A GABAergic network from AVP- to VIP-neurons in the suprachiasmatic nucleus sets the timing of circadian behavior rhythms

AVP neuron-specific Vgat-deficient mice (Avp-Vgat^−/−) showed a lengthened activity time (the interval between the locomotor activity onset and offset) of behavior rhythm with little change in the intracellular Ca²⁺ rhythm of AVP neurons, resulting in a misalignment of locomotor activity and AVP neuronal Ca²⁺ rhythm [25]. To elucidate how GABA released from AVP neurons regulates the activity of VIP neurons in the SCN, we first examined the temporal relationship between the circadian behavioral rhythm and the intracellular Ca²⁺ in VIP neurons (VIP-Ca²⁺) of Avp-Vgat^−/− mice in vivo using fiber photometry.

To specifically target the jGCaMP7s expression in VIP neurons of Avp-Vgat^−/− mice, we introduced a Vip-tTA allele [26] into Avp-Cre; Vgat^flox/flox (Avp-Vgat^−/−) or Avp-Cre; Vgat^wt/flox (control) mice. In Vip-tTA mice, the tetracycline transactivator (tTA) is expressed specifically in VIP neurons. We then specifically expressed the fluorescent Ca²⁺ indicator jGCaMP7s [27] in SCN VIP neurons by focally injecting a tTA-dependent AAV vector (AAV-TRE-jGCaMP7s) and implanted an optical fiber just above the SCN (Fig 1A and 1B) [12].

The VIP-Ca²⁺ rhythm was synchronized in antiphase with the locomotor activity rhythm in control (Avp-Cre; Vgat^wt/flox; Vip-tTA) mice (Figs 1C–1E, S1, and S2G) and showed high Ca²⁺ levels across the entire span of the behavioral rest period, as reported previously [12,28–30]. To quantitatively characterize the relationship between locomotor activity and VIP-Ca²⁺ rhythms, we set thresholds to define the onset and offset of the behavioral active phase and the high VIP-Ca²⁺ phase in the daily profiles of individual mice. For locomotor activity, the median daily locomotor activity level was regarded as the threshold (S1 Fig). For VIP-Ca²⁺, we examined three thresholds, i.e., 20%, 50%, and 80% of the peak-to-trough amplitude, because the shape of VIP-Ca²⁺ daily profiles was often skewed, especially in Avp-Vgat^−/− mice.

In Avp-Vgat^−/− mice (Avp-Cre; Vgat^flox/flox; Vip-tTA), the shape of the VIP-Ca²⁺ profile was altered, which correlated with the lengthening of behavioral activity time (onset-offset interval), as reported previously (Fig 1F and 1G) [25]. In LD, VIP-Ca²⁺ levels dropped earlier in the evening and began to rise earlier in the dawn (Figs 1F left and S2F left). In addition, VIP-Ca²⁺ took a longer time to increase and reach the plateau in both LD and DD (Fig 1J). Especially in DD, the circadian time (CT) at which VIP-Ca²⁺ rose to 50% or 80% of the maximum was significantly delayed, consistent with the later locomotor activity offset in Avp-Vgat^−/− mice (Figs 1E–1G right, and S2F right). In contrast, VIP-Ca²⁺ offset phases did not differ between genotypes in DD, although this might be due to CT12 being defined by the locomotor activity onset in individual mice (Figs 1F right and S2F right). Consequently, the duration of high VIP-Ca²⁺, when thresholds were set at 50% or 80% of the maximum, was significantly shortened (50%: Control, 11.87 ± 0.11 h; Avp-Vgat^−/−, 9.52 ± 0.67 h, P = 0.002; 80%: Control, 9.67 ± 0.30 h; Avp-Vgat^−/−, 5.71 ± 0.39 h, P < 0.0001), complementary to the expansion of behavioral activity time (Control, 13.66 ± 0.50 h; Avp-Vgat^−/−, 18.31 ± 0.26 h, P < 0.0001) in constant darkness (DD) (Fig 1H and 1I). Intriguingly, relative VIP-Ca²⁺ levels at locomotor activity onset and offset were similar between the two genotypes, indicating that the phase relationship between the VIP-Ca²⁺ rhythm and locomotor activity was essentially maintained in Avp-Vgat^−/− mice (Figs 1D, 1E, and S2B). These results suggest that GABA released from AVP neurons regulates the timing of high VIP-Ca²⁺. The delayed duration of high VIP-Ca²⁺ caused by VGAT deficiency in AVP neurons may allow the morning locomotor activity to occur later. In LD, light is likely to directly drive VIP neuronal activity in the absence of AVP neuronal GABA release, resulting in a less severe phenotype compared with that observed in DD.

Thus, SCN VIP neurons likely receive direct or indirect GABAergic regulation from AVP neurons, as well as GABA from other types of SCN neurons. To investigate the role of GABAergic signaling in VIP neurons in generating the behavior rhythm, we next aimed to eliminate ionotropic GABA_AR specifically in these neurons. GABA_AR is essentially a pentameric protein composed of α, β, and γ or δ subunits [31]. Each subunit has multiple subtype genes, but we had no information about which genes to delete. Considering that the β subunit is necessary for functional GABA_ARs and that only three subtype genes encode the β subunit (Gabrb1–3), we introduced indel mutations in all three simultaneously and specifically in VIP neurons by in vivo genome editing (Vip-GABA_AR^−/− mice). We first crossed Cre-dependent SpCas9-expressing (Rosa-LSL-SpCas9-2A-EGFP) mice [32] with VIP neuron-specific Cre driver (Vip-ires-Cre) mice [33]. Then we injected the mixture of three AAV vectors, each expressing gRNAs targeting one of the Gabrb genes (AAV-U6-gGabrb1, 2,3-EF1α-DIO-mCherry), into the SCN of these mice (Figs 2A, 2C, S3, and S4). Indeed, GABA_AR-mediated postsynaptic currents (GPSCs) disappeared almost completely in VIP neurons of SCN slices prepared from Vip-GABA_AR^−/− mice, confirming the effectiveness of this method (Fig 2D and 2E).

In LD, Vip-GABA_AR^−/− mice showed a daily locomotor activity rhythm comparable to controls. In DD, however, the morning locomotor activity of Vip-GABA_AR^−/− mice was significantly reduced with earlier and less obvious locomotor offset (Fig 2B, 2F, and 2G), resulting in a shortened activity time (Control, 13.64 ± 0.31 h; Vip-GABA_AR^−/−, 12.16 ± 0.15 h, P = 0.0001) (Fig 2H). In contrast, their free-running period in DD did not alter significantly (S5B Fig). These results suggest that the disinhibition of VIP neurons due to the absence of GABA_ARs may suppress the locomotor activity in the morning.

In Avp-Vgat^−/− mice, the high VIP-Ca²⁺ duration of the VIP-Ca²⁺ rhythm was shortened while the behavioral activity time was lengthened (Fig 1). Therefore, we next investigated how the VIP-Ca²⁺ rhythm alters in vivo in Vip-GABA_AR^−/− mice (Fig 3A and 3B).

The VIP-Ca²⁺ rhythm was synchronized in antiphase with the locomotor activity rhythm in both control and Vip-GABA_AR^−/− mice (Figs 3C–3E, S6, and S7H). In Vip-GABA_AR^−/− mice, the morning locomotor activity was reduced and ended earlier in DD (Figs 3G and S7F), as described above (Fig 2). Correspondingly, the shape of the VIP-Ca²⁺ profile was altered. VIP-Ca²⁺ levels began to rise earlier in both LD and DD and took a longer time to increase and reach the plateau (Figs 3F, 3J, and S7G). The speed of VIP-Ca²⁺ decline in the evening was also slower in Vip-GABA_AR^−/− than in control mice. Consequently, the duration of high VIP-Ca²⁺ was significantly extended in these mice when thresholds were set at 20% or 50% of the maximum in DD (20%: Control, 13.48 ± 0.20 h; Vip-GABA_AR^−/−, 15.05 ± 0.49 h, P = 0.0008; 50%: Control, 11.52 ± 0.19 h; Vip-GABA_AR^−/−, 12.44 ± 0.35 h, P = 0.015), consistent with the shortened behavioral activity time observed in these mice (Vip-GABA_AR^−/−, 12.09 ± 0.23 h; Control, 13.83 ± 0.23 h, P < 0.0001) (Fig 3H and 3I). Importantly, relative VIP-Ca²⁺ levels at locomotor activity onset and offset were similar between the two genotypes, except that the activity onset occurred at lower VIP-Ca²⁺ levels in Vip-GABA_AR^−/− mice in LD, suggesting that the phase relationship between the VIP-Ca²⁺ rhythm and locomotor activity was essentially maintained in Vip-GABA_AR^−/− mice (Figs 3D, 3E, and S7B). Thus, the earlier morning rise of VIP-Ca²⁺ may suppress the morning locomotor activity in Vip-GABA_AR^−/− mice, underscoring the involvement of GABAergic input to VIP neurons in regulating VIP-Ca²⁺ and behavior rhythms.

Notably, VIP-Ca²⁺ in Vip-GABA_AR^−/− mice showed greater fluctuations during the plateau phase, as indicated by an increased coefficient of variation (CV) (Figs 3F and S7C). The lack of GABA_AR might lead to an increase in basal Ca²⁺ levels in VIP neurons. However, the current methodology does not allow determination of absolute Ca²⁺ concentrations and therefore cannot directly test this possibility.

Previous studies have reported that AVP neuronal fibers make sparse contacts onto VIP neurons and that VIP neurons respond to the optogenetic activation of AVP neurons by increasing Ca²⁺ at around ZT22 in vivo [12,34]. Therefore, to further investigate the functional connectivity between AVP and VIP neurons, we next tested the time-of-day and GABA dependency of VIP-Ca² response to the optogenetic activation of AVP neurons in vivo.

To this end, ChrimsonR-mCherry, a red light-gated cation channel [35,36], and jGCaMP7s were expressed specifically in AVP and VIP neurons, respectively, by injecting AAV-CAG-FLEX-ChrimsonR-mCherry and AAV-TRE-FLEXoff-jGCaMP7s into the SCN of control Avp-Cre; Vgat^wt/flox; Vip-tTA mice (Fig 4A–4C). Interestingly, optogenetic stimulation of AVP neurons increased VIP-Ca²⁺ levels only at night, when basal Ca²⁺ levels are low, but not during the day, when basal Ca²⁺ levels are high (Fig 4D–4H). This lack of effect during the day may reflect a ceiling effect, potentially because AVP and VIP neuronal activities are already high under daytime conditions. We confirmed that such VIP-Ca²⁺ responses were not observed when mCherry alone was expressed instead of ChrimsonR-mCherry (S8C and S8D Fig). These results suggest that AVP neurons can activate VIP neurons in vivo.

To further examine the contribution of GABAergic transmission to the effects of AVP neuronal activation, we measured VIP-Ca²⁺ responses to ChrimsonR-mediated stimulation of AVP neurons in Avp-Vgat^−/− (Avp-Cre; Vgat^flox/flox; Vip-tTA) mice. In these mice, the nighttime increase in VIP-Ca²⁺ induced by ChrimsonR stimulation was significantly reduced compared with control mice (Fig 4D–4H). These results suggest that GABA released from AVP neurons is the primary mediator of AVP neuron-driven activation of VIP neurons.

To further understand how AVP neurons regulate VIP neurons in the SCN network, we next recorded the VIP-Ca²⁺ response to the optogenetic activation of AVP neurons in coronal slices of the middle SCN along the rostro-caudal axis. ChrimsonR-mCherry and jGCaMP7s were expressed in AVP and VIP neurons of Avp-Cre; Vip-tTA mice, respectively, by injecting AAV vectors into the SCN (Fig 5A and 5B). Then, SCN slices were prepared, and an optical fiber was placed close to the ChrimsonR-mCherry-positive region in the SCN for optogenetic activation (Fig 5B).

Optogenetic activation of AVP neurons around ZT22 increased VIP-Ca²⁺, as observed in vivo (Figs 5C, 5D, S9A, and S9B). Strikingly, the application of a GABA_AR antagonist, gabazine, alone raised the baseline VIP-Ca² (Figs 5E, 5F, and S9A), suggesting that VIP neurons are suppressed by GABA during the night. Moreover, the VIP-Ca²⁺ increase induced by the optogenetic activation of AVP neurons was largely inhibited in the presence of gabazine (Figs 5G, 5H, S9A, and S9C). These data may be explained by a circuit in which GABA released from AVP neurons indirectly activates VIP neurons by inhibiting intermediate GABA neurons that suppress VIP neurons. On the other hand, the fact that gabazine failed to completely inhibit the VIP-Ca²⁺ response may indicate the presence of parallel pathways mediated by AVP neuron-derived transmitters other than GABA, such as AVP and other neuropeptides. We also examined optogenetic stimulation around ZT14 and obtained results similar to those observed around ZT22 (S9D-S9G Fig).

If AVP neurons disinhibit VIP neurons, as suggested in the previous section, there should be some population of intermediate GABA neurons that are inhibited by AVP neurons. To test this possibility, we made a spatial map of the Ca²⁺ responses of non-AVP cells to the optogenetic activation of AVP neurons in SCN slices. To do this, we expressed ChrimsonR-mCherry and jGCaMP7s in AVP and non-AVP cells, respectively, by injecting AAV-CAG-FLEX-ChrimsonR-mCherry and AAV-EF1-rDIO (reverse DIO)-jGCaMP7s into the SCN of Avp-Cre mice (Fig 6A and 6B). Then, we prepared coronal slices of the middle SCN and monitored the jGCaMP7s signal throughout the slices.

The jGCaMP7s signals of non-AVP cells began to decrease shortly (~10 s) after the onset of optogenetic stimulation of AVP neurons in most areas of the SCN at around ZT22 (Figs 6C and S10A). As stimulation continued, the signals gradually recovered from this inhibitory response and eventually shifted to an increasing response in some areas including ventral region (Figs 6C–6E and S10). In contrast, cells in the intermediate region tended to exhibit stronger inhibitory responses with minimal excitation (Fig 6C–6E). Gabazine application alone increased non-AVP baseline Ca²⁺ levels in both the middle and ventral regions (Fig 6F and 6G), suggesting that most SCN cells are suppressed by GABA during the night. Moreover, the non-AVP Ca² responses induced by the optogenetic activation of AVP neurons were largely abolished by gabazine application in both the middle and ventral regions (Fig 6H–6J). A certain degree of variability in spatiotemporal Ca²⁺ response patterns between slices may derive from differences in the locations of coronal slice planes within the SCN, which are easily affected by slight difference in cutting angle and position and are thereby inevitable (S10A Fig). These results support the idea that AVP neurons indirectly activate VIP neurons via GABA by inhibiting another population of GABAergic neurons in the SCN network.

All experimental procedures were approved by the Kanazawa University Animal Experiment Committee (approval numbers: AP-224337) and the Kanazawa University Safety Committee for genetic recombinant experiments (approval numbers: Kindai 6-2626, Kindai 6-2891). The study was carried out in compliance with Fundamental Guidelines for Proper Conduct of Animal Experiment and Related Activities in Academic Research Institutions under the jurisdiction of the Ministry of Education, Culture, Sports, Science and Technology (Notice No. 71 of 2006), and Standards relating to the Care and Keeping and Reducing Pain of Laboratory Animals (Notice of the Ministry of the Environment No. 88 of 2006).

Avp-Cre BAC transgenic (C57BL/6J-Tg(Avp-icre)#Meid/Rbrc, RBRC12048) [11] and Vip-tTA knock-in (B6(Cg)-Vip^{em1(tTA2)Miem}/Rbrc, RBRC12109) [26] mice were reported previously. Vip-ires-Cre (Vip^tm1(cre)Zjh/J, JAX:010908) [33], Vgat^flox (Slc32a1^tm1Lowl/J, JAX:012897) [47], and Rosa26-LSL-SpCas9-2A-EGFP mice (B6J.129(B6N)-Gt(ROSA)26Sor^{tm1(CAG-cas9*,-EGFP)Fezh}/J, JAX:026175) [32] were obtained from Jackson Laboratory. All lines were congenic on C57BL/6. We compared the conditional knockouts with controls whose genetic backgrounds were comparable. Avp-Cre, Vip-ires-Cre, Vip-tTA, and Rosa26-LSL-SpCas9-2A-EGFP mice were used in hemizygous or heterozygous condition. We used both male and female mice in our experiment. Whether we pooled data from both sexes or analyzed them separately, the conclusions we reached remained the same. Mice were maintained under a strict 12-h light/12-h dark cycle in a temperature- and humidity-controlled room and fed ad libitum.

The AAV-2 ITR containing plasmids pGP-AAV-CAG-FLEX-jGCaMP7s-WPRE (Addgene plasmid #104495, a gift from Dr. Douglas Kim and GENIE Project) [27]. pAAV-TRE-jGCaMP7s was described previously [48]. In addition, we modified this plasmid to make an improved version by using an EcoRI-HindIII fragment of this plasmid containing jGCaMP7s sequence to replace an EcoRI-HindIII fragment containing ChrimsonR-mCherry from pAAV-TRE-ChrimsonR-mCherry (Addgene #92207, a gift from Alice Ting) [36]. pAAV-U6-gGabrb1~3-EF1α-DIO-mCherry, plasmids for CRISPR-Cas9-mediated Gabrb1~3 gene disruption, were generated as follows. The target sites for CRISPR-Cas9 were designed by CRISPOR (http://crispor.tefor.net/↗) [49]. Two sequences targeting each Gabrb gene were selected: Gabrb1, 5′-AAGGATATGACATTCGCTTG-3′ and 5′-CGCATCCCGACGTCCACCGG-3′; Gabrb2, 5′- TGACCCTAGTAATATGTCGC-3′ and 5′-ATGTTCATTCCTACGGCCAC-3′; Gabrb3, 5′- ATTCGCCTGAGACCCGACTT-3′ and 5′-CGACATCGCCAGCATCGACA-3′. Oligonucleotides encoding the guide sequences were cloned into the BbsI and BsaI sites of pX333 (Addgene #64073, a gift from Dr. Andrea Ventura) [50]. Then, a fragment containing two tandem units of U6-gRNA was amplified by PCR, using the following primers: 5′-agtacgcgTCGAGCATGCTCGAGAATGG-3′ and 5′-agtacgcgtCGGGTACCCCATTTGTCTGC-3′, and cloned into the MluI site of pAAV-EF1a-DIO-mCherry (a gift from Dr. Bryan Roth) as described previously [48]. pAAV-U6-gGabrb2-EF1α-DIO-mCherry happened to contain two copies of the amplified fragment. pAAV-U6-gControl-EF1α-DIO-mCherry contains spacer sequences from pX333 instead of gRNA sequences for Gabrb genes. pAAV-CAG-FLEX-ChrimsonR-mCherry was described previously [12]. As a negative control for the optogenetic study, we injected AAV-EF1α-DIO-mCherry or AAV-EF1α-DIO-hM3Dq-mCherry, which was generated with plasmids pAAV-EF1α-DIO-mCherry or pAAV-EF1α-DIO-hM3Dq-mCherry provided by Dr. Bryan Roth, University of North Carolina [51]. pAAV-TRE-FLEXoff-jGCaMP7s was made by replacing an EcoRI-SpeI fragment containing ChrimsonR-mCherry of pAAV-TRE-ChrimsonR-mCherry with an EcoRI-XbaI fragment containing jGCaMP7s from pGP-AAV-CAG-FLEX-jGCaMP7s-WPRE. pAAV-EF1-rDIO-jGCaMP7s was made by replacing an AscI-NheI fragment containing ChR2-EYFP of pAAV-DIO-hChR2(H134R)-EYFP-WPRE-pA (provided by Dr. Karl Deisseroth, Stanford University) with a jGCaMP7s cDNA fragment amplified by PCR from pGP-AAV-CAG-FLEX-jGCaMP7s-WPRE, using the following primers: 5′-ataggcgcGCCACCATGGGTTCTCATCA-3′ and 5′-gcgactagTCACTTCGCTGTCATCATTTG-3′. Note that gene expression from AAV-TRE-FLEXoff-jGCaMP7s and AAV-EF1-rDIO-jGCaMP7s stops upon Cre-mediated recombination. Recombinant AAV vectors (AAV2-rh10) were produced using a triple-transfection, helper-free method and purified as described previously [11]. The titers of recombinant AAV vectors were determined by quantitative PCR: AAV-CAG-DIO-jGCaMP7s, 3.4 × 10¹³; AAV-TRE-jGCaMP7s, 6.3 × 10¹¹; AAV-TRE-jGCaMP7s (improved), 5.8 × 10¹²; AAV-TRE-FLEXoff-jGCaMP7s, 1.48 × 10¹³; AAV-EF1-rDIO-jGCaMP7s, 1.5 × 10¹³; AAV-U6-gGabrb1-EF1α-DIO-mCherry, 6.06 × 10¹²; AAV-U6- gGabrb2-EF1α-DIO-mCherry, 7.42 × 10¹²; AAV-U6-gGabrb3-EF1α-DIO-mCherry, 6.56 × 10¹²; AAV-U6-gControl-EF1α-DIO-mCherry, 2.6 × 10¹²; AAV-EF1α-DIO-hM3Dq-mCherry, 4.5 × 10¹²; AAV-CAG-FLEX-ChrimsonR-mCherry, 1.5 × 10¹³; and AAV-EF1α-DIO-mCherry, 5.2 × 10¹² genome copies/ml. Stereotaxic injection of AAV vectors was performed as described previously [10]. Two weeks after surgery, we began monitoring the mice for their locomotor activity.

We used 8 Avp-Vgat^−/− × Vip-tTA (Avp-Cre; Vgat^flox/flox; Vip^wt/tTA) mice, 5 control mice (Avp-Cre; Vgat^wt/flox; Vip^wt/tTA), and 15 Vip-ires-Cre; Rosa26-LSL-SpCas9-2A-EGFP mice for the GABA_AR disruption study (6 or 9 for gGabrb1-3 or gControl). The mice were anesthetized by administering a cocktail of medetomidine (0.3 mg/kg), midazolam (4 mg/kg), and butorphanol (5 mg/kg) and were secured at the stereotaxic apparatus (Muromachi Kikai). Lidocaine (1%) was applied for local anesthesia before making the surgical incision. We drilled small hole in the exposed region of the skull using a dental drill. We injected 0.5–1.0 μL of the virus (AAV-U6-gGabrb1-3-EF1α-DIO-mCherry; AAV-CAG-FLEX-jGCaMP7s mixed (gGabrb1: gGabrb2: gGabrb3: jGCaMP7s = 1:1:1:2) or AAV-U6-gControl-EF1α-DIO-mCherry; AAV-CAG-FLEX-jGCaMP7s mixed or AAV-TRE-jGCaMP7s) (flow rate = 0.1 μL/min) at the right or bilateral of SCN (posterior: 0.5 mm, lateral: 0.25 mm, depth: 5.7 mm from the bregma) with a 33 G Hamilton Syringe (1701RN Neuros Syringe, Hamilton) to label VIP neurons. Subsequently, we placed an implantable optical fiber (400 μm core, N.A. 0.39, 6 mm, ferrule 2.5 mm, FT400EMT-CANNULA, Thorlabs or RWD) above the SCN (posterior: 0.2 mm, lateral: 0.2 mm, depth: 5.2–5.4 mm from the bregma) with self-adhesive resin cement (Super-bond C&B, Sun Medical or RelyX Unicem2 Automix, 3M ESPE). The cement was painted black. Atipamezole (0.3 mg/kg) was administered postoperatively to reduce the anesthetized period. The mice were used for experiments 2–8 weeks after the virus injection and optical fiber implantation. Their ages ranged from 3 to 14 months old, including both males and females.

A single-color fiber photometry system (COME2-FTR, Lucir) was used to record the Ca²⁺ signal of SCN neurons in freely moving Avp-Vgat^−/− mice (Figs 1 and S1) [12,52]. Fiber-coupled LED (M470F3, Thorlabs) with LED Driver (LEDD1B, Thorlabs) was used as an excitation blue light source. The light was reflected by a dichroic mirror (495 nm), went through an excitation bandpass filter (472/30 nm), and then to the animal via a custom-made patch cord (400 μm core, N.A. 0.39, ferrule 2.5 mm, length 50 cm, COME2-FTR/MF-F400, Lucir) and the implanted optical fiber. We detected the jGCaMP7s fluorescence signal by a photomultiplier through the same optical fibers and an emission bandpass filter (520/36 nm); furthermore, we recorded the signal using Power Lab (AD Instruments) with Lab Chart 8 software (AD Instruments). The excitation blue light intensity was 10–30 μW at the tip of the patch cord of the animal side. We recorded the same for 30 s every 10 min for 2 weeks to reduce photobleaching. During the recording, the mouse was housed in a 12-h light-dark cycle for more than 5 days (LD condition) and then moved to continuous darkness for ~10 days (DD condition) in a custom-made acrylic cage surrounded by a sound-attenuating chamber. A rotary joint for the patch cord was stopped during the recording to prevent artificial baseline fluctuation. The animal’s locomotor activity was monitored using an infrared sensor (Supermex PAT.P and CompACT AMS Ver. 3, Muromachi Kikai). The detected GCaMP signal was averaged within a 30-s session [25].

Another dual-color fiber photometry system (FP3002, Neurophotometrics) was used to record the calcium signal of SCN neurons in freely moving Vip-GABA_AR^−/− mice (Fig 3) [12,53,54]. Excitation light sources were a 470-nm LED for detecting calcium-dependent jGCaMP7s fluorescence signal (F470) and a 415-nm LED for calcium-independent isosbestic fluorescence signal (F415). The duration of excitation lights is 50 ms, and the onsets of the excitation timing of LEDs were interleaved. The lights passed through excitation bandpass filters, dichroic mirrors, and then to the animal via fiber-optic patch cords (BBP(4)_400/440/900-0.37_1m_FCM-4xFCM_LAF, MFP_400/440/LWMJ-0.37_1m_FCM-ZF2.5_LAF, Doric Lenses) and the implanted optical fiber. Subsequently, both signals were detected using a CMOS camera through the optical fibers, dichroic mirrors, and emission bandpass filters. The recorded signals were acquired using Bonsai software, with a sampling rate of 10 Hz for each color. The excitation intensities of the 470-nm and 415-nm LED at the animal side’s patch cord tip were from 60 μW to 110 μW. We recorded the same for 30 s every 10 min for 3 weeks to reduce photobleaching. The detected GCaMP signal was averaged within a 30-s session. Ratio (R) was defined as the ratio between F470 (Ca²⁺-dependent signal) and F415 (Ca²⁺-independent signal) (F470/F415) for calibration and reducing motion artifacts.

To detrend the gradual decrease of the signal during recording days, ±12 h average from the time (145 points) was calculated as 24-h reference (F_24h for the single-color system, R_24h for the dual-color system). The data were subsequently detrended by the subtraction of F_24h or R_24h (ΔF or ΔR). Then, the ΔF/F_24h or ΔR/R_24h value was calculated [12]. For the actogram and daily profile analyses of VIP-Ca²⁺ rhythms, which were performed via ClockLab (Actimetrics), all ΔF or ΔR values were converted to positive values by subtracting the minimum value of ΔF or ΔR. Subsequently, these values were multiplied by 100 or 1,000 and rounded off. For VIP-Ca²⁺ daily profiles, data during the last 4 (ΔF) or 5 (ΔR) days in the LD condition and last 5 (ΔF) or 7 (ΔR) days in the DD condition were used. To determine VIP-Ca²⁺ onset and VIP-Ca²⁺ offset, ΔF or ΔR was smoothened with a 11-point moving average before generating daily profiles. Then, the time points at which the signal crossed upward and downward thresholds corresponding to 20%, 50%, or 80% of the peak-to-trough amplitude (maximum – minimum ΔF or ΔR) were defined as the onset and offset of high VIP-Ca²⁺ activity, respectively, and the interval between VIP-Ca²⁺ onset and offset was defined as high VIP-Ca²⁺ duration. The resulting profiles were further normalized such that the minimum and maximum values were set to 0 and 1, respectively. CT12 was determined as the onset of locomotor activity, as described below. To quantify the speed of VIP-Ca²⁺ increase and decline in the morning and evening, the upward and downward slopes were calculated by dividing the magnitude of VIP-Ca²⁺ change (60% change: from 20% to 80% or from 80% to 20%) by the time (h) required for that change. To quantify the fluctuation of VIP-Ca²⁺ signal, the CV was calculated based on the deviation of the unsmoothed daily profile from the corresponding smoothed daily profile. Recorded Ca²⁺ rhythm patterns were consistent between single-color and dual-color fiber photometry systems, indicating that the overall conclusions were independent of the recording configuration.

During the fiber photometry recordings, the animal’s locomotor activity (home-cage activity) was monitored using an infrared sensor (Supermex PAT.P and CompACT AMS Ver. 3, Muromachi Kikai) in 1-min bins, then 10-min bins were made by analysis. Data were analyzed using ClockLab (Actimetrics). A double-plotted actogram of locomotor activity was prepared and overlaid on that of the GCaMP. The onset and offset of locomotor activity were calculated from the daily locomotor activity profiles of the same days used for VIP-Ca²⁺ profiles, regarding the median activity level as a threshold. The intervals between locomotor activity onset and offset were defined as locomotor activity time. The phase relationship between locomotor activity and VIP-Ca²⁺ onset/offset was determined in individual mice by comparing the daily profiles of locomotor activity and smoothened VIP-Ca²⁺ rhythm.

We confirmed the jGCaMP7s or gRNA (mCherry) expression and the position of the optical fiber by slicing the brains into 30 or 100 μm coronal sections using a cryostat (Leica). The sections were mounted on glass slides with a mounting medium (VECTASHIELD HardSet with DAPI, H-1500, Vector Laboratories or Dako Fluorescence Mounting Medium, Agilent Technologies) and observed via epifluorescence microscope (KEYENCE, BZ-9000E).

Male and female Vip-GABA_AR^−/− (Vip-ires-Cre; Rosa26-LSL-SpCas9-2A-EGFP injected bilaterally with AAV-U6-gGabrb1~3-EF1α-DIO-mCherry mixture) and control (injected with control AAV) (Fig 2C) mice, aged 8–20 weeks, were housed individually in a cage placed in a light-tight chamber (light intensity was approximately 100 lux). Spontaneous locomotor activity (home-cage activity) was monitored by infrared motion sensors (O’Hara) in 1-min bins as described previously [12]. Actogram, activity profile, and χ² periodogram analyses were performed via ClockLab (Actimetrics). The free-running period and amplitude (Qp values) were calculated for the last 7 days in DD by periodogram. The onset and offset of locomotor activity and the activity time were calculated from the daily activity profile of the same 7 days in DD as described above.

We used 10 Avp-Cre; Vip-tTA mice (n = 6 for ChrimsonR, n = 4 for control, both male and female), 4 Avp-Vgat^−/− × Vip-tTA (Avp-Cre; Vgat^flox/flox; Vip^wt/tTA), and 6 Avp-Vgat^+/− × Vip-tTA (Avp-Cre; Vgat^wt/flox; Vip^wt/tTA) mice. 8 Avp-Cre; Vip-tTA in this section are from a previously used cohort [12]. We injected 0.8–1.0 μL of the mixture of viruses (AAV-TRE-FLEXoff-jGCaMP7s with AAV-CAG-Flex-ChrimsonR-mCherry or AAV-EF1α-DIO-hM3Dq-mCherry and AAV-TRE-jGCaMP7s) into the right SCN (posterior: 0.5 mm, lateral: 0.25 mm, depth: 5.7 mm from the bregma) and then implanted an optical fiber (400 μm core, N.A. 0.39, 6 mm, Thorlabs) above the SCN (posterior: 0.2 mm, lateral: 0.2 mm, depth: 5.3 mm from the bregma) with dental cement. The mice were used for experiments more than 2 weeks after the surgery.

The dual-color fiber photometry system (FP3002, Neurophotometrics) was used to record the Ca²⁺ signal of SCN neurons with optogenetic stimulation in freely moving mice, as described above in “In vivo fiber photometry” (Fig 4) [53,54]. The excitation intensities of the 470-nm and 415-nm LEDs at the tip of patch cord on the animal side ranged 70–130 μW. Additionally, a 635-nm red laser (inside the fiber photometry system FP3002) was transmitted through the same optical fibers with an intensity of 2–3 mW. During the 360 s (6 min) recording in every 2 h, optical stimulation (635 nm, 50 ms pulse, 5 Hz, 120 s, 600 pulses) was applied in the middle of the recording. Throughout the experiment, the mice were housed in a custom-made acrylic cage surrounded by a sound-attenuating chamber and maintained in a 12-h LD cycle.

The recorded data were interleaved to eliminate artifacts caused by red laser stimulation, and half of it was discarded. Consequently, the final sampling rate for the jGCaMP7s fluorescence signals at the 470-nm light excitation (F470) and the 415-nm excitation (F415) was 5 Hz. Ratio (R) was defined as the ratio between F470 and F415 (F470/F415) for calibration and reducing motion artifacts. To extract circadian changes in baseline fluorescent ratio (R₀), R₀ was defined as the mean R-value during −30–0 s. Because recordings were obtained over 2 days, a 48-h reference (R_48h) was calculated as the average R₀ across the two recording days. The baseline signal (R₀) was detrended by subtracting R_48h and detrended R₀ values were expressed as ΔR/R_48h (%). For optogenetic responses, ΔR was defined as the difference between the mean R value during the late phase of stimulation period (90–120 s, R₁) and R₀. The response magnitude was calculated as ΔR/R₀ (%). After the recordings were completed, we confirmed the jGCaMP7s and ChrimsonR-mCherry expressions and the position of the optical fiber by histology.

Vip-GABA_AR^-/- (made as described above) mice were compared to control (Vip-ires-Cre; Rosa26-LSL-Cas9-2A-EGFP with AAV-EF1a-DIO-mCherry injected, 1.0 μL bilateral of SCN) mice. Both male and female mice aged 8–20 weeks were used. Coronal brain slices (250 µm thick), including the SCN, were prepared with a linear-slicer (NLS-MT, Dosaka EM), as described previously [55]. Under an upright fluorescence microscope (Olympus, BX51WI), we visually identified VIP neurons with EGFP fluorescence in the ventral SCN and non-VIP neurons without the fluorescence in the dorsal SCN. The gRNA AAV-infected VIP neurons were identified by additional mCherry fluorescence. For recording of non-glutamatergic spontaneous postsynaptic currents (PSCs), the slices were continuously perfused with an artificial CSF (ACSF) with the following composition (in mM): 125 NaCl, 2.5 KCl, 2 CaCl₂, 1 MgSO₄, 26 NaHCO₃, 1.25 NaH₂PO₄, 10 d-glucose, equilibrated with 95% O₂ and 5% CO₂, kept at 31 ± 1 °C, and further mixed with 10 μM 6-cyano-7-nitroquinoxaline-2,3 dione disodium (CNQX) and 25 μM D-(-)-2-amino-5-phosphonopentanoic acid (D-AP5) to block fast glutamatergic transmission. Whole-cell voltage-clamp recordings were made at −60 mV with borosilicate glass electrodes (5–6 MΩ) filled with an internal solution containing the following components (in mM): 87 CsCl, 20 TEA-Cl, 10 HEPES, 10 EGTA, 0.5 CaCl₂, 4 MgCl₂, 2 QX314, 4 Na-ATP, 0.4 Na-GTP, and 10 phosphocreatine (pH 7.3, adjusted with CsOH). A combination of an EPC10/2 amplifier and Patchmaster software (HEKA) was used to control membrane voltage and data acquisition. Series resistance was compensated routinely by 80%. The nonglutamatergic PSCs completely disappeared in the presence of 10 μM gabazine (SR95531, Tocris), showing that they were mediated by GABA_A receptors [25]. The GABAergic PSCs recorded for 1–3 min were analyzed using the MiniAnalysis program (Synaptosoft). The events were picked by an amplitude threshold of 5 pA and confirmed visually to have a typical PSC waveform.

We used 17 Avp-Cre; Vip-tTA mice (n = 6 (ZT22) or 5 (ZT14) for ChrimsonR, n = 6 (ZT22) for control) for VIP-Ca²⁺ recording and 9 Avp-Cre mice for recording Ca²⁺ in non-AVP neurons (aged 8–20 weeks, including both males and females). We injected 0.6 μL of the mixture of viruses (VIP-Ca²⁺: AAV-TRE-FLEXoff-jGCaMP7s with AAV-CAG-Flex-ChrimsonR-mCherry or AAV-EF1α-DIO-mCherry, non-AVP-Ca²⁺: AAV-EF1α-rDIO-jGCaMP7s with AAV-CAG-Flex-ChrimsonR-mCherry) into the SCN bilateral (posterior: 0.5 mm, lateral: 0.25 mm, depth: 5.7 mm from the bregma). Coronal brain slices (300 µm thick), including the SCN, were prepared as described in “Slice electrophygiology” For this experiment, mice were sacrificed at ZT19–20 in darkness under a red light or at ZT11–12 in light. The slices were placed in an imaging chamber and continuously perfused with the ACSF. Before the recordings started around projected ZT22 or ZT14, the slices were pre-incubated in the experimental environment for at least one hour, which was critical to observe VIP-Ca²⁺ response to the optogenetic stimulation. Under the upright fluorescence microscope (Olympus, BX51WI), we visually identified AVP neurons with ChrimsonR-mCherry or mCherry (control) fluorescence in the dorsomedial SCN and VIP neurons with jGCaMP7s fluorescence in the ventral SCN. We imaged jGCaMP7s fluorescence every 5 s through an optical filter set (an excitation filter 470–495 nm, a dichroic mirror 505 nm and an emission filter 510–550 nm) and a digital CMOS camera (Prime BSI Express, Teledyne Vision Solutions) with MetaFluor software.

After the jGCaMP7s signal had stabilized, we optogenetically stimulated SCN AVP neurons for 2 min (617 nm, 10 Hz, 40 ms duration) through an optic fiber (200 μm core, N.A. 0.39, 6 mm, ferrule 1.25 mm, FT200EMT-CANNULA; Thorlabs) positioned near the AVP neurons and connected to a high-powered LED under the control of Digital Stimulator (WPI DS8000B) and LED driver (THORLABS). Gabazine (10 μM) was applied to ACSF to verify GABA_AR involvement. For VIP-Ca²⁺, we measured the fluorescence values of the entire jGCaMP7s-expressing ROIs using MetaFluor software, detrended the traces by division, and calculated the change in fluorescence over baseline fluorescence (ΔF/F₀%) as (Fi-F₀) * 100/ F₀. For time-course analyses, fluorescence signals were expressed as ΔF/F₀ (%), where F₀ was defined according to the type of manipulation. For optogenetic stimulation analyses, F₀ was defined as the mean F value during −60 to −30 s relative to stimulation onset. For drug application analyses, F₀ was defined as the mean F value during −120 to −60 s to avoid overlap with subsequent comparison windows. ΔF was first calculated at each time point as the difference between the fluorescence signal (F) and the corresponding F₀. Then, the ΔF/F₀ (%) value was calculated. To quantify response magnitude, ΔF/F₀ values were then averaged within specified analysis time windows.

For non-AVP-Ca²⁺, we selected unilateral SCN regions (150 × 260 pixels, 2.6 μm/pixel). Images were analyzed using MATLAB. The resolution of images was then adjusted to 20.8 μm/pixel. Fluorescence signals were analyzed in a two-step procedure consisting of exploratory (qualitative) pixel-wise analysis followed by ROI-based quantitative analysis. In the exploratory analysis, fluorescence signals were quantified on a pixel-by-pixel basis to generate ΔF/F₀ (%) heatmaps, which were used to visualize the spatial distribution of Ca²⁺ responses. For this purpose, F₀ was defined as the mean F value during −120–0 s. Based on these heatmaps, two representative regions exhibiting prominent responses (middle and ventral SCN) were selected for further quantitative analysis. For subsequent analyses, square regions of interest (ROIs; 4 × 4 pixels) were defined in the middle and ventral of the SCN regions. ROI-based fluorescence signals were then analyzed using the same analytical framework as described above. For these analyses, F₀ was defined as the mean F value during −60 to −30 s for both optogenetic stimulation and drug application experiments.

All results are expressed as mean ± SEM. For comparisons of two groups, two-tailed Student t test or Welch’s t test was performed. For comparisons of multiple groups with no difference of variance, two-way mixed-design ANOVA with time or lighting conditions as a within-subject (repeated-measures) factor and genotype as a between-subject (independent design) factor, followed by post-hoc two-tailed Student’s or Welch’s t test with Bonferroni correction were performed. For comparisons of multiple groups with difference of variance, nonparametric tests, Kruskal–Wallis test with post-hoc Dunn’s test were performed. All P values less than 0.05 were considered as statistically significant. Only relevant information from the statistical analysis was indicated in the text and figures. For circular data, Watson–Williams test was performed with Circstat MATLAB Toolbox for Circular Statistics [56]