Ancestral Circuits for the Coordinated Modulation of Brain State.

Detailed methods are provided in the online version of this paper and include the following: KEY RESOURCES TABLE REAGENT OR RESOURCE SOURCE IDENTIFIER Antibodies mouse anti-tyrosine hydroxylase Immunostar 22941; RRID: AB_572268 rabbit anti-serotonin Immunostar 20080; RRID: AB_572263 rabbit anti-orexin Millipore 22941 goat anti-orexin Santa Cruz Biotechnology sc-8070; RRID: AB_653610 rabbit anti-CART Abcam ab192364 rabbit anti-galanin Millipore ab5909; RRID: AB_2108517 rabbit anti-neuropeptideVF Sigma-Aldrich HPA041733; RRID: AB_10794524 rabbit anti-neuropeptideY Sigma-Aldrich N9528; RRID: AB_260814 rat anti-somatostatin Abcam ab30788; RRID: AB_778010 rabbit anti-dopamine beta-hydroxylase Immunostar 22806; RRID: AB_572229 goat anti-choline acetyltransferase Millipore A144P donkey anti-mouse Alexa647 Jackson ImmunoResearch 715-606-151; RRID: AB_2340866 donkey anti-rabbit Alexa 594 Jackson ImmunoResearch 711-586-152; RRID: AB_2340622 donkey anti-rat DyLight405 Jackson ImmunoResearch 712-475-153; RRID: AB_2340681 donkey anti-rat Alexa647 Jackson ImmunoResearch 712-606-153; RRID: AB_2340696 donkey anti-goat Alexa647 Jackson ImmunoResearch 712-606-147 donkey anti-goat DyLight405 Jackson ImmunoResearch 705-475-147; RRID: AB_2340427 Bacterial and Virus Strains AAVDJ-Ef1a-DIO-GCaMP6m Stanford Gene Vector and Virus Core N/A AAV1-Ef1a-DIO-ChR2-eYFP Stanford Gene Vector and Virus Core N/A AAVDJ-Ef1a-DIO-eNpHR3.0-eYFP Stanford Gene Vector and Virus Core N/A AAVDJ-Ef1a-DIO-eYFP Stanford Gene Vector and Virus Core N/A Experimental Models: Organisms/Strains Zebrafish: Tg(elavl3:H2B-GCaMP6s) Vladimirov et al., 2014 jf5 Zebrafish: Tg(cmlc2:GFP) Huang et al., 2003 N/A Mouse: TH-IRES-Cre EMMA, Ted Ebebdal N/A Mouse: SERT-Cre, Slc6a4-Cre MMRRC #017260-UCD Mouse: ChAT-Cre MMRRC #017269-UCD Mouse: SST-IRES-Cre The Jackson Laboratory Jax stock # 013044 Mouse: AGRP-IRES-Cre The Jackson Laboratory Jax stock: # 012899 Mouse: CART-IRES2-Cre The Jackson Laboratory Jax stock: # 028533 Software and Algorithms Python 2.7 Python https://www.python.org Statsmodels 0.8.0 Seabold and Perktold, 2010 http://www.statsmodels.org/ Scikit-Learn 0.17.0 Pedregosa et al., 2011 http://scikit-learn.org/ MATLAB The MathWorks https://www.mathworks.com/products/matlab Fiji (ImageJ) NIH http://fiji.sc CMTK Rohlfing and Maurer, 2003 https://www.nitrc.org/projects/cmtk/ CONTACT FOR REAGENTS AND RESOURCE SHARING Further information and requests for reagents should be directed to and will be fulfilled by the Lead Contact, Karl Deisseroth ( deissero@stanford.edu ).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

All procedures were approved by the Stanford University Institutional Animal Care and Use Committee. Zebrafish We used 7–10 days post fertilization larval zebrafish for this study, bred on a Nacre or Casper background. No statistical methods were used to predetermine sample size, and animal selection was not randomized or blinded. We used Tg(elavl3:H2B-GCaMP6s) fish ( Vladimirov et al., 2014 ) for imaging, registration, and behavioral experiments, and Tg(cmlc2:GFP) fish ( Huang et al., 2003 ) for heart-rate imaging. Fish were group-housed under a 14:10 light:dark cycle until the day of experiments, and were fed with paramecia (Parameciavap) twice daily from 5–6 days post fertilization onward. All testing occurred during the late morning and afternoon.

Mice

We used adult mice for this study (male and female, > P45), consisting of TH-IRES-Cre mice (EMMA, Ted Ebebdal; Lindeberg et al., 2004 ), SERT-Cre mice (MMRRC, Slc6a4-cre, #017260-UCD), ChAT-Cre mice (MMRRC, #017269-UCD), SST-IRES-Cre mice (The Jackson Laboratory, #013044), AGRP-IRES-Cre mice (The Jackson Laboratory, #012899), and CART-IRES2-Cre mice (The Jackson Laboratory, #028533). All mice were injected unilaterally, bilaterally, or in the midline with 400 nL of either AAVDJ-Ef1α-DIO-GCaMP6m (3 × 10 12 viral particles/ml; Chen et al., 2013 ), AAV1-Ef1α-DIO-ChR2-eYFP (3 × 10 12 viral particles/ml), AAVDJ-Ef1α-DIO-eYFP (3 × 10 12 viral particles/ml), or AAVDJ-Ef1α-DIO-eNpHR3.0-eYFP (3 × 10 12 viral particles/ml). GCaMP6m (Addgene plasmid # 40754) was a gift from Douglas Kim (Janelia Research Campus) and modified in our lab. No statistical methods were used to predetermine sample size, and animal selection was not randomized. For freely moving behavior in the open field test, the experimenter was blinded. We did not systematically test for differences between male and female mice in this study, due to low and/or uneven numbers of age-matched male and females in each experimental condition. Mice were either group-or singly housed and maintained under a 14:10 reverse light:dark cycle, and had ad libitum access to food and water, until water restriction began (see below). All testing occurred during the morning and early afternoon.

METHOD DETAILS Zebrafish behavior

For all experiments, larval zebrafish were embedded in 2.5% low melting point agarose (Millipore) in the lid of a 35 mm Petri dish (E&K Scientific). Fish were embedded in agarose for 2–6 hours prior to beginning experiments. Once agarose solidified after embedding, agarose posterior to the pectoral fins was carefully removed with a scalpel, so that the tail was free to move. Petri dishes were placed on the lid of a 100 mm Petri dish (Fisher), with a diffuser screen attached to the bottom with vacuum grease (with a hole cut out for the tail). Behavioral monitoring, stimulus generation, and recording of frame times from the two-photon microscope were achieved with custom software written in MATLAB (MathWorks). Visual stimuli were projected at 60 Hz using a Laser Pico Projector (MicroVision), surrounded by three Red Wratten filters (Kodak) to prevent interference with the two-photon detection system. Visual stimuli were directed off a dichroic mirror to project on the bottom of the Petri dish with the embedded fish. The tail of the fish was illuminated by IR lights from above and behind the fish, and tail movements were filmed at 120 frames per second from below with an AVT Manta G 031 camera (Allied Vision) through a AF-S DX Micro Nikkor 85 mm f/3.5G ED VR macro lens (Nikon). Looming dot stimuli were black disks on a red background, and were presented ~0.5 mm to the right of the fish’s right eye, were initiated as ~1° (0.01 cm diameter) and expanded to ~100°(1.3 cm diameter) over the course of 3.5 s (~0.37 cm/s). Each stimulus in a behavioral session was identical, and were presented with a pseudorandom inter stimulus interval (range = 8 to 12 s). Optomotor stimuli were a black and red checkerboard pattern, presented to a fish ~0.1 cm away (squares = 0.25 cm across; ~100°), that drifted in the caudal to rostral direction (0.4 cm/s), for 10 s at a time, with a pseudorandom inter stimulus interval (range = 8 to 12 s). For exogenous manipulations of alertness, fish were either treated with caffeine (50 mg/L caffeine (Sigma), dissolved in fish water, 45 minutes prior to beginning of behavior), or one night of sleep deprivation (a Petri dish of 5–10 fish was placed on a slowly rotating shaker at room temp overnight in constant light).

Zebrafish two-photon microscopy

Two-photon imaging was performed using an Olympus FVMPE multiphoton microscope (Olympus Corporation), with a resonant scanner, in either unidirectional or bidirectional scanning mode. We used a 16x objective (0.8 NA; Nikon) for most experiments. Heart rate imaging was performed at 2x zoom in a single z-plane with 6x frame averaging, at 8.3 frames/second. Functional brain imaging was performed at 1.1x zoom in 14–16 z-planes, separated by 15 μm, at 1.3–2.8 volumes/second (5000–8000 volumes). After completion of behavior and functional brain imaging, a structural stack was obtained at 1 μm spacing and 16x frame averaging, starting 15 μm above the first z-plane, ending 15 μm below the last z-plane, and repeated 10 times. Fast axial scanning was achieved with a piezo motor coupled to the objective. Fast functional imaging of GCaMP and GFP were conducted at 920 nm. Structural scans in live fish were conducted at 860 nm, in order to obtain the structure of GCaMP+ cells independent of calcium-dependent fluorescence. Structural scans in fixed fish were conducted at 810 nm (DyLight 405 dye), 920 nm (GCaMP), and 1150 nm (Alexa594 and Alexa647 dyes), and split into four spectrally separable PMTs to detect each dye independently. Fixed samples were imaged at each wavelength at 1 μm spacing and 16x frame averaging, and repeated twice. In both live and fixed structural stacks, voxels were 1.4 × 1.4 × 1 μm in size. All images were corrected for axial chromatic aberrations induced by the objective, as determined by measurements of axial displacement from imaging multi-colored beads across each of the four detectors (data not shown). For brain imaging of caffeine-treated fish, fish were fully embedded in agarose and imaged with a 25x objective (1.05 NA; Olympus) at 2.6 volumes/second, for ~6 minutes before and 6 minutes after acute caffeine (50 mg/L) application. For heart imaging under caffeine and sleep deprivation, fish were fully embedded in agarose (right side up or upside down) and resting heart rate was imaged at 15–20 Hz. Fish were either sleep-deprived, treated with caffeine 45 minutes prior to imaging, or treated with vehicle (fish water).

Zebrafish immunohistochemistry

After completion of behavior and imaging, a small block of agarose containing the fish was cut out, submerged in ice-cold PBS, then placed into 4% PFA in phosphate-buffered saline (PBS; pH 7.4, Life Technologies) with 0.2% Triton-X (PBST) in a 1.5 mL tube, overnight on a shaker at 4 °C. Samples were then washed with PBST and left on a shaker at room temperature for 2–4 hours (repeated 2–3 times). Primary antibodies were then applied at 1:200 in PBST, and samples were left on a shaker at 4 °C for 48 hours. Samples were then washed with PBST and left on a shaker at room temperature for 2–4 hours (repeated 2–3 times). Secondary antibodies were then applied at 1:200 in PBST, and samples were left on a shaker at 4 °C for 48 hours. Samples were then washed with PBST and left on a shaker at room temperature for 2–4 hours (repeated 2–3 times). At this point, whole-mount zebrafish samples (still in their agarose block) were re-mounted in agarose on a Petri dish lid, and imaged again under the two-photon microscope. The primary antibodies used were: mouse anti-tyrosine hydroxylase (Immunostar, 22941), rabbit anti-serotonin (Immunostar, 20080), rabbit anti-orexin (Millipore, AB3704), goat anti-orexin (Santa Cruz Biotechnology, sc-8070), rabbit anti-CART (Abcam, ab192364), rabbit anti-galanin (Millipore, ab5909), rabbit anti-neuropeptideVF (Sigma, HPA041733), rabbit anti-neuropeptideY (Sigma, N9528), rat anti-somatostatin (Abcam, ab30788), rabbit anti-dopamine beta-hydroxylase (Immunostar, 22806), and goat anti-choline acetyltransferase (Millipore, A144P). Secondary antibodies used were donkey anti-mouse Alexa647 (Jackson Immuno Research, 715-606-151), donkey anti-rabbit Alexa 594 (Jackson Immuno Research, 711-586-152), donkey anti-rat DyLight405 (Jackson Immuno Research, 712-475-153), donkey anti-rat Alexa647 (Jackson Immuno Research, 712-606-153), donkey anti-goat DyLight405 (Jackson Immuno Research, 705-475-147), and donkey anti-goat Alexa647 (Jackson Immuno Research, 705-606-147). The labeling pattern observed with these antibodies closely resembled expression patterns of genes encoding these proteins or other neurotransmitter-related proteins, as previously observed with in situ hybridization (antibody target, gene ): hypocretin- hcrt ( Prober et al., 2006 ), cocaine- and amphetamine-regulated transcript (CART)- cart2 ( Thisse and Thisse, 2004 ), galanin- gal ( Podlasz et al., 2012 ), neuropeptide Y- npy ( Mathieu et al., 2002 ), neuropeptide VF- npvf ( Yelin-Bekerman et al., 2015 ), somatostatin- sst1.1 ( Herget and Ryu, 2015 ), serotonin- slc6a4a/b ( Norton et al., 2008 ), tyrosine hydroxylase- th1 ( Ryu et al., 2006 ; Filippi et al., 2010 ), dopamine beta-hydroxylase- dbh ( Holzschuh et al., 2003 ; Filippi et al., 2010 ), and choline acetyltransferase- vachta ( Hong et al., 2013 ).

Zebrafish in situ hybridization

After completion of live imaging (zoomed in, focused on hypothalamus), the fish was removed from agarose, submerged in ice-cold PBS, then placed into 4% PFA in PBS overnight at 4 °C. Afterward, protocols were followed as detailed in Choi et al. (2016) , using the hybridization chain reaction (HCR) technique. Twelve DNA 20-mer probes were designed against hypocretin ( hcrt ) mRNA ( Yelin-Bekerman et al., 2015 ) using Stellaris software (LGC Biosearch Technologies), with added ‘B1’ amplifiers to the 5′ and 3′ ends of each probe, and B1-linked Alexa 647 hairpins (Molecular Instruments, Caltech). Fish were then re-mounted in agarose on a Petri dish lid (in SSCT), and imaged again under the two-photon microscope. Endogenous GCaMP6s fluorescence was sufficiently preserved under these conditions, and therefore no anti-GFP stain was required. Volume registration and cell-type identification Volumes were first created by averaging across imaging cycles (2 cycles for fixed samples, 10 cycles for live samples), and saved as .nrrd files in a mm scale. These volumes were live GCaMP, fixed GCaMP, each antibody channel, Z-brain atlas Tg(elavl3:H2B-RFP) volume, and each Z-brain mask (294 in total; Randlett et al., 2015 ). Z-brain volumes were generated by exporting sparse MATLAB arrays as .nrrd files. Volumes were moved into Amazon Web Service’s cloud computing environment (c3.8xlarge instance), where the Computational Morphometry Toolkit (CMTK; Rohlfing and Maurer, 2003 ) was installed. The fixed GCaMP volume was aligned to the live GCaMP volume using rigid affine and non-rigid b-splines approaches ( Figure S2A ). Once the final transformation was determined, the transformation coordinates were applied to the fixed GCaMP volume and each of the antibody volumes. In a similar manner, the Z-brain atlas Tg(elavl3:H2B-RFP) volume was aligned to the live GCaMP volume, and the resulting transformation was applied to the Z-brain Tg(elavl3:H2B-RFP) volume and each of the Z-brain mask volumes. From each of these volumes now aligned to the live GCaMP volume, z-planes were extracted that correspond to the z-planes with activity recorded (every 15 μm, from 15 μm below the dorsal extent and 15 μm above the ventral extent). For each antibody channel z-plane, antibody+ cell bodies were manually selected in Fiji/ImageJ ( Schindelin et al., 2012 ), and a binary image was saved. ROIs identified from live-imaged z-planes were identified as belonging to a neuromodulatory group by the intersection of antibody+ cells and z-brain masks (see Figure S2D ). An ROI was included in a neuromodulatory group of interest if 100% of the pixels in the ROI overlapped with the relevant Z-brain mask (indicating an anatomical region) and > 75% of pixels overlapped with antibody+ cell in the relevant antibody channel. Dopaminergic and noradrenergic neurons were distinguishable within a single tyrosine hydroxylase stain, using segmentation with the Z-brain atlas (confirmed with dual tyrosine hydroxylase and dopamine beta-hydroxylase staining, data not shown). For spinal projection neuron labeling, Texas Red Dextran (10000 mW, lysine fixable; Invitrogen, D1863) was pressure-injected into the spinal cord of 6 days post fertilization zebrafish anesthetized with 0.1% MS-222 (Sigma) and fully embedded in agarose. Fish were cut out of agarose and placed in normal fish system water to recover for 24–36 hours before live imaging. Cell distance among spinal projection neurons was defined by identifying the center of each soma in eight neurons per fish (in z-projections), in cells manually identifiable in all three images (live, fixed, fixed+registered). The measured cells were the Mauthner cell, MeLr, MeLc, and CaD, on both sides ( Orger et al., 2008 ). The x-y distance between the same cell in each image was measured in Fiji/ImageJ.

Selection of cell types in mice

We used gene expression atlases ( Lein et al., 2007 ) and published papers to find homologous or putatively homologous cell types between larval zebrafish and mice: Dopaminergic posterior tuberculum/A11: In both fish and mammals these nuclei contain the only dopaminergic cell type that projects to the spinal cord ( Tay et al., 2011 ; Koblinger et al., 2014 ), and are specified by the expression of common transcription factors ( Löhr et al., 2009 ). Serotonergic dorsal raphe neurons: In both fish and mammals these cells release serotonin, reside in the midline of the hind-brain/brainstem ( Abrams et al., 2004 ; McLean and Fetcho, 2004 ), express the serotonin transporter gene slc6a4 ( Norton et al., 2008 ), and project broadly throughout the forebrain, midbrain, and hindbrain ( Abrams et al., 2004 ; McLean and Fetcho, 2004 ). Noradrenergic locus coeruleus: In both fish and mammals these cells reside in the hindbrain/brainstem, express dopamine beta hydroxylase and tyrosine hydroxylase ( McLean and Fetcho, 2004 ), and project diffusely throughout the brain, including extensive forebrain innervation ( McLean and Fetcho, 2004 ; Aston-Jones and Cohen, 2005 ; Tay et al., 2011 ). Somatostatin neurons in the hypothalamus: Somatostatin neurons with interesting functional properties were found in the zebrafish hypothalamus ( Figures 3C and S3 ), and therefore we searched for somatostatin-expressing neurons in the mammalian hypothalamus. We found the highest levels of expression in the arcuate nucleus ( Palkovits et al., 1976 ; Herget and Ryu, 2015 ), and therefore chose to investigate these cells as a putative homologous cell type. We also noted that NPY cells and somatostatin cells were close to one another in fish ( Figure S3 ), and are both localized to the arcuate nucleus in mammals ( Palkovits et al., 1976 ; Atasoy et al., 2012 ; Campbell et al., 2017 ),. Cholinergic tegmentum/lateral dorsal tegmentum: In both fish and mammals, these cells are nearby but medial to the noradrenergic locus coeruleus and lateral to the midline serotonergic dorsal raphe in the hindbrain/brainstem ( Figures 2C and S3 ; Clemente et al., 2004 ; Wang and Morales, 2009 ; Lee and Dan, 2012 ). Neuropeptide-Y (NPY) cells in the hypothalamus: In both fish and mammals, NPY-expressing cells can be found in the hypothalamus ( Figure S3 ; Mathieu et al., 2002 ; Atasoy et al., 2012 ). While NPY cells can be found in multiple hypothalamic nuclei in mammals (in addition to other areas), we targeted the area of densest expression, the arcuate nucleus, where these neurons also express agouti-related protein (AGRP) and GABA ( Atasoy et al., 2012 ). Cocaine and amphetamine-related transcript (CART) cells in the tegmentum/Edinger-Westphal nucleus: In fish, we found prominent expression of CART in a small midbrain cluster of midline neurons in the tegmentum ( Figures 3B and S3 ). While CART is expressed in many regions of the mammalian brain ( Couceyro et al., 1997 ), it is prominent in the small midline Edinger-Westphal nucleus in the midbrain ( Giardino et al., 2012 ). In both fish and mice, this region of CART expression is relatively close to the cholinergic tegmentum, noradrenergic locus coeruleus, and serotonergic dorsal raphe ( Figure S3 ). CART neurons in the mammalian Edinger-Westphal nucleus are centrally projecting, in contrast to the preganglionic ChAT+ neurons in this nucleus that mediate premotor pupillary responses ( Giardino et al., 2012 ).

Mouse surgery

Mice were anesthetized under 3% isofluorane, and maintained with 1%–2% isofluorane in a stereotactic frame. Ophthalmic ointment was applied to prevent eyes from drying. The skull was exposed under aseptic conditions, and a small craniotomy was performed over the injection site. 300–400 nL virus was injected with a 10 μL syringe and a beveled metal needle (World Precision Instruments), at a rate of 150 nl/min, controlled by an injection pump (Harvard Apparatus). The syringe was slowly withdrawn 5–10 minutes after completion of injection. Following virus injection a 400 μm (for photometry) or 200 μm (for optogenetics) fiber (Doric) was slowly lowered to within 50–100 μm of the injection site and, together with a custom stainless steel headplate extending in the medial-lateral orientation, was fixed in place using tissue adhesive (3M Vetbond) and adhesive cement (C&B Metabond, Parkell). For mice used in optogenetic inactivation studies, viral injections and 200 μm fiber implants were performed bilaterally, with the exception of the midline Edinger-Westphal nucleus, where a single midline fiber sufficed. Mice were administered 0.05 mg/Kg Buprenorphine for analgesia, and placed on a warming pad to recover from isofluorane anesthesia. After recovery of coordinated locomotion, mice were returned to a clean home cage, and their health was monitored daily for the subsequent week. Mice begin water restriction one to two weeks after surgery. Injection coordinates were: −2.0 A/P, 0.5 M/L, −4.2 D/V (nucleus A11), −5.45 A/P, 1.0 M/L, −3.2 D/V (locus coeruleus), −5.0 A/P, 0.6 M/L, −3.0 D/V (lateral dorsal tegmentum), −4.35 A/P, 0.2 M/L, −2.1 D/V (dorsal raphe nucleus), −1.5 A/P, 0.1 M/L, −5.6 D/V (arcuate nucleus), −3.65 A/P, 0.0 M/L, −3.0 D/V (Edinger-Westphal nucleus). eYFP control mice were derived from multiple genotypes; GCaMP controls: 1 TH-IRES-Cre (locus coeruleus), and 2 SST-IRES-Cre (arcuate nucleus); head-fixed ChR2 controls: 3 ChAT-Cre (lateral dorsal tegmentum), 1 TH-IRES-Cre (locus coeruleus), and 1 CART-IRES2-Cre (Edinger-Westphal nucleus); open field test ChR2 controls: 2 ChAT-Cre (lateral dorsal tegmentum) 1 TH-IRES-Cre (locus coeruleus), and 1 CART-IRES2-Cre (Edinger-Westphal nucleus); NpHR controls: 3 TH-IRES-Cre (locus coeruleus) and 1 CART-IRES2-Cre (Edinger-Westphal nucleus).

Mouse behavior

Mice were water restricted for 1 day prior to beginning training, and were maintained at > 90% of their pre-deprivation weight throughout training. Mice were habituated to head fixation and trained to obtain auditory stimulus-triggered water rewards by contacting a lick port with their tongue, over the course of 4–5 days. Mice were head-fixed in a plastic tube using the implanted custom stainless steel headplate and mounting system. Behavioral responses were recorded using a custom lickometer built using a capacitive touch sensor (TinkerKit). The behavioral protocol was run on a real-time microcontroller (Arduino Due) using software communicating with MATLAB (MathWorks). The behavioral protocols were implemented using Bpod (courtesy of Josh Sanders, Cold Spring Harbor Lab). Sounds were delivered using a calibrated electrostatic speaker (Tucker Davis Technology, ES1) positioned ~10 cm from the mouse’s head, and were generated by custom MATLAB software and played through a data acquisition board (National Instruments, NI PCIe-6321). Tones were 5 kHz, 68 dB, 500 ms in duration, and presented at a pseudorandom inter-stimulus interval (range = 2.1–5 s). For caffeine experiments, trained mice were injected with caffeine (20 mg/Kg, i.p., Sigma) dissolved in saline or saline alone 20 min before behavioral testing. For head-fixed optogenetic experiments, mice received 20 Hz stimulation for 2 s prior to stimulus onset, delivered with a 473 nm DPSS laser (OEM Laser Systems) or constant illumination for 3 s (beginning 2 s prior to stimulus onset) with a 589 nm DPSS laser (OEM Laser Systems) and a Master-8 pulse generator (A.M.P.I.). Stimulation occurred on 50% of the trials in pseudorandom order. Light intensity was 7–10 mW from fiber tip. For freely moving optogenetic experiments, the experimenter was blinded to the condition. Mice acclimated to the behavior room for 1 hour before testing. Mice were connected to patch cords, allowed to recover for 3–5 minutes, then placed in the center of the open field chamber (50 cm × 50 cm, white). After a 5-minute acclimation period, 5-minute serial laser OFF and ON epochs commenced (OFF, tonicON, OFF, phasicON, OFF). For ChR2 animals, a 473 nm laser was pulsed (5 ms pulse with) tonically (5 Hz) or phasically (20 Hz for 500 ms every 2 s), during the 1st and 2nd stimulate epochs, respectively (with 5 minute laser off trials interspersed). For NpHR animals, the 589 nm laser constantly illuminated during the single stimulation epoch, with laser off epochs before and after (OFF, ON, OFF). Light intensity was 7–10 mW from the fiber tip. Behavior was recorded and automatically scored using BiObserve Viewer software.

Mouse pupil tracking

A high-speed camera (AVT Manta, G-032B) with a 24 mm/F1.4 lens (Edmund Optics #59871) was focused on the eye. Fiber-coupled infrared illumination (Thorlabs M780F2) was positioned near the eye so as to maximize contrast of the imaged pupil and to eliminate or minimize the size of the spectral reflection. In most cases, the single spectral reflection spot could not be eliminated and was instead positioned at the center of the eye, such that it was almost always contained within the pupil (and thus served only as a DC bias on the computed area of the pupil). Video was acquired at 60 Hz using pymba, a Python wrapper for AVT camera control, and frame times were recorded and synchronized to the behavioral task using a National Instruments DAQ (NI PCIe-6323). Video data were analyzed in MATLAB to extract time series corresponding to pupil area versus time during task. Manual calibration was used to crop a region-of-interest around the eye and to determine a threshold intensity value that distinguished between the inside and outside of the pupil. The size of the pupil at each time point was computed as the total number of pixels within the pupil divided by the total number of pixels covering the eye. For optogenetic experiments, mice received 20–30 trials of laser stimulation (ChR2 experiments: 20 Hz 473 nm stimulation for 2 s; NpHR experiments: constant 589 nm stimulation for 2 s) while recording the eye. A black rubber tube was wrapped around the fiber implant on the head to limit the pupillary dilation response to light from the fiber implant. We restricted our analyses to trials where the pupil was between the 5 th and 90 th percentile of all pupil values recorded in that mouse to avoid ceiling and floor effects. Change in pupil was quantified as the mean pupil size in the 0.5–2.0 s after laser onset, divided by the mean pupil size the −2.0-0 s prior to laser onset.

Mouse fiber photometry

The fiber photometry setup was similar to that described in Gunaydin et al. (2014) . Briefly, a fiber-coupled 470 nm LED (Thorlabs, M470F3) was reflected off of a dichroic mirror (Semrock, FF495) into a 600 μm, 0.48 NA optical fiber patch-cord (Doric). The patch-cord was then coupled to a 400 μm, 0.48 NA fiber (Doric, MFC_400/430-0.48_MF2.5_FLT) implanted in the mouse’s brain. The resulting GCaMP or GFP fluorescence was transmitted through the dichroic mirror and band-pass filter (Semrock, FF01-F20/35), and focused via a convex lens (Thorlabs, LA1255A) onto a photo-detector (Newport, 2151). A lock-in amplifier (Stanford Research Systems, SR810) was used to modulate the LED at 400 Hz, and then demodulate the resulting signal collected from the photo-detector. The demodulated signal was digitized at 1 kHz using a data acquisition board (National Instruments, NI PCIe-6321).

Mouse immunohistochemistry

Mice were deeply anesthetized and transcardially perfused with ice-cold PBS, followed by 4% PFA in PBS. Brains were removed and placed in 4% PFA in PBS for post-fixation. Brains were then placed in a 30% sucrose solution for 24 hours at 4 °C. Brains were then mounted, and placed on a freezing microtome for sectioning into 60 μm coronal sections. Sections were placed into PBS in 24 well plates, and washed at room temperature in PBST. Sections were incubated with primary antibodies (1:500 in PBST) overnight at 4 °C on a shaker, washed in PBST for 1 hour at room temperature on a shaker (repeated 4–5 times), then incubated with secondary antibodies (1:500 in PBST) overnight at 4 °C on a shaker. Sections were washed in PBST for 1 hour at room temperature on a shaker (repeated 4–5 times) then mounted on slides using PVA-DABCO, and imaged using an SP5 confocal microscope (Leica). The primary antibodies used were: mouse anti-tyrosine hydroxylase (Immunostar, 22941), rabbit anti-serotonin (Immunostar, 20080), rabbit anti-dopamine beta-hydroxylase (Immunostar, 22806), rat anti-somatostatin (Millipore, MAB354), rabbit anti-CART (Abcam, ab192364), rabbit anti-neuropeptideY (Sigma, N9528), and goat anti-choline acetyltransferase (Millipore, A144P). Secondary antibodies used were donkey anti-mouse Alexa647 (Jackson Immuno Research, 715-606-151), donkey anti-rabbit Alexa 594 (Jackson Immuno Research, 711-586-152), donkey anti-rat Alexa647 (Jackson Immuno Research, 712-606-153), and donkey anti-goat Alexa647 (Jackson Immuno Research, 705-606-147).

QUANTIFICATION AND STATISTICAL ANALYSIS

All analyses and visualizations were performed with custom code written in Python, using NumPy, Scipy, Matplotlib, IPython, Seaborn, Statsmodels, Pandas, Scikit-image, and Scikit-learn libraries ( Jones et al., 2001 ; Hunter, 2007 ; Perez and Granger, 2007 ; McKinney, 2010 ; Seabold and Perktold, 2010 ; Pedregosa et al., 2011 ; van der Walt et al., 2011 ; van der Walt et al., 2014 ).

Zebrafish analyses

The pixels containing the fish tail were determined in each frame of the tail monitoring videos using adaptive thresholding and blob detection algorithms in Scikit-image ( van der Walt et al., 2014 ; Madelaine et al., 2017 ). Tail movements were identified by analyzing the number of tail-containing pixels that did not overlap between adjacent frames. The mean and standard deviation of this value when the tail was motionless was estimated as the median of these statistics within all 300 ms time bins. Tail movements were identified as frames when this value remained 4 standard deviations above baseline for at least 40 ms. Tail movements that were separated by less than 50 ms were merged. To classify movement types, the orientation of the tail was computed for every frame as the angle between neutral tail position and the major-axis of the ellipse fit to second-moment of the pixels containing the tail. Movements were then classified as large-angle turns or escape-like if the maximum angle of deflection exceeded 23 degrees and the maximum velocity of deflection exceeded 1.4 degrees/ms. These movements were also classified in terms of their peak tail angle (95 th percentile of all measured angles within movement) and peak tail velocity (95 th percentile of tail orientation changes between frames, divided by the inter-frame interval (8 ms)). Further analyses were conducted on a binary array that spanned the entire behavioral session, with the times of turns/escape onsets noted as ones. RTs were defined as the time to first large angle tail turn after stimulus onset, within 3 s of stimulus onset. Trials were excluded if fish made any movement within the 500 ms prior to stimulus onset. Trials between the 2.5 th and 97.5 th percentile of RTs were used. For analysis of coarser trial response types, trials were defined as responses (all trials included for RT analysis, as described above), omissions (all trials with no movement of any kind from 3 s prior to stimulus onset until 6 s after stimulus onset), or premature responses (all trials with any kind of movement in the 500 ms prior to stimulus onset). Heart rate was obtained from heart recordings by extracting fluorescence time series from the center of the ventricle, where GFP+ cardiac tissue moved in and out with each beat. This resulting time series contained peaks at regular intervals—approximately 1.5–3 peaks per second. Peaks were identified, and a time series of heart rate was generated by smoothing instantaneous heart rate signal by a moving average with a width of 20 s. A correlation coefficient was defined using the RTs and the mean of the heart rate for 1.2 s before each stimulus onset, pooled for all fish tested. Correlation between total movement rate and heart rate were conducted by collecting mean movement rate values (for all types of movement) and mean heart rate in 20 s bins spanning the imaging session. Live GCaMP brain imaging was corrected for x-y motion using affine corrections in MoCo ( Dubbs et al., 2016 ), running in Fiji/ImageJ ( Schindelin et al., 2012 ). Fish were excluded if there was drift in the z-direction. From motion corrected frames, regions of interest (ROIs) were identified by segmenting nuclei from the 90 th percentile image of each plane taken across the entire imaging session. These images were auto-leveled and template-matched to a nucleus-sized disk before using local peak finding and the watershed algorithm to extract ROIs. Raw fluorescence from each ROI was extracted, turned into a dF/F trace (baseline F = 5 th percentile of fluorescence), and converted to a z-score. Pearson’s correlation coefficients were defined for the trials that were not excluded, using the RT (or peak tail angle, or peak tail velocity) and the mean of the neural activity for 1.2 s before stimulus onset. Correlations of neurons to visual or motor stimuli were performed by obtaining the Pearson’s correlation coefficient between the activity trace of a cell over the entire behavioral session and a constructed behavioral regressor. Regressors were constructed by convolving a binary array of event onsets for sensory stimuli or turning behavior with a 4 s decay to approximate GCaMP6s fluorescence decays. For analysis of cell types in visual regions, we analyzed neurons identified with the Z-brain atlas ( Randlett et al., 2015 ). Within visual areas (AF6, AF8, tectum), we analyzed neurons classified as being looming stimulus-tuned, as defined by cells with a significant (p < 0.05) correlation to visual regressors. These cells were then used to assess the correlation between RT and the peak neural response in the 4 s after stimulus onset. Prediction of trial type from neural activity was assessed using logistic regression. Behavioral trials were assigned as responses or omissions (as described above). For each fish, an array of binary values was generated to denote trials that did (1) or did not (0) fit into each category, resulting in two behavior classification arrays. For prediction, we used the mean of the neural activity for 1.2 s before stimulus onset. Data were randomly split into training and testing sets (training set = 50% of trials), and neurons were classified as predictive of trial type if the area under the ROC curve exceeded chance (0.5). For analysis of caffeine-activated cells, the mean and standard deviation of the time series was determined for the 3 minutes before caffeine application and 3 minutes during caffeine exposure (starting 3 minutes after initial caffeine application). For analysis of correlations between neuromodulatory cell types, we used the spontaneous activity recordings from fully embedded fish in the absence of sensory stimulation (5–6 minute recording at 3 Hz). For analysis of correlation during the task, we used fish with greater than 20 RTs (n = 18) to analyze the difference between fast RTs (bottom 25 th percentile of RTs) and slow RTs (top 25 th percentile of RTs). For each trial type, we assessed correlations between cell types from 4 s prior to stimulus onset until 6 s after stimulus onset. For prediction analyses using individual neurons, we used fish with at least 35 RTs and greater than 20 alertness+ and greater than 20 alertness− neurons simultaneously recorded (n = 4). For prediction analysis using unique cell classes, we used fish where at least three alertness+ cell types were simultaneously recorded, and all neurons within a group were averaged to generate one time series per neuromodulatory group. The pre-stimulus neural activity (1.2 s) was used to predict the RT, using one to 15 simultaneously recorded cells, or one to three simultaneously recorded mean cell types, as inputs. Data were randomly split into training and testing sets (training set = 80% of trials), and the accuracy of prediction was assessed by the Pearson’s correlation coefficient of the true and predicted RT values. As a control, a matched number of alertness− neuromodulatory cells or cell types recorded from the same fish were subjected to the same analysis.

Mouse analyses

Mouse behavior was determined from the licks recorded on the lickometer, and the timing of auditory stimulus onset. RTs were defined as the time to first lick after stimulus onset, within 800 ms of stimulus onset. Mouse photometry datasets were comprised of two consecutive days of recording, and optogenetic experiments comprised one day. For each type of experiment, the first 10 trials and final 20 trials of each daily session were excluded to mitigate the effects of experimenter presence at the beginning of the session and satiation at the end. Trials were also excluded if mice initiated licking in the 150 ms before, and 100 ms after stimulus onset. For analysis of coarser trial response types, trials were defined as responses (all trials included for RT analysis, as described above) or omissions (all trials with no licks from 500 ms prior to stimulus onset until 1 s after stimulus onset). For freely moving optogenetics experiments, behavioral variables were scored as the mean of first 3 minutes of each behavioral epoch. For each mouse, values were established as the differences between the light ON trial and the preceding light OFF epoch. Neural activity was quantified as dF/F for each trial. Baseline F was defined for each trial as the 10 th percentile of all activity in a sliding window of five trials. Pearson’s correlation coefficients were defined between the RT and the max of the neural activity for 1.2 s before stimulus onset. Pupil area was also analyzed in the 1.2 before a stimulus, to determine the correlations with RT.

Statistical tests

Statistical tests were conducted using NumPy, Scipy, Statsmodels, and Scikit-learn packages in Python. Distributions were tested for normality using the Shapiro-Wilk test and equality of variance using Levene’s test. For data with normal distributions and equal variance, one-way ANOVAs, two-sided Student’s t tests, paired t tests, or permutation tests (10,000 permutations) were used. Otherwise, non-parametric tests were used for comparing multiple groups (Kruskal-Wallis test), two independent groups (two-sided Mann-Whitney rank test), two related groups (two-sided Wilcoxon signed-rank test), or assessing the difference of single groups from zero (one-sample Wilcoxon signed-rank test). To correct for multiple comparisons (multiple one-way tests or multiple comparisons with a common control group), p values were subjected to a false discovery rate correction.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

All procedures were approved by the Stanford University Institutional Animal Care and Use Committee. Zebrafish We used 7–10 days post fertilization larval zebrafish for this study, bred on a Nacre or Casper background. No statistical methods were used to predetermine sample size, and animal selection was not randomized or blinded. We used Tg(elavl3:H2B-GCaMP6s) fish ( Vladimirov et al., 2014 ) for imaging, registration, and behavioral experiments, and Tg(cmlc2:GFP) fish ( Huang et al., 2003 ) for heart-rate imaging. Fish were group-housed under a 14:10 light:dark cycle until the day of experiments, and were fed with paramecia (Parameciavap) twice daily from 5–6 days post fertilization onward. All testing occurred during the late morning and afternoon.

Mice

We used adult mice for this study (male and female, > P45), consisting of TH-IRES-Cre mice (EMMA, Ted Ebebdal; Lindeberg et al., 2004 ), SERT-Cre mice (MMRRC, Slc6a4-cre, #017260-UCD), ChAT-Cre mice (MMRRC, #017269-UCD), SST-IRES-Cre mice (The Jackson Laboratory, #013044), AGRP-IRES-Cre mice (The Jackson Laboratory, #012899), and CART-IRES2-Cre mice (The Jackson Laboratory, #028533). All mice were injected unilaterally, bilaterally, or in the midline with 400 nL of either AAVDJ-Ef1α-DIO-GCaMP6m (3 × 10 12 viral particles/ml; Chen et al., 2013 ), AAV1-Ef1α-DIO-ChR2-eYFP (3 × 10 12 viral particles/ml), AAVDJ-Ef1α-DIO-eYFP (3 × 10 12 viral particles/ml), or AAVDJ-Ef1α-DIO-eNpHR3.0-eYFP (3 × 10 12 viral particles/ml). GCaMP6m (Addgene plasmid # 40754) was a gift from Douglas Kim (Janelia Research Campus) and modified in our lab. No statistical methods were used to predetermine sample size, and animal selection was not randomized. For freely moving behavior in the open field test, the experimenter was blinded. We did not systematically test for differences between male and female mice in this study, due to low and/or uneven numbers of age-matched male and females in each experimental condition. Mice were either group-or singly housed and maintained under a 14:10 reverse light:dark cycle, and had ad libitum access to food and water, until water restriction began (see below). All testing occurred during the morning and early afternoon.

METHOD DETAILS Zebrafish behavior

For all experiments, larval zebrafish were embedded in 2.5% low melting point agarose (Millipore) in the lid of a 35 mm Petri dish (E&K Scientific). Fish were embedded in agarose for 2–6 hours prior to beginning experiments. Once agarose solidified after embedding, agarose posterior to the pectoral fins was carefully removed with a scalpel, so that the tail was free to move. Petri dishes were placed on the lid of a 100 mm Petri dish (Fisher), with a diffuser screen attached to the bottom with vacuum grease (with a hole cut out for the tail). Behavioral monitoring, stimulus generation, and recording of frame times from the two-photon microscope were achieved with custom software written in MATLAB (MathWorks). Visual stimuli were projected at 60 Hz using a Laser Pico Projector (MicroVision), surrounded by three Red Wratten filters (Kodak) to prevent interference with the two-photon detection system. Visual stimuli were directed off a dichroic mirror to project on the bottom of the Petri dish with the embedded fish. The tail of the fish was illuminated by IR lights from above and behind the fish, and tail movements were filmed at 120 frames per second from below with an AVT Manta G 031 camera (Allied Vision) through a AF-S DX Micro Nikkor 85 mm f/3.5G ED VR macro lens (Nikon). Looming dot stimuli were black disks on a red background, and were presented ~0.5 mm to the right of the fish’s right eye, were initiated as ~1° (0.01 cm diameter) and expanded to ~100°(1.3 cm diameter) over the course of 3.5 s (~0.37 cm/s). Each stimulus in a behavioral session was identical, and were presented with a pseudorandom inter stimulus interval (range = 8 to 12 s). Optomotor stimuli were a black and red checkerboard pattern, presented to a fish ~0.1 cm away (squares = 0.25 cm across; ~100°), that drifted in the caudal to rostral direction (0.4 cm/s), for 10 s at a time, with a pseudorandom inter stimulus interval (range = 8 to 12 s). For exogenous manipulations of alertness, fish were either treated with caffeine (50 mg/L caffeine (Sigma), dissolved in fish water, 45 minutes prior to beginning of behavior), or one night of sleep deprivation (a Petri dish of 5–10 fish was placed on a slowly rotating shaker at room temp overnight in constant light).

Zebrafish two-photon microscopy

Two-photon imaging was performed using an Olympus FVMPE multiphoton microscope (Olympus Corporation), with a resonant scanner, in either unidirectional or bidirectional scanning mode. We used a 16x objective (0.8 NA; Nikon) for most experiments. Heart rate imaging was performed at 2x zoom in a single z-plane with 6x frame averaging, at 8.3 frames/second. Functional brain imaging was performed at 1.1x zoom in 14–16 z-planes, separated by 15 μm, at 1.3–2.8 volumes/second (5000–8000 volumes). After completion of behavior and functional brain imaging, a structural stack was obtained at 1 μm spacing and 16x frame averaging, starting 15 μm above the first z-plane, ending 15 μm below the last z-plane, and repeated 10 times. Fast axial scanning was achieved with a piezo motor coupled to the objective. Fast functional imaging of GCaMP and GFP were conducted at 920 nm. Structural scans in live fish were conducted at 860 nm, in order to obtain the structure of GCaMP+ cells independent of calcium-dependent fluorescence. Structural scans in fixed fish were conducted at 810 nm (DyLight 405 dye), 920 nm (GCaMP), and 1150 nm (Alexa594 and Alexa647 dyes), and split into four spectrally separable PMTs to detect each dye independently. Fixed samples were imaged at each wavelength at 1 μm spacing and 16x frame averaging, and repeated twice. In both live and fixed structural stacks, voxels were 1.4 × 1.4 × 1 μm in size. All images were corrected for axial chromatic aberrations induced by the objective, as determined by measurements of axial displacement from imaging multi-colored beads across each of the four detectors (data not shown). For brain imaging of caffeine-treated fish, fish were fully embedded in agarose and imaged with a 25x objective (1.05 NA; Olympus) at 2.6 volumes/second, for ~6 minutes before and 6 minutes after acute caffeine (50 mg/L) application. For heart imaging under caffeine and sleep deprivation, fish were fully embedded in agarose (right side up or upside down) and resting heart rate was imaged at 15–20 Hz. Fish were either sleep-deprived, treated with caffeine 45 minutes prior to imaging, or treated with vehicle (fish water).

Zebrafish immunohistochemistry

After completion of behavior and imaging, a small block of agarose containing the fish was cut out, submerged in ice-cold PBS, then placed into 4% PFA in phosphate-buffered saline (PBS; pH 7.4, Life Technologies) with 0.2% Triton-X (PBST) in a 1.5 mL tube, overnight on a shaker at 4 °C. Samples were then washed with PBST and left on a shaker at room temperature for 2–4 hours (repeated 2–3 times). Primary antibodies were then applied at 1:200 in PBST, and samples were left on a shaker at 4 °C for 48 hours. Samples were then washed with PBST and left on a shaker at room temperature for 2–4 hours (repeated 2–3 times). Secondary antibodies were then applied at 1:200 in PBST, and samples were left on a shaker at 4 °C for 48 hours. Samples were then washed with PBST and left on a shaker at room temperature for 2–4 hours (repeated 2–3 times). At this point, whole-mount zebrafish samples (still in their agarose block) were re-mounted in agarose on a Petri dish lid, and imaged again under the two-photon microscope. The primary antibodies used were: mouse anti-tyrosine hydroxylase (Immunostar, 22941), rabbit anti-serotonin (Immunostar, 20080), rabbit anti-orexin (Millipore, AB3704), goat anti-orexin (Santa Cruz Biotechnology, sc-8070), rabbit anti-CART (Abcam, ab192364), rabbit anti-galanin (Millipore, ab5909), rabbit anti-neuropeptideVF (Sigma, HPA041733), rabbit anti-neuropeptideY (Sigma, N9528), rat anti-somatostatin (Abcam, ab30788), rabbit anti-dopamine beta-hydroxylase (Immunostar, 22806), and goat anti-choline acetyltransferase (Millipore, A144P). Secondary antibodies used were donkey anti-mouse Alexa647 (Jackson Immuno Research, 715-606-151), donkey anti-rabbit Alexa 594 (Jackson Immuno Research, 711-586-152), donkey anti-rat DyLight405 (Jackson Immuno Research, 712-475-153), donkey anti-rat Alexa647 (Jackson Immuno Research, 712-606-153), donkey anti-goat DyLight405 (Jackson Immuno Research, 705-475-147), and donkey anti-goat Alexa647 (Jackson Immuno Research, 705-606-147). The labeling pattern observed with these antibodies closely resembled expression patterns of genes encoding these proteins or other neurotransmitter-related proteins, as previously observed with in situ hybridization (antibody target, gene ): hypocretin- hcrt ( Prober et al., 2006 ), cocaine- and amphetamine-regulated transcript (CART)- cart2 ( Thisse and Thisse, 2004 ), galanin- gal ( Podlasz et al., 2012 ), neuropeptide Y- npy ( Mathieu et al., 2002 ), neuropeptide VF- npvf ( Yelin-Bekerman et al., 2015 ), somatostatin- sst1.1 ( Herget and Ryu, 2015 ), serotonin- slc6a4a/b ( Norton et al., 2008 ), tyrosine hydroxylase- th1 ( Ryu et al., 2006 ; Filippi et al., 2010 ), dopamine beta-hydroxylase- dbh ( Holzschuh et al., 2003 ; Filippi et al., 2010 ), and choline acetyltransferase- vachta ( Hong et al., 2013 ).

Zebrafish in situ hybridization

After completion of live imaging (zoomed in, focused on hypothalamus), the fish was removed from agarose, submerged in ice-cold PBS, then placed into 4% PFA in PBS overnight at 4 °C. Afterward, protocols were followed as detailed in Choi et al. (2016) , using the hybridization chain reaction (HCR) technique. Twelve DNA 20-mer probes were designed against hypocretin ( hcrt ) mRNA ( Yelin-Bekerman et al., 2015 ) using Stellaris software (LGC Biosearch Technologies), with added ‘B1’ amplifiers to the 5′ and 3′ ends of each probe, and B1-linked Alexa 647 hairpins (Molecular Instruments, Caltech). Fish were then re-mounted in agarose on a Petri dish lid (in SSCT), and imaged again under the two-photon microscope. Endogenous GCaMP6s fluorescence was sufficiently preserved under these conditions, and therefore no anti-GFP stain was required. Volume registration and cell-type identification Volumes were first created by averaging across imaging cycles (2 cycles for fixed samples, 10 cycles for live samples), and saved as .nrrd files in a mm scale. These volumes were live GCaMP, fixed GCaMP, each antibody channel, Z-brain atlas Tg(elavl3:H2B-RFP) volume, and each Z-brain mask (294 in total; Randlett et al., 2015 ). Z-brain volumes were generated by exporting sparse MATLAB arrays as .nrrd files. Volumes were moved into Amazon Web Service’s cloud computing environment (c3.8xlarge instance), where the Computational Morphometry Toolkit (CMTK; Rohlfing and Maurer, 2003 ) was installed. The fixed GCaMP volume was aligned to the live GCaMP volume using rigid affine and non-rigid b-splines approaches ( Figure S2A ). Once the final transformation was determined, the transformation coordinates were applied to the fixed GCaMP volume and each of the antibody volumes. In a similar manner, the Z-brain atlas Tg(elavl3:H2B-RFP) volume was aligned to the live GCaMP volume, and the resulting transformation was applied to the Z-brain Tg(elavl3:H2B-RFP) volume and each of the Z-brain mask volumes. From each of these volumes now aligned to the live GCaMP volume, z-planes were extracted that correspond to the z-planes with activity recorded (every 15 μm, from 15 μm below the dorsal extent and 15 μm above the ventral extent). For each antibody channel z-plane, antibody+ cell bodies were manually selected in Fiji/ImageJ ( Schindelin et al., 2012 ), and a binary image was saved. ROIs identified from live-imaged z-planes were identified as belonging to a neuromodulatory group by the intersection of antibody+ cells and z-brain masks (see Figure S2D ). An ROI was included in a neuromodulatory group of interest if 100% of the pixels in the ROI overlapped with the relevant Z-brain mask (indicating an anatomical region) and > 75% of pixels overlapped with antibody+ cell in the relevant antibody channel. Dopaminergic and noradrenergic neurons were distinguishable within a single tyrosine hydroxylase stain, using segmentation with the Z-brain atlas (confirmed with dual tyrosine hydroxylase and dopamine beta-hydroxylase staining, data not shown). For spinal projection neuron labeling, Texas Red Dextran (10000 mW, lysine fixable; Invitrogen, D1863) was pressure-injected into the spinal cord of 6 days post fertilization zebrafish anesthetized with 0.1% MS-222 (Sigma) and fully embedded in agarose. Fish were cut out of agarose and placed in normal fish system water to recover for 24–36 hours before live imaging. Cell distance among spinal projection neurons was defined by identifying the center of each soma in eight neurons per fish (in z-projections), in cells manually identifiable in all three images (live, fixed, fixed+registered). The measured cells were the Mauthner cell, MeLr, MeLc, and CaD, on both sides ( Orger et al., 2008 ). The x-y distance between the same cell in each image was measured in Fiji/ImageJ.

Selection of cell types in mice

We used gene expression atlases ( Lein et al., 2007 ) and published papers to find homologous or putatively homologous cell types between larval zebrafish and mice: Dopaminergic posterior tuberculum/A11: In both fish and mammals these nuclei contain the only dopaminergic cell type that projects to the spinal cord ( Tay et al., 2011 ; Koblinger et al., 2014 ), and are specified by the expression of common transcription factors ( Löhr et al., 2009 ). Serotonergic dorsal raphe neurons: In both fish and mammals these cells release serotonin, reside in the midline of the hind-brain/brainstem ( Abrams et al., 2004 ; McLean and Fetcho, 2004 ), express the serotonin transporter gene slc6a4 ( Norton et al., 2008 ), and project broadly throughout the forebrain, midbrain, and hindbrain ( Abrams et al., 2004 ; McLean and Fetcho, 2004 ). Noradrenergic locus coeruleus: In both fish and mammals these cells reside in the hindbrain/brainstem, express dopamine beta hydroxylase and tyrosine hydroxylase ( McLean and Fetcho, 2004 ), and project diffusely throughout the brain, including extensive forebrain innervation ( McLean and Fetcho, 2004 ; Aston-Jones and Cohen, 2005 ; Tay et al., 2011 ). Somatostatin neurons in the hypothalamus: Somatostatin neurons with interesting functional properties were found in the zebrafish hypothalamus ( Figures 3C and S3 ), and therefore we searched for somatostatin-expressing neurons in the mammalian hypothalamus. We found the highest levels of expression in the arcuate nucleus ( Palkovits et al., 1976 ; Herget and Ryu, 2015 ), and therefore chose to investigate these cells as a putative homologous cell type. We also noted that NPY cells and somatostatin cells were close to one another in fish ( Figure S3 ), and are both localized to the arcuate nucleus in mammals ( Palkovits et al., 1976 ; Atasoy et al., 2012 ; Campbell et al., 2017 ),. Cholinergic tegmentum/lateral dorsal tegmentum: In both fish and mammals, these cells are nearby but medial to the noradrenergic locus coeruleus and lateral to the midline serotonergic dorsal raphe in the hindbrain/brainstem ( Figures 2C and S3 ; Clemente et al., 2004 ; Wang and Morales, 2009 ; Lee and Dan, 2012 ). Neuropeptide-Y (NPY) cells in the hypothalamus: In both fish and mammals, NPY-expressing cells can be found in the hypothalamus ( Figure S3 ; Mathieu et al., 2002 ; Atasoy et al., 2012 ). While NPY cells can be found in multiple hypothalamic nuclei in mammals (in addition to other areas), we targeted the area of densest expression, the arcuate nucleus, where these neurons also express agouti-related protein (AGRP) and GABA ( Atasoy et al., 2012 ). Cocaine and amphetamine-related transcript (CART) cells in the tegmentum/Edinger-Westphal nucleus: In fish, we found prominent expression of CART in a small midbrain cluster of midline neurons in the tegmentum ( Figures 3B and S3 ). While CART is expressed in many regions of the mammalian brain ( Couceyro et al., 1997 ), it is prominent in the small midline Edinger-Westphal nucleus in the midbrain ( Giardino et al., 2012 ). In both fish and mice, this region of CART expression is relatively close to the cholinergic tegmentum, noradrenergic locus coeruleus, and serotonergic dorsal raphe ( Figure S3 ). CART neurons in the mammalian Edinger-Westphal nucleus are centrally projecting, in contrast to the preganglionic ChAT+ neurons in this nucleus that mediate premotor pupillary responses ( Giardino et al., 2012 ).

Mouse surgery

Mice were anesthetized under 3% isofluorane, and maintained with 1%–2% isofluorane in a stereotactic frame. Ophthalmic ointment was applied to prevent eyes from drying. The skull was exposed under aseptic conditions, and a small craniotomy was performed over the injection site. 300–400 nL virus was injected with a 10 μL syringe and a beveled metal needle (World Precision Instruments), at a rate of 150 nl/min, controlled by an injection pump (Harvard Apparatus). The syringe was slowly withdrawn 5–10 minutes after completion of injection. Following virus injection a 400 μm (for photometry) or 200 μm (for optogenetics) fiber (Doric) was slowly lowered to within 50–100 μm of the injection site and, together with a custom stainless steel headplate extending in the medial-lateral orientation, was fixed in place using tissue adhesive (3M Vetbond) and adhesive cement (C&B Metabond, Parkell). For mice used in optogenetic inactivation studies, viral injections and 200 μm fiber implants were performed bilaterally, with the exception of the midline Edinger-Westphal nucleus, where a single midline fiber sufficed. Mice were administered 0.05 mg/Kg Buprenorphine for analgesia, and placed on a warming pad to recover from isofluorane anesthesia. After recovery of coordinated locomotion, mice were returned to a clean home cage, and their health was monitored daily for the subsequent week. Mice begin water restriction one to two weeks after surgery. Injection coordinates were: −2.0 A/P, 0.5 M/L, −4.2 D/V (nucleus A11), −5.45 A/P, 1.0 M/L, −3.2 D/V (locus coeruleus), −5.0 A/P, 0.6 M/L, −3.0 D/V (lateral dorsal tegmentum), −4.35 A/P, 0.2 M/L, −2.1 D/V (dorsal raphe nucleus), −1.5 A/P, 0.1 M/L, −5.6 D/V (arcuate nucleus), −3.65 A/P, 0.0 M/L, −3.0 D/V (Edinger-Westphal nucleus). eYFP control mice were derived from multiple genotypes; GCaMP controls: 1 TH-IRES-Cre (locus coeruleus), and 2 SST-IRES-Cre (arcuate nucleus); head-fixed ChR2 controls: 3 ChAT-Cre (lateral dorsal tegmentum), 1 TH-IRES-Cre (locus coeruleus), and 1 CART-IRES2-Cre (Edinger-Westphal nucleus); open field test ChR2 controls: 2 ChAT-Cre (lateral dorsal tegmentum) 1 TH-IRES-Cre (locus coeruleus), and 1 CART-IRES2-Cre (Edinger-Westphal nucleus); NpHR controls: 3 TH-IRES-Cre (locus coeruleus) and 1 CART-IRES2-Cre (Edinger-Westphal nucleus).

Mouse behavior

Mice were water restricted for 1 day prior to beginning training, and were maintained at > 90% of their pre-deprivation weight throughout training. Mice were habituated to head fixation and trained to obtain auditory stimulus-triggered water rewards by contacting a lick port with their tongue, over the course of 4–5 days. Mice were head-fixed in a plastic tube using the implanted custom stainless steel headplate and mounting system. Behavioral responses were recorded using a custom lickometer built using a capacitive touch sensor (TinkerKit). The behavioral protocol was run on a real-time microcontroller (Arduino Due) using software communicating with MATLAB (MathWorks). The behavioral protocols were implemented using Bpod (courtesy of Josh Sanders, Cold Spring Harbor Lab). Sounds were delivered using a calibrated electrostatic speaker (Tucker Davis Technology, ES1) positioned ~10 cm from the mouse’s head, and were generated by custom MATLAB software and played through a data acquisition board (National Instruments, NI PCIe-6321). Tones were 5 kHz, 68 dB, 500 ms in duration, and presented at a pseudorandom inter-stimulus interval (range = 2.1–5 s). For caffeine experiments, trained mice were injected with caffeine (20 mg/Kg, i.p., Sigma) dissolved in saline or saline alone 20 min before behavioral testing. For head-fixed optogenetic experiments, mice received 20 Hz stimulation for 2 s prior to stimulus onset, delivered with a 473 nm DPSS laser (OEM Laser Systems) or constant illumination for 3 s (beginning 2 s prior to stimulus onset) with a 589 nm DPSS laser (OEM Laser Systems) and a Master-8 pulse generator (A.M.P.I.). Stimulation occurred on 50% of the trials in pseudorandom order. Light intensity was 7–10 mW from fiber tip. For freely moving optogenetic experiments, the experimenter was blinded to the condition. Mice acclimated to the behavior room for 1 hour before testing. Mice were connected to patch cords, allowed to recover for 3–5 minutes, then placed in the center of the open field chamber (50 cm × 50 cm, white). After a 5-minute acclimation period, 5-minute serial laser OFF and ON epochs commenced (OFF, tonicON, OFF, phasicON, OFF). For ChR2 animals, a 473 nm laser was pulsed (5 ms pulse with) tonically (5 Hz) or phasically (20 Hz for 500 ms every 2 s), during the 1st and 2nd stimulate epochs, respectively (with 5 minute laser off trials interspersed). For NpHR animals, the 589 nm laser constantly illuminated during the single stimulation epoch, with laser off epochs before and after (OFF, ON, OFF). Light intensity was 7–10 mW from the fiber tip. Behavior was recorded and automatically scored using BiObserve Viewer software.

Mouse pupil tracking

A high-speed camera (AVT Manta, G-032B) with a 24 mm/F1.4 lens (Edmund Optics #59871) was focused on the eye. Fiber-coupled infrared illumination (Thorlabs M780F2) was positioned near the eye so as to maximize contrast of the imaged pupil and to eliminate or minimize the size of the spectral reflection. In most cases, the single spectral reflection spot could not be eliminated and was instead positioned at the center of the eye, such that it was almost always contained within the pupil (and thus served only as a DC bias on the computed area of the pupil). Video was acquired at 60 Hz using pymba, a Python wrapper for AVT camera control, and frame times were recorded and synchronized to the behavioral task using a National Instruments DAQ (NI PCIe-6323). Video data were analyzed in MATLAB to extract time series corresponding to pupil area versus time during task. Manual calibration was used to crop a region-of-interest around the eye and to determine a threshold intensity value that distinguished between the inside and outside of the pupil. The size of the pupil at each time point was computed as the total number of pixels within the pupil divided by the total number of pixels covering the eye. For optogenetic experiments, mice received 20–30 trials of laser stimulation (ChR2 experiments: 20 Hz 473 nm stimulation for 2 s; NpHR experiments: constant 589 nm stimulation for 2 s) while recording the eye. A black rubber tube was wrapped around the fiber implant on the head to limit the pupillary dilation response to light from the fiber implant. We restricted our analyses to trials where the pupil was between the 5 th and 90 th percentile of all pupil values recorded in that mouse to avoid ceiling and floor effects. Change in pupil was quantified as the mean pupil size in the 0.5–2.0 s after laser onset, divided by the mean pupil size the −2.0-0 s prior to laser onset.

Mouse fiber photometry

The fiber photometry setup was similar to that described in Gunaydin et al. (2014) . Briefly, a fiber-coupled 470 nm LED (Thorlabs, M470F3) was reflected off of a dichroic mirror (Semrock, FF495) into a 600 μm, 0.48 NA optical fiber patch-cord (Doric). The patch-cord was then coupled to a 400 μm, 0.48 NA fiber (Doric, MFC_400/430-0.48_MF2.5_FLT) implanted in the mouse’s brain. The resulting GCaMP or GFP fluorescence was transmitted through the dichroic mirror and band-pass filter (Semrock, FF01-F20/35), and focused via a convex lens (Thorlabs, LA1255A) onto a photo-detector (Newport, 2151). A lock-in amplifier (Stanford Research Systems, SR810) was used to modulate the LED at 400 Hz, and then demodulate the resulting signal collected from the photo-detector. The demodulated signal was digitized at 1 kHz using a data acquisition board (National Instruments, NI PCIe-6321).

Mouse immunohistochemistry

Mice were deeply anesthetized and transcardially perfused with ice-cold PBS, followed by 4% PFA in PBS. Brains were removed and placed in 4% PFA in PBS for post-fixation. Brains were then placed in a 30% sucrose solution for 24 hours at 4 °C. Brains were then mounted, and placed on a freezing microtome for sectioning into 60 μm coronal sections. Sections were placed into PBS in 24 well plates, and washed at room temperature in PBST. Sections were incubated with primary antibodies (1:500 in PBST) overnight at 4 °C on a shaker, washed in PBST for 1 hour at room temperature on a shaker (repeated 4–5 times), then incubated with secondary antibodies (1:500 in PBST) overnight at 4 °C on a shaker. Sections were washed in PBST for 1 hour at room temperature on a shaker (repeated 4–5 times) then mounted on slides using PVA-DABCO, and imaged using an SP5 confocal microscope (Leica). The primary antibodies used were: mouse anti-tyrosine hydroxylase (Immunostar, 22941), rabbit anti-serotonin (Immunostar, 20080), rabbit anti-dopamine beta-hydroxylase (Immunostar, 22806), rat anti-somatostatin (Millipore, MAB354), rabbit anti-CART (Abcam, ab192364), rabbit anti-neuropeptideY (Sigma, N9528), and goat anti-choline acetyltransferase (Millipore, A144P). Secondary antibodies used were donkey anti-mouse Alexa647 (Jackson Immuno Research, 715-606-151), donkey anti-rabbit Alexa 594 (Jackson Immuno Research, 711-586-152), donkey anti-rat Alexa647 (Jackson Immuno Research, 712-606-153), and donkey anti-goat Alexa647 (Jackson Immuno Research, 705-606-147).