The VIP-VPAC2 neuropeptidergic axis is a cellular pacemaking hub of the suprachiasmatic nucleus circadian circuit

The hypothalamic suprachiasmatic nuclei (SCN) are the principal mammalian circadian timekeeper, co-ordinating organism-wide daily and seasonal rhythms. To achieve this, cell-autonomous circadian timing by the ~20,000 SCN cells is welded into a tight circuit-wide ensemble oscillation. This creates essential, network-level emergent properties of precise, high-amplitude oscillation with tightly defined ensemble period and phase. Although synchronised, regional cell groups exhibit differentially phased activity, creating stereotypical spatiotemporal circadian waves of cellular activation across the circuit. The cellular circuit pacemaking components that generate these critical emergent properties are unknown. Using intersectional genetics and real-time imaging, we show that SCN cells expressing vasoactive intestinal polypeptide (VIP) or its cognate receptor, VPAC2, are neurochemically and electrophysiologically distinct, but together they control de novo rhythmicity, setting ensemble period and phase with circuit-level spatiotemporal complexity. The VIP/VPAC2 cellular axis is therefore a neurochemically and topologically specific pacemaker hub that determines the emergent properties of the SCN timekeeper. Circadian activity modulation in the suprachiasmatic nucleus (SCN) is a network-level emergent property that requires neuropeptide VIP signaling, yet the precise cellular mechanisms are unknown. Patton et al. show that cells expressing VIP or its receptor VPAC2 together determine these emergent properties of the SCN.

C ircadian clocks enable organisms to anticipate and adapt to daily and seasonal environments 1 . The principal circadian clock in mammals is the retinorecipient, hypothalamic suprachiasmatic nucleus (SCN) 2 . It generates a molecular representation of solar time via a cell-autonomous transcriptional/translational negative feedback loop (TTFL), whereby self-sustained circadian oscillations in expression of the negative transcriptional regulators Period (Per) and Cryptochrome (Cry) are driven by heterodimers of the positive regulators Clock and Bmal1 3 . Almost all cells of the body contain similar TTFLs, but rely on daily systemic cues controlled by the SCN to maintain their oscillation 4 and to synchronise them with solar time 5 . As the central circadian co-ordinator, the SCN is unique in its intrinsic ability to generate robust and precise circadian oscillations of TTFL-dependent gene expression and electrophysiological activity, even in the absence of environmental cues. These autonomous oscillations are so powerful that they can persist indefinitely in vivo and ex vivo in organotypic slice culture 6,7 .
This robustness and precision of the SCN are emergent (i.e. circuit-level) properties arising from cellular interactions. When the network is compromised pharmacologically 8 or by dispersal in culture 9 , the individual cells lose circadian coherence and are more sensitive to genetic perturbation. In the intact circuit, however, robustness is accompanied by cellular synchrony across the SCN, with tightly defined ensemble period and ensemble phase. Importantly, the phases of cellular TTFLs across the SCN are staggered as localised cell groups exhibit peak circadian activities in a distinct sequence. This generates stereotypical spatio-temporal waves of gene expression and intracellular calcium levels ([Ca 2+ ] i ) that progress from the dorsomedial to ventral SCN 10 . These waves are thought to be important in allowing the SCN to broadcast distinct and appropriate circadian signals to various targets in the brain 2 and also for circuit-level encoding of daylength, whereby the SCN cues seasonal rhythms 11,12 .
Its emergent properties are critical to the paramount circadian role of the SCN, but the cellular circuit elements and functions that create them are not known 2 . The SCN consists of two subdivisions. The core contains cells that express vasoactive intestinal polypeptide (VIP) or gastrin releasing peptide (GRP), and is the entry point for retinal entrainment 13 , while the shell contains cells expressing arginine-vasopressin (AVP) and the VIP receptor VPAC2. Loss of genes encoding VIP or VPAC2 compromises cellular synchrony and circadian amplitude 14 , while direct optogenetic activation of VIP cells is sufficient to reset the ensemble phase of the SCN 13 , likely via VIP-dependent changes in electrical activity and/or gene expression in VPAC2-expressing cells 15 . In addition, cells expressing neuromedin S (NMS) and D1a dopamine receptors (DRD1A) extend across core and shell 16,17 , while single-cell transcriptomics has revealed further cellular heterogeneity, highlighting potential topological features of the SCN circuit that include several signalling axes, each consisting of neuropeptide-and receptor-expressing cellular populations 18 . Intersectional genetics has recently been used in an attempt to determine the particular contributions of these defined cellular elements to SCN network properties. From this, it is clear that the cell-autonomous circadian properties of NMS-and DRD1Aexpressing neurons 16,17 are important determinants of SCN circuit functions but, paradoxically, signalling by NMS or dopamine is not necessary for robust time-keeping. The contributions of these cells (or sub-groups of them) to circuit functions, therefore, is likely mediated by other signalling factors, which may also include astrocyte-derived signals 19 . Thus, a consensus view on the cellular, neurochemical and topological origin of network properties of the SCN remains elusive. To put it simply, which cell population(s) constituting these signalling axes are the SCN pacemakers? To address this, we focussed on the cells of the neurochemically explicit VIP/VPAC2 axis for reasons outlined above 2,13 , and the observation that paracrine VIP-ergic signalling can re-programme circuit function 14,15,20 . To what extent do the cell-autonomous circadian properties of VIP and VPAC2 cells, alone or in combination, determine the circuit-level ensemble properties of the SCN? Our results show that the VIP/VPAC2 cellular axis is a neurochemically and topologically specific pacemaker hub that determines the emergent properties of the SCN timekeeper.

Results
VIP/VPAC2 cells contribute to SCN spatiotemporal complexity. We characterised mouse lines expressing Cre-recombinase in either VIP or VPAC2 cells, mapping cellular distributions by inter-crossing with R26R-EYFP genomic reporter mice. EYFPlabelled cells exhibited contrasting SCN distributions: VIP EYFP cells clustered ventrally within the core whereas VPAC2 EYFP cells localised around the dorsal, lateral and medial shell (Supplementary Fig. 1a, b). Immunostaining and in situ hybridisation of SCN sections showed clear segregation of VIP and VPAC2 (or AVP) cells ( Supplementary Fig. 1c-f), with minimal evidence for mutual VIP-VIP cell signalling via VPAC2. These lines therefore provided selective genetic access to distinct VIP or VPAC2 cellular compartments.
The emergent network properties of the SCN are evident in CCD recordings of Per2::Luciferase bioluminescence in organotypic slices. These reveal striking spatio-temporal waves whereby the SCN can be mapped into~6 temporally defined, regionally specific phase-clusters (Fig. 1a, Supplementary Movie 1, Supplementary Fig. 2). This phase-map shows earlier activity in the extreme dorsomedial lip (region 1, cells defined, at least in part, by RGS16 expression 21 ) progressing into the ventral core before spreading into the dorsal, lateral and medial SCN shell. Regional distribution of VIP Cre and VPAC2 Cre cells, as revealed by Credependent EYFP fluorescence (delivered by adeno-associated virus (AAV)), mapped to different phase clusters: VIP EYFP neurons were within the ventral phase-advanced core, straddling regions 2 to 4, and with a ventral-to-dorsal wave progression (Fig. 1a, Supplementary Fig. 3a). Conversely, the VPAC2 EYFP neurons mapped dorsally across the shell, straddling delayed phase regions 4-6, with a predominantly lateral wave progression (Fig. 1a, Supplementary Fig. 3b). This coarse mapping suggested that the two cell types are differentially phased with contrasting spatiotemporal dynamics of circadian activation. To examine this directly, we monitored the circadian cytosolic [Ca 2+ ] i rhythms of VIP and VPAC2 cells as reported by GCaMP6f 6,22,23 . When expressed pan-neuronally (via AAV under the synapsin promoter), GCaMP6f reported strong circadian oscillations across the SCN peaking at ca. CT7 (CT12 is the peak of the accompanying Per2:: Luciferase signal), and with a spatio-temporal wave comparable to Per2::Luciferase: emerging in the dorsal SCN, entering the ventral core and with a final dorsal, lateral and medial spread into the SCN shell ( Supplementary Fig. 4a-e). Cre-conditionally expressed GCaMP6f reported strong [Ca 2+ ] i rhythms in both VIP GCaMP6f and VPAC2 GCaMP6f cells (Fig. 1b-d), but with different waveforms: VIP GCaMP6f oscillations showed a sharper rise and broader peak than VPAC2 GCaMP6f cells (Fig. 1e), indicative of population-specific cellular properties. Importantly, when registered against the Per2::Luciferase rhythm, the cycle of VIP GCaMP6f cells was significantly phase-advanced relative to pan-neuronal and VPAC2 GCaMP6f oscillations (Fig. 1e, (Fig. 1g, Supplementary Fig. 4f, g). Furthermore, within these domains they exhibited contrasting local spatio-temporal dynamics: VIP GCaMP6f neurons displayed a strong ventral-todorsal wave, whereas VPAC2 GCaMP6f neurons displayed two features: an early rise in a small number of dorsomedial lip neurons, followed by the bulk of the wave emanating from the centre of the slice medially and laterally (Fig. 1g, Supplementary  Fig. 4e-g)    Fig. 1 Regional distribution and circadian phase of VIP Cre and VPAC2 Cre cells in SCN. a Registration of VIP Cre and VPAC2 Cre cells to corresponding phase maps of SCN spatiotemporal bioluminescent wave. For both cell types, left panel shows projections of average bioluminescence emissions recorded by CCD; middle panel shows the corresponding phase map categorically coloured as 1-h phase bins overlaid with manually annotated sequential phase clusters; and right panel shows distribution of Cre-expressing cells revealed by Cre-conditional EYFP intensity. The perimeter of the SCN is delineated by the white outline, and the location of the third ventricle is denoted by asterisks. Note registration of VIP Cre and VPAC2 Cre cells with earlier and later phase clusters, respectively. These images are representative of a set of three experiments for VIP EYFP and VPAC2 EYFP each, which are presented in Supplementary Fig. 3. Scale bars = 50 µm. b Representative micrographs of Per2::Luciferase in register with GCaMP6f conditionally targetted to VIP (upper) or VPAC2 (lower) cells. Scale bars = 250 µm. c Representative detrended Per2::Luciferase (purple) and conditionally targetted GCaMP6f traces (green) from recordings from VIP Cre (upper) and VPAC2 Cre (lower) slices. d Paired period measures of Per2::Luciferase (purple) and conditional GCaMP6f (green) rhythms. e Phase-aligned normalised single cycles of Per2::Luciferase (two overlaid purple) versus corresponding VIP GCaMP6f (blue) and VPAC2 GCaMP6f (pink) cycles. Lines and shading are mean ± SEM. f Circadian-normalised phase measures for pan-neuronal GCaMP6f (grey), VIP GCaMP6f (blue) and VPAC2 GCaMP6f (pink), (**p = 0.004, ***p = 0.0005). g Circadian-normalised phase-maps of neuronal GCaMP6f (left), VIP GCaMP6f (middle) and VPAC2 GCaMP6f (right). SCN outlines from corresponding Per2::Luciferase phase-maps, asterisks mark third ventricle. In all histograms, individual points represent individual SCN slices and bars are mean ± SEM. Statistics: d paired two-tailed t-tests; f ordinary one-way ANOVA with Holm-Sidak's correction for multiple comparisons. For GCaMP6f imaging (d-f): n = 6, 9 and 10 for neuronal, VIP GCaMP6f and VPAC2 GCaMP6f . Only significant comparisons (p < 0.05) are shown. Source data are provided as a source data file.
When compared between circadian day and night, the contrasting regularity of firing in VIP and VPAC2 cells (p < 0.0001) neither changed with time, nor was there an interaction (p = 0.14) between time and cell type ( Supplementary Fig. 5d, e). Regularity of firing is, therefore, not circadian, but an intrinsic property distinguishing VIP and VPAC2 neurons. In contrast, the temporal pattern of RMP differed significantly between VIP and VPAC2 cells (Fig. 2c-e) (ANOVA, time: p < 0.05, genotype: p < 0.001, interaction: p < 0.0001). VPAC2 TdTom cells were consistently depolarised compared to VIP TdTom cells and showed significant hyperpolarisation at night versus day (Fig. 2d, e), whereas VIP TdTom cells lacked measurable circadian RMP changes (Fig. 2d). In both populations, SFR was significantly higher in circadian day (CT0-8) versus night (CT16-20) (p < 0.0001) (Fig. 2f, g), but with a significant interaction between time and cell type (p < 0.05). Whereas SFR of VIP TdTom cells peaked at CT0-4 ( Fig. 2f), VPAC2 TdTom cells peaked at CT4-8 indicating that electrical activity in VIP cells is phase advanced relative to VPAC2 cells. To obtain complementary analyses of circuit-level electrophysiological activity, we transduced SCN with the voltage sensor AAV synapsin-ArcLight, expressed either panneuronally 19,22 (Supplementary Fig. 5f-h) or conditionally in VIP Cre or VPAC2 Cre cells (Fig. 2h), and phased-registered the fluorescence to simultaneous whole-field Per2::Luciferase recordings. Both VIP ArcLight and VPAC2 ArcLight neurons demonstrated stable circadian rhythms in electrical activity (Fig. 2i, j). Importantly, the peak circadian phase was significantly phase advanced in VIP ArcLight cells by~1 h relative to pan-neuronal and VPAC2 ArcLight cells (Fig. 2k) Supplementary Fig. 6a). Cry1-driven bioluminescence from VIP Cre and VPAC2 Cre SCN slices was strongly circadian (Supplementary Fig. 6b-f). We then determined TTFL phase by co-registration with pan-neuronal synapsin-GCaMP6f ( Fig. 3a-c). Combined bioluminescent/fluorescent CCD recordings confirmed spatially appropriate Cry1-Luciferase signals, and phase-aligned bioluminescence curves revealed significant phase differences: VIP Cry1-Luc cells peaking~1.8 h earlier than VPAC2 Cry1-Luc cells (VIP: CT11.0 ± 0.2 h; VPAC2: CT12.8 ± 0.3 h) (Fig. 3d, e). Interestingly, Rayleigh measurements indicated more phase dispersal in VPAC2 Cry1-Luc cells than VIP Cry1-Luc cells (Fig. 3f, g, Supplementary Fig. 6g, h), echoing their corresponding local [Ca 2+ ] i waves. We therefore quantified the local bioluminescent wave by centre-of-luminescence (CoL) measurements 7,17,19 (Fig. 3h, Supplementary Fig. 6g, h). Similar to the contrasting spatial dynamics of VIP and VPAC2 cells revealed by [Ca 2+ ] i recording, CoL measurements showed celltype-specific properties to the regional spatio-temporal wave: VIP Cry1-Luc bioluminescence exhibited a precisely ordered ventro-dorsal sweep, while VPAC2 Cry1-Luc bioluminescence spread widely across the SCN shell, with limited spatio-temporal coherence as evidenced by the restricted CoL area (Fig. 3i). Together, these data show that VIP and VPAC2 cells form spatially, neurochemically and electrophysiologically distinct populations, with VIP cells phase advanced to VPAC2 cells by all measures: electrical activity, cytosolic calcium and TTFL. Furthermore, the populations exhibit distinct local stereotypical waves, and individually correspond to separate components of the SCN ensemble wave, suggesting that serial signalling through VIP to VPAC2 cells determines a specific temporal segment of the network, and that the two populations may encode and broadcast different variants of circadian information.
Control of circuit coherence and phase by VIP/VPAC2 cells. VIP-and VPAC2-expressing cells have different intrinsic circadian properties. We therefore tested their particular contributions to SCN network function by selective ablation using Creconditional diphtheria toxin receptor (DtR-P2A-mCherry). SCN slices transduced with DtR highlighted VIP Cre or VPAC2 Cre cells with spatially appropriate mCherry signals, >99% of which were lost following specific ablation by addition of diphtheria toxin (Dtx) when compared with vehicle controls (Supplementary Fig. 7a-d). Circuit-level circadian function monitored as Per2:: Luciferase rhythms (Fig. 4a, b) was unperturbed in the absence of DtR and/or Dtx. Dtx-mediated ablation of VIP Cre or VPAC2 Cre cells did not affect ensemble period ( Supplementary Fig. 7e), but did reduce rhythm amplitude and coherence (Fig. 4a, b, Fig. 2 Electrophysiological characterisation of VIP Cre and VPAC2 Cre neurons. a VIP and VPAC2 cells in SCN slices, targetted for recording by Creconditionally expressed TdTomato, which was further confirmed by Biocytin-Alexa488 filling of the cells. Scale bar = 250 µm. b Summary generic cellular electrophysiological properties from current clamp recordings (pooled by slice) made from VIP (blue) or VPAC2 (pink) cells pooled across all circadian times. From left to right: resting membrane potential (RMP), input resistance (R input ), after hyperpolarisation half duration (AHPD 50 ), spontaneous firing rate (SFR) and firing regularity (L v R) followed by representative individual action potential waveforms. Each point represents the mean value for an individual slice (n = 30 slices per genotype, mean of 12 cells per slice), and lines represent median of the slices ±IQR (*p = 0.045, **p = 0.003, ****p < 0.0001). c Example current clamp recordings for VIP (blue) and VPAC2 (pink) cells made during day-time (CT0-4, CT4-8) and night-time (CT16-20) (dotted line: −55 mV). d-g Phase-aligned recordings of: d, e RMP of VIP (d) and VPAC2 cells; f, g SFR for VIP (f) and VPAC2 (g); Blue and pink represent VIP and VPAC2 respectively (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). h Representative micrographs of Per2::Luciferase in register with ArcLight conditionally targetted to VIP (upper) or VPAC2 (lower) cells. Scale bars = 250 µm. i Representative detrended Per2::Luciferase (purple) and conditionally targetted inverted ArcLight traces (green) from recordings from VIP Cre (upper) and VPAC2 Cre (lower) slices. j Paired period measures of Per2::Luciferase and conditional ArcLight rhythms. k Circadian-normalised phase measures of peak depolarisation for pan-neuronal ArcLight (grey), VIP GCaMP6f (blue) and VPAC2 GCaMP6f (pink) (**p < 0.01). In all histograms, each point represents the mean value for an individual slice, while the histogram bar/line represents the mean for all of the slices ±SEM. For time-aligned electrophysiology (d-g): pooled slice level data for n = 10 slices per genotype per time point (except for VIP CT16-18, VIP CT18-20, and VPAC2 CT4-6, where n = 9 slices per genotype per time point); for ArcLight imaging (j, k): n = 6, 8 and 9 for panneuronal, VIP ArcLight and VPAC2 ArcLight . Statistics: b, d-g linear mixed model followed by: b Two-tailed Mann-Whitney U test; d-g two-way ANOVA with Tukey's correction for multiple comparisons; j paired two-tailed t-test; k ordinary one-way ANOVA with Holm-Sidak's correction for multiple comparisons. Only significant comparisons (p < 0.05) are shown. Source data are provided as a source data file.
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17110-x ARTICLE NATURE COMMUNICATIONS | (2020) 11:3394 | https://doi.org/10.1038/s41467-020-17110-x | www.nature.com/naturecommunications Supplementary Fig. 7f, g). CCD recordings to assess cellular and network function revealed that VIP Cre cell ablation restricted the most remaining bioluminescence dorsally, while VPAC2 Cre ablation restricted remaining bioluminescence to the central and core SCN (Fig. 4c, d). Furthermore, raster-and Rayleigh-plots of VIP-ablated SCN revealed progressively desynchronised cellular oscillations with reduced amplitude, and divergence of individual cellular periods well beyond those of the intact slice. In contrast, VPAC2-ablated SCN showed significantly less phase and period divergence ( Fig. 4e-g, Supplementary Fig. 7h-j). Thus, ablation of VIP cells dramatically reduced amplitude and circadian precision, increasing cellular desynchrony and period divergence, while ablation of VPAC2 cells resulted in a largely synchronous SCN with limited phase and period dispersal. VIP and VPAC2 cells therefore contribute differentially to circuit-level functions: loss of VIP cells compromises SCN synchrony more than does loss of VPAC2 cells.
Statistics: e unpaired two-tailed t-test; g, i unpaired two-tailed Mann-Whitney U test. Only significant comparisons (p < 0.05) are shown. Source data are provided as a source data file.
ensemble phase, whereas electrical activation of their VPAC2 target cells was not. If resetting by VIP cells cannot be explained solely by increased firing of their VPAC2 targets, molecular and/or metabolic actions of VIP additional to electrical activation likely effect resetting of the TTFLs of VPAC2 cells. The most direct way would be VIPdependent acute induction of Per1 and Per2, mediated by cAMP/ calcium responsive elements (CREs) 15 . To explore this, we monitored acute changes in Per2::Luciferase bioluminescence in arrhythmic Cry1,2-null (Cry-null) SCN, which provided a stable baseline to observe TTFL responses induced by activation of VIP cells. In Cry-null SCN, acute pan-neuronal optogenetic activation elevated bioluminescence, peaking after~3-4 h and returning to baseline after~10-11 h (Fig. 5g). Stimulation solely of VIP ChR2::EYFP cells in Cry-null SCN resulted in a comparable bioluminescent induction (Fig. 5h). Furthermore, using ChR2::EYFP to delineate the localisation of VIP cells and cross-correlating maps of bioluminescent change with EYFP intensity revealed that optogenetic activation induced Per2 expression both inside and outside the territory of VIP cells, with the same magnitude, kinetics and distribution of responses (both p > 0.2) ( Fig. 5i-l, Supplementary   Fig. 8b). Thus, output from VIP cells activates and recruits pan-SCN cellular TTFLs, with serial activation along the VIP/VPAC2 cellular axis re-setting ensemble phase.
VIP/VPAC2 cells control period and ab initio oscillation. The VIP/VPAC2 cellular axis contributes to SCN spatio-temporal dynamics, maintains cellular synchrony and controls ensemble phase. The final emergent properties are circadian oscillation and ensemble period. The ability of VIP and/or VPAC2 cells to determine ensemble period was tested by genetic complementation of short-period Cry1-null SCN using conditional AAV expression of Cry1 (pCry1-DIO-Cry1::EGFP) to lengthen their cell-autonomous period [26][27][28] . If VIP Cre or VPAC2 Cre cells have pace-setting properties, then following transduction the ensemble period should lengthen from~22 h to~24 h to match their newly established, Cry1-dependent periods 27 . Expression of Cry1 in VIP Cre cells, VPAC2 Cre cells, or in both populations simultaneously was confirmed by appropriate nuclear EGFP signal, with cell-type specificity confirmed by co-localisation with Credependent mCherry expression (pSyn-DIO-mCherry). The transduction efficiency for both AAVs was over 90% of targetted  Fig. 9). Expression of Cry1 in VIP Cre cells caused a small, but significant, period lengthening by~1 h (Fig. 6a, d) whilst expression in VPAC2 Cre cells was ineffective (Fig. 6b, d). In contrast, expression in both populations progressively lengthened period, stabilising at~24.5 h (Fig. 6c, (Fig. 7a). Activation of both groups, however, rapidly (~4 days) initiated rhythmicity, and with a long period comparable to rhythms initiated by pan-neuronal (pSyn-Cremediated) expression of Cry1 (Fig. 7a, b). The oscillation was of high quality, assessed by goodness-of-fit (GOF), again approaching that of SCN subject to pan-neuronal expression of Cry1 27 (Fig. 7c). Importantly, the oscillation is not simply due to mass action by recruitment of a larger population of cells: transduction of comparable numbers of VPAC2 cells alone did not improve rhythm quality (Fig. 7d). Rather, it was due to the specific recruitment of both cellular constituents of the VIP neuropeptidergic axis. Initiation of ensemble rhythmicity demonstrated the clear pacemaking property of the VIP/VPAC2 cellular axis. We then used CCD recordings to test whether it also directed further network features, including synchrony and spatio-temporal dynamics. CoL measurements revealed the absence of a spatiotemporal wave in Cry1-transduced, single-Cre SCN (Fig. 7e). In contrast, joint activation of cell-autonomous TTFLs in VIP Cre and VPAC2 Cre cells initiated strong Per2::Luciferase bioluminescent waves, approaching wild-type trajectories and accompanied by tight phase coupling of oscillators across the network (Fig. 7e,  f, Supplementary Fig. 10). Furthermore, phase maps of Per2:: Luciferase bioluminescence in Cry1-initiated slices recapitulated many of the spatiotemporal dynamics of the wild-type slice, including region-specific phasing (Fig. 1), with the induced activity starting dorsomedially and progressing to the core before moving dorsally, medially and laterally into the phase-delayed shell (Fig. 7g, Supplementary Fig. 10). Targetted optogenetic activation of VIP cells sets ensemble phase, whereas activation of VPAC2 cells does not. a Representative merged fluorescence/brightfield micrographs of VIP ChR2::EYFP (left) and VPAC2 ChR2::EYFP (right). b Representative 470 nm (blue) or 625 nm (red) stimulated normalised Per2::Luciferase trace from VIP ChR2::EYFP SCN. Inset: phase-delay third cycle post-stimulation. c Phase response curve (PRC) for VIP ChR2::EYFP SCN stimulated at 470 nm (blue) or 625 nm (red/pink) (*p = 0.032, ****p < 0.0001). d, e As in (b, c), but for VPAC2 ChR2::EYFP SCN. f Combined PRCs for 470 nm-stimulated VIP ChR2::EYFP (blue) and VPAC2 ChR2::EYFP (pink) SCN (***p = 0.0002). g, h Normalised bioluminescence traces for Cry-null SCN expressing (g) pan-neuronal ChR2::EYFP or (h) VIP ChR2::EYFP stimulated at 470 nm (blue) or 625 nm (red) alongside paired summary data (right) (**p = 0.008, ***p = 0.0009). i Micrographs showing VIP ChR2::EYFP Cry-null SCN (Per2::Luciferase, ChR2::EYFP and merge) (left) alongside Per2::Luciferase timeseries (upper) and ChR2::EYFP (lower) before and after optogenetic stimulation. j Example spatial bioluminescence change (left) and VIP Cre -dependent EYFP intensity (right) in VIP ChR2::EYFP SCN slice. k Averaged bioluminescence traces within (EYFP intensity > 1, orange) or outside (EYFP intensity < 1, blue) EYFP regions (left) in VIP ChR2::EYFP SCN, with paired summaries (right) measured over interval indicated by horizontal line. l Cumulative distributions of bioluminescent changes within (orange) or outside (grey) EYFP regions in VIP ChR2::EYFP SCN. Scale bars = 250 µm. Arrowheads: stimulation. PRCs: small points represent individual slices; large points represent mean ± SEM. Line plots: lines and shading represent mean ± SEM. Histograms: paired points represent individual slices; bars represent mean ± SEM. PRCs: n ≥ 5 for each phase, explicit n values are given in Supplementary Table 1; Cry-null recordings: PMT: neuronal ChR2::EYFP n = 7; VIP ChR2::EYFP n = 6 SCN; Imaging: n = 5 SCN. Statistics: c, b, f two-way ANOVA; g, h, k, paired two-tailed t-test; l repeated-measures two-way ANOVA. Only significant comparisons (p < 0.05) are shown. Source data are provided as a source data file. representative raw Per2::Luciferase PMT traces of Cry1-null SCN before (black) and during (green) initiation of conditional Cry1-expression, and following AAV washout (WO) (coloured by genotype: VIP Cre (blue), VPAC2 Cre (pink) and VIP Cre ;VPAC2 Cre (purple)). Lower: corresponding peak-to-peak period plots (mean ± SEM) for VIP Cre , VPAC2 Cre and VIP Cre ;VPAC2 Cre SCN. Arrow indicates addition of conditional Cry1 AAV. For peak-to-peak period plots, joined grey points represent individual slices; coloured points represent mean ± SEM with colours denoting treatment interval in traces as above. d Summary period data for: pre-AAV-Cry1 baseline (black) and post-AAV-Cry1 washout for VIP Cre (blue), VPAC2 Cre (pink) and VIP Cre ;VPAC2 Cre (purple) SCN (**p = 0.007, ****p < 0.0001). Histograms: paired points represent individual slices; bars represent mean ± SEM (n = 7, 9 and 10 for VIP Cre , VPAC2 Cre and VIP Cre ;VPAC2 Cre ). Statistics: d repeated-measures two-way ANOVA with Sidak's correction for multiple comparisons. Only significant comparisons (p < 0.05) are shown. Source data are provided as a source data file. Joint recruitment of the TTFLs of the VIP and VPAC2 cells alone set the pace and initiated the ex vivo SCN ensemble rhythm. We therefore asked if these populations are sufficient to transmit rhythmic behaviour to the animal. Male and female Crynull mice, expressing Cre in VIP and VPAC2 cells, were arrhythmic under constant darkness (DD), and no period could be assigned (Fig. 8a). They then received bilateral stereotaxic injections of the pCry1-DIO-Cry1::EGFP AAV, and post hoc histology showed that of the eight mice receiving stereotaxic injections, six were successfully targetted, with Cry1::EGFP expression that spanned across the ventral core and dorsal shell of the SCN (Fig. 8b). The remaining two mice exhibited limited Cry1::EGFP expression in the very ventral SCN alone (Supplementary Fig. 11a). Following administration of the Cry1expressing AAV and return to DD, the targetted mice, with one exception, initiated a significant circadian behavioural rhythm, with a long period typical of a Cry1-dependent SCN (27.0 ± 0.1 h, n = 5). This was associated with a significant increase in the periodogram amplitude and non-parametrically determined relative amplitude of behavioural rhythms (Fig. 8c, d). Circadian behaviour was not initiated in the two mis-targetted mice. Of the five of six targetted mice in which circadian behaviour was initiated, their long period was directly comparable to that of Crynull SCN slices transduced with AAV-Cry1::EGFP (Figs. 7b, 8e). In addition, when the behaviour of these mice was plotted as a normalised activity profile and aligned to activity onset across animals, there was a clear division of activity between subjective day and subjective night (% total activity: Day 35.9 ± 0.04 vs Night 64.1 ± 0.04) (Fig. 8f) compared to pre-surgery where arrhythmic animals lacked this definitive structure (% total activity: Day 53.6 ± 0.03 vs Night 46.4 ± 0.03) ( Supplementary  Fig. 11b). Finally, the breeding cohorts generated two Cry1-null, Cry2-proficient mice which had characteristic short period behavioural rhythms. Following successful targetting of their SCN by AAV-Cry1::EGFP ( Supplementary Fig. 12), their freerunning period lengthened in a manner comparable to that seen in ex vivo Cry1-null SCN slices. These experiments therefore reveal that not only are the VIP and VPAC2 cells, together, driving components of the SCN network ex vivo, but also they are sufficient, in the absence of any other cellular clocks, to direct the circadian behaviour of the animal.
This study has provided a comprehensive characterisation of the role of cells within the VIP/VPAC2 neuropeptidergic axis in determining the emergent properties of the SCN circuit. VIP and VPAC2 cells are neurochemically and electrophysiologically distinct with differentially phased population rhythms that contribute to SCN spatio-temporal dynamics (Figs. 1-4, 9). Cell-specific ablation or activation of VIP or VPAC2 revealed that signals from the VIP core are much more important in maintaining network function and ensemble phase than are signals derived from VPAC2 cells (Figs. [5][6]9). Finally, although VIP cells exert weak pacemaking control, full period control and de novo initiation of stable high-amplitude rhythms require both compartments of the VIP-ergic cellular axis: showing that the TTFL of VPAC2 cells is necessary for circuit function (Figs. 7-9). Finally, these effects seen ex vivo in the slice can be recapitulated in animals to initiate circadian behaviour: when they contain the only cell-autonomous clocks in the animal, VIP and VPAC2 cells a * Pre-surgery Post-surgery 16 36 Period, hours 16 36 Period, hours . Scale bar is 100 µm. c Summary behavioural amplitude for VIP;VPAC2 Cre Cry-null mice pre-and post-surgery (n = 6 mice) (*p = 0.011). d Summary behavioural relative amplitude for VIP;VPAC2 Cre Cry-null mice pre-and post-surgery (n = 6 mice) (*p = 0.034). e Summary free-running periods for VIP;VPAC2 Cre Cry-null post-surgery plotted alongside the periods of initiated VIP;VPAC2 Cre Cry-null ex vivo SCN (replotted from Fig. 7b) (n = 5 mice, 12 SCN). f Average normalised activity profiles for the last 7 days of recording post-surgery shown as mean + SEM. The activity of each mouse is aligned at activity onset (indicated by the black arrow), and the mean activity from the pre-surgery condition (subjected to the same relative alignment, shown in Supplementary Fig. 11b) is plotted as a horizontal line (pink) (n = 5 mice). In all histograms individual points (paired/unpaired) represent individual mice/SCN slices; bars represent mean ± SEM. Statistics: c, d paired one-tailed t-test, e unpaired two-tailed t-test. Only significant comparisons (p < 0.05) are shown. Source data are provided as a source data file.
together can direct circadian behaviour (Fig. 8). Thus, emergent, ensemble properties of the SCN network are conferred by circadian signalling via the VIP/VPAC2 axis, which is a cellular pacemaking hub within the SCN circadian circuit.

Discussion
Recent single-cell SCN transcriptomic analysis has identified putative topological elements involving peptides (AVP, Prok2, PACAP, and VIP) and their cognate receptors 18 that mediate paracrine and/or autocrine signalling 29,30 . Through direct experimental manipulation, we now show that one of these molecularly defined elements, the VIP/VPAC2 cellular axis, consists of serially active topological components of the SCN circuit and determines circuit-level emergent properties of the network. VIP and VPAC2 cells represent diverse neuronal populations 31,32 and they have distinct electrical signatures that recapitulate those reported for VIP and generic non-VIP cells 13,33 , and the difference between VIP and VPAC2 neurons in their intrinsic cellular property of firing regularity is circadian invariant 13 . In both VIP and VPAC2 cells, however, RMP and SFR are rhythmic, peaking during the middle of circadian day, but VIP cells are more hyperpolarised with a lower input resistance and longer AHP than VPAC2 cells. Thus, these two cell populations, when recorded within an intact circuit, have distinctive intrinsic electrophysiological signatures. Furthermore, although the VIP and VPAC2 cells are rhythmically active 13,33 , they occupy different phases within the SCN cycle, as inferred from co-registration with phase maps of Per2::Luciferase and demonstrated directly by electrophysiological recordings. VIP cells are phase advanced compared to VPAC2 cells, an effect also seen for RMP, [Ca 2+ ] i , and the TTFL cycle (pCry1-Luciferase report). Thus, VIP and VPAC2 cells are serial components within a more complete ensemble phase wave. Moreover, within each cell population there are local waves, most evident as a dorsoventral trajectory in tightly synchronised VIP cells, whereas VPAC2 cells are less synchronised with a weaker local wave that possibly reflects heterogeneity within the VPAC2 population (e.g. AVP and non-AVP cells). Further analysis of cell-type specific circadian activity, focusing on neuropeptidergically defined topological elements 18 could reveal how the complete ensemble phase wave is generated by serial local activations. This would then inform how circadian information is generated within, and propagated across, the multi-layered SCN cellular network 2 .
Previous reports have established that calcium rhythms across the SCN network are phase dispersed 22,23,34 , but inconsistent with our findings, voltage across the network has been reported as simultaneous 22 . These results were based on ArcLight recordings where VIP-cells, AVP-cells or coarse ventral/dorsal ROIs were used as proxies for core and shell SCN, respectively 22 . Here, we utilised two functionally relevant components (VIP and VPAC2) to delineate core and shell to reveal differential phasing between these two compartments, an observation that is consistent with reported phase staggering of the electrical activities of individual neurons across the circadian cycle 12 . The difference we observed is of a smaller magnitude than that reported by calcium, but follows the same pattern: VIP cells are phase-advanced relative to VPAC2 cells 22,34 . This difference in phasing between cytosolic and membrane rhythms may generate differential calcium and voltage relationships that reflect population-specific inputs and outputs.
The importance of the VIP/VPAC2 serial connection is evident when the link is broken in mice lacking genes encoding either VIP or VPAC2, with disrupted network oscillation and circadian behaviour 35,36 . VIP and VPAC2 cells are not, however, equal contributors, because ablation of VIP cells reduced amplitude and disrupted ensemble synchrony, whereas loss of VPAC2 cells only affected amplitude. This suggests a model whereby VIP cells synchronise the rest of the SCN, including VPAC2 cells, while non-VPAC2 cells (including VIP cells) can synchronise in the absence of VPAC2-expressing cells. The mechanism mediating VPAC2-independent synchronisation awaits clarification, although it may involve Prok2, GRP and GABA signalling 11,18,20 . In addition, astrocytes, which are active when SCN neurons are quiescent 19 , have recently been shown to play a central role in driving the SCN circuit 26,37 and may well contribute to the spatio-temporal wave. Ultimately, understanding SCN circuit topology will inform analysis of the origin of the wave, while more detailed knowledge of wave structure will highlight underlying circuit topology.
VIP cells mediate retinal entrainment and thereby ensemble phase. Correspondingly, their optogenetic activation, alone or as part of the DRD1A population, can reset the SCN slice in culture 13 (and this study) and the mouse in vivo 38 , while their inhibition blocks light-induced resetting in vivo 39 . These effects must involve trans-synaptic activation of target cells, which was evidenced directly as Per2::Luciferase induction in SCN regions outside the VIP ChR2::EYFP core. Moreover, optogenetic resetting of the SCN slice is attenuated by VPAC2 antagonism 13 , and so resetting may act through the same VIP-to-VPAC2 pathway that directs the spatio-temporal wave. It is striking, however, that optogenetically increasing the firing rate of VPAC2 cells did not,  of itself, control ensemble phase, even though exogenous VIP stimulates SCN firing 40 . The phase-setting effect of VIP cells must therefore be via additional effects on VPAC2 cells, for example induction of CRE-dependent transcription, which in turn regulates their cell-autonomous TTFL 15 . Ultimately, changes in phase of VPAC2 cells, some of which express AVP and/or Prok2, will be transmitted to the rest of the SCN network and distal targets, via paracrine and/or autocrine signalling 18,29 . Disruption of network function allows cell-autonomous oscillations to continue but their periods rapidly diverge 6,8 , suggesting that ensemble period is determined by circuit-wide computations 2 . Identification of which cells set ensemble pace has relied on intersectional approaches to create temporally chimeric SCN whereby the cell-autonomous period of a defined subset of SCN neurons has been genetically altered. Two partially overlapping cell populations expressing NMS 16 or DRD1A 17 have been shown to control SCN circadian period as a function of their own TTFL period. Both populations are, however, heterogeneous containing AVP, VIP and other cells. Selective targetting of VIP cells alone had no effect on behavioural circadian period 16 , while AVP cells can impose their period on behaviour but not the isolated SCN slice 41,42 . Here we show, using Cry1-complementation, that VIPto-VPAC2-mediated synchronisation draws all SCN cells to a common period, and VIP and VPAC2 cells together were able to impose their cell-autonomous period on the SCN almost as effectively as when all cells were complemented with Cry1 27 . This was also evident in vivo, further demonstrating the power of the VIP/VPAC2 cellular axis. The effectiveness of DRD1A and NMS cells in setting ensemble period may therefore be because they contain the VIP and VPAC2 populations. Such re-assignment of the VIP/VPAC2 cellular axis as the effective pace-setter is consistent with the disruption and/or short-period oscillations seen in VIP-null mice 36 and SCN 43 . In contrast, deletion of NMS is without effect on circadian timekeeping 16 : NMS signalling per se, therefore, is unlikely to contribute significantly to circuit computations mediated by VIP and VPAC2 cells.
Finally, the definitive emergent property is de novo oscillation. In contrast to commonly adopted loss-of-function approaches which test the necessity of cell-autonomous clocks in cell groups 16,41 , we used Cry1-complementation of Cry1,2-null SCN to test for their sufficiency. Pan-neuronal expression of Cry1 initiated de novo circadian oscillations in Cry1,2-null SCN 26,28 (and current study), but restricted expression in either VIP cells or VPAC2 was ineffective. Notably, joint expression of Cry1 in VIP and VPAC2 cells rapidly initiated molecular cycles with appropriate ensemble period, cellular synchronisation and phase divergence. This circuit-wide effect was driven by Cry1 expression in ca. 50% of SCN cells. It is therefore possible that neither VIP cells nor VPAC2 cells alone were sufficiently abundant to exert control. Against this quantitative interpretation is the observation that with untargetted Cry1 expression, transduction of between 10 and 35% of neurons is sufficient to initiate robust behavioural rhythms in Cry-null mice 26,28 . It appears, therefore, that the intrinsic properties rather than the abundance of the VIP and VPAC2 cells are critical to circuit-level organisation of the SCN. Moreover, this organisation was readily transmitted to the control of circadian behaviour.
The current study has shown that neurons within the neurochemically explicit and topologically defined VIP-to-VPAC2 axis are able to define the emergent network-level properties essential to the role of the SCN as the circadian pacemaker, in vitro and in vivo. This network is not, however, restricted to neurons: SCN astrocytes also have a cell-autonomous TTFL, are able to set circadian period 19,37 and maintain SCN network function (albeit less potently than neurons) 26 . The challenge, now, is to understand how these individual elements, neuronal and astrocytic, assemble into a coherent, multi-layered network 2 . The current study provides a platform onto which we can assign circadian properties to functional SCN units with the ambition to generate a complete topological network map. Linking network topology, function and emergent properties in the spatiotemporal domain in this way will have important implications not only for our understanding of the circadian clock, but also across other circuits of the brain 44 . The following AAVs backbones were produced in-house and packaged by Penn Vector Core as AAV1 serotype: Cre-conditional pCry1-luciferease (pCry1.DIO.luc) and Cre-conditional Cry1::EGFP (pCry1.DIO.Cry1::EGFP).
The resulting PCR fragments were purified and incubated with the linearised vector backbone and Gibson mix (NEB) for 30 min at 50°C. The reaction was chilled on ice before adding 5 µl of the mix to chemically competent cells, kindly donated by Dr. Ernesto Ciabatti (MRC Laboratory of Molecular Biology, Cambridge, UK). This was left on ice for 2 min and then plated onto LB plates containing ampicillin. Colonies were selected, mini-prepped and confirmed by sequencing. Following validation in cells, the Ef1a.DiO.mCherry-P2A-DtR plasmid was packaged into AAV1 serotype vectors by Penn Vector Core. NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-17110-x ARTICLE NATURE COMMUNICATIONS | (2020) 11:3394 | https://doi.org/10.1038/s41467-020-17110-x | www.nature.com/naturecommunications SCN explant culture, bioluminescence and fluorescence imaging. Mice (P8-10) were sacrificed according to local and Home Office rules, and the suprachiasmatic nucleus (SCN) was removed and cultured as an explant. Briefly, coronal hypothalamic slices were cut at 300 µm and the SCN was dissected free using a razor blade in ice-cold GBSS supplemented with (in mM): 5 mg/ml glucose, 50 µM D-AP5, 100 nM MK-801 and 3 mM MgCl 2 . Slices were maintained in the interface method for 2-3 h in media containing: 50% Eagle's Basal Medium (Gibco), 25% EBSS (Gibco), and 25% Horse Serum supplemented with 5 mg/ml Glucose, 2 mM GlutaMAX (Gibco), 1:100 dilution of Penicillin/Streptomycin (Gibco), 50 µM D-AP5, 100 nM MK-801, and 3 mM MgCl 2 . Following 2-3 h in culture, slices were incubated in the same media without the addition of D-AP5, MK-801 and MgCl 2 for a week. After a week, culture medium was changed, and 1-µl AAVs (between 1 × 10 12 and 1 × 10 13 GC/ml in PBS) were added drop-wise to the surface of the slice 24 h later. Transduced slices were left for one week before AAVs were washed out by fresh culture medium and, in most cases, successful transduction was assessed by imaging.
For bioluminescent photomultiplier tube (PMT) recordings, slices were transferred to DMEM-based (Sigma-Aldrich) recording medium supplemented with: 4.17 mM NaHCO 3 , 5 mg/ml glucose, 1:100 dilution of Penicillin/ Streptomycin (Gibco), 10 mM HEPES, 5% FCS, 2 mM GlutaMAX, and 100 µM luciferin in 35-mm dishes. The dishes were then sealed with glass coverslips and vacuum grease before being transferred to a custom built PMT (H9319-11 photon counting head, Hamamatsu) array within a light-tight incubator at 37°C. Bioluminescent emissions were collected in real time and binned into 6-min intervals before analysis. For bioluminescent imaging via CCD camera, slices were sealed into 35-mm dishes and transferred to the heated stage of an inverted microscope and CCD camera (Hamamatsu) setup. Bioluminescent time-lapse images were taken over 1-h intervals.
For combined bioluminescent and fluorescent imaging, slices were sealed into 35-mm dishes with glass bottoms (Mattek) and transferred to the heated stage of an LV200 microscope system (Olympus) running Olympus proprietary acquisition software (CellM, xcellence rt or cellSens) and equipped with an EM-CCD camera (Hamamatsu). Bioluminescence (PER2::Luciferase and pCry1-luc) and fluorescence (EYFP, GCaMP6f and ArcLight) images were taken once every 30 min, and recorded for at least 5 cycles. Exposure times ranged between 9.5 and 29.5 min for bioluminescence and 25 and 100 ms for fluorescent reporters (EYFP: 25-100 ms; GCaMP/ArcLight: 100 ms) dependent on the configuration of the experiment.
Analysis of circadian imaging data. Images were analysed in FIJI 46 , IgorPro (Wavemetrics) using the SARFIA plugin, Excel (Microsoft) and Graphpad Prism (Graphpad). To generate phase maps and maps of bioluminescence/fluorescence intensity, slices were thresholded to remove extra-SCN signals before a custom FIJI plugin was used to extract signals from a continuous grid of ROIs placed over the SCN. In the case of phase maps, signals were detrended and smoothed by a 2.5 h moving average before peaks were identified and the peak-to-peak period of the PER2::LUC signal was used to calculate and normalise the phase of each cycle to allow the cycle to cycle and mean phase maps to be plotted for PER2 and GCaMP signals. For changes in normalised bioluminescence intensity, the change within each ROI was normalised to the baseline bioluminescence intensity within the same ROI. In order to define regions delineated by fluorescent markers, the average intensity projection was overlaid with the same grid and the intensity of the signal was normalised to the mean intensity across all ROIs. Areas with a normalised intensity of <1 were defined as outside the region, while areas with a normalised intensity of >1 were defined as within the region. Continuous phase maps were assembled and coloured in Graphpad Prism using the heat-map plot and were coloured using the "rainbow" colour setting. In the case of cluster phase-maps, phases were classified by manually assigning colours to any ROIs falling within the following phase ranges: <CT10, CT10-11, CT11-12, CT12-13, CT13-14 and >CT14.
CoL was measured using IgorPro scripts utilising the SARFIA toolset. CoL trajectory area was calculated from the perimeter of the CoL excursion and expressed as a percentage of the area of the temporally integrated bioluminescence signal. Circadian bioluminescence and fluorescence signals were analysed using a manual peak identification and peak-to-peak period approach in the case of phasemapping electrophysiology experiments (Fig. 2), fluorescence/bioluminescence experiments (Fig. 3) and optogenetics experiments (Fig. 6). Differences in phasing comparing pan-neuronal or targeted ArcLight and GCaMP6f were assessed statistically by ordinary one-way ANOVA with Holm-Sidak's correction for multiple comparisons. Potential differences in aggregate period between Per2:: Luciferase and pan-neuronal or targeted (VIP Cre and VPAC2 Cre ) ArcLight/ GCaMP6f were assessed statistically by paired t-tests. Differences in phasing, Rayleigh vector length, CoL trajectory, baseline and amplitude for pCry1-luciferase rhythms were assessed by unpaired t-tests. Differences in period for pCry1luciferase rhythms between pan-neuronal (synapsin-Cre) and targeted (VIP Cre and VPAC2 Cre ) conditions were assessed by ordinary one-way ANOVA with Holm-Sidak's correction for multiple comparisons. Bioluminescence recordings from slice cultures in long-term PMT recordings were analysed to calculate circadian period length, amplitude, relative amplitude error (RAE) and goodness of fit (GOF) using the Fast Fourier Transform-Non-Linear Least Squares (FFT-NLLS) function in the BioDARE software 47 (www.biodare2.ed.ac.uk).
Sectioning/immunostaining/confocal microscopy. Mouse brains were postfixed in 4% PFA (in phosphate buffer) and cryoprotected in 20% sucrose (in PBS) prior to sectioning. Brains were sectioned on a freezing microtome (Bright Instruments, UK) at 40-µm thickness. Sections containing the SCN were blocked with 5% NGS prior to immunostaining. Primary antisera used were: Rabbit anti-AVP (Peninsula, T-4563) (used at 1:1000 dilution), rabbit anti-VIP (Immunostar, 20077) (used at 1:750 dilution) and guinea pig anti-VIP (Peninsula T-5030) (used at 1:1000 dilution). Fixed brain sections and slices were imaged using the Zeiss 780 inverted system (Zeiss, Germany) with either ×10 or ×20 objective for low power images or ×63 objective for high power images, used for cell counting. Cell counting for co-localisation was done manually using the FIJI Nucleus Counter and Cell Counter plugins. Transduction efficiency of AAVs was calculated using AAVdependent fluorescence cell counting in conjunction co-localisation. Efficiency of AAV1.pCry1.DIO.Cry1::EGFP transduction was expressed as a percentage of Cre+ cell counts, where Cre+ cells were revealed using AAV8.pSyn.DIO.mCherry transduction. AAV8.pSyn.DIO.mCherry transduction efficiency was then assessed by calculating the percentage of Cry1::EGFP-positive cells where mCherry is absent, termed "over-spill" expression.
Riboprobe generation. Due to spatial patterning of VIP-ir in the projections of VIP neurons, it can be technically difficult to perform cell counts. To confirm specificity of Cre-recombinase expression in the VIP-cells and perform cell counts, fluorescent in situ hybridisation (FISH) was performed against Cre-recombinase and VIP. Labelled riboprobes were generated via a one-step reverse-transcriptase PCR (RT-PCR) protocol using gene-specific primers to generate cDNA, followed by an in vitro transcription (IVT) to create the riboprobe. Primer sequences were sourced from the Allen Brain Atlas (VIP forward: CCTGGCATTCCTGATACT CTTC, VIP reverse: ATTCTCTGATTTCAGCTCTGCC; Cre forward: CCAATT TACTGACCGTACACCA, Cre reverse: TATTTACATTGGTCCAGCCACC) and synthesised (Sigma) containing either the bacterial promoter T7 sequence (TAA TACGACTCACTATAGGGAGA; forward primer) or T3 sequence (AATTAAC CCTCACTAAAGGGAGA; reverse primer) 5′ to the rest of the sequence to generate sense and antisense probes, respectively, via IVT. A one-step reverse-transcriptase (RT)-PCR was carried out (SuperScript III One-Step RT-PCR System with Platinum Taq kit, Invitrogen) to reverse-transcribe RNA extracted from SCN slices (RNeasy Micro, Qiagen) and amplify the resultant cDNA. 50-100 ng RNA was used as a template. Once the resultant cDNA was purified (QIAquick PCR Purification kit, Qiagen), 500 ng was in vitro transcribed using either T3 or T7 polymerase (Roche), in a reaction mix containing 1X transcription buffer (Roche), 0.01 M DTT, digoxigenin (DIG) or fluorescein (FLU) labelling mix (Roche), RNasin, and polymerase. This was left at 37°C for 2 h in a thermocycler and was subsequently cleaned up using Micro Bio-Spin 6 columns (Bio-Rad) before diluting 1:5 in nuclease-free water and storing at −20°C. RNA probes were diluted 1:100 in hybridisation buffer, heated at 60°C for 10 min and then immediately chilled on ice for 5 min. Eighty microlitres of probe solution was added to each slide and a coverslip was placed on top of the solution to ensure uniform spreading and reduce evaporation. Slides were placed in a hybridisation chamber (saturated with 50% formamide and 50% PBS) and incubated overnight in a ventilated oven at 58°C. Following hybridisation, coverslips were removed by washing in 4× saline-sodium citrate buffer (SSC) before treating with 20 µg/ml RNase A for 30 min at 37°C. The sections were subsequently washed in decreasing SSC concentrations (2×, 1×, 0.5×) before a final wash in 0.1x SSC at 60°C. Sections were then washed and equilibrated in Tris-NaCl buffer (TNB; 0.1 M Tris (pH 8), 150 mM NaCl) before blocking in 5% normal sheep serum (in TNB) for 30 min. The primary antibody (anti-DIG-POD) was diluted 1:100 in this blocking buffer and 80 µl of this antibody solution to each slide, cover-slipping and incubating at 4°C overnight. Sections were washed three times in TNB to remove the antibody solution and were then treated with TSA working solution (TSA Plus Cy3 diluted in TSA Diluent 1:50; Perkin Elmer): 80 µl was added to each slide, which were then cover-slipped and incubated in the dark at room temperature for 1 h to produce a fluorescent precipitate. Three washes in TNB followed before incubating the slides in 3% H 2 O 2 (in TNB) for 30 min. Sections were again blocked and treated with the remaining antibody (anti-FLU-POD) at a 1:100 dilution and incubated overnight as before. The washes and TSA treatment were also repeated, using TSA plus FLU. Three washes in TNB and a final rinse in water were carried out before immediately cover-slipping with Vectashield Hardset mounting medium with DAPI.
Electrophysiology. For electrophysiological recordings, SCN slices were maintained in PMTs and their bioluminescent output was monitored to determine phase and extrapolate time-stamped recordings in circadian time. Slices were transferred to the stage of an upright (BX51WI, Olympus) microscope by excision of the supporting membrane and maintained in ACSF (in mM: 125 NaCl, 25 NaHCO 3 , 1.25 NaH 2 PO 4 , 3 KCl, 25 Glucose, 2 CaCl 2 , 1 MgCl 2 ) bubbled with 95% CO 2 /5% O 2 , constantly perfused at~1.5 ml/min at 32°C via an inline heater. For cell-attached recordings, borosilicate glass pipettes (3-6MOhms) were filled with 150 mM NaCl and recordings were made in voltage clamp mode using a Cornerstone BVC-700A amplifier (Dagan). Signals were digitised at 10 kHz and filtered at 1 kHz. For electrophysiological optogenetic characterisation, cells recorded in cell-attached voltage clamp mode were identified by their ChR2::EYFP tag. Each cell recorded was stimulated with a train of 60 equally spaced pulses (15 ms) delivered at 1, 5, 10, 15 and 20 Hz using a computer-controlled TTL-triggered LED array (pE-4000, 470 nm LED, CoolLED) via the widefield standard EGFP fluorescence cube. Firing index was calculated by dividing the number of evoked spikes by 60 (the number of light pulses delivered), with values of <1 representing failure of the stimulation pattern to induce firing at the desired rate.
For whole-cell recordings, borosilicate glass pipettes (4-8 MOhms) were filled with a K-gluconate based internal solution (in mM: 135 K-gluconate, 7 NaCl, 10 HEPES, 2 Na 2 -ATP, 0.3 Na 2 -GTP, 0.01 Biocytin-Alexa488, 2 MgCl 2 ) and recordings were made using a Cornerstone BVC-700A amplifier (Dagan). Targeted recordings were made via visualisation of the tdTomato signal excited using widefield illumination via LED illumination (pE-4000, 550 nm LED, CoolLED) via a standard Texas Red fluorescence filter cube. Signals were digitised at 20 kHz and filtered at 3 kHz. Series resistance for whole-cell recordings was monitored at break in and at the end of the recording, and if it changed by >25%, recordings were omitted. Input resistance was measured by applying a series of 500 ms long hyperpolarising current steps (−30pA to 0pA, 10pA steps) to neurons and measuring the steady-state voltage deflection. Input resistance was defined as the gradient of the linear relationship between the voltage deflection and the injected current. All recordings were completed within 3-4 min of break-in to minimise effects of intracellular perfusion, but cells were allowed to perfuse for up to 10 min to confirm successful targeting by registration of the green Alexa-488 signal with the red tdTomato signal and images were acquired using micromanager (ImageJ). For all whole-cell recordings, correction of the junction potential was applied a posteriori.
Analysis was performed in AxoGraphX (AxoGraph), Igor Pro (Wavemetrics) using the Neuromatic plugin 48 , Excel (Microsoft), Graphpad Prism (Graphpad) and R (version 3.6.1, courtesy of R Foundation for Statistical Computing, Vienna, Austria. https://www.R-project.org) using RStudio (version 1.2.1335, courtesy RStudio Team, Boston, MA. http://www.rstudio.com). The regularity of firing was assessed quantitatively using the LvR metric (reproduced from ref. 25 ): where I i and I i+1 represent consecutive interspike intervals (ISIs), n is the number of ISIs and R represents the refractoriness constant (defined as 5 ms 25 ) to correct for the refractory period within an ISI directly following a spike and define regularity as a variable independent of firing rate. For pooled and time-aligned electrophysiological data, a linear mixed model was fitted to the data to take into account differences in the numbers for neurons recorded between slices. For pooled electrophysiological data, to examine the underlying differences between the two genotypes (VIP Cre or VPAC2 Cre ), the linear mixed model was fit using genotype as a fixed factor with slice as a random factor. In the time-aligned datasets, the linear mixed model was fit using genotype, time and an interaction between the two as fixed factors with slice as a random factor. The models were fit using the nlme package in R (nlme version 3.1.143) (courtesy R Core Team nlme: https://CRAN.R-project.org/package=nlme). In the case of the pooled electrophysiological data, the significance of the genotype effect was assessed by applying an unpaired, two-tailed Welch's t-test. For the timealigned data, the significance of the fixed factors in the model were determined through application of a two-way ANOVA. The two-way ANOVA was followed by pair-wise comparisons made on the estimated marginal means derived from the model and Tukey's correction was applied to these multiple comparisons made using the emmeans package in R (emmeans version 1. Optogenetics. SCN slices were recorded in specially adapted PMTs with fibre optic cables running into the base of the chamber to allow for optogenetic stimulation (1 h, 10 Hz) in situ via computer-controlled LEDs (in-house hardware and software, Electronics Workshop, MRC Laboratory of Molecular Biology). Stimulation strength of both LED wavelengths (470 nm: M470F, Thor Labs/625 nm: M625F, Thor Labs) was measured to be at least 1 mW/mm 2 . Slices were recorded for 4 cycles before stimulation and this was used to extrapolate the phase of stimulation and to predict the peaks post-stimulation. Slices were recorded for 4 cycles post-stimulation and phase shifts were calculated as the difference between the actual peak and the predicted peak normalised in circadian time to the baseline period and expressed as the mean of the three cycles following stimulation. For the Cry-null optogenetic experiments, stimulation was given after recording a 24 h stable PER2::Luciferase baseline. In the case of direct imaging, slices were checked for a stable baseline in PMTs before being transferred to an LV200 bioluminescence/fluorescence imaging system (Olympus). Bioluminescence and EYFP signals were captured every 30 min before optogenetic stimulation (1 h, 10 Hz) was delivered to the slice via the GFP cube. Slices were then recorded for a further 24 h. Significance was assessed in the PRCs via two-way ANOVA with Sidak's correction for multiple comparisons. Bioluminescent changes in Cry-null SCN receiving panneuronal or VIP-targeted optogenetic stimulation were analysed via paired t-tests or repeated-measures two-way ANOVAs with Sidak's correction for multiple comparisons.
Toxin ablation. Following baseline recording of SCN slices for >5 days, Dtx or vehicle (recording media) was added to the bulk media. Dtx (Sigma) was bathapplied to SCN slices at a final concentration of 20 ng/ml. Following 5-7 days of treatment, Dtx or vehicle was washed off via media change and slices were recorded further for >5 days. Recordings were analysed in BioDare, values were calculated based on rhythms 4 days pre-treatment and 4 days post-treatment excepting the first 36 h immediately after treatment application. Data were analysed via either a two-way ANOVA or three-way ANOVA, followed by Sidak's correction for multiple comparisons.
AAV-mediated Cry1 complementation. Cry1-null SCN slices were recorded in PMTs for 7 days to confirm arrhythmicity. They were then transduced with 1 μL of AAV.pCry1.DIO.Cry1EGFP and returned to the PMTs for a further 7 days, followed by a media change and return to PMTs for a final 7 days. Pre-and postanalysis of Per2::Luciferase rhythms used the first and last 7 days respectively. In order to follow transient period changes from cycle to cycle as Cry1::EGFP expression increased following transduction, peak-to-peak period was calculated using the area under a curve feature in Graphpad Prism to automatically identify the time of PER2::LUC peaks. Time intervals between these automatically identified peaks were expressed as the period for that cycle. Period comparisons were made via a repeated-measures two-way ANOVAs with Sidak's correction for multiple comparisons. Cry1,2-null SCN slices were recorded from in PMTs for 7 days to confirm arrhythmicity. They were then transduced with 1 μL of AAV.pCry1.DIO. Cry1::EGFP and returned to the PMTs for a further 7 days, followed by a media change and return to PMTs for a final 7 days. Pre-and post-analysis of Per2:: Luciferase rhythms used the first and last 7 days respectively. After PMT recordings, slices were imaged on CCD camera. Finally, the slices were transduced with AAV8.Syn1.DIO.mCherry to confirm Cre-recombinase expression for one week before being fixed (4% PFA in phosphate buffer) and mounted using Vectashield with DAPI (Vectorlabs, U.S.A) on slides ready for confocal imaging. Cell counting was performed in FIJI using the "nucleus counter" plugin. %EGFP cells was calculated relative to DAPI cell counts. Data were analysed using BioDARE 49 . Comparisons of period between pan-neuronal and VIP Cre /VPAC2 Cre initiations were made via unpaired two-tailed t-tests. GOF measures were compared via a repeated-measures two-way ANOVA with Sidak's correction for multiple comparisons. Trajectory area and Rayleigh vector data was compared by ordinary oneway ANOVA with Kruskal-Wallis correction for multiple comparisons.
In vivo stereotaxic injection of AAVs. The sex of SCN slices was not known. Therefore, both male and female adult (aged between 10 to 14 weeks) mice expressing both VIP Cre and VPAC2 Cre on a Cry1,2-null background from two separate cohorts (n = 4 male, 4 female; N = 8) were individually housed in cages equipped with running wheels with food and water available ad libitum. As part of the breeding strategy, Cry1-null, Cry2-heterozygous animals were also produced and subjected to the same stereotaxic surgery to determine the in vivo pacemaking efficiency the VIP and VPAC2 cells (n = 3 male, 2 female; N = 5). Mice were singly caged in cages equipped with running wheels with environmental conditions complying with NC3Rs guidelines of ambient temperatures of between 20-24°C and relative humidity between 45 and 65%. Mice were entrained to a 12-h light:12h dark schedule (lights on at 07:00) for ca. 10 days, followed by 10 days in continuous darkness (DD) to assess their endogenous free-running period. The mice then received bilateral stereotaxic injections of AAV1 pCry1-DIO-Cry1::EGFP (0.3 µl per site) into the region of the SCN (±0.25 mm medio-lateral to Bregma, 5.5 mm deep to dural surface) under halothane anaesthesia. Following surgery, the mice were maintained on a 12-h light:12-h dark schedule for several days for recovery before transferring back to DD for re-assessment of their endogenous behavioural period. To confirm that administration of AAV particles to the SCN led to expression in the VIP and VPAC2 cell populations within the SCN, mice were culled and brain harvested, fixed in 4% PFA, cryo-preserved overnight in 20% sucrose and sectioned (40 µm) on a sledge freezing-microtome. Confocal microscopy of native EGFP signal was used to identify successful targeting of the SCN. Due to inefficient targeting, two animals were removed from further analysis in the Cry1,2-null cohort, and three animals were removed from the Cry1-null, Cry2heterozygous cohort.
Behavioural data were analysed using ClockLab (Actimetrics, Inc., USA) running within Matlab (Mathworks, USA). Activity profiles were generated by outputting the last 7-days of recordings from ClockLab binned at 12-min intervals across a 27 h day (the average post-surgery period) for both pre-and post surgery. Profiles were then averaged to generate a single 27 h profile for each animal which was then normalised to the highest peak in the activity trace. Traces were then aligned so that their activity onsets were centered at the mid-point of the 27 h day, and averaged across all animals to generate an average activity profile.
For Cry1,2-null cohorts, paired comparisons were made between pre-and postsurgery conditions using a paired one-tailed Student's t-test. A one-tailed test was deemed appropriate due to a priori assumptions that can be made from the ex vivo SCN data. Comparisons were made between slice and behavioural data using a two-tailed unpaired t-test. For Cry1-null,Cry2-heterozygous cohort, comparisons were not made against slice data because the final number of animals (n = 2) was deemed too low for the statistical analysis to be reliable.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The source data underlying Figs. 1d, f, 2b-g, j, k, 3e, g, i, 4g, 5c, e-h, k, l, 6a-d, 7b-f, and 8c-f and the source data underlying Supplementary Figs. 4c, d, 5b, d, e, h, I, 6c-e, 7d-j, 8a, 9b-d, 11b and 12b are provided as a Source Data file. Other data and biological materials are available from the corresponding author on reasonable request.