Geniculohypothalamic GABAergic projections gate suprachiasmatic nucleus responses to retinal input

Lydia Hanna1, Lauren Walmsley1, Abigail Pienaar1, Michael Howarth1, Timothy M. Brown1*

1Faculty of Biology Medicine and Health, University of Manchester, UK

*For correspondence:AV Hill Building, University of Manchester, Oxford Road, Manchester, M13 9PTtimothy.brown@manchester.ac.uk

Running Title: Geniculohypothalamic signals gate SCN responses

Keywords: circadian, electrophysiology, intergeniculate leaflet, mouse, channelrhodopsin

TOC category: Neuroscience – behavioural/systems/cognitive

Key Points Summary

Visual input to the suprachiasmatic nucleus circadian clock is critical for animals to adapt their physiology and behaviour in line with the solar day.

In addition to direct retinal projections, the clock receives input from the visual thalamus but the role of this geniculohypothalamic pathway in circadian photoreception is poorly understood.

Here we develop a novel brain slice preparation which preserves the geniculohypothalamic pathway to show that GABAergic thalamic neurons inhibit retinally-driven activity in the central clock in a circadian time-dependent manner.

We also show that in vivo manipulation of thalamic signalling adjusts specific features of the hypothalamic light response, indicating that the geniculohypothalamic pathway is primarily activated by crossed retinal inputs.

Our data provide a mechanism by which geniculohypothalamic signals can adjust the magnitude of circadian and more acute hypothalamic light responses according to time-of-day and establish an important new model for future investigations of the circadian visual system.

12

34

5678

9

10

11

12

13

14

151617181920212223242526272829

Abstract

Sensory input to the master mammalian circadian clock, the suprachiasmatic nucleus (SCN), is vital in allowing animals to optimise physiology and behaviour alongside daily changes in the environment. Retinal inputs encoding changes in external illumination provide the principle source of such information. The SCN also receives input from other retinorecipient brain regions, primarily via the geniculohypothalamic tract (GHT), but the contribution of these indirect projections to circadian photoreception are currently poorly understood. To address this deficit, here we established an in vitro mouse brain slice preparation which retains connectivity across the extended circadian system. Using multi-electrode recordings, we first confirm that this preparation retains intact optic projections to the SCN, thalamus and pretectum and a functional GHT. We next show that optogenetic activation of GHT neurons selectively suppresses SCN responses to retinal input, that this effect exhibits a pronounced day-night variation and involves a GABAergic mechanism. This inhibitory action was not associated with overt circadian rhythmicity in GHT output, indicating modulation at the SCN level. Finally, we use in vivo electrophysiological recordings alongside pharmacological inactivation or optogenetic excitation to show that GHT signalling actively modulates specific features of the SCN light response, indicating that GHT cells are primarily activated by crossed retinal projections. Together, our data establish a new model for studying network communication in the extended circadian system and provide novel insight into the roles of GHT-signalling, revealing a mechanism by which thalamic activity can help gate retinal input to the SCN according to time of day.

Abbreviations

ChR2, Channelrhodopsin-2; GAD2, glutamate decarboxylase 2; GHT, geniculohypothalamic tract; IGL, intergeniculate leaflet; ipRGCs, intrinsically photosensitive retinal ganglion cells; OC, optic chiasm; OT, optic tract; PCA, principle component analysis; PON, pretectal olivary nucleus; SCN, suprachiasmatic nucleus; vLGN, ventral lateral geniculate nucleus; vSPZ, ventral subparaventricular zone; ZT, Zeitgeber time.

Introduction

Visual input to the suprachiasmatic nucleus (SCN) circadian clock plays an essential role, allowing mammals to optimise physiology and behaviour across the solar day. This function is supported by direct retinal projections to the SCN from intrinsically photosensitive, melanopsin-expressing, retinal ganglion cells (ipRGCs; Hattar et al., 2006). While ipRGCs are thus essential for circadian photoentrainment (Güler et al., 2007; Chen et al., 2011), the SCN is also reciprocally connected to several other ipRGC target regions: the thalamic intergeniculate leaflet/ventral lateral geniculate nucleus (IGL/vLGN) and the pretectal olivary nucleus (PON) (Watts et al., 1987; Mikkelsen, 1992; Kalsbeek et al., 1993; Morin et al., 1994; Horvath, 1998; Moore et al., 2000).

So far, the IGL/vLGN projection to the SCN (the geniculohypothalamic tract; GHT) is the only component of this extended circadian system that has been investigated in detail. IGL/vLGN cells are primarily GABAergic and express a variety of peptide co-transmitters, with cells expressing neuropeptide Y (NPY) providing a major (but not exclusive) contribution to the GHT (Moore & Card,

30

31

32333435363738394041424344454647484950

51

5253545556

57

5859606162636465

66676869

1994; Morin & Blanchard, 1995; Morin & Blanchard, 2001). In line with the presence of substantial input to the IGL/vLGN from the major brain arousal systems (Moore & Card, 1994), at present, the best-defined role for the GHT is in mediating circadian responses to arousal-inducing (‘non-photic’) stimuli (Harrington, 1997; Morin & Allen, 2006). Hence, circadian responses to non-photic stimuli (which alter the timing of rodent behavioural rhythms in a manner essentially opposite to light) are blocked by lesions of the GHT (Johnson et al., 1988; Janik & Mrosovsky, 1994; Wickland & Turek, 1994; Marchant et al., 1997) and can be mimicked by electrical stimulation of the GHT (Rusak et al., 1989), in vivo/in vitro application of NPY (Shibata & Moore, 1993; Huhman & Albers, 1994; Huhman et al., 1996) or GABA agonists (Smith et al., 1989; Gillespie et al., 1996; Huhman et al., 1997; Novak et al., 2004).

By contrast, despite the substantial retinal input to the IGL/vLGN (and its connections to the visual centres in the pretectum), the contribution of the GHT to circadian photoreception remains uncertain. Data from studies using pharmacological manipulation of either GABA (Gillespie et al., 1996; Gillespie et al., 1997; Novak & Albers, 2004) or NPY signalling suggest that neurochemicals released from the GHT could actively inhibit light-driven phase shifting (Yannielli & Harrington, 2000; Lall & Biello, 2002, 2003; Yannielli et al., 2004). The extent to which these reported effects reflect the physiological contribution of the GHT to photic phase-resetting or whether GABA and/or NPY released from the GHT directly modulates light-evoked firing within the SCN remains uncertain, however. This is especially true in the case of GABA signalling, where the pharmacological approaches used above are unable to specifically distinguish effects arising due to GHT-derived GABA release from those due to the action of local SCN neurons or other long-range inputs.

Moreover, while lesion studies are equivocal as to how the GHT influences circadian photoreception, on balance, these in fact seem to favour the view that circadian light responses are diminished, rather than enhanced without a functional GHT (Harrington & Rusak, 1986; Pickard et al., 1987; Johnson et al., 1989; Maywood et al., 1997; Edelstein & Amir, 1999a, b; Morin & Pace, 2002; Muscat & Morin, 2006; Evans et al., 2012). Similarly, existing attempts to define the nature of the visual signals conveyed by the GHT provide no clear consensus as to whether these are primarily activated, suppressed or unresponsive to increases in illumination (Harrington & Rusak, 1989; Zhang & Rusak, 1989; Thankachan & Rusak, 2005; Blasiak & Lewandowski, 2013).

Here then, we set out to resolve these questions as to how the GHT influences SCN visual processing. We first generate and validate an in vitro model that retains an intact extended circadian system, including a functional GHT. We next use this model, alongside optogenetic manipulations, to define the impact of GHT signals on SCN responses to optic input. Finally, we perform in vivo recordings coupled with local pharmacological inhibition or optogenetic excitation to reveal the physiological impact of the GHT on light-evoked neural activity within the SCN.

Materials and Methods

Animals

All animal use was in accordance with the Animals, Scientific Procedures, Act of 1986 (UK) and received institutional ethics committee and UK Home Office approval. In vitro electrophysiological

70717273747576777879

8081828384858687888990

9192939495969798

99100101102103104

105

106

107

108109

experiments were performed on adult (40-130 day old) male and female Opn4 +/tau-lacZ mice and their wild-type littermates (Hattar et al., 2006). For optogenetic control of GABAergic neurons, glutamate decarboxylase 2 (GAD2)-IRES-Cre mice (Taniguchi et al., 2011) were crossed with a second line (Ai32) exhibiting Cre recombinase-dependent Channelrhodopsin-2/EYFP (ChR2/EYFP) fusion protein expression (Madisen et al., 2012). The resulting double heterozygous offspring thus exhibited ChR2/EYFP expression restricted to GABAergic (GAD-2 expressing) neurons. In vivo electrophysiological experiments were performed on adult male C57/Bl6 wild-type or GAD2-cre; Ai32 mice under urethane anaesthesia. Animals were housed under a 12 h light:dark cycle at a temperature of 22°C with food and water available ad libitum. Zeitgeber time (ZT) was used to define the projected time of the SCN circadian clock, with lights-off defined as ZT 12.

X-gal labelling and slice reconstruction

In order to trace ipRGC projections to the extended circadian system, Opn4 +/tau-lacZ mice were euthanized with an overdose of urethane (intraperitoneal injection: 0.3ml of 20% urethane) and were then transcardially perfused with 0.1 M phosphate-buffered saline (PBS; pH 7.4) followed by cold, 4% paraformaldehyde (PFA) in PBS for 10 min. Brains were removed, post-fixed in 4% PFA solution and then cryoprotected in 30% sucrose in PBS overnight. Coronal sections (60 µm thick) were then obtained using a freezing sledge microtome (8000 Sledge; Bright Instruments, UK). To visualise β–galactosidase expression, X-gal staining was then performed as described previously (Sakhi et al., 2014). In brief, sections were washed twice for 10 min in Buffer B (0.1 M PBS at pH 7.4, 2 mM MgCl2, 0.01% Na-desoxycholate and 0.02% [octylphenoxyl] polyethoxyethanol [IGEPAL]) and then incubated for 18 h in staining solution (Buffer B with 5 mM potassium ferricyanide, 5 mM potassium ferrocyanide, and X-gal (1 mg/ml, Bioline Reagents Ltd., UK)) at 37ᵒC in darkness. Following X-gal staining, sections were washed twice for 5 min in 0.1 M PBS and mounted onto glass slides. Nuclear fast red counterstain (Sigma-Aldrich, UK) was applied to the sections for 5 min, followed by blot-drying, dehydration steps (in 50, 70 and 100% ethanol: 5 min each), and finally a clearing agent was added for 5 min (HistoChoice, Sigma-Aldrich, UK). Slides were then mounted with DPX (Sigma-Aldrich, UK) and coverslipped. Images of serial sections were acquired with a Plan-Apochromat (20x/0.80) objective using a Pannoramic 250 Flash II slide scanner (3DHISTECH, Hungary). Resulting images of each brain section were then manually aligned and imported into MATLAB (The Mathworks, Inc., USA) where RGB colour ratios were used to distinguish X-gal staining from background (fast red) and produce a 3D map of relative ipRGC fibre/terminal density.

Immunofluorescence processing

For visualisation of ChR2 expression, GAD2-Cre; Ai32 mice were euthanized and brains were cryoprotected and sectioned as indicated above for X-gal labelling studies (except in this case 50 µm thick). Resulting brain sections were washed three times in 1% triton-X 100 in PBS (PBS-T) for 5 min each and then incubated in blocking solution (10% normal donkey serum, 0.2% PBS-T) for 1 h at room temperature (21ᵒC). Sections were subsequently incubated in chicken anti-GFP primary antibody (Abcam, ab13970; 1:1000, diluted in blocking solution) overnight at 4˚C. After the incubation period, sections were washed three times in 0.2% PBS-T for 15 min each and then incubated in Alexa Fluor 488-AffiniPure donkey anti-chicken secondary antibody (Stratech, 703-545-155-JIR; 1:5000, diluted in blocking solution) for 1 h at room temperature. Sections were washed three times in PBS for 15 min, mounted onto glass slides with DAPI-containing Vectashield HardSet

110111112113114115116117118119

120

121122123124125126127128129130131132133134135136137138139140

141

142143144145146147148149150151

(Vector Labs, H-1500) and coverslipped. Epifluorescence images of immunostaining were then acquired by Pannoramic 250 Flash II slide scanner as described above.

In vitro multi-electrode recordings

Mice were sacrificed by cervical dislocation followed by decapitation (ZT2-4). Brains were then rapidly removed and placed into a custom-made adult mouse brain matrix (Zivic Instruments, USA). The base of the brains (just rostral to the cerebellum) were cut at 30° off the coronal plane and immediately mounted onto a metal stage and sectioned using a 7000 smz-2 vibrating microtome (Campden Instrument, UK) in ice-cold (4°C) sucrose-based slicing solution composed of (in mM): sucrose (189); D-glucose (10); NaHCO3 (26); KCl (3); MgSO4 (5); CaCl2 (0.1); NaH2PO4 (1.25); oxygenated with 95% O2 / 5% CO2 mixture. Once the optic chiasm/rostral pole of the SCN became visible we then cut a 600 μm thick section which contained the SCN, IGL/vLGN and PON. Only one slice was produced per animal. The resulting slice was then transferred into a petri dish containing oxygenated artificial cerebrospinal fluid (aCSF) composed of (in mM): NaCl (124); KCl (3); NaHCO3

(24); NaH2PO4 (1.25); MgSO4 (1); glucose (10); CaCl2 (2); slices were then left to rest for 1 h at room temperature (22°C).

Multi-electrode array recordings from acute brain slices were performed as previously described (Walmsley et al., 2015). Slices were placed, recording side down, onto 60pMEA100/30iR-Ti-gr perforated multi-electrode arrays (pMEAs; Multi Channel Systems, MCS GmbH, Germany), comprising 59 electrodes (site area: 707 µm2) spaced 100 µm apart with one dedicated low impedance reference electrode. Perforated arrays were chosen due to their improved slice perfusion rate, low signal-to-noise ratio and long-term recording stability (Reinhard et al., 2014). Confirmation of optimal slice placement over the pMEA electrode sites was confirmed by overlaying images taken with a GXCAM-1.3 camera (GX Optical, UK). Where relevant, anatomical placement was also confirmed post hoc with X-gal staining (for Opn4+/tau-lacZ) or endogenous EYFP fluorescence (for GAD2-Cre; Ai32) as discussed below.

During recordings, slices were held in place by both the bottom flow suction (via the MEA perforations) driven by a constant vacuum pump (MCS GmbH, Germany) and a weighted grid. The pMEA recording chamber was continuously perfused with pre-warmed oxygenated aCSF (34 ± 1°C) to both slice surfaces at a combined rate of 2.5–3 ml/min. For long-term recordings (>26 h), 0.001% gentamicin (Sigma-Aldrich, UK) was added directly to the aCSF. Neural signals were acquired with MC_Rack software as time-stamped action potential waveforms using a USB-ME64 system and MEA1060UP-BC amplifier (MCS GmbH, Germany). Signals were sampled at 50 kHz (12.5 kHz for long-term recordings), high-pass filtered at 200 Hz (second order Butterworth) and multiunit spikes crossing a threshold (usually set at -16.5 μV) were extracted for further analysis. In some cases we also isolated single unit activity from these multiunit recordings, using principle components analysis (PCA)-based spike sorting (Offline sorter V3; Plexon, USA). Reliable single unit isolation was confirmed by reference to MANOVA F statistics, J3 and Davies-Bouldin validity metrics (Offline sorter) and the presence of a distinct refractory period (>20ms for SCN neurons) in the interspike interval distribution.

For post-hoc histological analysis, slices were submerged in 4% PFA at room temperature for 40 min and then kept overnight at 4ᵒC in 0.1 M PBS. In the case of X-gal staining, sections were then additionally washed twice for 15 min in Buffer B and incubated overnight in staining solution at 37ᵒC

152153

154

155156157158159160161162163164165166

167168169170171172173174175176

177178179180181182183184185186187188189190

191192193

in darkness (see solution details in X-gal labelling section above). All sections were subsequently washed twice for 5 min in 0.1 M PBS and mounted onto glass slides with Fluoromount (Sigma-Aldrich, UK). Images of X-gal staining were acquired with ZEN digital imaging software (Carl Zeiss, Germany) and EYFP fluorescence was captured on an Olympus BX51 upright microscope (4x/0.30 Plan Fln) using a Coolsnap ES camera (Photometrics) with MetaVue Software (Molecular Devices).

Electrical and optogenetic stimulation

To assess responses to optic input, retinal projections were activated by electrical stimulation of the optic chiasm (OC) using two methods: either biphasic stimulation via a neighbouring pair of pMEA electrodes visually confirmed to be situated within the OC (2500 mV, 100 µs duration, 0.5 Hz) or via an independently positioned stimulating concentric bipolar electrode (FHC, USA; 0.3 - 1.5 mA, 200 µs duration, 0.5 Hz). Similar stimulation parameters were used to activate the GHT (in this case with the stimulating electrode positioned within the IGL-region at least ~200 μm lateral to the optic tract). Stimuli were applied as single pulses (0.5 Hz) or as a train of five stimuli, separated at 20 or 40 Hz.

For optogenetic stimulation, blue light flashes (5 - 10 ms duration, 0.5 Hz) were delivered by a 200 µm core fibre (465 nm LED, ~800 mW/mm2 at fibre tip; PlexBright; Plexon) placed 100-200 µm above the slice surface over the IGL or SCN region. Based on the numerical aperture of the fibre (0.66), this stimulus illuminates a ~500 μm diameter portion of the slice (although due to light scatter the effective stimulated region will be larger). In order to assess functional convergence of retinal and thalamic inputs in the SCN and GHT signalling mechanisms, optical stimulation of the IGL/vLGN was synchronously delivered, or preceded pMEA electrical stimulation (2500 mV, 100 µs duration) of the OC at various latencies (25 ms - 1 s, 10 – 60 s between each OC/GHT pair: 5 - 6 different latencies were tested per experiment and delivered in an interleaved fashion). For long-term recordings of thalamic evoked and spontaneous activity, we alternately stimulated optic input via a concentric electrode placed in the OC (0.6 - 1.2 mA, 200 µs duration, 0.02 Hz) and directly activated GHT GABAergic neurons via the optical fibre as above (5 ms duration, 0.02 Hz).

Pharmacological manipulation

Pharmacological separation of the synaptic components of evoked responses was achieved by bath application of ionotropic NMDA (D-(-)-2-Amino-5-phosphonopentanoic acid; D-AP5, 50 µM) and AMPA/kainate (6-Cyano-7-nitroquinoxaline-2,3-dione disodium; CNQX, 20 µM) glutamate receptor blockers. Contributions of GABAergic transmission were evaluated by bath perfusion with (+)-bicuculline (BIC, 20 µM), an antagonist at ionotropic GABAA receptors, and CGP55845 hydrochloride (CGP, 3 µM), an antagonist at metabotropic GABAB receptors. At the end of all experiments, slices were treated with bath application of N-Methyl-D-aspartic acid (NMDA, 20 μM) to confirm maintained cell responsiveness, followed by tetrodotoxin citrate (TTX, 1 μM), to confirm that acquired signals exclusively reflected Na+-dependent action potentials. All drugs were purchased from Sigma-Aldrich (UK) or Tocris (UK), dissolved in ddH2O and kept as stock solutions at -20°C (with the exception of BIC and CGP which were dissolved in DMSO) and were diluted to their respective final concentrations in pre-warmed, oxygenated aCSF.

Visual stimuli

194195196197198

199

200

201202203204205206207

208209210211212213214215216217218219

220

221222223224225226227228229230231232

233

Visual stimuli were delivered and quantified as described previously (Walmsley & Brown, 2015). Full field visual stimuli were generated via two LEDs (λmax 405 nm; half-width: ±7 nm; Thorlabs, USA), controlled via LabVIEW (National Instruments, USA) and neutral density filter wheels (Thorlabs, USA). Light was supplied via flexible fibre optic light guides (Edmund Optics, UK; 7 mm diameter) terminating in internally reflective plastic cones that fit snugly over the mouse’s eye to prevent off-target effects to stimulation of the opposing eye. Light measurements (at the level of the mouse eye) were performed using a calibrated spectroradiometer (Bentham instruments, UK), with effective photoreceptor excitations calculated using established measures (Govardovskii et al., 2000; Jacobs & Williams, 2007; Walmsley et al., 2015). Owing to our choice of wavelength, apparent brightness of our stimuli differed by no more than 0.2 log units for any of the known mouse opsins, with values reported in this manuscript reflecting effective irradiance for rod opsin.

In vivo multi-electrode recordings

Mice were removed from their housing environment 1-2h before lights off (ZT10-11), anaesthetised by intraperitoneal injection of urethane (1.55g/kg in 0.9% physiological saline) and then placed in a stereotaxic frame (SR-15M; Narishige International Ltd., UK), where their temperature was maintained at 37˚C using a homeothermic mat (Harvard Apparatus, UK). The surface of the skull was exposed and two holes were drilled using stereotaxic co-ordinates (0.95 mm lateral to, and 0.3 mm posterior to bregma for SCN recording electrode; 2.3 mm lateral to, and 3.5 mm posterior to bregma for IGL probe) obtained from the mouse atlas (Paxinos & Franklin, 2001). Pupils were dilated with topical application of atropine (1% in saline) and a thin coating of mineral oil was then applied to prevent drying of the cornea (both sourced from Sigma-Aldrich, UK).

For SCN recordings 32-channel electrodes were then coated with fluorescent dye (DiI; Invitrogen, UK) and inserted into the brain 0.95 mm lateral, and 0.3 mm caudal to bregma (angled at 9° relative to the dorsal-ventral axis). We employed two different electrode configurations for these studies: Buszaki 32L (Neuronexus, USA) or 2x16Parallel (Cambridge Neurotech; UK) both of which produced equivalent data. Buszaki32L consisted of four shanks (200 μm spacing) each with eight closely spaced recordings sites in diamond formation (inter-site distance of 22 - 42 μm; site area: 160 μm2). The 2x16Parallel electrodes had two shanks (250 μm spacing) each with two rows of 8 recording sites (25 μm spacing; site area: 160 μm2). For manipulation of thalamic activity a second neural probe (see details below), again coated in DiI, was then inserted into the brain 2.3 mm lateral, and 3.5 mm caudal to bregma (angled at 16° relative to the dorsal-ventral axis) such that the tip of the electrode reached the IGL-region.

In all cases, electrodes were then lowered into position using a fluid-filled micromanipulator (MO-10, Narishige International Ltd., UK). After allowing at least 30 min for neural activity to stabilise following probe insertion, wideband neural signals were acquired using a Recorder64 system (Plexon), amplified (x3000) and digitized at 40 kHz. Timestamped multiunit spikes (1 ms segments of data) were then extracted from these recordings by threshold crossing (-40 µV) for subsequent analysis.

At the end of these in vivo experiments, brains were removed and post-fixed in 4% PFA and then cryoprotected and sectioned (100 µm) as described above for X-gal labelling studies. DiI-marked

234235236237238239240241242243244

245

246

247248249250251252253254255

256257258259260261262263264265266

267268269270271272

273274

probe position (and where relevant endogenous GAD-driven EYFP fluorescence) was then visualised and captured on an Olympus BX51 upright microscope (4x/0.30 Plan Fln) using a Coolsnap ES camera (Photometrics) with MetaVue Software (Molecular Devices).

Pharmacological inactivation of the visual thalamus

For these experiments we implanted a 16-site linear recording array (100 μm spacing; site area: 177 μm2; E16-20mm-100-177; Neuronexus, USA) with a drug cannula (outer diameter 165μm; fused to the underside of the array) protruding 100 μm ventral to the probe tip into the IGL/vLGN region. To facilitate drug infusion into the visual thalamus, the cannula underlying the recording probe (100 µm aperture) was connected via flexible narrow bore tubing to a syringe pump (Harvard Apparatus, Cambridge, UK) preloaded with muscimol (1 mM, Sigma).

After allowing neural activity to stabilise, we then evaluated baseline visual responsiveness across the SCN and visual thalamus. Mice were maintained in darkness and 5 s light steps were applied in an interleaved fashion to contralateral and/or ipsilateral eyes for a total of 10 repeats at logarithmically increasing intensities spanning 11.4 - 15.4 log photons/cm2/s (inter-stimulus interval: 20 s).

Once we had collected this baseline data we then began to infuse muscimol into the thalamus in 1 μl increments (delivered over 1 min) while monitoring responses to visual stimuli (5 s binocular light steps 15.4 log photons/cm2/s). Drug onset (observed after nominal injection volumes of 2-3 μl) was easily identifiable by a rapid loss of spontaneous and evoked neural activity, alongside a small increase in electrical noise. Following confirmation of abolition of thalamic activity by muscimol application, mice were then dark adapted (30 min) and the protocol above was repeated. In no case did we observe a reversal of drug effects during our recording sessions.

Optogenetic activation of the visual thalamus

These experiments were performed in GAD2-cre; Ai32 mice. We implanted a 32-site recording array (3 parallel rows of 10-12 sites at 50 μm spacing; site area: 177 μm2; OALPPoly3; Neuronexus) into the thalamus with an optical fibre (110 μm core; 0.66 Numerical Aperture) terminating 200μm above the dorsal-most recording site. For optogenetic stimulation, the optical fibre was connected to a 465nm PlexBright LED (as above), providing ~630mW/mm2 light energy at the fibre tip.

After allowing neural activity to stabilise, we first confirmed that we were able to manipulate GABAergic neuronal activity within the visual thalamus by monitoring multiunit responses to a train of 10ms light flashes (100 repeats at 0.5Hz). We then evaluated neural responses across the SCN and visual thalamus to bright light steps (15.4 log photons/cm2/s) targeting either the contralateral, ipsilateral or both eyes in the presence or absence of repeated optogenetic stimulation (5s trains of 10ms flashes at 5, 10 and 20Hz). Each test stimulus was repeated 10 times in an interleaved fashion (with 40s inter-stimulus interval).

Data analysis

Interpretation of multiunit data

275276277

278

279280281282283284

285286287288289

290291292293294295296

297

298299300301302

303304305306307308309

310

311

Here we evaluate neural responses to a range of stimulus conditions by monitoring ensemble activity detected across multiple recording electrodes within a single preparation. Multiunit recording has been used extensively to monitor SCN activity (Kubota et al., 1981; Tcheng & Gillette, 1996; Gribkoff et al., 1998; Yamazaki et al., 1998; Mrugala et al., 2000; Tousson & Meissl, 2004; Brown et al., 2006; VanderLeest et al., 2007; Nakamura et al., 2008) and provides a reliable indication of overall population activity within the nucleus (Schaap et al., 2003; Brown & Piggins, 2009) that matches well that attained using other single-cell recording approaches (Green & Gillette, 1982; Groos & Hendriks, 1982; Shibata et al., 1982; Schaap et al., 1999; Pennartz et al., 2002; Itri et al., 2005; Nygård et al., 2005). Nonetheless, an important first consideration in interpreting the data from our multi-electrode recordings is the size of the neural ensemble such recordings represent and the extent to which signals detected at different electrodes can be considered independent.

Based on the typical peak SCN multiunit firing rates we detect in vitro for recordings spanning a full circadian cycle (mean ± SD: 4.4±6.4 spikes/s across each channels most active 1hr epoch; n=144 channels from 5 slices), relative to equivalent data obtained for single SCN cells (mean = 2.5 spikes/s; (Brown & Piggins, 2009), we consider it highly unlikely that any single electrode samples from more than 7 neurons. Given the small size and dense packing of SCN cells (diameter <10μm; average inter-soma distance ~14 μm; Abrahamson & Moore, 2001), this upper limit of 7 neurons would be contained within a hemisphere with a radius of 17 μm relative to the centre of each recording site. We thus consider it most unlikely that we detect cells further than 17μm from each electrode (substantially smaller than the inter-electrode distance of these arrays: 100μm).

Similarly, comparing the typical light evoked multiunit activity we detect in vivo (mean ± SD: 14.1±15.5 spikes/s for a 5s, 405nm light step providing 15.4 log photons/cm2/s to both eyes; n=295 channels from 20 mice) relative to typical single cell responses to an equivalent stimulus (mean=9.7 spikes/s; Walmsley & Brown, 2015), we estimate our multiunit recordings derive from no more than 5 cells/channel. This suggests an upper limit to the listening range of each electrode of ~15 μm (approximately half the inter-electrode distance for these arrays: 22-42μm).

To provide additional confidence that the multiunit signals we record at neighbouring electrodes are truly independent, we also calculated the incidence of synchronous spike detection (defined here as spikes occurring within ±200μs) at adjacent channels across subsets of our in vivo or in vitro SCN recordings. Since some SCN neurons are known to communicate using gap junctions (Long et al., 2005), the results of this analysis should be considered as an upper bound to the extent to which we may detect the same neuron at more than one electrode.

In the case of in vitro recordings, we analysed data from 794 pairs of horizontally or vertically adjacent channels that showed detectable neural activity (from 5 slices recorded over a full circadian cycle) and found that the incidence of synchronous spikes was very low (median = 1.2% of spikes at each channel, interquartile range=0.4-2.9% of spikes). An equivalent analysis performed on data from in vivo recordings (560 pairs of adjacent channels from 10 recordings) revealed a broadly similar incidence of synchronous spikes (median = 2.1% of spikes at each channel, interquartile range=1.2-3.7% of spikes). Thus for both recording modalities, we consider the signals detected at individual channels as almost exclusively representing the activity of distinct groups of neurons.

Anatomical locations of recorded ensembles

312313314315316317318319320321322

323324325326327328329330331

332333334335336337

338339340341342343

344345346347348349350351

352

As indicated in the relevant sections above, placement of recording electrodes in all experiments was confirmed by overlaid reconstructions of the known recording array geometry and histological images (using ventricles and fibre tracts as landmarks and, where available, X-gal staining or EYFP fluorescence). The anatomical extent of our recording arrays ensured that in many experiments some recording sites were located outside our regions of interest. For most of the analysis presented here, where electrode sites situated formally just outside these regions of interest exhibited functional responses to stimulation of retinal afferents pathways (see below), such data was included in our overall analysis. In particular, for hypothalamic recordings, retinally-driven responses occasionally identified at recording sites located in the ventral subparaventricular zone (vSPZ), just dorsal to the SCN, were also included in our analysis. Exceptions to the rule above were our analysis of circadian rhythms in SCN, vSPZ and IGL/vLGN output (analysis presented in Fig. 3 and Fig. 6), where we excluded data from sites located outside the borders of these regions evident in the anatomical images. Small errors in our anatomical classification of recording site location here are most unlikely to have influenced our conclusions. Hence, including or excluding data from sites located at the projected anatomical borders of the regions in question produced essentially identical results to those reported in this manuscript.

Identification of responding channels

Acute changes in neural activity evoked by electrical or optogenetic stimulations were classified as excitatory or inhibitory when the average spike counts across multiple stimulus repeats (typically substantially >100) respectively exceeded the upper or lower bounds of the 99% confidence limits for prestimulus spike counts (within at least one 10ms bin <100ms after stimulus onset). For the majority of data presented in this manuscript (fast excitatory responses which are relatively time locked to the stimulus), representative multiunit data illustrating such responses are presented as perievent histograms reporting spike probability per unit time relative to stimulus onset. The exceptions here are data for GHT-driven inhibitory responses where perievent histograms are instead converted to average firing rates per unit time across trials.

Where relevant (analysis presented in Figs. 6 and 7), channels identified as responding to electrical stimulation of the OC were further categorised as receiving modulatory input from the GHT. For this analysis, channels were considered ‘GHT responsive’ when the OC-evoked response following prior optogenetic stimulation of the GHT (at any latency) was significantly different (P<0.05) to the response evoked following synchronous stimulation (paired t-tests with Bonferroni correction, based on responses to at least 120 trials at each latency).

For in vivo recordings, recorded ensembles at each site were considered light responsive when mean firing rates over the course of 5s binocular light exposure (15.4 log photons/cm 2/s) were significantly different to the firing rate in the 5s epoch immediately prior to light exposure (paired t-test from responses to 10 trials).

For analysis of responsiveness to pharmacological agents (NMDA; Fig. 3), channels were considered responsive based on previously established criteria (Brown et al., 2008): when the mean firing rate changed by more than 20% during the 3-8min after the start of drug application relative to pre-drug levels. To account for the possible confounding effects of baseline drift, pre-drug firing was extrapolated based on a linear fit to the mean firing rates observed over the 5 min epoch immediately prior to the start of drug application.

353354355356357358359360361362363364365366367368

369

370371372373374375376377378

379380381382383384

385386387388

389390391392393394

Analysis of multiunit responses

Where we report average time-course data for excitatory responses in the presence and absence of pharmacological blocking agents (Fig. 2.), we combined data from all channels identified (as above) as responding across each of the 2-6 slices tested. For this analysis we first calculated, for each channel, perievent response histograms (5ms bin size) for the 6min epochs immediately prior to drug application and 3-9min immediately following the start of drug application (corresponding to 180 trials in either epoch). These histograms were then normalised by subtracting pre-stimulus spike counts (averaged across the 50ms preceding stimulation) and divided by the maximal response observed under baseline conditions. For comparison of pre- and post-drug responses the resulting datasets were then analysed by two-way repeated measures (RM) ANOVA with Sidak’s test for multiple comparisons (GraphPad Prism, GraphPad Software, USA).

For quantification of the time-course and magnitude of SCN responses to GHT-stimulation we produced peri-event histograms (responses to at least 300 trials, bin size=1ms) and smoothed with a 10ms boxcar filter. Latency was measured as time to the largest positive poststimulus deflection and response magnitude measured as the percent change in spike probability relative the average spike probability during the 50ms epoch preceding stimulation. The effects of pharmacological agents on this latter parameter were compared by paired t-tests of equivalent analysis performed on responses occurring 0-10 min prior to drug application vs. 5-15 min after the start of drug perfusion.

For comparisons of activity evoked during combined GHT/OC stimulation (Figs 6 and 7), responses were quantified as the change in spike probability between the 25ms epochs immediately before and after OC stimulation. For analysis, the corresponding data for each channel was then normalised according to the maximal response observed at any GHT-OC stimulation delay. For experiments involving pharmacological blocking agents this normalisation was applied separately for baseline and each of the drug treatments (25min epochs starting immediately prior to or 5 to 30 min after the start of each 30min drug application; corresponding to 50 trials/epoch). Statistical significance was assessed using two-way RM ANOVA and Sidak’s test, as above.

For most analyses of in vivo visual response magnitude, the change in firing at each light responsive channel (relative to the 5s epoch immediately prior to light stimulation) was normalised according to the largest value obtained for that site under all stimulus conditions tested (i.e. across each intensity/eye/degree of GHT manipulation). Changes in amplitude/sensitivity (Fig. 8), were then assessed by fitting irradiance response relationships to the normalised data using 4-parameter sigmoidal functions, with F-test (GraphPad) used to determine whether the fitted parameters differed significantly between pre and post-muscimol application for each stimulus condition. Where this analysis indicated significant differences in the irradiance response relationship pre- vs. post-muscimol for a given condition, we renormalised the two curves to have the same maximal value and repeated the F-test (to specifically assess whether there were differences in sensitivity). For assessment of changes in visual response amplitude as a function of GHT stimulation frequency (Fig. 9F), data were normalised separately for each eye and analysed by two-way RM ANOVA. Equivalent data for LGN recordings are analysed and presented in raw (not normalised) form.

Analysis of circadian rhythmicity

395

396397398399400401402403404405

406407408409410411412

413414415416417418419420

421422423424425426427428429430431432433

434

For analysis of circadian rhythmicity in long-term hypothalamic recordings (data presented in Fig. 3), data from each channel was considered to exhibit circadian variation when it could be better fit by a sinusoidal function (constrained to have a period between 20 and 28h) vs. a first order polynomial (calculated using custom MATLAB routines). Data from a small number of channels where we did not detect multiunit activity (n=37/295 from 5 slices; primarily channels located in the optic tract or over the 3rd ventricle) were excluded from this analysis. Of the remaining channels, identified best fit functions described the data well (median±SD of Pearson’s r values = 0.81±0.16, n=258). For those channels considered to exhibit circadian variation (median±SD of Pearson’s r values for sinusoidal fit= 0.83±0.15, n=168) we next calculated the timing of peak firing. For the analysis presented in the manuscript, multiunit time series were binned (60 s) and smoothed (2 h boxcar filter), with the ZT of bin exhibiting maximal values taken as the timing of the peak.

The phase and co-ordination of the daily peak in electrical activity was then calculated by Rayleigh analysis using custom MATLAB scripts as described previously (Walmsley et al., 2015). Comparable analysis performed using peak ZT’s estimated directly from the sinusoidal fits produced essentially identical results. Differences in clustering strength and phasing were assessed by first linearising the data such that, within each dataset, each observation was no more than 12h displaced from the overall circular median (i.e. subtracting 24h from those observations that occurred >12h later and adding 24h to those observations occurring >12h earlier than the median). We then applied conventional analysis to identify differences in variance (Browne-Forsythe’s test) and median (Mann-Witney U-test).

For other analysis of circadian variation presented here (Fig. 6), spontaneous or evoked responses at each responding recording site within the region of interest were 1 h-binned and then normalised according to the peak value observed across all bins. We then tested whether the resulting normalised population datasets were better fit by a sinusoidal function (with 24h periodicity) or a first order polynomial (via F-test; GraphPad). Where data were better fit by a sinusoidal function we rebinned/normalised the data into 6h blocks and performed one-way RM ANOVA with Tukey’s post hoc tests.

Results

Generation and validation of a functional in vitro model of the extended circadian system

In order to better understand the physiological roles of GHT signals in modulating SCN responses to retinal input, we first set out to develop an in vitro model which preserved functional connectivity across the extended circadian system. Since projections between the SCN, IGL/vLGN and PON all run via the optic tract (OT; Watts et al., 1987; Mikkelsen, 1992; Kalsbeek et al., 1993; Morin et al., 1994; Horvath, 1998; Moore et al., 2000), we initially aimed to determine what angle and thickness of brain section would be required to preserve OT-projections to all of these nuclei largely intact. To achieve this aim, we exploited the fact that M1-type ipRGC projections specifically identify nuclei of the extended circadian system in the mouse brain (Hattar et al., 2006). Using a selective reporter mouse line, Opn4tau-LacZ, we were thus able to readily visualise retinal projections to all three nuclei in X-gal stained coronal serial sections (Fig. 1A). After aligning the resulting sections, we then generated a 3-dimensional map of M1-ipRGC central projections. Based on a flattened sagittal plane view of

435436437438439440441442443444445

446447448449450451452453454

455456457458459460461

462

463

464

465466467468469470471472473474475

this projection map (Fig. 1B), we determined that a 600 µm thick brain section cut at 30° off the coronal plane would preserve OT projections and large portions of the SCN, IGL/vLGN and PON intact (for coronal plane reconstruction of the resulting slice see Fig. 1C).

Using the results of this modelling we next prepared live brain sections from Opn4+/tau-LacZ mice, with the aid of a custom matrix that allowed us to accurately block the brain at the required angle for subsequent sectioning. From initial light microscopic visualisation of the resulting slices, we could reliably (>90% of slices) observe a continuous OT extending far beyond the thalamus and easily identifiable SCN, IGL/vLGN and pretectal regions (Fig. 2A), subsequently confirmed by post hoc X-gal staining (Fig. 2B). We next set out to confirm that optic tract connectivity across these visual nuclei remained functional in our slice preparation via pMEA recording.

As expected, electrical stimulation of the optic chiasm (OC) in these angled slices reliably evoked spike firing at subsets of recording sites located in the pretectum (Fig. 2C,D; n=21 responding sites identified in 3 slices), visual thalamus (Fig. 2C,E; n=45 sites responding from 6 slices) and SCN region (Fig. 2C,F; n=68 sites from 2 slices). In order to confirm that the OC-evoked changes in spike rate we recorded represented genuine synaptically-driven responses, we then bath applied ionotropic glutamate receptor antagonists (CNQX, 20 µM and D-AP5, 50 µM). This manipulation completely and reversibly abolished the evoked responses at all responding channels in the pretectum, thalamus and SCN/peri-SCN region (Fig. 2G-I; two-way RM ANOVA: Time x Treatment effects all P<0.001). Together, these data confirm that our angled slice preparations retain a functional OT projection to all major subcortical visual nuclei.

Before using this model to investigate the impact of GHT signalling in more detail, however, we first aimed to confirm that the relatively thick (600 µm) slices required would remain viable for long enough in vitro to facilitate analysis of circadian variation in circadian network function. In fact, we found that slice viability could be routinely maintained for longer than 24 h in vitro (Fig. 3A). Hence, of 144 recording sites (from 5 slices) located within the SCN, 100 (~70%) displayed evidence of circadian rhythmicity (Fig. 3B,C; see methods for analysis details). Previous work indicates that ex vivo SCN firing rate rhythms (recorded using a wide variety of different methodologies) reliably report phasing of the SCN in vivo and are independent of time of preparation, with peak firing consistently occurring during the mid-late projected day (e.g. vanderLeest et al., 2009; see also Brown & Piggins, 2007 for a detailed discussion of this point). Consistent with this earlier work, we found that the timing of peak multiunit activity at SCN channels identified as rhythmic was strongly clustered around the mid-late projected day (Fig. 3D).

As expected for a major target of SCN output (Abrahamson & Moore, 2001), we also identified electrode sites located just dorsal to the SCN (in the vSPZ) exhibiting evidence of circadian variation in firing activity. Among this group, however, the proportion of sites identified as rhythmic was significantly lower than for SCN located electrodes (Fig. 3C; n=61/114 vSPZ recording sites; Fisher’s exact test P<0.01). Moreover, the timing of peak firing for rhythmic vSPZ channels was significantly less well clustered and had a median phase significantly later than observed in the SCN (Fig. 3D; P<0.01 and P<0.001 respectively; see methods for details of analysis).

To further evaluate the viability of these angled slices, at the end of these long term recordings (>26h after slice preparation) we bath applied NMDA (20μM, 5min application), a manipulation which should drive increases in firing in most SCN neurons (~80% of cells in mice of a similar strain

476477478

479480481482483484485

486487488489490491492493494495

496497498499500501502503504505506507

508509510511512513514

515516517

background; Cutler et al., 2003). As expected, this treatment drove robust changes in firing (see methods) in >70% of channels located across the SCN and vSPZ (Fig. 3E,F; n=180/258 channels, no significant difference in responses between SCN and vSPZ; Fisher’s exact test P=0.22). Collectively then, these data establish that our preparation remains viable for greater than 1 circadian cycle in vitro.

Impact of GHT signals on spontaneous SCN activity

Since our angled slice preparations retain intact OT projections across the extended circadian system, we expected that connections between the SCN, IGL/vLGN and PON (which follow the same trajectory) should be similarly preserved in this model. Here we specifically focused on the GHT, although in pilot experiments (6 slices, data not shown) we were also able to readily detect communication between the IGL/vLGN and PON.

To test for the retention of a functional GHT, we next electrically stimulated the IGL/vLGN region via concentric electrodes while recording from the SCN and vSPZ (Fig. 4A). Based on the predominately GABAergic nature of IGL/vLGN neurons (Moore & Card, 1994; Morin & Blanchard, 2001), to maximise our chances of observing inhibitory SCN responses to GHT stimulation, we performed these recordings across the mid-late projected day when spontaneous SCN firing is highest (ZT 4-11). In line with data obtained following in vivo stimulation of the IGL/vLGN (Roig et al., 1997; González & Dyball, 2006), we found that single pulse stimulation of the GHT could evoke either inhibitory (Fig. 4B) or excitatory responses (Fig. 4C) across individual electrode sites within the SCN (n=21 and n=27 sites respectively from 12 slices).

Of note, we observed differences in the latency of increases vs. decreases in SCN firing following thalamic stimulation, with inhibitions typically displaying longer and more variable latency relative to the usually faster excitatory responses (Fig. 4D; median±SD time to peak : 53±32 ms vs. 18±10 ms respectively; Mann-Whitney test: P<0.001). These values represent time to peak response rather than response onset (which is not always easy to accurately assess for inhibitory responses) and are equivalent those reported by previous in vivo studies (Roig et al., 1997; González & Dyball, 2006).

Since we observed a relatively low proportion of sites showing inhibitory responses to GHT stimulation in the experiments above, in some cases, we also evaluated the possibility that single pulse stimulation was simply insufficient to evoke a measureable inhibition in SCN firing. We found no evidence, however, that trains of GHT stimulation (5 stimuli at 20 or 40 Hz), designed to approximate light-driven burst firing seen in the visual thalamus (Brown et al., 2010; Howarth et al., 2014), resulted in a greater magnitude or proportion of sites exhibiting inhibitory responses (Fig. 4E; n=12 sites; one-way ANOVA; P>0.05).

In keeping with previous in vivo data then, thalamic stimulation in our angled slice preparation evokes a mixture of excitatory and inhibitory responses. While the reported excitatory actions of GABA within the SCN (Wagner et al., 1997; Choi et al., 2008; Irwin & Allen, 2009; Farajnia et al., 2014) could, in principle, provide an explanation for the relatively high number of electrode sites showing excitatory responses to thalamic stimulation, our subsequent experiments rule this out as a primary explanation. Hence, where tested (n=4 recording sites), excitatory responses were completely abolished by bath application of ionotropic glutamate receptor blockers (Fig. 4C,G). By contrast, whereas inhibitory responses were reliably blocked by the GABAA receptor antagonist (+)-

518519520521522

523

524525526527528

529530531532533534535536537

538539540541542543

544545546547548549550

551552553554555556557558

bicuculline (BIC; 20µM), excitatory responses were consistently unaffected (Fig. 4F-H; n= 5 and 23 sites tested respectively).

While the presence of a very small number of glutamatergic cells in the mouse IGL-region (experiment: 73818754; (Lein et al., 2007)) provides one possible origin for the excitatory responses we observed above, it is important to also note that many of the ipRGCs innervating the IGL send axon collaterals to the SCN (Pickard, 1985; Morin et al., 2003). Accordingly, we think it more likely that antidromic activation of this ipRGC population accounts for the excitatory responses we observed in some SCN neurons, as described previously for increased SCN Fos-like immunoreactivity following electrical stimulation of the thalamus (Treep et al., 1995).

To circumvent this issue, we next turned to an optogenetics-based approach to selectively activate GABAergic IGL/vLGN neurons in a temporally controlled manner. To this end, we employed a mouse line (GAD2-cre; Ai32) where Cre-recombinase specifically directs ChR2/EYFP expression to GABAergic (GAD2-expressing) neurons (Taniguchi et al., 2011). As expected, the antibody enhanced fluorescent signal was highly enriched throughout the IGL/vLGN region (Fig. 5A,B) and other predominantly GABAergic central nuclei (data not shown) in GAD2-Cre; Ai32 mice, indicating robust expression of the ChR2/EYFP. Consistent with this anatomical localisation, targeted optical stimulation (Fig. 5B; 465 nm LED; ~800mW/mm2 at fibre tip, 10 ms flash) over the IGL/vLGN (Fig. 5C-E) evoked fast and reliable bursts of spike firing from neurons across the stimulated region.

Having established that our approach allows us to robustly activate IGL/vLGN neurons, we next evaluated the impact of GABAergic GHT projection neuron activity on spontaneous firing within the SCN region (Fig. 5F). In line with our data above, while in no case did we observe any excitatory responses in the SCN following selective stimulation of GABAergic GHT cells, we did find a small proportion of sites (n=12 sites from 4 slices) that displayed statistically significant decreases in firing (Fig. 5G,H; see methods). The magnitude and timecourse of the resulting inhibitory responses were broadly similar to those following electrical stimulation (mean±SEM inhibition: 43.8±5.5%; time to peak, median±SD: 60±25 ms) and, where tested (n=6 sites), these were abolished by the bath application of 20 µM BIC (Fig. 5I; paired t-test P<0.01), consistent with their GABAergic origin. Together then, our data suggest that, while GABAergic GHT projections are clearly intact in our angled slice preparation, they appear to have, at most, quite modest effects on spontaneous firing in the SCN.

Impact of GHT signals on SCN responses to optic input

Insofar as our ability to detect inhibitory GHT-driven responses relies on the recipient cell to be firing action potentials, we speculated that the modest effects observed above reflected the fact that the primary role of this projection is to modulate responses to excitatory input from the retina. Indeed, anatomical studies suggest that GHT projections are concentrated within retinorecipient portions of the SCN (Mikkelsen, 1990; Moore et al., 2000; Abrahamson & Moore, 2001). To establish the influence of GHT signals on retinohypothalamic signalling, we next monitored SCN responses to electrical stimulation of the OC (simulating retinal input) concurrent with or following optical activation of the GABAergic GHT projection (as in Fig. 5F). Given the relatively long latency of GHT-driven inhibitory responses, we thus expected that activation of this pathway would only be able to modulate SCN responses if this occurred before (but not synchronous with) electrical stimulation of the OC. Moreover, to evaluate the possibility that the functional impact of GHT signals on the OC-

559560

561562563564565566567

568569570571572573574575576

577578579580581582583584585586587588

589

590591592593594595596597598599600

evoked response might change across the circadian day, we ran this protocol (1 OC pulse/min) over a full 24 h recording epoch (Fig. 6A).

We first calculated, for each independent recording site within the SCN, the average response to the several hundred stimulus presentations across the full 24 h recording epoch. We found that, in fact, the majority (n=45/71 sites from 3 slices) of hypothalamic recording sites at which we observed OC-evoked spiking also showed a statistically significant reduction in this response when GHT stimulation occurred prior to OC stimulation (see methods). The apparent magnitude of the effect at these ‘GHT-responsive’ sites was relatively modest (~20% reduction) but long-lasting, being evident for all tested delays between 50 - 500 ms (Fig. 6B; two-way RM ANOVA with Sidak’s post-tests). Consistent with our findings above, very few of these sites showed detectable changes in spontaneous firing following GHT stimulation (n=4/45 sites passing criteria for responsiveness, see methods for details). Moreover, in line with anatomical studies indicating stronger GHT input to the ipsilateral vs. contralateral SCN (Pickard, 1982; Mikkelsen, 1990; Moore et al., 2000), we also found that the magnitude of the GHT-driven inhibition of OC-evoked responses (but not the proportion of GHT-responsive sites) was significantly higher for sites located ipsilateral to the stimulated GHT (Fig. 6C; t-test, P<0.05).

We also sought to determine the extent to which the multiunit responses described above reflected the activity of individual units within the SCN. To this end we performed PCA-based spike sorting on data from channels were we observed OC-evoked changes in firing. From these 71 channels we were able to isolate 23 single units that exhibited clear excitatory responses to OC-stimulation. More than half of these units (n=12/23) exhibited a significant GHT-driven modulation in the amplitude of OC-evoked responses, a proportion that was not statistically different from the proportion of channels identified as GHT responsive (Fishers exact test, P=0.24). Importantly, the majority of these GHT-responsive units (n=10/12) were isolated from electrode sites where multiunit data was similarly modulated by GHT input, with only a few instances where single units isolated from GHT-responsive channels lacked a significant modulation (n=3 units from 13 channels) or vice versa (n=2 units from 10 channels). Moreover, the magnitude of the GHT-driven modulations in the responding single units were not significantly different from those we report for multiunit data (analysis as in Fig 6B; two-way RM ANOVA; P=0.47 for population and P=0.43 for interactionxGHT stimulation latency). In sum, these data indicate that our analysis based on multiunit responses does not lead us to substantially underestimate the impact of GHT-inputs on individual SCN neurons.

We next then evaluated the extent to which these GHT-driven modulations of hypothalamic responses to optic input varied across the circadian cycle. As expected, given the nocturnal increase in SCN sensitivity to light (Meijer et al., 1996; Meijer et al., 1998; Brown et al., 2011; van Oosterhout et al., 2012), OC-evoked multiunit responses themselves exhibited a very pronounced circadian variation with peak activity occurring during the early projected night, both following synchronous activation of the GHT and following GHT pre-stimulation (Fig. 6A,D,E). Importantly, however, we could detect clear evidence of a circadian variation in the functional impact of GHT signals (Fig. 6F; one-way RM ANOVA of 6 h binned data; P<0.001): thus, the percentage reduction in OC-evoked responses was ~two-fold greater during the projected day vs. the early-mid projected night (Fig. 6F).Based on these data, we next asked whether this variation in the functional impact of GHT stimulation reflected a change in the output/excitability of GABAergic neurons in the IGL/vLGN region, or might instead reflect an SCN intrinsic mechanism. To this end, we directly recorded OC-

601602

603604605606607608609610611612613614615616

617618619620621622623624625626627628629630631

632633634635636637638639640641642643

and optically-evoked responses from the IGL/vLGN region over a 24 h recording epoch (Fig. 6G,H; n=34 sites from 3 slices). While, on average, we observed a slight decline in optically and electrically evoked activity towards the end of these recordings (Fig. 6H), we saw no clear evidence for any pronounced circadian variation in either spontaneous or evoked activity across the IGL/vLGN (see methods for analysis details). In summary then, we infer that daily variation in the functional impact of GHT signals on SCN responses involves a mechanism local to the SCN itself.

Focusing on the projected day, where GHT signals exert the most substantial effects on SCN response to OT input, we next sought to better understand the properties and mechanistic basis of this influence. Hence, although by definition this mechanism is likely to involve GABA, many IGL cells co-express NPY and/or other neuromodulators (Moore & Card, 1994; Morin & Blanchard, 1995; Morin & Blanchard, 2001). To this end, employing similar approaches to those described above, we investigated OC-evoked responses in the SCN following optical stimulation of GHT GABAergic cells at variable latencies, in the presence and absence of GABA receptor blocking drugs (Fig. 7A; data analysed by two-way RM ANOVA).

As before, we observed a very long-lasting inhibition of OC-evoked responses at many sites (n=44/81 OC-responsive sites from 3 slices) following optical stimulation of the IGL/vLGN. This effect was maximal at GHT pre-stimulation latencies of 50 - 100 ms and began to diminish at longer latencies (Fig. 7B; 1s timepoint P<0.01 vs. 50 and 100ms; Sidak’s post-test). Of note, blockade of GABA A

receptors (20 µM BIC) significantly attenuated, but did not completely block, the inhibitory influence of GHT pre-stimulation, the effect being most pronounced for short latency inhibitions (Fig. 7B). By contrast, when we then co-applied BIC with the GABAB receptor antagonist CGP55845 (3 µM), GHT-driven inhibitory responses were almost completely abolished at all latencies (Fig. 7B).

We also noted that blocking GABA receptor signalling produced a general enhancement in the magnitude of OC-driven responses at these GHT-responsive channels (mean±SEM: 1.0±0.1, 1.6±0.2 and 2.5±0.3 evoked spikes at 0ms delay for baseline, BIC and BIC+CGP respectively; one-way RM ANOVA with Sidak’s multiple comparison test, P<0.001 for baseline vs. BIC and BIC vs. BIC+CGP). Similar results were also obtained for GHT-unresponsive channels (mean±SEM: 1.2±0.1, 2.0±0.2 and 2.9±0.3 evoked spikes for baseline, BIC and BIC+CGP respectively, P<0.001 for baseline vs. BIC and BIC vs. BIC+CGP).

Collectively, our data above indicate that GABA signalling is a major regulator of SCN responses to retinal input, at least a portion of which derives from the activity of afferents from the GHT. Indeed, the modulatory activity of the GHT revealed here seems to arise almost entirely via GABA signalling, with GABAA receptors mediating initial components of the response and GABAB receptors driving the longer lasting elements. We subsequently confirmed this view with an additional experiment focusing on early components (<100ms) of the GHT-mediated inhibition (Fig. 7C; n=14 GHT responsive sites from 3 slices). Consistent with the 20 – 30 ms latency reported by antidromic activation of IGL cells from the SCN (Zhang & Rusak, 1989; Blasiak & Lewandowski, 2013), we observed statistically significant reductions in OC-evoked spiking with GHT-prestimulation latencies as low as 25 ms (two-way RM ANOVA with Sidak’s post-tests). Importantly, the inhibitory effects of GHT-stimulation at all latencies tested here were almost completely blocked by BIC alone at all delays tested (Fig. 7C).

644645646647648649

650651652653654655656657

658659660661662663664665

666667668669670671672

673674675676677678679680681682683684

Manipulating GHT signalling modulates specific features of SCN-driven light responses

Collectively, our data above suggest that GABAergic cells contributing to the GHT exert very little effect on spontaneous activity across the SCN but play a more substantial role in gating SCN responses to optic input. Based on these findings we next sought to understand the extent to which light-driven changes in GHT output might modulate SCN responses to visual signals. Superficially, one might assume that because the route for any potential light-driven GHT output to reach the SCN is longer than that for direct retinal input to this region, these GHT-derived inhibitory signals would be too slow to modulate SCN visual responses. On the other hand, visual stimuli relevant to the circadian system would be expected to produce sustained firing in RGCs (e.g. Wong, 2012) which, given the very persistent nature of the GHT-driven inhibition (Fig. 7), could quite substantially modulate on-going SCN light responses. To determine then how/whether GHT activity contributes to SCN visual signalling, we monitored light responses directly from the SCN via multi-electrode recordings in anaesthetised mice, before and after unilateral inhibition of spiking in the visual thalamus.

To achieve robust inhibition of GHT signalling we performed relatively large infusions of the GABA A

receptor agonist muscimol (1 mM; 2-3 μl injection at 1 μl/min) directly into the visual thalamus while simultaneously monitoring multiunit firing via a linear multi-electrode array attached to the drug-delivery cannula (Fig. 8A,B). This manipulation rapidly abolished spontaneous and light-driven spiking activity at recording sites located up to 750 μm dorsal to the tip of the drug delivery cannula (Fig. 8C). Assuming an approximately spherical spread of the drug (as observed for dye used to mark the electrode track, Fig. 8A), we inferred then that this method of drug delivery inhibits spike firing across the vast majority of the visual thalamus (including that of cells contributing to the GHT).

We next turned our attention to the impact of GHT inhibition on light-evoked activity within the SCN. Although our in vitro data suggested that GHT-driven signals exert a stronger inhibitory effect during the projected day, we here instead chose to perform these in vivo experiments during the early-mid projected night (ZT 12-17). Hence, across this epoch, light-driven SCN activity is more readily quantifiable and is associated with more robust circadian phase resetting effects (Meijer et al., 1996; Meijer et al., 1998; Brown et al., 2011; van Oosterhout et al., 2012). In fact, despite performing these recordings at a time when the GHT exerts its most modest effects, we still observed pronounced changes in multiunit SCN light responses following inhibition of the ipsilateral GHT projection (Fig. 8D,G). In particular, there were substantially enhanced responses when light pulses (405 nm, 5 s, 11.4 - 15.4 log effective photons/cm2/s) were applied to just the contralateral or to both eyes (Fig. 8G; F-test for identical irradiance response relationship pre vs. post muscimol, P=0.001 and P=0.002 respectively; n=63 light responsive sites from 5 mice). This effect stemmed primarily from an increase in maximal response amplitude rather than sensitivity, since in both cases, sigmoidal fit parameters for pre- vs. post-GHT inhibition data were identical when we normalised those datasets to exhibit a common maximal value (F-test; P=0.999 and P=0.984; data not shown). We did not, however, detect any statistically significant changes in the irradiance response relationship for stimulation of the ipsilateral retina following muscimol application (Fig. 8G; F-test; P=0.118). Together then, these data suggest that the (inhibitory) GHT projection is primarily activated by crossed retinal inputs.

685

686

687688689690691692693694695696697698699

700701702703704705706707

708709710711712713714715716717718719720721722723724725726

In additional sets of experiments, we evaluated the impact of inhibiting the contralateral GHT projection (Fig. 8E,H) and assessed the possibility that our data above may have been influenced by spontaneous/circadian changes in SCN photic sensitivity over the tested recording epochs (Fig. 8F,I; identical protocol to the above but with no muscimol injection). Importantly, when we did not infuse muscimol, there was no significant change in responsiveness to any of the stimuli we employed over the tested recording epochs (Fig. 8I; F-test; P=0.250, P=0.101 and P=0.994 respectively for changes in irradiance response relationship to contralateral, ipsilateral and binocular light steps; n=127 light responsive sites from 6 mice). Despite a trend towards increased SCN responses to stimulation of the ipsilateral retina, there was a similar lack of statistically significant differences in responsiveness following inhibition of the contralateral GHT (Fig. 8F, H; F-test; P=0.481, P=0.06, P=0.318 respectively for contralateral, ipsilateral and binocular light steps; n=33 light responsive sites from 3 mice). These later data thus broadly align with findings from anatomical studies (Pickard, 1982; Mikkelsen, 1990; Moore et al., 2000) and our in vitro experiments (Fig. 6C) indicating that the contralateral GHT projection is weaker than that arising from the ipsilateral IGL/vLGN.

We also investigated whether inhibition of GHT signalling changed spontaneous activity within the SCN in any way. Although we found statistically significant increases in spontaneous firing under all conditions tested in Fig. 8D-F (mean±SEM: 123±38 vs. 45±15 and 9±3% above baseline for ipsilateral, contralateral or no GHT inhibition respectively, paired t-tests, all P<0.05), the magnitude of this effect was substantially greater following inhibition of the contralateral GHT (one-way ANOVA with Dunnett’s post-hoc test, P<0.001 vs. no inhibition). It is also worth noting here that the average change in spontaneous population activity following inhibition of the ipsilateral GHT (6.1±2.4 spikes/s) was significantly smaller than the change in light-evoked firing evoked by this manipulation (19.6±2.4 spikes/s for 15.4 log photons/cm2/s presented to both eyes; paired t-test, P<0.001).

Collectively, the data above suggest that, in vivo, inhibitory input from the ipsilateral GHT is active under low light conditions and is further increased by the activity of crossed retinal projections. This conclusion, that GHT input is primarily stimulated by crossed retinal inputs, derives from the lack of overt changes in the irradiance response relationship for SCN responses to ipsilateral retinal stimulation following thalamic inhibition (Fig. 8G). However, since we have previously shown that many SCN neurons preferentially receive input from just one eye (Walmsley & Brown, 2015), an alternate explanation for these findings is that GHT inputs preferentially converge on SCN neurons that receive contralateral retinal input.

To evaluate this latter possibility, we next asked whether artificially stimulating GHT activity (by optogenetic stimulation of GABAergic neurons in the IGL/vLGN region) could adjust SCN responses to contralateral and/or ipsilateral retinal stimulation. Accordingly, we implanted recording probes with attached optical fibres (‘optrodes’; see methods for further details) into the IGL region of GAD2-Cre;Ai32 mice, allowing us to illuminate large portions of the IGL/vLGN and to monitor the resulting activity (Fig. 9A-C). Using this approach, we then evaluated SCN multiunit responses (Fig. 9D) to retinal stimulation (5s steps, 15.4 log effective photons/cm2/s) in the absence or presence of trains of optogenetic stimulation (10ms flashes at 5-20Hz for 5s).

Since there is a delay of 30-40ms before visually evoked activity becomes first detectable within the SCN and LGN (van Diepen et al., 2013; Howarth et al., 2014; Walmsley & Brown, 2015), we reasoned that the first pulse of these stimulus trains should be sufficient to increase the inhibitory influence of

727728729730731732733734735736737738739740

741742743744745746747748749

750751752753754755756757

758759760761762763764765

766767768

the GHT prior to the arrival of any retinal input. Thus, assuming that GHT inputs do not preferentially converge on SCN neurons receiving contralateral retinal input, we expected that initial components of the SCN visual response should be inhibited regardless of which eye we stimulated. By contrast, we reasoned that the effect of continued optogenetic stimulation on SCN responses should be much greater under conditions when GHT activity is normally low (i.e. during ipsilateral visual stimulation).

Consistent with this reasoning, we found that the earliest components of the light-evoked multiunit firing (0-250ms after visual stimulation), were reliably reduced by concomitant optogenetic stimulation of the GHT regardless of which eye or stimulation frequency we used (Fig. 9E,F; n=72 light responsive SCN sites from 6 mice; two-way RM ANOVA with Sidak’s post-tests). These data thus confirm that the GHT is capable of modulating SCN responses to input from either retina. We also found that, while optogenetic stimulation at 5 and 10Hz could also suppress later components (3-5s after stimulus onset) of the SCN response to ipsilateral retinal inputs, there were no statistically significant reductions in the responses to visual stimuli that also involved the contralateral retina at any frequency tested (Fig. 9E,G; two-way RM ANOVA with Sidak’s post-tests).

We interpret these latter data as reflecting the fact that the GHT is already much more active during contralateral retinal stimulation and, as such, further increasing this activity with optogenetic stimuli has negligible effects on the inhibitory signals sent to the SCN. Analysis of optogenetically-driven changes in thalamic activity broadly supported this hypothesis (Fig. 9H,I). Here we specifically focused on recording sites in the IGL/vLGN region where we observed both visual and optogenetically evoked responses (n=41 sites from 6 mice). Since only a subset of IGL/vLGN neurons contribute to the GHT, the population activity observed here is not necessarily directly reflective of the signals being sent to the SCN. Nonetheless, while all sites exhibited very pronounced light-induced firing during contralateral retinal stimulation, activity during stimulation of the ipsilateral retina was negligible (and in fact, on average, slightly suppressed; Fig. 9H,I).

Moreover, while optogenetic stimuli could significantly elevate thalamic activity spike firing during stimulation of either or both retinas (Fig. 9H; two-way RM ANOVA with Sidak’s post-tests), the magnitude of the resulting changes were always substantially smaller than the nominal stimulation frequencies tested (mean±SEM = 5.1±0.4 spikes/s for 20Hz stimulation during binocular illumination – the largest change we observed). We assume these observations reflect the fact that our optogenetic stimuli also substantially increase local inhibition within the thalamus, limiting the amount that we are able to increase overall firing rates. Regardless, the net result was that optogenetic stimulation produced only a marginal increase in the already high activity when paired with illumination of the contralateral retina but resulted in a qualitative change in (the much lower) thalamic activity when paired with stimulation of just the ipsilateral retina (Fig. 9H,I).

As expected based on our in vitro data, optogenetic stimulation in the absence of concomitant retinal stimulation did not significantly alter SCN firing rates at any frequency tested (two-way RM ANOVA with Sidak’s post-test; all P >0.05; not shown) despite producing significant, albeit modest changes in LGN firing activity (two-way RM ANOVA with Sidak’s post-test; P >0.001 at all frequencies; maximal change in firing = 1.5±0.3 spikes/s for 20Hz stimulation).

In summary then, the data above support our view that the primary action of the GHT is to modulate SCN responses to on-going retinal input and establish that, while the GHT is capable of modulating

769770771772773

774775776777778779780781782

783784785786787788789790791792

793794795796797798799800801802

803804805806807

808809

SCN responses to input from either eye, the primary sensory signal driving GHT output is light detected by the contralateral retina.

Discussion

Here we show that GHT neurons play an important role in gating SCN neuronal responses to optic input and actively modulates specific features of the SCN light response. Although existing data imply that GHT activity can oppose light-induced phase resetting (Yannielli & Harrington, 2000; Lall & Biello, 2002, 2003; Yannielli et al., 2004), to our knowledge, this is the first report that the GHT directly modulates acute electrophysiological responses to light/retinal input. In addition, we establish a novel in vitro model for studying network processing within the circadian system. Using that model we show that GABA released from the GHT specifically modulates SCN responses to retinal input with little effect on spontaneous activity. Moreover, we show that this modulatory action exhibits a pronounced circadian variation. Collectively, then, our data (summarised in Fig. 10) provide a mechanistic basis by which the GHT can shape temporal variations in the SCN response to retinal input and influence both circadian photoentrainment as well as much more rapid light-dependent changes in downstream physiology (e.g. hormone secretion).

Existing attempts to define how GHT signals act within the central clock have been hampered by the lack of an in vitro model that allows for experimental manipulation of this pathway. As such, in vitro studies have focused on pharmacological manipulation of NPY signalling as a proxy for GHT activity (Shibata & Moore, 1993; Biello et al., 1997; Gribkoff et al., 1998; Yannielli & Harrington, 2000). Although such information is useful, this approach provides at best an incomplete picture of the influence of the GHT. Hence, not all GHT cells produce NPY (Moore & Card, 1994; Morin & Blanchard, 2001), nor are the circumstances which trigger NPY release well understood. Similarly, while studies using pharmacological manipulation of GABA signalling provide data consistent with the idea that GHT-derived GABA release may modulate circadian photic phase resetting (Gillespie et al., 1996; Gillespie et al., 1997; Novak & Albers, 2004), such approaches are unable to distinguish the specific role of the GHT vs. the many other possible sources of GABAergic modulation. Our angled slice model now provides a direct way to assess the mechanisms by which GHT signals influence SCN activity.

Here we employed two approaches to manipulate GHT signalling: electrical stimulation and targeted optogenetic activation of GAD2-positive neurons. The former has the advantage that it should activate any cell contributing to the GHT, regardless of neurochemical phenotype. The downside of this approach, however, is that it carries the risk of indirectly triggering retinal input to the SCN, due to antidromic activation of ipRGCs that send axon collaterals to the hypothalamus (Pickard, 1985; Treep et al., 1995; Morin et al., 2003). We believe that the modest number of excitatory SCN responses we observed following electrical stimulation of the GHT were likely driven by this route. Hence, whereas ipRGC inputs to the SCN are glutamatergic (Gompf et al., 2015), almost all IGL/vLGN neurons are GABAergic (Moore & Card, 1994; Harrington, 1997; Lein et al., 2007) and the excitatory responses observed here were blocked by glutamate, but not GABA receptor antagonists.

As such, here we instead used optogenetics to selectively activate GABAergic GHT projections. In line with the results obtained using electrical stimulation, changes in spontaneous SCN firing were very

810811

812

813

814815816817818819820821822823824825

826827828829830831832833834835836837838

839840841842843844845846847848

849850

rarely detectable following optogenetic activation of the GABAergic GHT projection. By contrast, an influence of this pathway was readily apparent as a long-lasting inhibition of evoked spiking (persisting for at least 1 s) at the majority of hypothalamic sites (~60%) responding to stimulation of retinal afferents. In line with this in vitro data, optogenetic activation of the GHT in vivo could modulate visual responses in the SCN but did not result in any detectable changes in spontaneous firing. Similarly, pharmacological inhibition of the GHT in vivo produced much larger changes in light-evoked vs. spontaneous SCN activity. Collectively then, these observations indicate that our in vitro preparation preserves a substantial proportion of the GHT intact and suggest that the principle action of the GHT, at least under the conditions studied here, is to modulate SCN neuronal responses to excitatory input.

A number of features of our experimental design here are worth further discussion. Firstly, in order to experimentally demonstrate GHT-driven modulatory effects in vitro, we stimulated the GHT before activating retinal afferents. Here we find that GHT-driven inhibition of OT-evoked responses becomes detectable within at least 25 ms, consistent with what one would predict from the latency of antidromically evoked IGL spikes following hypothalamic stimulation (Zhang & Rusak, 1989; Blasiak & Lewandowski, 2013). In the intact animal, RGC impulses will of course actually arrive in the SCN slightly earlier than in the IGL/vLGN (latencies for visual responses ~30 vs. 40 ms: present study; van Diepen et al., 2013; Howarth et al., 2014; Walmsley & Brown, 2015). As such, the SCN response to a single RGC impulse should have terminated before any visually driven GHT-dependent signal arrives. Importantly, however, visual stimuli sufficient to trigger meaningful circadian responses must necessarily generate prolonged trains of RGC impulses. Indeed, even after many hours of continuous light exposure, ipRGCs continue to fire at around 10 spikes/s (Wong, 2012). Accordingly, during physiologically relevant trains of retinal input to the SCN, the GHT-driven inhibition we demonstrate here is easily long-lasting enough (hundreds of milliseconds) to modulate SCN responses to all but the very earliest components of the SCN response to a visual stimulus (illustrated in schematic form in Fig. 10).

Secondly, most of the analysis presented here derives from multiunit recordings. It is formally possible that by combining signals from a heterogeneous population neurons, only some of which receive GHT input, this use of multiunit data leads us to underestimate the impact of GHT signals on the recipient neurons. Analysis of single cell responses form a subset of our recordings indicates that such a possibility is very unlikely to result in us substantially underestimating the impact of GHT input. Hence, there were very few instances where we could observe an effect of GHT stimulation at the multiunit but not single unit level and, among single cells where we detected a significant influence of GHT-input, the magnitude of the resulting responses were very similar to those we report for multiunit data.

By contrast, perhaps a more significant factor that could result in our in vitro data underrepresenting the magnitude of GHT-driven inhibitory input to SCN neurons is that we focused on investigating the effects of single-pulse GHT stimulation. It is formally possible that GHT-driven signals (arising via GABA and/or NPY release) could summate during high frequency stimulation to produce a greater inhibition. This possibility could explain then why we find that in vivo inhibition of GHT activity (discussed in more detail below) produces a greater increase in visual response magnitude (~100%) than one would predict from our in vitro data at an equivalent circadian phase (~25%). Alternatively/additionally, it is worth noting that because the thalamic input to the SCN runs via the

851852853854855856857858859860

861862863864865866867868869870871872873874875876

877878879880881882883884885

886887888889890891892893

OT, the electrical stimulation used in our in vitro studies to activate retinal afferents will likely also directly stimulate GHT terminals, potentially occluding the effect of prior optical stimulation of the GHT.

In line with the idea that electrical stimulation of the OC does itself trigger some baseline level of GHT activity, we and others (Moldavan & Allen, 2013) find that GABA receptor blockade does indeed result in a general increase in the magnitude of OC-evoked responses. On the other hand, the GHT is far from the only source of GABA release in the SCN. Hence, facilitatory effects of GABA receptor blockade on OC-evoked responses do not necessarily confirm an endogenous GHT-contribution of this kind. In this regard, it is noteworthy that the reductions in early visual responses we observe following optical activation of the GHT in vivo (where we expect one optogenetically-driven GHT impulse to reach the SCN before any light-driven GHT signals do) are broadly similar in magnitude to those we report in vitro at an equivalent circadian phase.

On balance then, although our present work leaves these latter issues somewhat unresolved, the inhibitory actions of GHT-stimulation we report in vitro should perhaps be considered a conservative estimate of the extent to which these inputs regulate SCN activity. Despite this caveat, it is abundantly clear from both our in vitro and in vivo data that GHT-derived signals exert a much more robust effect on SCN responses to retinal input than on spontaneous firing. Such an arrangement may, in principle arise simply because the low spontaneous activity of SCN neurons ensures that the functional impact of GHT-dependent inhibitory signals is hard to detect in the absence of a driving input. We think it most unlikely that this constitutes the primary explanation for the apparent lack of effect of GHT-stimulation on spontaneous firing. Indeed the design of these experiments (where we evaluated responses to hundreds of stimulus repeats) was specifically tailored to enable us to detect inhibitory responses even when spontaneous activity was relatively low.

An alternative explanation then is that GHT inputs might be arranged such that they can specifically gate SCN responses to retinal signals without exerting major effects on spontaneous firing. In this regard, NPY cells reportedly form axodendritic contacts with SCN neurons (Pelletier, 1990). Moreover, it has previously been demonstrated that retinal and GABAergic terminals (potentially including those originating from the GHT) are typically found in very close proximity within the SCN, often converging onto the same postsynaptic dendrite (Castel et al., 1993). Since spontaneous SCN firing is primarily generated within the soma, such an arrangement should indeed allow GABA released by this population of GHT cells to preferentially gate responses to excitatory input. In addition, the presence of GABAB receptors almost exclusively located on (nearby) retinal terminals leaves open the possibility that GABA spillover might also presynaptically modulate retinal input (Jiang et al., 1995; Moldavan et al., 2006; Belenky et al., 2008). In any of the above situations, however, the net effect is that the functional impacts of GHT signals are most clearly evident in the face of on-going retinal transmission.

It is worth noting here that pharmacological activation of NPY receptors is able to suppress spontaneous firing in the SCN (Gribkoff et al., 1998). By contrast, we found that inhibitory responses to optogenetic activation of the GHT could be almost entirely blocked by GABA receptor antagonists. It appears then that, under our experimental conditions, NPY released from the GHT does not exert a major effect. In principle, it is formally possible that the approaches we used preferentially activate non-NPY cells. This seems unlikely, however, since almost all GABAergic neurons produce GAD2

894895896

897898899900901902903904905

906907908909910911912913914915916

917918919920921922923924925926927928929

930931932933934935

(Zhao et al., 2013), expression of which is itself very reliably reported by the GAD2-Cre line (Taniguchi et al., 2011). Certainly our data indicate very robust ChR2/EYFP expression in the IGL/vLGN region. Instead then, we suggest that specific patterns of presynaptic terminal activity may be required to evoke endogenous NPY release, with the short bursts of firing evoked by our test stimuli primarily acting via triggering GABA release.

Significantly, here we identify a pronounced rhythm in the functional impact of this GHT-derived GABA release, with ~two-fold greater inhibition of OC-evoked spiking during the mid-late projected day vs. the early projected night. As discussed below then, these observations provide a potential mechanism by which signals from the GHT could contribute to shaping the characteristic temporal gating of the SCN light response/clock resetting (Meijer et al., 1996; Meijer et al., 1998; Brown et al., 2011; van Oosterhout et al., 2012). The effect we observe in vitro is not, however, associated with any appreciable change in responses recorded directly from the IGL/vLGN region. Accordingly, we interpret these data as reflecting a circadian variation in the impact of GHT-derived GABA release at the level of the SCN neurons and/or retinal terminals. In line with this latter view, one previous report does indeed indicate that GABAB receptor mediated presynaptic inhibition of retinohypothalamic transmission is enhanced during the projected day (Moldavan & Allen, 2013). We should note, however, that it is formally possible that GHT-derived GABA release might exhibit circadian variation even though IGL/vLGN firing does not itself change over the circadian day.

In interpreting the functional significance of the in vitro data discussed above, a key area to consider is the nature of the sensory signals that actually activate GHT transmission. Owing to the challenging nature of directly identifying GHT cells in electrophysiological recordings, existing attempts to address this deficit have not provided a clear answer (Harrington & Rusak, 1989; Zhang & Rusak, 1989; Thankachan & Rusak, 2005; Blasiak & Lewandowski, 2013). Although we do not directly investigate this question here, our in vivo experiments employing thalamic inhibition clearly indicate that, during visual stimulation, activity of the ipsilateral GHT reduces SCN population responses specifically to stimuli that target the contralateral retina.

Due to the extensive interconnections among the SCN, IGL/vLGN and pretectum (Morin & Allen, 2006), thalamic inhibition could influence SCN light responses via various routes. Nonetheless, the simplest interpretation of our data is that the inhibitory GABAergic neurons that form the GHT themselves primarily receive ON-excitatory input from crossed retinal projections. Consistent with this view, we rule out one important alternative explanation for the above data (that the GHT preferentially impinges on SCN neurons receiving contralateral retinal input; Walmsley & Brown, 2015). Hence, we show that optogenetic activation of the GHT in vivo is capable of reducing early components (first 250ms) of the SCN response to both ipsilateral and contralateral input. We also show that maintained optogenetic stimulation has negligible effects on later components of the SCN response to contralateral retinal stimulation, as expected if endogenous activity of the GHT is already high. By contrast, maintained optogenetic stimulation at 5 and 10Hz produced continued inhibition of responses driven by the ipsilateral retina, consistent with a much lower level of endogenous GHT activity under these conditions.

Regarding interpretation of this latter element of our data, we should note that at a higher optogenetic stimulation frequency (20Hz) this inhibition of on-going ipsilateral retinal-driven activity disappeared. This is despite the fact that our analysis of population activity in the IGL/vLGN region

936937938939940

941942943944945946947948949950951952953

954955956957958959960961

962963964965966967968969970971972973974

975976977

during the same recordings revealed significantly increased spiking activity (of a broadly similar magnitude to that observed at lower frequencies). Since only a subset of neurons in the IGL/vLGN contribute to the GHT, we assume then that the thalamic data we record here is not entirely representative of the activity of GHT neurons. Alternatively, it is formally possible that the specific spiking pattern of GHT afferents is also an important determinant of the overall impact on the SCN. In either case, our general conclusion that GHT signalling is primarily activated by the illumination of the contralateral retina is certainly supported by our observations of light evoked activity in the IGL/vLGN (present study and previous large scale recordings of sensory responses in this region: Howarth et al., 2014), which indicate that excitatory responses to stimulation of the ipsilateral retina are rare (present in only ~25% of cells).

Of note, the enhancement of SCN visual responses we observed following thalamic inhibition showed no signs of diminishing across the duration of the visual stimulations tested here. Again, this is consistent with both the long-lasting effects of GHT activity we observe in vitro and the sustained visual responses driven by ipRGCs which provide a major source of input to the IGL/vLGN (Brown et al., 2010; Wong, 2012). As such, it seems that the GHT modulates the irradiance coding properties of SCN neurons. Based on the relationship between SCN firing and irradiance, the magnitude of this effect is large for relatively bright stimuli (>1014 photons/cm2/s; equivalent to those encountered around sunrise/sunset). Indeed, for contralateral light steps, we found that ~10-fold lower light levels were required to produce SCN population firing responses equivalent to those attained before inhibition of the GHT.

Most significantly, the in vivo recordings discussed above occurred across a time epoch when the GHT was expected to exert its weakest effects. Indeed, assuming a two-fold greater effect of the GHT input during the day (as predicted by our in vitro data), it seems that input from the GHT could be sufficient to account for a substantial component of the circadian modulation in light-driven SCN activity observed previously (3-4-fold variation in magnitude at submaximal intensity; Meijer et al., 1998; Brown et al., 2011; van Oosterhout et al. 2012). An intriguing question then is to what extent does the GHT helps to define the phase response curve for behavioural photoentrainment?

Certainly, based on data from IGL-lesioned hamsters, the GHT is not an absolute requirement for photoentrainment under most circumstances (Harrington & Rusak, 1986; Pickard et al., 1987; Johnson et al., 1989; Maywood et al., 1997; Edelstein & Amir, 1999a, b; Morin & Pace, 2002; Muscat & Morin, 2006; Evans et al., 2012). Nonetheless, we consider it noteworthy that, in mice, behavioural phase resetting responses to light are largest during the first half of the projected night (Daan & Pittendrigh, 1976), a portion of the day when the inhibitory effects of GHT-activation are at their lowest ebb. Insofar as we have previously shown that the magnitude of light-evoked SCN firing is predictive of mouse circadian phase resetting responses (Brown et al., 2011), our data are thus consistent with the possibility that GHT inputs contribute to reducing circadian responses to light around dawn and across the projected day.

In summary, then, our data provide new insight into the contributions of the GHT to visual processing within the SCN. We show that the GHT actively modulates SCN responses to retinal input and demonstrate a circadian variation in the magnitude of this effect which is consistent with an important role in the temporal gating of photic responses, a fundamental property of the circadian system. We also establish a new model system for studying geniculohypothalamic communication,

978979980981982983984985986987

988989990991992993994995996997

998999

10001001100210031004

1005100610071008100910101011101210131014

10151016101710181019

which we expect to have widespread utility for many further investigations of how network interactions across the extended circadian system contribute to daily rhythms in physiology and behaviour.

Additional information

Competing interests

None declared.

Author contributions

T.M.B., L.H. and L.W. designed the experiments, L.H., L.W., A.P., M.H. and T.M.B. collected and analysed the data, T.M.B. and L.H. wrote the manuscript. All authors approved the final version of the manuscript, agree to be accountable for all aspects of the work and all persons designated as authors qualify for authorship. All experiments were performed at the University of Manchester, UK.

Funding

This work was funded by a Biotechnology and Biological Sciences Research Council (BBSRC, UK) David Philips Fellowship to T.M.B. (BB/I017836/1). L.W. was funded by a BBSRC doctoral training grant and strategic skills award.

References

Abrahamson EE & Moore RY. (2001). Suprachiasmatic nucleus in the mouse: retinal innervation, intrinsic organization and efferent projections. Brain Res 916, 172-191.

Belenky MA, Yarom Y & Pickard GE. (2008). Heterogeneous expression of gamma-aminobutyric acid and gamma-aminobutyric acid-associated receptors and transporters in the rat suprachiasmatic nucleus. J Comp Neurol 506, 708-732.

Biello SM, Golombek DA & Harrington ME. (1997). Neuropeptide Y and glutamate block each other's phase shifts in the suprachiasmatic nucleus in vitro. Neuroscience 77, 1049-1057.

Blasiak T & Lewandowski MH. (2013). Differential firing pattern and response to lighting conditions of rat intergeniculate leaflet neurons projecting to suprachiasmatic nucleus or contralateral intergeniculate leaflet. Neuroscience 228, 315-324.

Brown TM, Banks JR & Piggins HD. (2006). A novel suction electrode recording technique for monitoring circadian rhythms in single and multiunit discharge from brain slices. J Neurosci Methods 156, 173-181.

Brown TM, Gias C, Hatori M, Keding SR, Semo M, Coffey PJ, Gigg J, Piggins HD, Panda S & Lucas RJ. (2010). Melanopsin contributions to irradiance coding in the thalamo-cortical visual system. PLoS Biol 8, e1000558.

102010211022

1023

1024

1025

1026

1027102810291030

1031

103210331034

1035

1036

10371038

1039104010411042

104310441045

1046104710481049

1050105110521053

1054105510561057

Brown TM, McLachlan E & Piggins HD. (2008). Angiotensin II regulates the activity of mouse suprachiasmatic nuclei neurons. Neuroscience 154, 839-847.

Brown TM & Piggins HD. (2007). Electrophysiology of the suprachiasmatic circadian clock. Prog Neurobiol 82, 229-255.

Brown TM & Piggins HD. (2009). Spatiotemporal heterogeneity in the electrical activity of suprachiasmatic nuclei neurons and their response to photoperiod. J Biol Rhythms 24, 44-54.

Brown TM, Wynne J, Piggins HD & Lucas RJ. (2011). Multiple hypothalamic cell populations encoding distinct visual information. J Physiol 589, 1173-1194.

Castel M, Belenky M, Cohen S, Ottersen OP & Storm-Mathisen J. (1993). Glutamate-like immunoreactivity in retinal terminals of the mouse suprachiasmatic nucleus. Eur J Neurosci. 5, 368-381.

Chen SK, Badea TC & Hattar S. (2011). Photoentrainment and pupillary light reflex are mediated by distinct populations of ipRGCs. Nature 476, 92-95.

Choi HJ, Lee CJ, Schroeder A, Kim YS, Jung SH, Kim JS, Kim dY, Son EJ, Han HC, Hong SK, Colwell CS & Kim YI. (2008). Excitatory actions of GABA in the suprachiasmatic nucleus. J Neurosci 28, 5450-5459.

Cutler DJ, Haraura M, Reed HE, Shen S, Sheward WJ, Morrison CF, Marston HM, Harmar AJ & Piggins HD. (2003). The mouse VPAC2 receptor confers suprachiasmatic nuclei cellular rhythmicity and responsiveness to vasoactive intestinal polypeptide in vitro. Eur J Neurosci 17, 197-204.

Daan, S & Pittendrigh, C. (1976). A Functional Analysis of Circadian Pacemakers in Nocturnal Rodents: II. The Variability of Phase Response Curves. J Comp Physiol 106, 253-266.

Edelstein K & Amir S. (1999a). The intergeniculate leaflet does not mediate the disruptive effects of constant light on circadian rhythms in the rat. Neuroscience 90, 1093-1101.

Edelstein K & Amir S. (1999b). The role of the intergeniculate leaflet in entrainment of circadian rhythms to a skeleton photoperiod. J Neurosci 19, 372-380.

Evans JA, Carter SN, Freeman DA & Gorman MR. (2012). Dim nighttime illumination alters photoperiodic responses of hamsters through the intergeniculate leaflet and other photic pathways. Neuroscience 202, 300-308.

Farajnia S, van Westering TL, Meijer JH & Michel S. (2014). Seasonal induction of GABAergic excitation in the central mammalian clock. Proc Natl Acad Sci U S A 111, 9627-9632.

105810591060

106110621063

106410651066

106710681069

107010711072107310741075

1076107710781079

1080108110821083

10841085108610871088

108910901091

1092109310941095

109610971098

1099

Gillespie CF, Huhman KL, Babagbemi TO & Albers HE. (1996). Bicuculline increases and muscimol reduces the phase-delaying effects of light and VIP/PHI/GRP in the suprachiasmatic region. J Biol Rhythms 11, 137-144.

Gillespie CF, Mintz EM, Marvel CL, Huhman KL & Albers HE. (1997). GABA(A) and GABA(B) agonists and antagonists alter the phase-shifting effects of light when microinjected into the suprachiasmatic region. Brain Res 759, 181-189.

Gompf HS, Fuller PM, Hattar S, Saper CB & Lu J. (2015). Impaired circadian photosensitivity in mice lacking glutamate transmission from retinal melanopsin cells. J Biol Rhythms 30, 35-41.

González JA & Dyball RE. (2006). Pinealectomy reduces optic nerve but not intergeniculate leaflet input to the suprachiasmatic nucleus at night. J Neuroendocrinol 18, 146-153.

Govardovskii VI, Fyhrquist N, Reuter T, Kuzmin DG & Donner K. (2000). In search of the visual pigment template. Vis Neurosci 17, 509-528.

Green DJ & Gillette R. (1982). Circadian rhythm of firing rate recorded from single cells in the rat suprachiasmatic brain slice. Brain Res 245, 198-200.

Gribkoff VK, Pieschl RL, Wisialowski TA, van den Pol AN & Yocca FD. (1998). Phase shifting of circadian rhythms and depression of neuronal activity in the rat suprachiasmatic nucleus by neuropeptide Y: mediation by different receptor subtypes. J Neurosci 18, 3014-3022.

Groos G & Hendriks J. (1982). Circadian rhythms in electrical discharge of rat suprachiasmatic neurones recorded in vitro. Neurosci Lett 34, 283-288.

Güler AD, Altimus CM, Ecker JL & Hattar S. (2007). Multiple photoreceptors contribute to nonimage-forming visual functions predominantly through melanopsin-containing retinal ganglion cells. Cold Spring Harb Symp Quant Biol 72, 509-515.

Harrington ME. (1997). The ventral lateral geniculate nucleus and the intergeniculate leaflet: interrelated structures in the visual and circadian systems. Neurosci Biobehav Rev 21, 705-727.

Harrington ME & Rusak B. (1986). Lesions of the thalamic intergeniculate leaflet alter hamster circadian rhythms. J Biol Rhythms 1, 309-325.

Harrington ME & Rusak B. (1989). Photic responses of geniculo-hypothalamic tract neurons in the Syrian hamster. Vis Neurosci 2, 367-375.

110011011102

1103110411051106

110711081109

111011111112

111311141115

111611171118

1119112011211122

112311241125

1126112711281129

1130113111321133

113411351136

113711381139

1140

Hattar S, Kumar M, Park A, Tong P, Tung J, Yau KW & Berson DM. (2006). Central projections of melanopsin-expressing retinal ganglion cells in the mouse. J Comp Neurol 497, 326-349.

Horvath TL. (1998). An alternate pathway for visual signal integration into the hypothalamo-pituitary axis: retinorecipient intergeniculate neurons project to various regions of the hypothalamus and innervate neuroendocrine cells including those producing dopamine. J Neurosci 18, 1546-1558.

Howarth M, Walmsley L & Brown TM. (2014). Binocular integration in the mouse lateral geniculate nuclei. Curr Biol 24, 1241-1247.

Huhman KL & Albers HE. (1994). Neuropeptide Y microinjected into the suprachiasmatic region phase shifts circadian rhythms in constant darkness. Peptides 15, 1475-1478.

Huhman KL, Gillespie CF, Marvel CL & Albers HE. (1996). Neuropeptide Y phase shifts circadian rhythms in vivo via a Y2 receptor. Neuroreport 7, 1249-1252.

Huhman KL, Marvel CL, Gillespie CF, Mintz EM & Albers HE. (1997). Tetrodotoxin blocks NPY-induced but not muscimol-induced phase advances of wheel-running activity in Syrian hamsters. Brain Res 772, 176-180.

Irwin RP & Allen CN. (2009). GABAergic signaling induces divergent neuronal Ca2+ responses in the suprachiasmatic nucleus network. Eur J Neurosci 30, 1462-1475.

Itri JN, Michel S, Vansteensel MJ, Meijer JH & Colwell CS. (2005). Fast delayed rectifier potassium current is required for circadian neural activity. Nat Neurosci 8, 650-656.

Jacobs GH & Williams GA. (2007). Contributions of the mouse UV photopigment to the ERG and to vision. Doc Ophthalmol 115, 137-144.

Janik D & Mrosovsky N. (1994). Intergeniculate leaflet lesions and behaviorally-induced shifts of circadian rhythms. Brain Res 651, 174-182.

Jiang ZG, Allen CN & North RA. (1995). Presynaptic inhibition by baclofen of retinohypothalamic excitatory synaptic transmission in rat suprachiasmatic nucleus. Neuroscience 64, 813-819.

Johnson RF, Moore RY & Morin LP. (1989). Lateral geniculate lesions alter circadian activity rhythms in the hamster. Brain Res Bull 22, 411-422.

Johnson RF, Smale L, Moore RY & Morin LP. (1988). Lateral geniculate lesions block circadian phase-shift responses to a benzodiazepine. Proc Natl Acad Sci U S A 85, 5301-5304.

11411142

11431144114511461147

114811491150

115111521153

115411551156

1157115811591160

116111621163

116411651166

116711681169

117011711172

117311741175

117611771178

117911801181

1182

Kalsbeek A, Teclemariam-Mesbah R & Pévet P. (1993). Efferent projections of the suprachiasmatic nucleus in the golden hamster (Mesocricetus auratus). J Comp Neurol 332, 293-314.

Kubota A, Inouye ST & Kawamura H. (1981). Reversal of multiunit activity within and outside the suprachiasmatic nucleus in the rat. Neurosci Lett 27, 303-308.

Lall GS & Biello SM. (2002). Attenuation of phase shifts to light by activity or neuropeptide Y: a time course study. Brain Res 957, 109-116.

Lall GS & Biello SM. (2003). Attenuation of circadian light induced phase advances and delays by neuropeptide Y and a neuropeptide Y Y1/Y5 receptor agonist. Neuroscience 119, 611-618.

Lein ES, Hawrylycz MJ, Ao N, Ayres M, Bensinger A, Bernard A, Boe AF, Boguski MS, Brockway KS, Byrnes EJ, Chen L, Chen TM, Chin MC, Chong J, Crook BE, Czaplinska A, Dang CN, Datta S, Dee NR, Desaki AL, Desta T, Diep E, Dolbeare TA, Donelan MJ, Dong HW, Dougherty JG, Duncan BJ, Ebbert AJ, Eichele G, Estin LK, Faber C, Facer BA, Fields R, Fischer SR, Fliss TP, Frensley C, Gates SN, Glattfelder KJ, Halverson KR, Hart MR, Hohmann JG, Howell MP, Jeung DP, Johnson RA, Karr PT, Kawal R, Kidney JM, Knapik RH, Kuan CL, Lake JH, Laramee AR, Larsen KD, Lau C, Lemon TA, Liang AJ, Liu Y, Luong LT, Michaels J, Morgan JJ, Morgan RJ, Mortrud MT, Mosqueda NF, Ng LL, Ng R, Orta GJ, Overly CC, Pak TH, Parry SE, Pathak SD, Pearson OC, Puchalski RB, Riley ZL, Rockett HR, Rowland SA, Royall JJ, Ruiz MJ, Sarno NR, Schaffnit K, Shapovalova NV, Sivisay T, Slaughterbeck CR, Smith SC, Smith KA, Smith BI, Sodt AJ, Stewart NN, Stumpf KR, Sunkin SM, Sutram M, Tam A, Teemer CD, Thaller C, Thompson CL, Varnam LR, Visel A, Whitlock RM, Wohnoutka PE, Wolkey CK, Wong VY, Wood M, Yaylaoglu MB, Young RC, Youngstrom BL, Yuan XF, Zhang B, Zwingman TA & Jones AR. (2007). Genome-wide atlas of gene expression in the adult mouse brain. Nature 445, 168-176.

Long MA, Jutras MJ, Connors BW & Burwell RD. (2005). Electrical synapses coordinate activity in the suprachiasmatic nucleus. Nat Neurosci 8, 61-66.

Madisen L, Mao T, Koch H, Zhuo JM, Berenyi A, Fujisawa S, Hsu YW, Garcia AJ, Gu X, Zanella S, Kidney J, Gu H, Mao Y, Hooks BM, Boyden ES, Buzsáki G, Ramirez JM, Jones AR, Svoboda K, Han X, Turner EE & Zeng H. (2012). A toolbox of Cre-dependent optogenetic transgenic mice for light-induced activation and silencing. Nat Neurosci 15, 793-802.

Marchant EG, Watson NV & Mistlberger RE. (1997). Both neuropeptide Y and serotonin are necessary for entrainment of circadian rhythms in mice by daily treadmill running schedules. J Neurosci 17, 7974-7987.

Maywood ES, Smith E, Hall SJ & Hastings MH. (1997). A thalamic contribution to arousal-induced, non-photic entrainment of the circadian clock of the Syrian hamster. Eur J Neurosci 9, 1739-1747.

Meijer JH, Watanabe K, Détàri L & Schaap J. (1996). Circadian rhythm in light response in suprachiasmatic nucleus neurons of freely moving rats. Brain Res 741, 352-355.

11831184

118511861187

118811891190

119111921193

119411951196119711981199120012011202120312041205120612071208

120912101211

12121213121412151216

1217121812191220

1221122212231224

122512261227

Meijer JH, Watanabe K, Schaap J, Albus H & Détári L. (1998). Light responsiveness of the suprachiasmatic nucleus: long-term multiunit and single-unit recordings in freely moving rats. J Neurosci 18, 9078-9087.

Mikkelsen JD. (1990). Projections from the lateral geniculate nucleus to the hypothalamus of the Mongolian gerbil (Meriones unguiculatus): an anterograde and retrograde tracing study. J Comp Neurol 299, 493-508.

Mikkelsen JD. (1992). The organization of the crossed geniculogeniculate pathway of the rat: a Phaseolus vulgaris-leucoagglutinin study. Neuroscience 48, 953-962.

Moldavan MG & Allen CN. (2013). GABAB receptor-mediated frequency-dependent and circadian changes in synaptic plasticity modulate retinal input to the suprachiasmatic nucleus. J Physiol 591, 2475-2490.

Moldavan MG, Irwin RP & Allen CN. (2006). Presynaptic GABA(B) receptors regulate retinohypothalamic tract synaptic transmission by inhibiting voltage-gated Ca2+ channels. J Neurophysiol 95, 3727-3741.

Moore RY & Card JP. (1994). Intergeniculate leaflet: an anatomically and functionally distinct subdivision of the lateral geniculate complex. J Comp Neurol 344, 403-430.

Moore RY, Weis R & Moga MM. (2000). Efferent projections of the intergeniculate leaflet and the ventral lateral geniculate nucleus in the rat. J Comp Neurol 420, 398-418.

Morin LP & Allen CN. (2006). The circadian visual system, 2005. Brain Res Rev 51, 1-60.

Morin LP & Blanchard J. (1995). Organization of the hamster intergeniculate leaflet: NPY and ENK projections to the suprachiasmatic nucleus, intergeniculate leaflet and posterior limitans nucleus. Vis Neurosci 12, 57-67.

Morin LP & Blanchard JH. (2001). Neuromodulator content of hamster intergeniculate leaflet neurons and their projection to the suprachiasmatic nucleus or visual midbrain. J Comp Neurol 437, 79-90.

Morin LP, Blanchard JH & Provencio I. (2003). Retinal ganglion cell projections to the hamster suprachiasmatic nucleus, intergeniculate leaflet, and visual midbrain: bifurcation and melanopsin immunoreactivity. J Comp Neurol 465, 401-416.

Morin LP, Goodless-Sanchez N, Smale L & Moore RY. (1994). Projections of the suprachiasmatic nuclei, subparaventricular zone and retrochiasmatic area in the golden hamster. Neuroscience 61, 391-410.

1228122912301231

1232123312341235

123612371238

1239124012411242

1243124412451246

124712481249

125012511252

12531254

1255125612571258

1259126012611262

1263126412651266

1267126812691270

Morin LP & Pace L. (2002). The intergeniculate leaflet, but not the visual midbrain, mediates hamster circadian rhythm response to constant light. J Biol Rhythms 17, 217-226.

Mrugala M, Zlomanczuk P, Jagota A & Schwartz WJ. (2000). Rhythmic multiunit neural activity in slices of hamster suprachiasmatic nucleus reflect prior photoperiod. Am J Physiol Regul Integr Comp Physiol 278, R987-994.

Muscat L & Morin LP. (2006). Intergeniculate leaflet: contributions to photic and non-photic responsiveness of the hamster circadian system. Neuroscience 140, 305-320.

Nakamura W, Yamazaki S, Nakamura TJ, Shirakawa T, Block GD & Takumi T. (2008). In vivo monitoring of circadian timing in freely moving mice. Curr Biol 18, 381-385.

Novak CM & Albers HE. (2004). Circadian phase alteration by GABA and light differs in diurnal and nocturnal rodents during the day. Behav Neurosci 118, 498-504.

Novak CM, Ehlen JC, Huhman KL & Albers HE. (2004). GABA(B) receptor activation in the suprachiasmatic nucleus of diurnal and nocturnal rodents. Brain Res Bull 63, 531-535.

Nygård M, Hill RH, Wikström MA & Kristensson K. (2005). Age-related changes in electrophysiological properties of the mouse suprachiasmatic nucleus in vitro. Brain Res Bull 65, 149-154.

Paxinos G & Franklin K. (2001). The Mouse Brain in Stereotaxic Coordinates, Second Edition. Academic Press, San Diego.

Pelletier G. (1990). Ultrastructural localization of neuropeptide Y in the hypothalamus. Ann N Y Acad Sci 611, 232-246.

Pennartz CM, de Jeu MT, Bos NP, Schaap J & Geurtsen AM. (2002). Diurnal modulation of pacemaker potentials and calcium current in the mammalian circadian clock. Nature 416, 286-290.

Pickard GE. (1982). The afferent connections of the suprachiasmatic nucleus of the golden hamster with emphasis on the retinohypothalamic projection. J Comp Neurol 211, 65-83.

Pickard GE. (1985). Bifurcating axons of retinal ganglion cells terminate in the hypothalamic suprachiasmatic nucleus and the intergeniculate leaflet of the thalamus. Neurosci Lett 55, 211-217.

Pickard GE, Ralph MR & Menaker M. (1987). The intergeniculate leaflet partially mediates effects of light on circadian rhythms. J Biol Rhythms 2, 35-56.

127112721273

1274127512761277

127812791280

128112821283

128412851286

128712881289

1290129112921293

129412951296

129712981299

130013011302

130313041305

1306130713081309

131013111312

Reinhard K, Tikidji-Hamburyan A, Seitter H, Idrees S, Mutter M, Benkner B & Münch TA. (2014). Step-by-step instructions for retina recordings with perforated multi electrode arrays. PLoS One 9, e106148.

Roig JA, Granados-Fuentes D & Aguilar-Roblero R. (1997). Neuronal subpopulations in the suprachiasmatic nuclei based on their response to retinal and intergeniculate leaflet stimulation. Neuroreport 8, 885-889.

Rusak B, Meijer JH & Harrington ME. (1989). Hamster circadian rhythms are phase-shifted by electrical stimulation of the geniculo-hypothalamic tract. Brain Res 493, 283-291.

Sakhi K, Wegner S, Belle MD, Howarth M, Delagrange P, Brown TM & Piggins HD. (2014). Intrinsic and extrinsic cues regulate the daily profile of mouse lateral habenula neuronal activity. J Physiol 592, 5025-5045.

Schaap J, Albus H, VanderLeest HT, Eilers PH, Détári L & Meijer JH. (2003). Heterogeneity of rhythmic suprachiasmatic nucleus neurons: Implications for circadian waveform and photoperiodic encoding. Proc Natl Acad Sci U S A 100, 15994-15999.

Schaap J, Bos NP, de Jeu MT, Geurtsen AM, Meijer JH & Pennartz CM. (1999). Neurons of the rat suprachiasmatic nucleus show a circadian rhythm in membrane properties that is lost during prolonged whole-cell recording. Brain Res 815, 154-166.

Shibata S & Moore RY. (1993). Neuropeptide Y and optic chiasm stimulation affect suprachiasmatic nucleus circadian function in vitro. Brain Res 615, 95-100.

Shibata S, Oomura Y, Kita H & Hattori K. (1982). Circadian rhythmic changes of neuronal activity in the suprachiasmatic nucleus of the rat hypothalamic slice. Brain Res 247, 154-158.

Smith RD, Inouye S & Turek FW. (1989). Central administration of muscimol phase-shifts the mammalian circadian clock. J Comp Physiol A 164, 805-814.

Taniguchi H, He M, Wu P, Kim S, Paik R, Sugino K, Kvitsiani D, Kvitsani D, Fu Y, Lu J, Lin Y, Miyoshi G, Shima Y, Fishell G, Nelson SB & Huang ZJ. (2011). A resource of Cre driver lines for genetic targeting of GABAergic neurons in cerebral cortex. Neuron 71, 995-1013.

Tcheng TK & Gillette MU. (1996). A novel carbon fiber bundle microelectrode and modified brain slice chamber for recording long-term multiunit activity from brain slices. J Neurosci Methods 69, 163-169.

Thankachan S & Rusak B. (2005). Juxtacellular recording/labeling analysis of physiological and anatomical characteristics of rat intergeniculate leaflet neurons. J Neurosci 25, 9195-9204.

1313131413151316

1317131813191320

132113221323

1324132513261327

1328132913301331

1332133313341335

133613371338

133913401341

134213431344

1345134613471348

1349135013511352

135313541355

Tousson E & Meissl H. (2004). Suprachiasmatic nuclei grafts restore the circadian rhythm in the paraventricular nucleus of the hypothalamus. J Neurosci 24, 2983-2988.

Treep JA, Abe H, Rusak B & Goguen DM. (1995). Two distinct retinal projections to the hamster suprachiasmatic nucleus. J Biol Rhythms 10, 299-307.

van Diepen HC, Ramkisoensing A, Peirson SN, Foster RG & Meijer JH. (2013). Irradiance encoding in the suprachiasmatic nuclei by rod and cone photoreceptors. FASEB J 27, 4204-4212.

van Oosterhout F, Fisher SP, van Diepen HC, Watson TS, Houben T, VanderLeest HT, Thompson S, Peirson SN, Foster RG & Meijer JH. (2012). Ultraviolet light provides a major input to non-image-forming light detection in mice. Curr Biol 22, 1397-1402.

VanderLeest HT, Houben T, Michel S, Deboer T, Albus H, Vansteensel MJ, Block GD & Meijer JH. (2007). Seasonal encoding by the circadian pacemaker of the SCN. Curr Biol 17, 468-473.

vanderLeest HT, Vansteensel MJ, Duindam H, Michel S & Meijer JH. (2009). Phase of the electrical activity rhythm in the SCN in vitro not influenced by preparation time. Chronobiol Int 26, 1075-1089.

Wagner S, Castel M, Gainer H & Yarom Y. (1997). GABA in the mammalian suprachiasmatic nucleus and its role in diurnal rhythmicity. Nature 387, 598-603.

Walmsley L & Brown TM. (2015). Eye-specific visual processing in the mouse suprachiasmatic nuclei. J Physiol 593, 1731-1743.

Walmsley L, Hanna L, Mouland J, Martial F, West A, Smedley AR, Bechtold DA, Webb AR, Lucas RJ & Brown TM. (2015). Colour as a signal for entraining the mammalian circadian clock. PLoS Biol 13, e1002127.

Watts AG, Swanson LW & Sanchez-Watts G. (1987). Efferent projections of the suprachiasmatic nucleus: I. Studies using anterograde transport of Phaseolus vulgaris leucoagglutinin in the rat. J Comp Neurol 258, 204-229.

Wickland C & Turek FW. (1994). Lesions of the thalamic intergeniculate leaflet block activity-induced phase shifts in the circadian activity rhythm of the golden hamster. Brain Res 660, 293-300.

Wong KY. (2012). A retinal ganglion cell that can signal irradiance continuously for 10 hours. J Neurosci 32, 11478-11485.

135613571358

135913601361

136213631364

1365136613671368

136913701371

1372137313741375

137613771378

137913801381

1382138313841385

1386138713881389

139013911392

139313941395

1396

Yamazaki S, Kerbeshian MC, Hocker CG, Block GD & Menaker M. (1998). Rhythmic properties of the hamster suprachiasmatic nucleus in vivo. J Neurosci 18, 10709-10723.

Yannielli PC, Brewer JM & Harrington ME. (2004). Blockade of the NPY Y5 receptor potentiates circadian responses to light: complementary in vivo and in vitro studies. Eur J Neurosci 19, 891-897.

Yannielli PC & Harrington ME. (2000). Neuropeptide Y applied in vitro can block the phase shifts induced by light in vivo. Neuroreport 11, 1587-1591.

Zhang DX & Rusak B. (1989). Photic sensitivity of geniculate neurons that project to the suprachiasmatic nuclei or the contralateral geniculate. Brain Res 504, 161-164.

Zhao C, Eisinger B & Gammie SC. (2013). Characterization of GABAergic neurons in the mouse lateral septum: a double fluorescence in situ hybridization and immunohistochemical study using tyramide signal amplification. PLoS One 8, e73750.

Figure Legends

13971398

1399140014011402

140314041405

140614071408

1409141014111412

1413

1414

1415

Figure 1. Modelling the non-image forming visual projection. (A) Central targets of M1-type ipRGCs visualised via X-gal stained sections from Opn4+/tau-LacZ mice. (B) Sagittal reconstruction of forebrain M1-type ipRGC axon/terminal labelling intensity, indicating chosen slice angle and thickness: 30° off the coronal plane, 600 µm thickness. (C) Projected ‘coronal’ view of slice depicted in B flattened across z-axis. Abbreviations: pretectal olivary nucleus, PON; intergeniculate leaflet/ventral lateral geniculate nucleus, IGL/vLGN; suprachiasmatic nucleus, SCN.

1416

141714181419142014211422

1423

Figure 2. Angled slice preparations retain optic tract connectivity across all major non-image forming nuclei. (A) Image of the recorded side (i.e. side placed facing recording sites) of an acute angled slice prepared from an Opn4+/tau-lacZ reporter mouse before and (B) after X-gal staining. (C) Schematic of the slice preparation showing stimulating electrode and pMEA placements used. (D-F) Electrical stimulation of the optic chiasm (OC) produces widespread excitatory responses across the pretectum (D), LGN (E) and SCN/peri-SCN (F). Traces represent probability of spike occurrence (from 600 trials) relative to stimulation onset (arrows) at each site on the 59-electrode recording array for an illustrative slice. (G-I) Bath application of ionotropic glutamate receptor blockers CNQX (20 µM) and D-AP5 (50 µM) completely abolished the evoked excitatory responses in the pretectum ( G; n=21 sites responding to OC stimulation from 3 slices), LGN (H; n=45 sites from 6 slices) and SCN (I; n=68 sites from 2 slices). Left-hand panels show normalised responses to OC stimulation before ( ) and after ( ) drug application (see methods). Shaded region indicates timepoints where the post-drug response is significantly reduced (***; P<0.001) relative to baseline (two-way RM ANOVA with Sidak’s multiple comparison test). Right-hand panels display perievent rasters of multiunit activity from representative recording sites (green shading = 6 min drug application).

1424

142514261427142814291430143114321433143414351436143714381439

Figure 3. Angled slices remain viable for more than 24 h and display circadian rhythms in electrical activity. (A) Image of a representative angled slice preparation illustrating the distribution of pMEA recording sites within and surrounding the SCN. Red circles indicate sites categorised as being located in the SCN, blue circles indicate sites categorised as located in the vSPZ. (B) Multiunit firing rate profiles recorded across the 59 sites of the electrode array from slice in A. Data were binned (60s) and smoothed with a 2h boxcar filter. (C) Bar chart illustrating the percentage of channels within the SCN and vSPZ considered rhythmic (based on data from 5 slices; see methods for details of analysis). Difference in the proportions of rhythmic channels here was analysed by Fisher’s exact test. (D) Rayleigh vector plots displaying the timing of peak SCN firing activity detected at rhythmic channels in the SCN (n=100) and vSPZ (n=61). Arrows indicate the median phase and strength of clustering (r value). Differences in phase and clustering strength were analysed by Mann-Whitney U-test and Browne-Forsythe’s test respectively (analysis performed on linearised data; see methods). (E) Proportion of channels across the SCN and vSPZ from the above experiments exhibiting significant changes in firing in response to a 5 min application of 20μM NMDA, applied >26h after slice preparation. (F) Mean±SEM firing rate across all NMDA-activated channels (n=180), illustrating robust changes in firing. Grey shaded areas in B and D correspond to the projected timing of the light/dark transition. ** and *** = P<0.01 and P<0.001 respectively.

1440

14411442144314441445144614471448144914501451145214531454145514561457

Figure 4. Electrical stimulation of the geniculohypothalamic tract evokes inhibitory and excitatory responses in the SCN. (A) Schematic of the slice preparation showing stimulation electrode and pMEA placement used. (B) Electrical stimulation (concentric electrode; 1.2 mA, 200 μs) of the IGL region evoked inhibitory responses at a subset of SCN recording sites (n=21 sites from 12 slices; representative recording site shown). (C) Electrical stimulation of the geniculohypothalamic tract (GHT) also evoked excitatory responses in a second group of SCN recording sites from these 12 slices (n=27 sites; left panel, representative recording site shown) which could be abolished by bath application of ionotropic glutamate receptor blockers (CNQX, 20 µM and D-AP5, 50 µM; right panel, representative recording site shown of 4 SCN recording sites tested). (D) GHT-evoked inhibitory responses (time to peak, 10 ms moving window) were typically slower and exhibited more variable timing than excitatory responses. (E) Where tested (n=12 sites), short trains of pulses (5 stimuli at 50 ms, i.e. 20 Hz, or 5 stimuli at 25 ms, 40 Hz) did not evoked significantly larger inhibitions than single pulse stimulation (one-way ANOVA; P=0.98). (F) GHT-evoked (1.2 mA, 200 µs; 5 x 40 Hz train) inhibitory responses (middle and bottom panels) but not excitatory responses (top panels) were blocked by bath application of the GABAAR antagonist, bicuculline (BIC 20 μM; representative SCN sites shown from a total of 23 excitatory and 5 inhibitory responses tested). (G,H) Quantification of the effect of BIC and CNQX/D-AP5 relative to predrug values (black bars) for excitatory (G) and inhibitory responses (H). Data were analysed by paired t-test (except CNQX in G, where only 4 cells were tested). **=P<0.01; n.s.=P>0.05.

1458

1459146014611462146314641465146614671468146914701471147214731474147514761477

1478

Figure 5. Optogenetics-based selective activation of geniculohypothalamic signalling. (A) GFP-immunostaining enhanced ChR2/EYFP expression in the thalamus of GAD2-Cre; Ai32 mice, revealed robust expression in neural processes across the predominantly GABAergic IGL/vLGN region. (B) Central region of A showing EYFP signal across the dLGN and IGL/vLGN at higher magnification. (C) Live image of slice showing optical fibre positioning over IGL/LGN (left) and schematic of slice preparation with optical fibre and pMEA placements used (right). (D, E) Direct optical stimulation (465 nm LED; ~800mW/mm2 at fibre tip, 10 ms flash – represented by blue shaded area in traces) evoked robust and rapid excitatory responses across the IGL/vLGN and much weaker responses in the dLGN. Traces in D represent probability of spike occurrence (from 200 trials) relative to stimulation onset at each site on the 59-electrode recording array for the slice shown in C. Data in Ei,ii and iii show perievent rasters and larger spike probability histograms for the traces indicated in D. (F) Schematic of slice preparation showing optical fibre and pMEA placements used for investigation of GHT signalling. (G) Optical stimulation of GABAergic GHT projection neurons evoked modest inhibitory responses in a small subset of SCN recording sites (n=12 sites from 4 slices; inhibition, mean±SEM: 43.8±5.5% of pre-stimulation; latency, median±SD: 60±25 ms: representative SCN recording sites are shown in G and H). (H,I) Where tested (n=6), optogenetic-evoked inhibitory responses were abolished by bath application of the GABAAR antagonist, (+)-bicuculline (20 μM). (H) data from a representative SCN recording site pre- and post-BIC (upper and lower panels respectively). (I) mean±SEM response magnitude pre- and post-BIC (n=6). Data were analysed by paired t-test. **=P<0.01.

1479

14801481148214831484148514861487148814891490149114921493149414951496149714981499

1500

Figure 6. Geniculohypothalamic modulation of SCN responses to optic input. (A) Perievent raster of SCN multiunit responses to electrical stimulation of the OC; synchronous with (left panel), or 250 ms after (right panel), optogenetic activation of GABAergic GHT projection neurons (see Fig. 5F for recording setup). (B) Normalised mean (±SEM) OC-driven responses of SCN subpopulations as a function of GHT pre-stimulation latency: at most recording sites (n=45/71 OC-responsive sites from 3 slices; see methods for classification details), GHT pre-stimulation evoked long-lasting suppression of OC-driven responses (two-way RM ANOVA: GHT pre-stimulation, Group and interaction all P<0.001; asterisks indicate significance differences between groups in Sidak’s post-tests). (C) GHT-driven inhibition of OC-evoked responses was significantly greater ipsilateral to the optogenetically stimulated LGN (n=22 vs. n=23 contralateral sites; t-test, P<0.05). (D) Perievent histograms of SCN responses (perievent rasters are depicted in A) showing enhanced OC-driven responses during the late day/early projected evening. (E) Normalised mean (±SEM) response of GHT-responsive sites (n=45) to OC-stimulation: synchronous with ( ), or subsequent to ( ), optical activation of GABAergic GHT projection neurons. In both cases F-test indicated a significantly better fit to a sinusoidal (24h) vs. linear function (Both P<0.001 that a linear process generated the data). (F) GHT-driven inhibition of OC-evoked responses (data from E, expressed as a percentage of the response observed under

1501

1502150315041505150615071508150915101511151215131514151515161517

synchronous stimulation) varies significantly across the 24 h in vitro recording period (one-way RM ANOVA of 6 h binned data; P<0.001; asterisks indicate significance differences between timepoints in Tukey’s post-hoc tests). (G) Perievent raster of IGL/vLGN multiunit responses to OC and optical stimulation (as in Fig. 5C). (H) Normalised mean (±SEM) IGL/vLGN population activity (n=34 responding sites from 3 slices) under basal conditions (top panel), following direct ChR2 (middle panel) or OC stimulation (bottom panel). In none of the above cases did we observe evidence of pronounced day-night variation (F-test: all P=0.999 that were more likely to have been generated by a linear vs. sinusoidal model). Shaded areas in A and E-H represent projected night. For B and F: *, ** and *** represent P<0.05, P<0.01 and P<0.001 respectively.

151815191520152115221523152415251526

1527

Figure 7. Geniculohypothalamic-driven inhibition of SCN responses to optic input involves GABAA

and GABAB receptors. (A) Perievent histograms of representative SCN multiunit responses to electrical stimulation of the OC at varying latencies of GABAergic GHT neuron pre-stimulation; under control conditions (top panel), following bath application of the GABAAR antagonist bicuculline (20 µm, BIC; middle panel), or co-application of BIC (20 µM) & the GABABR antagonist CGP55845 (3 µM, CGP; bottom panel). (B) Normalised mean (±SEM) population responses to OC-stimulation as a function of GHT pre-stimulation latency (n=44 GHT-responsive sites from a total of 81 OC-responsive sites in 3 slices). Bath application of BIC substantially attenuated GHT-driven inhibition at short latencies; residual responses under BIC were further attenuated when co-applied with CGP. (C) Normalised mean (±SEM) population responses to OC-stimulation in a separate, second experiment investigating the effects of shorter GHT pre-stimulation latencies (n=14 GHT responsive/21 OC-responsive sites from 3 slices): note that the initial components of the GHT-evoked inhibition were almost completely attenuated by BIC application alone. Data in B and C were analysed by two-way RM ANOVA with Sidak’s post-hoc tests (P<0.01 for GHT pre-stimulation, treatment and interaction in both cases). Black symbols in B and C represent significant differences relative to 0ms delay (***=P<0.001), Red symbols represent significant differences between BIC and baseline, blue symbols (in B) represent differences between BIC and BIC+CGP (#=P<0.01, $=P<0.05).

1528

15291530153115321533153415351536153715381539154015411542154315441545

1546

Figure 8. Geniculohypothalamic signalling modulates specific features of the SCN light response in vivo. (A) Histological reconstruction of a DiI labelled recording electrode/drug cannula placement in the LGN (light microscopic image, pseudo-coloured green) and (B) schematic showing recording electrode/drug cannula placement in LGN and projected spread of drug infusion. (C) Multiunit responses detected at level of probe in A-B to 5 s light step (405 nm, 15.3 log photons/cm2/s) presented to both eyes before and after infusion of the GABAA receptor agonist muscimol (1 mM; 2-3 μl injection at 1 μl/min) completely abolished spiking activity at recording sites up to 750 μm from cannula tip. (D) Normalised mean (±SEM) SCN population responses to 5 s light steps (405 nm, 15.3 log photons/cm2/s) presented to the contralateral (left eye), ipsilateral (right eye) or both eyes, before and after infusion of muscimol into the ipsilateral LGN (n=63 sites from 5 animals); schematic (left panel) shows recording electrode placement in the SCN and projected spread of drug infusion in the LGN. (E-F) Data as in D for responses following infusion of muscimol into the contralateral LGN (E; n=33 sites from 3 mice) or under control conditions (F; identical protocol with no drug delivery, n=127 sites from 6 mice); schematics of the recording set-up are shown on the left. (G-I) Normalised mean (±SEM) SCN population responses to 5 s light steps of varying intensity presented to the contralateral (left), ipsilateral (right) or both eyes (left, middle and right panels respectively), before (

) and after muscimol (or no drug) infusion ( ) into the ipsilateral LGN (G), contralateral LGN (H) or under control conditions (I). For each panel in G-I an F-test was performed to determine whether a single sigmoidal function could explain the irradiance response relationship pre-and post-muscimol. This was the case (P>0.05) in all but panel G contralateral and binocular (P=0.001 and 0.002 respectively).

1547

154815491550155115521553155415551556155715581559156015611562156315641565156615671568

1569

Figure 9. Optogenetic activation of the geniculohypothalamic tract inhibits specific features of the SCN light response in vivo. (A; upper pannel) Histological reconstruction of a DiI labelled optrode placement in the LGN of a GAD2-Cre; Ai32 mouse (DiI in red, ChR2-EYFP in green, DAPI in blue) and (lower panel) schematic showing optrode placement in LGN and projected spread of optical illumination (based on 0.66 numerical aperture fibre). (B) Representative perievent histograms showing responses from optrode in A (responses to 100 trials, for clarity responses shown only for central 12 sites) to 10ms light flashes (465 nm LED; ~630mW/mm2 at fibre tip, represented by blue shading). (C) perievent raster for optogenetic responses identified at the site in B indicated by *. ( D) Schematic of recording setup for data in E-G. (E) Normalised mean (±SEM) SCN population responses to 5 s light steps (405 nm, 15.3 log photons/cm2/s) presented to the contralateral, ipsilateral or both eyes, in the absence (red) and presence (blue) of optogenetic stimulation of the GHT (for clarity data for 10Hz trains only are shown). Data are derived from 72 light responsive channels (from 6 mice). Left panels show initial components of the response (10ms bin size, smoothed with a 50ms boxcar filter), right panels show full light response (200ms bin size). (F, G) Normalised mean (±SEM) change in light-evoked SCN activity as a function of GHT stimulation frequency for stimulation of contralateral, ipsilateral or both eyes, either during early (F, 0-250ms) or later (G, 3-5s after start of visual stimulation) components of the response. Data were analysed by two-way RM ANOVA with Sidak’s post-test. For F, ANOVA revealed a significant effect of stimulation frequency (P<0.001) but not eye or interaction, post-tests indicated all frequencies were significantly reduced compared to

1570

1571157215731574157515761577157815791580158115821583158415851586158715881589

no GHT stimulation. For G, ANOVA revealed a significant interaction between eye and stimulation frequency, coloured asterisks represent significant differences relative to the no-GHT stimulation group (0 Hz) for that eye. (H) Mean (±SEM) light-evoked IGL/vLGN activity from experiments above. Data are based on changes across full 5s stimulation (from 41 light and optically responsive sites from 6 mice), analysis performed on the early and later epochs (as in F and G; not shown) produced qualitatively similar results: two-way RM ANOVA revealed significant interaction (eye x stimulation frequency; P<0.001). Coloured asterisks represent significant differences relative to the no-GHT stimulation group for that eye (Sidak’s post-test). (I) Normalised mean (±SEM) IGL/vLGN responses from dataset above for contralateral and ipsilateral visual stimuli in the absence (red) or presence (blue) of optogenetic stimulation (for clarity data for 10Hz trains only are shown). Grey shaded areas in E and I represent darkness. In G and H, ** and *** represent =P<0.01 and P<0.001 respectively.

15901591159215931594159515961597159815991600

1601

Figure 10. Schematic of proposed geniculohypothalamic circuitry and its impact on SCN output. (A) Based on the data presented in this manuscript, we propose that GABAergic neurons contributing to the GHT are primarily excited by contralateral retina inputs and convey the resulting inhibitory signals to SCN neurons that receive input from either (or both) eyes. An SCN-intrinsic mechanism then adjusts SCN responses to retinal and GHT inputs according to time of day. ( B,C) Schematic illustrating inferred excitatory and inhibitory synaptic influences on SCN neurons (B) and the resulting spike output under varying illumination (C; upper and lower panels reflect SCN cells receiving input from just the contra- or ipsilateral eye respectively). Following experimental illumination of just the contralateral retina (left panels), GHT input is activated and inhibitory inputs reach the SCN 50-100ms after arrival of excitatory retinal inputs. For SCN cells that receive contralateral retinal inputs, the resulting long-lasting inhibitory influence attenuates the excitatory retinal drive leading to a suppression of steady state light-driven firing rates (especially during the projected day when SCN neurons are more sensitive to GHT signals). By contrast, in the absence of any driving excitatory input (SCN cells receiving only ipsilateral input), spontaneous firing is unaffected. When only the ipsilateral retina is illuminated, activity of the GHT is very low and therefore doesn’t influence SCN responses (middle panels). Under the more commonly encountered situation in the natural world where both eyes are illuminated (right panels) inhibitory GHT signals provided by the GHT are able to reduce SCN responses regardless of which eye a cell receives input from.

1602

1603160416051606160716081609161016111612161316141615161616171618161916201621

1622