN cells that lacked rapid excitatory responses to optogenetic stimulation (Supplementary Fig. 3b), the majority which (n = 87/110 non-VIP cells) also exhibited circadian variations in spontaneous firing (Fig. 1d–f). Indeed, the basic properties of VIP and non-VIP neurons were very similar (Fig. 1f–h). Importantly, however, there were differences in the timing of activity across the two populations. In particular, while peak VIP cell activity was strongly clustered across the mid-late projected day, the non-VIP population exhibited a significantly broader distribution of peak times that was skewed towards earlier portions of the day (Fig. 1e). Analysis of within-slice estimates of clustering and phasing for the two subpopulations produced qualitatively similar results (Supplementary Fig. 3c–f). Thus, our data indicate that VIP-expressing neurons constitute a functionally distinct subgroup within the SCN, whose output primarily occupies a specific temporal window across the mid-late day.

To confirm that our experimental procedures did not alter the timing of SCN cell activity, we also compared overall SCN population output to equivalent data from control mice (VIP^+/+; Ai32^+/−; Supplementary Fig. 3a). Importantly, there were no significant differences in the phase or clustering of SCN cellular rhythms, the proportion of neurons showing circadian variation nor the single cell circadian waveforms (Supplementary Fig. 4a–d). We did, however, find that absolute amplitudes of VIP-ChR2 SCN firing rate rhythms were modestly reduced relative to control (Supplementary Fig. 4e), perhaps reflecting the reduced VIP expression observed in VIP^+/cre mice⁴¹. Notably, circadian rhythms in behaviour are not noticeably impaired in this line, a finding which is consistent with our own observations that the basic rhythmic properties are intact in VIP-ChR2 slices.

We next sought to identify downstream neurons that received input from SCN VIP cells. Accordingly, we evaluated data from long-term (≥26 h) pMEA recordings spanning known target regions (SPZ, PVN, and ventral thalamus) while optogenetically stimulating the SCN (Fig. 2a, b; n = 44 VIP-ChR2 slices). Across these targets, we identified a subset of neurons (herein termed VIPⁱⁿ) that exhibited rapid attenuation of firing following optogenetic stimulation, consistent with inhibitory input from SCN VIP neurons (Fig. 2b–d; n = 51 cells). More rarely, neurons exhibiting modest increases in firing in response to optogenetic stimulation of the SCN were also identified (Fig. 2b, e; n = 12/718 total cells). Such responses were qualitatively different from those of SCN VIP cells, exhibiting sluggish kinetics indicative of a polysynaptic origin (Fig. 2c; referred to here as slow-activation). Collectively then, these data indicate that VIP cells convey primarily inhibitory signals to downstream neurons.

Since SCN VIP cells exhibit maximal firing across the mid-late subjective day, we next asked whether this translated to an antiphase rhythm in firing for the VIPⁱⁿ cells that receive this signal. Accordingly, the majority of VIPⁱⁿ cells (n = 43/51) displayed clear evidence of circadian variation in spontaneous firing, with peak firing times distributed across the projected night and early projected day (Fig. 3b–d; Supplementary Fig. 5a). The phase clustering in VIPⁱⁿ cells was weaker than that of the SCN VIP cells themselves (Rayleigh R = 0.21 vs. 0.69, P < 0.0001; Browne–Forsythe’s test) and was not overtly dependent on the anatomical location of the recorded neurons (including significant clusters in the SPZ region and around the dorsal cap of the PVN) or the time of slice preparation (Supplementary Fig. 5b–f). Nevertheless, collectively there was a clear absence of VIPⁱⁿ cells whose peak firing occurred during the mid-late projected day (Fig. 3c) and, across the population, average firing rates fell significantly below the daily mean during this epoch (Fig. 3d). This decrease in VIPⁱⁿ cell activity, which aligns with the daily peak in SCN VIP cell population firing, is therefore consistent with a rhythmic inhibitory drive imposed by the SCN VIP cells.

Interestingly, the majority of non-VIPⁱⁿ cells identified in the same experiments (n = 568/667) also displayed daily variations in firing, although in this case we found cells exhibiting peak firing at all phases across the circadian cycle (Fig. 3b, d, e). Thus, across this population, phase clustering was significantly weaker than that observed for VIPⁱⁿ cells (Fig. 3b), with many non-VIP cells exhibiting peak firing during the mid to late day (Fig. 3d). As expected from the lack of phase clustering in the aggregated cell population, within-slice phase estimates for non-VIPⁱⁿ cells were similarly distributed throughout the circadian cycle (Rayleigh R = 0.1; n = 44 slices). There were also differences in the circadian properties of the non-VIPⁱⁿ cells; specifically, peak widths were narrower and peak to trough amplitudes and mean firing rates lower than observed for VIPⁱⁿ cells (Fig. 3g, h).

In sum, input from SCN VIP cells is specifically associated with high amplitude circadian variation in the firing of downstream target cells, which characteristically exhibit low activity during mid-late portions of the day and high activity during the night or early day.

We next investigated the mechanisms by which activation of VIP neurons modulates the firing rate of VIPⁱⁿ cells. For practical reasons, here we used penetrating multisite probes (rather than pMEA) to sample neurons across VIP cell target regions in ex vivo VIP-ChR2 slices while optogenetically stimulating the SCN (Fig. 4a). Data obtained using this closely approach matched the results of pMEA recordings, however. Hence, we again identified a subset of neurons across the VIP cell target regions that exhibited reliable decreases in firing following optogenetic stimulation (Fig. 4a; n = 37 VIPⁱⁿ) and a smaller subset that exhibited slow activations (n = 3/561 isolated neurons). Stimulus-driven decreases in VIPⁱⁿ cell firing were prevented by (+)-bicuculline, with no further change in activity following co-application of the VIP antagonist (Fig. 4b, c; n = 19 VIPⁱⁿ cells tested). Similarly, (+)-bicuculline attenuated responses in the 3 cells which exhibited slow-activation (Fig. 4b, c), consistent with the polysynaptic origin of these responses suggested above.

In sum, our data suggest that SCN VIP cells primarily influence the firing of their downstream target neurons via GABA rather than VIP signalling. Given recent data indicating that paired high-frequency stimuli are required for VIP-dependent phase shifts in VIP-ChR2 mice³⁰, we next evaluated the impact of paired optogenetic stimulation at instantaneous frequencies spanning the range exhibited by VIP neurons (Supplementary Fig. 6a; 0.5–40 Hz). Across this range, VIPⁱⁿ cell responses were reliably predicted by a linear sum of the response to low frequency (0.5 Hz) single pulse stimulation, as were responses of those showing slow-activation (Supplementary Fig. 6b, c). Moreover, in all cases tested, responses were essentially abolished by (+)-bicuculline treatment as for single pulse stimuli (Supplementary Fig. 6d, e).

We also examined the impact of sustained SCN VIP cell stimulation (Fig. 4d; 5s blocks of 0.5–40 Hz stimulation). As expected, this manipulation evoked a rate-dependent decrease in the steady-state firing of VIPⁱⁿ cells (Fig. 4e). The maximal inhibition (57 ± 8% decrease in spontaneous firing) was consistently observed following stimulation at 20 Hz (Fig. 4e, f)—generally considered as the upper limit for spontaneous SCN cell firing²⁰. As above, however, responses to sustained optogenetic stimulation were reliably abolished by (+)-bicuculline across all frequencies (Fig. 4f).

Our data, therefore, confirm that SCN VIP cells can exert a powerful influence over the activity of their downstream target neurons and that this effect is primarily driven by GABA rather than VIP signalling. Since, in our experiments (+)-bicuculline preceded VIP antagonist application, we do not exclude the possibility that VIP release could exert some modulatory influence over these GABA-dependent changes in VIPⁱⁿ cell firing. Nonetheless, we also found that bath application of exogenous VIP exerted negligible effects on the firing rate of VIPⁱⁿ and neighbouring non-VIPⁱⁿ neurons (Supplementary Fig. 6f, g). Most likely then, the mouse SCN primarily employs VIP for internal communication rather than output (at least to those regions we recorded from), an idea supported by the paucity of VIP receptors across the recorded regions in mice^42,43.

To establish how GABAergic output from VIP cells influences daily rhythms in their postsynaptic target neurons, we next assessed the impact of optogenetically inhibiting SCN VIP cell firing. Accordingly, we established a mouse line in which VIP cells express the light-gated proton pump Archaerhosopsin-3 (VIP^+/cre; Ai40^+/−; termed here VIP-Arch) and performed long-term pMEA recordings from the SCN and target regions (n = 23 slices).

As expected, illumination of VIP-Arch slices with yellow light (10 s steps) rapidly decreased firing of a subset of SCN neurons, typically followed by rebound excitation on termination (Fig. 5a; n = 16/67 SCN neurons). Just as in VIP-ChR2 mice, SCN VIP cells identified in this manner reliably exhibited peak spontaneous firing across the mid-late projected day (Fig. 5b). Similarly, recordings across VIP-cell target regions revealed a subset of neurons that exhibited rapid increases in firing in response to illumination of the SCN, as expected for a disinhibition of VIPⁱⁿ cell activity (Fig. 5d; n = 16/248 cells). These VIPⁱⁿ cells exhibited the expected circadian variations characterised by peak firing during the night or early day (Fig. 5e).

We subsequently analysed data from these experiments to determine the impact of acute Arch-inhibition on the circadian activity of VIP cells and their downstream targets (see Methods). This manipulation did not completely silence VIP cells, which continued to express daily variations in firing during Arch-inhibition, albeit with substantially diminished amplitude (Fig. 5c). As a consequence, the portions of the day when VIP cells were firing at a high rate (>50% of the spontaneous maximum) was severely truncated under these conditions (Fig. 5g). In line with these changes, optogenetic inhibition of SCN VIP cell activity produced a corresponding increase in the duration of the VIPⁱⁿ cells’ active epoch (Fig. 5g). These data, therefore, support the view that inhibitory input from SCN VIP cells substantially influences the circadian firing patterns VIPⁱⁿ cells.

We further found asymmetrical changes in circadian waveform of VIPⁱⁿ cell firing. Specifically, analysis of both absolute and proportional changes in spike output revealed a phase-dependent modulation of VIPⁱⁿ cell responses, with consistently robust increases in activity when Arch-stimulation coincided with the daily rise rather than decline in spontaneous firing (Fig. 6h, i). By contrast, Arch-mediated inhibition of SCN VIP cells did not display an equivalent phase-dependency (Fig. 5h, i). This indicates that the asymmetrical increases in VIPⁱⁿ cell firing reflect circadian variation in the degree of spontaneous inhibitory input from SCN VIP cells and highlights a primary role for SCN cell group in controlling the daily rise in VIPⁱⁿ cell firing.

Having established that circadian output from SCN VIP cells shapes rhythmic activity in downstream recipient neurons, we next investigated the physiological significance of this clock output pathway. Here, we employed a chemogenetic strategy to acutely modulate VIP cell output, by microinjecting viral constructs encoding Cre-dependent Gq- or Gi-coupled DREADDs or control vector into the SCN of VIP^+/cre mice (Fig. 6a). As expected^31,44, this resulted in strong transfection of neurons within the ventral, VIP-cell rich, region of the SCN in VIP^+/cre mice but no transfection in VIP^+/+ animals (Supplementary Fig 7a–c).

We then used this approach to examine the impact of VIP cell activity on circulating corticosterone, a major clock-controlled endocrine signal where a potential regulatory influence of SCN VIP neurons has previously been postulated¹¹. To this end, we compared circulating corticosterone in virally transfected mice before and 90 min following injection of vehicle or a DREADD-selective⁴⁵ dose of clozapine (CLZ; 0.1 mg/kg; see Methods). Based on our neurophysiological data and the endogenous diurnal profile of circulating corticosterone in mice (Supplementary Fig. 8), we performed these studies over three different epochs (Fig. 6c), where endogenous corticosterone levels were stable and sub-maximal but spontaneous VIP cell activity was high (mid-day) or low (early-day and early-night).

Vehicle administration did not significantly alter circulating corticosterone levels at any time-point for any of the experimental groups (Supplementary Fig. 8b–d). Similarly, in Gq-DREADD-expressing mice we did not find any significant effect of activating SCN VIP cells across any of the test epochs (Fig. 6c), nor did CLZ injection result in significant changes in circulating CORT in control vector expressing mice (Fig. 6e). By contrast, chemogenetic inhibition of VIP cells in Gi-DREADD-expressing animals significantly increased circulating corticosterone, with particularly robust effects at the mid-day epoch (Fig. 6d). Accordingly, the observed changes in circulating CORT (relative to pre-injection levels) in these Gi-DREADD-transfected animals were significantly larger than those for the vector control group (two-way mixed effects ANOVA; virus: F_1,14 = 14.8, P = 0.002; Epoch: F_2,28 = 2.05, P = 0.15; Epoch X virus: F_2,16 = 1.49, P = 0.24; Sidak’s post-test at mid-day timepoint, P = 0.01). This mid-day time-point, where greatest effects were observed, corresponds to the normal peak in spontaneous SCN VIP cell activity. These data, therefore, suggest that the endogenous activity of SCN VIP cells is especially important for supressing day-time corticosterone release, ensuring that the daily rise is timed to occur just-prior to activity onset.

To further validate these findings, we also confirmed that virally delivered Gi- and Gq-DREADDs effectively modulate the activity of SCN VIP cells. Gi-DREADD-mediated inhibition of SCN VIP neurons was previously shown to prevent neuronal activation following nocturnal light pulses³⁶. Accordingly, we here found in ex vivo recordings from Gi-DREADD-transfected VIP-ChR2 mice (where the DREADD-mCherry and ChR2-EYFP signals strongly co-localise; Supplementary Fig. 7d) that CLZ application during the projected day significantly inhibited the spontaneous firing of optogenetically identified SCN VIP cells (Supplementary Fig. 7e, f). Moreover, that effect was associated with significant increases in the firing of VIPⁱⁿ cells located outside the SCN (Supplementary Fig. 7e, f). Similarly, we found that in vivo activation of chemogenetic constructs in Gq-DREADD transfected VIP^+/cre mice at ZT14.5 was sufficient to significantly elevate c-Fos expression in the SCN (Supplementary Fig. 7g, h). These data, therefore, confirm that our DREADD-based strategy effectively modulates SCN VIP cell output and highlights an important role for this pathways in regulating endogenous rhythms of circulating CORT.

We next used the same approaches to investigate the contribution of SCN VIP cells to regulating other key physiological outputs under clock control, namely heart rate and locomotor activity. Thus, a subset of Gq- (n = 5), Gi-DREADD (n = 5) and control vector (n = 6) transfected VIP^+/cre mice were implanted with radiotelemetry remotes, allowing untethered, home cage, monitoring of heart rate and activity. The impact of chemogenetic activation or inhibition of SCN VIP cells was then investigated at equivalent time-points to those used for assessment of circulating corticosterone.

Inhibition of SCN VIP cells did not significantly alter heart rate (Fig. 7b, d) or activity levels (Supplementary Fig. 9b, e) across any of the analysed time-points. Similarly CLZ injection did not significantly impact heart rate or activity in control vector-transfected VIP^+/cre mice (Fig. 7c, f; Supplementary Fig. 9c, f). By contrast, Gq-DREADD-driven activation of SCN VIP cells significantly reduced heart rate relative to matched vehicle injections, with most reliable effects observed during the early-day a time-course consistent with previously reported DREADD effects⁴⁶ (Fig. 7a, d). Importantly, this effect on heart rate was not due to suppression of activity (Supplementary Fig. 7a). Hence, while analysis of the simultaneously acquired activity data did reveal a significant effect of CLZ in this group, changes observed in activity levels specifically during this early-day epoch were essentially identical for vehicle and CLZ (Supplementary Fig. 9d, Sidak’s post-test: P = 0.47). We did, however, find a modest suppression of activity following activation of SCN VIP cells during the early night (Sidak’s post-test P = 0.04), consistent with a prior report that optogenetic activation of VIP cells during this time-point suppresses locomotor activity³⁰. Accordingly, while our data support this view that VIP cell output can acutely regulate locomotor activity they also indicate a specific role in regulating daily rhythms in cardiac output that is independent of this effect.

All animals were used in accordance with the Animals, Scientific Procedures, Act of 1986 (UK) and received both institutional ethics committee and UK Home Office approval. Experimental animals were generated by crossing VIP-IRES-Cre mice (JAX #010908)⁶⁸ and animals bearing Cre-dependent channelrhodopsin2 (ChR2)-EYFP (Ai32; JAX #0102569)⁶⁹ or Archaerhodopsin 3-EYFP (Ai40D; JAX #021188) constructs, as appropriate. Mice were housed in 12:12-h light/dark cycles in a temperature controlled environment (22 °C). Zeitgeber time (ZT) 0 was designated as time of lights-on and ZT12 as lights-off. Food and water were provided ad libitum.

Electrophysiological experiments were performed on male and female mice (age 41–164 days), that were heterozygous for VIP-IRES-Cre and either Ai32 (VIP^+/cre; Ai32^+/−, termed VIP-ChR2; experiments presented in Figs. 1–4 and Supplementary Figs. 1–7) or Ai40 (VIP^+/cre; Ai40^+/−, termed VIP-Arch; experiments presented in Fig. 5). Where relevant, control data was obtained from VIP^+/+; Ai32^+/− animals (Supplementary Fig 4).

Mice were removed from the home cage at ZT0-1 or ZT11-12 and culled via cervical dislocation, followed by decapitation and brain removal. Coronal slices (350 μm thickness) containing the mid-rostrocaudal extent of the SCN (where VIP cells are concentrated⁸) and the PVN were prepared using a 7000 smz-2 vibrating microtome (Campden Instrument, UK). Slicing was performed in an ice cold sucrose slicing solution (4 °C; sucrose (189 mM); D-glucose (10 mM); NaHCO₃ (26 mM); KCl (3 mM); MgSO₄ (5 mM); CaCl₂ (0.1 mM); NaH2PO4 (1.25 mM)), oxygenated with 95% O₂/5% CO₂. Slices were subsequently placed in aCSF solution (NaCl (124 mM); KCl (3 mM); NaHCO₃ (24 mM); NaH₂PO₄ (1.25 mM); MgSO₄ (1 mM); glucose (10 mM); CaCl₂ (2 mM), 0.0001% gentamicin (Sigma-Aldrich, UK)), oxygenated with 95% O₂/5% CO₂ and left to rest at room temperature for ~20 min. Slices were further equilibrated within the recording setups for 1–2 h prior to the start of electrophysiological data collection.

For multielectrode array recordings, procedures were similar to those described by Hanna et al.⁷⁰. Slices were aligned on 6 × 10 perforated multielectrode arrays (pMEAs; Multichannel Systems, MCS, Germany; Fig. 1), such that recording sites spanned either the SCN and SPZ, or the PVN, SPZ and ventral thalamus. Slice position was visualised under white light trans-illumination using a GXCAM-1.3 camera attached to a dissecting microscope (GX optical, UK) and snapshots were taken to verify placement over electrode sites. Whilst in the recording chamber both slice surfaces were constantly perfused with warm oxygenated aCSF (33 °C ± 1 °C) at a rate of ~3 ml a minute. Consistent flow was driven by a constant vacuum pump (MCS gmbH, Germany) removing waste aCSF, and a peristaltic pump (120S Watson-Marlow, Falmouth, UK; flow rate ~3 ml/min) delivering fresh oxygenated aCSF. Slice position was maintained using a weighted harp (ALA Scientific Instruments, NY, USA), alongside the suction caused by the flow of aCSF through the perforated MEA array. Neural activity was then acquired continuously over >26 h using MC_Rack software via a USB-ME64 system and MEA1060UP-BC amplifier (MCS GmbH, Germany). Signals were sampled at 50 kHz and high pass filtered at 200 Hz (Second order Butterworth). Spikes crossing a threshold (normally set at −16.5 µV) were then extracted as timestamped waveforms (1.5 ms duration) and single unit activity was isolated offline by principle components based spike sorting using Offline Sorter (V3; Plexon, TX, USA) and NeuroExplorer (V5; Nex Technologies).

For ex vivo recordings using penetrating multisite electrodes, coronal brain sections (prepared as described above) were transferred to a heated submerged slice chamber (BSC-1; Digitimer Ltd., Welwyn Garden City, UK) and held down by a weighted harp (ALA Scientific Instruments). Bath temperature was maintained at ~33 °C and fresh, pre-warmed and oxygenated aCSF was supplied by peristaltic pump (120S Watson-Marlow; flow rate ~1.5 ml/min). Silicon 32-channel electrodes (Buszaki 32L, NeuroNexus, MI, USA), each consisting of 4 shanks with 8 closely spaced electrodes (see Fig. 4), were then positioned within the brain slice (with the aid of a dissecting microscope) such that recordings sites either spanned the SPZ, PVN and ventral thalamus or (for Gi-DREADD validation in Supplementary Fig 6d–f) the SCN, SPZ and PVN. Neural signals were acquired by a SmartBox system (Neuronexus) at 20 KHz with spike extraction and single unit isolation performed on virtual tetrode waveforms as described in Howarth et al.⁷¹ using custom Matlab (Mathworks, MA, USA) scripts and Offline sorter software.

At the end of all experiments slices received 5 min bath applications of N-methyl-D-aspartic acid (NMDA, 20 μM) to confirm viability, followed by tetrodotoxin citrate (TTX, 1 μM), to confirm that acquired signals exclusively reflected Na+-dependent action potentials. Where relevant, slices also received one or more of the following compounds, singly or in combination: VIP (100 nM), the VIP receptor antagonist [D-p-Cl-Phe6,Leu17]-VIP (1 µM), (+)-bicuculline (BIC, 20 µM) to block GABA_A receptors and/or D-(-)-2-amino-5-phosphonopentanoic acid (D-AP5, 50 µM) and 6-cyano-7-nitroquinoxaline-2,3-dione disodium (CNQX, 20 µM) to block ionotropic glutamate receptors. For Gi-DREADD validation experiments, slices received clozapine (CLZ, 100 nM). All compounds were obtained from Sigma-Aldrich (Poole, UK) or Tocris (Abingdon, UK). Drugs were stored as 1000X stock solutions at −20 °C (In ddH₂O or, in the case of BIC and CLZ, DMSO) and were diluted to working concentrations in pre-warmed, oxygenated, aCSF just prior to use.

For all experiments, an optical fibre (200 µm core; 0.66NA; Plexon) was positioned directly above the SCN (~100–200 µm above slice surface). For studies employing ChR2, illumination was provided by a PlexBright tabletop 460 nm module (Plexon), providing ~792 mW/m² output at the fibre tip. For long-term pMEA experiments, we applied ‘ramped’ stimuli (consisting of a sequence of 60–140 µs flashes for a total of 3 ms illumination over a 9 ms epoch) at a rate of 1/min for the duration of the recording (>26 h). For acute experiments using penetrating electrodes, square wave 10 ms pulses were applied at varying interstimulus intervals; either as single pulses (0.5 Hz), pairs with ISIs spanning 0.5–40 Hz (repeating interleaved sequence at 2 s/pair) or stimulus trains with rates from 0.5–40 Hz (repeating 4 s interleaved blocks separated by 5 s of no stimulation). For long-term pMEA experiments employing Arch, illumination was provided by a fibre coupled 560 nm LED module (Thorlabs, Ely, UK), providing ~140 mW/mm² at the fibre tip. Here light was provided as 10 s steps every 5 min for the duration of the recording.

Analysis of long-term pMEA recordings was performed using Matlab routines, as described previously⁷⁰. Isolated neuron firing rates were considered to exhibit circadian variation when better fit by a sinusoidal function (constrained to a periodicity of 20–28 h) than a first order polynomial. For rhythmic channels, the projected ZT of peak firing was then determined from a 60 s binned time-series (smoothed with a 2 h boxcar filter). Peak width and peak trough amplitude were determined from this smoothed time-series, with the former representing the duration over which cell firing was >50% of the peak within a single 24 h cycle. For comparison of cell phasing between populations, Rayleigh analysis was used to determine circular median and data were subsequently adjusted to a relative phase such that, within each dataset, each observation was no more than 12 h displaced from the population median. This linearisation procedure allowed for conventional tests to assess differences in variance across the population (Browne–Forsythe’s test) and the corresponding mean phase. Since we identified variable numbers of cells corresponding to the populations of interest from each slice recording, for the latter (and all other group-wise comparisons between derived measures of individual cell firing activity) we used a multilevel mixed-effects linear model that included the slice that each cell was recorded from as a 2nd-level random effect (SPSS, IBM, NY, USA).

Acute changes in neural activity evoked by optogenetic stimuli were classified as excitatory or inhibitory when the average spike counts across multiple stimulus repeats (typically substantially >100) respectively exceeded the upper or lower bounds of the 99% confidence limits for prestimulus spike counts (NeuroExplorer v4; Nex technologies, CO, USA). Latency was determined from the first post-stimulus bin to exceed this boundary. Where cells exhibited fast excitatory responses to ChR2 excitation (peak response detected within 10 ms of the light stimulus) they were considered VIP-expressing cells. Cells matching these criteria (exclusively located in the SCN) were easily distinguishable from rarely encountered cells in SCN target regions showing weaker/more sluggish excitatory responses (slow-activation). To asses synaptic contributions to optogenetically driven responses, within each relevant experimental block (25 min epochs prior to and during drug application) we calculated the mean stimulus-driven changes in spike firing within an appropriate temporal window (25 ms for single pulse stimuli, 5s for trains, individually tailored for paired stimuli based on the duration over which significant changes in spike rate were observed). Responses were subsequently compared using a multilevel mixed-effects linear model that included the slice that each cell was recorded from as a 2nd-level random effect (as above) and stimulation frequency and/or drug as repeated fixed factors as required. Planned post-hoc comparisons were then applied as appropriate where significant main effects of the fixed factors were detected.

For comparison of observed vs. predicted responses to paired optogenetic stimulation (Supplementary Fig.), we first calculated post-stimulus spike count histograms for each cell from 0 ms to 500 ms following low frequency stimulation (LFS, 0.5 Hz) of SCN VIP cells. Predicted response profiles for paired stimulation were then calculated as a simple linear sum of the initial response histogram (extracted above) and an appropriately time-shifted version of the same (with the caveat that the resulting spike counts firing rates were constrained to have a minimal possible value of zero). We then calculated the mean change in firing rate for each cell between response onset and offset (first and last bins were firing rates were significantly below baseline) for both predicted and measured response to LFS and stimulus pairs and normalised the resulting data series to have maximal value of one. Group data processed in this way were then analysed by multilevel mixed-effects linear modelling as above. 6c

For analysis of data from VIP-Arch slices (Fig. 5), cell classification and analysis procedures were essentially identical to those use for VIP-ChR2 with minor modifications. Specifically, to compare circadian profiles in the presence/absence of Arch mediated-inhibition of SCN VIP cells, we first took the raw spike counts over time, extracted from these epochs with and without illumination, and then used spline based interpolation to produce the corresponding time-series of mean firing rate (spikes/s) with 60s resolution, before subsequent processing as above. In addition, we specifically compared firing rates in the presence and absence of Arch mediated-inhibition across specific portions of the circadian profile for each cell (trough, peak, rising and falling; defined respectively by the lower and upper quartiles of the cells firing profile and the corresponding intervening regions). For these comparisons, we assessed both absolute and relative changes in firing rate. In the case of the latter we calculated a standard modulation index of the form (Arch−spontaneous)/(Arch+spontaneous), The resulting measure was therefore bounded between −1 and 1 (for Arch-mediated decreases and increases in firing rate respectively).

Estimates of the anatomical localisation of recorded neurons were obtained from snapshot images of the slice/recording electrode placements, with recording electrodes (and units detected therein) being assigned as located within or outside the SCN based on the observable anatomical boundaries of the nucleus for each slice, evident in the white light trans-illuminated images. Prior analysis⁷⁰ indicates that the actual positions of recorded neurons are likely to lie within ±17 µm of the recording site centres. Coupled with the fact that the precise anatomical boundaries between the extra-SCN regions studied here were not always as unambiguously identifiable in the slice images as for the SCN (in particular the PVN/SPZ border), we did not categorise these extra-SCN neurons to specific target structures. Instead, we estimated anatomical positions based the known electrode array geometries relative to fixed landmarks within the slice; specifically the distance between each recording site and the intersection of best fit lines through the 3rd ventricle and the ventral border of the SCN/optic tract. We then present these data as either binned proportions of identified cells across the dorsal-ventral axis of the slice (Figs. 2b and 4a) or as estimated positions (Supplementary Fig 5b) mapped onto a standardised anatomical template derived from the Mouse Brain Atlas⁷².

Viral injections were performed on male and female mice (49–157 days old). Experiments were primarily performed in VIP^+/cre individuals (Figs. 6 and 7 and Supplementary Figs. 8 and 9). For additional validation of viral specificity (Supplementary Fig. 7b, c), injections were performed in VIP^+/+ animals. For electrophysiological validation of the Gi-DREADD approach (Supplementary Fig. 7d–f) mice were heterozygous for both VIP-IRES-Cre and Ai32 (i.e. VIP^+/cre; Ai32^+/−).

Mice were anaesthetised using 1% isoflurane in O2, and placed into a stereotaxic frame following which the skull was exposed. Craniotomy and bilateral injection of virus was completed using a motorised injector under computer control (Drill and 473 Microinjection Robot: Neurostar, Tϋbingen, Germany). Subsequently, 69 nl of viral vector encoding an activating or inhibiting DREADD or control construct (AAV2-hSyn-DIO-hM3D(Gq)-mCherry, AAV2-hSyn-DIO-hM4D(Gi)-mCherry or AAV2-hSyn-DIO-mCherry respectively; Addgene, MA, USA) was delivered at 25 nl/s. Micropipettes were left in place for 5 min following injection, and were then retracted in 100 µm/5s increments until above the SCN.

Tail vein blood samples (5 µl) were collected, following needle puncture, into heparinised capillary tubes (Drummond Microcaps, 5 μl). Samples were dispensed into 20 µl of PBS and placed on ice before centrifuging (VWR Micro Star 17r; 10,000× for 10 min 4 °C), with the resulting supernatant stored at −80° until analysis.

To assess diurnal variation in cortisol level, samples were collected at 9 time-points across the day-night cycle. Collections completed during the dark phase of the animals were performed under dim red light and animals were allowed sufficient time to recover between sample collections to allow tail healing and stress-induced corticosterone dissipation (minimum 3 days between sampling). Tail blood sampling for DREADD expressing animals was performed at defined epochs across the diurnal cycle (early day, mid-day and early night). Blood was initially sampled at ZT1, 6 or 14.5, immediately followed by injection of a DREADD-selective⁴⁵ dose of clozapine (0.1 mg/kg i.p.; Tocris) or vehicle (0.01% DMSO in saline). This dose of clozapine has previously been shown to produce comparable effects to those commonly used for the original DREADD activator (10 mg/kg), which does not cross the blood brain barrier and therefore requires metabolic conversion to clozapine in order to act centrally⁴⁵. Moreover, at 0.1 mg/kg, the clozapine dose used here is ~10-fold lower than the threshold for non-DREADD mediated effects in other systems^45,73.

Ninety min post injection a second blood sample was collected (a timeframe chosen to encompass the expected peak in DREADD effects). Testing at each epoch was separated by at least 5-days to avoid any lingering effect of the prior manipulations.

In all cases, serum samples were subsequently analysed using pre-coated ELISA plates (ABCAM, ab108821; Cambridge, UK). Concentration changes were assessed using two-way RM ANOVA with Sidak’s multiple comparisons (GraphPad Prism, CA, USA).

Mice were anaesthetised using 1% isoflurane in O₂. Once deeply anaesthetised, an incision was made into the peritoneal cavity and ETA-F10 radio telemetry devices (Data Sciences International; MN, USA) inserted. The muscle lining was sutured and the electrodes of the ECG telemetry remote were routed subcutaneously and sutured to lie on either side of the heart. Once electrodes were in position the wound was closed. Animals were administered analgesia (Buprenorphine 0.01 mg/kg) and left to recover for one week before the start of experiments. During the experiment, mice were singly housed (with ad libitum food) under a 12:12 h light-dark cycle. ECG/activity signals were acquired by a receiver pads placed under the mouse’s home cage (RPC-1; Data sciences international). Activity counts and heart rate (bpm) were acquired in 5 min intervals for the duration of the experiment. For assessment of DREADD-based manipulation of SCN VIP cell activity commenced, animals received vehicle injection (0.01% DMSO in saline, i.p.) at one of three time-points (ZT 1, 6 or 14.5) and subsequently a DREADD-selective clozapine injection (0.1 mg/kg, i.p) at an equivalent timepoint the following day. Initial testing was preceded by a 3-day baseline recording epoch and at least five days drug free days were left between testing at each timepoint.

For visualisation of responses to chemogenetic manipulations, data (5 min bins) were smoothed with a Gaussian function (σ = 30 min). For subsequent analysis, we extracted the change in HR or activity 1–5 h post i.p. injection, relative to the mean value in the preceding 2 h epoch for each animal. This time range was chose to avoid acute changes in heart rate/activity induced by the injection and to encompass that over which DREADD-based responses were reported previously^45,46. The resulting data were then analysed by two-way RM ANOVA with Sidak’s multiple comparisons (GraphPad Prism).

At the end of experimental procedures, virally injected mice were deeply anaesthetised with urethane (10%; i.p. injection) and transcardially perfused with 4% paraformaldehyde. Brains were removed and post-fixed overnight in 4% paraformaldehyde, before being dehydrated in 30% sucrose. Sections were then cut at 50 μm on a freezing sledge microtome (8000 Sledge; Bright Instruments, UK) and blocked for 3 hs (10% Donkey serum, Sigma Aldrich; 0.2%PBS-T) at room temperature. To confirm viral expression, the intrinsic mCherry tag was amplified by incubation with primary antibody (rabbit anti-ds red; 1:2500; TaKaRa Bio, CA, USA; 2.5% donkey serum, 0.2% PBS-T) overnight at 4 °C. Following a 30 min wash (0.2%PBS-T), sections were incubated with secondary antibody (Alexa 546 donkey anti rabbit; 1:1000; Thermo Fisher Scientific; 2.5% donkey serum, 0.2%PBS-T) for 4 h at room temperature. Sections were then mounted using vectashield hardset with DAPI (Vector Labs, H-1500).

A subset of Gq transfected animals underwent concurrent staining for c-Fos activation. Clozapine (0.1 mg/kg, i.p) was delivered 90 min prior to transcardial perfusion, which occurred at ZT16. Protocol was as above, with additional primary antibody (chicken anti c-Fos; 1:200; Sigma Aldrich), and additional secondary antibody (Alexa 488 donkey anti chicken; 1:500; Thermo Fisher Scientific).

For visualisation of VIP cell terminal fields, VIP^+/cre; Ai32^+/− mice were transcardially perfused, sectioned and processed as above, substituting the primary (chicken anti-GFP; 1:10,000; Abcam, Cambridge, MA, USA), and secondary antibodies (Alexa 488 donkey anti chicken; 1:500; Thermo Fisher Scientific) to allow visualisation of the intrinsic EYFP tag expressed by VIP cells in these animals. For DREADD virus expression (Supplementary Fig. 7d–f) these antibodies were used concurrently with those above for mCherry visualisation.

Images were collected on a Zeiss Axioimager.D2 upright microscope using a 10×/0.5 EC Plan-neofluor objective and captured using a Coolsnap HQ2 camera (Photometrics; AZ, USA) through Micromanager software v1.4.23. Images were then processed and analysed using Fiji ImageJ.

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

Materials, algorithms and data generated in these studies are available from the authors upon reasonable request.