Mice. All the biosafety procedures for handling and sacrificing animals were approved by the bioethics committee from the Consejo Superior de Investigaciones Científicas (CSIC), Comunidad de Madrid (PROEX 50.8/23) and followed the European Commission guidelines for the welfare of experimental animals (2010/63/EU, 86/609/EEC). C57BL/6J wild-type mice between 1-3 months of both sexes were used, and were housed in standard laboratory cages with ad libitum access to food under a 12:12 h dark-light cycle in temperature-controlled rooms. All animals were obtained from the animal facility of the Cajal Institute (registration number ES280790000184).
AstroLight viral vectors. pAAV-TRE-ChR2-YFP, pAAV-M13-TEV-C-P2A-TdTomato and pAAV-TM-CaM-NES-TEV-N-AsLOV2-TEVseq-tTA plasmids were acquired from Addgene (Items # 171622, #92391 and # 92392 respectively). The hSyn promoter was substituted by the GFAP promoter short version GFAP104 and cloned into AAV9 particles by Unitat de Producció de Vectors (UPV), Universitat Autonoma Barcelona.
AAV injection and cannula implants. All mice used in this study underwent AAV injection and cannula implantation. For initial characterization (Fig. 1), 2 combinations of AAVs were used: (1) animals co-infected with AAV9-Syn-ChrimsonR-tdTomato in mPFC and AAV5-GFAP-cyto-GCaMP6f in the NAc; (2) animals co-infected with AAV9-Syn-ChrimsonR-tdTomato in mPFC and AstroLight AAVs in the NAc. For the operant conditioning behavioral experiments (Fig. 2, 3 and 4), 3 different AAVs were used: (1) animals injected with AstroLight AAVs in the NAc, (2) animals infected with AAV5-GFAP-cyto-GCaMP6f in the NAc and (3) animals infected with AAV5-GFAP-hChR2(H134R)-mCherry in the NAc.
We used the following viral constructs: AstroLight AAVs [AAV9-GFAP-M13-TEV-C-P2a-TdTomato (viral titer 1.01 x 1013 gc/ml); AAV9-GFAP-TM-CaM-NES-TEV-N-AsLOV2 (viral titer 1.23 x 1013 gc/ml) AAV9-TRE-ChR2-YFP (viral titer 2.5 x 1012 gc/ml] mixed in a 1:1:2 ratio, AAV5-GFAP-cyto-GCaMP6f (viral titer 6.13 x 1013 gc/ml; PENN Vector Core), AAV5-GFAP-hChR2(H134R)-mCherry (viral titer 2.4 x 1012 gc/ml, UNC vector core) and AAV9-Syn-ChrimsonR-tdTomato (viral titer 4.1 x 1012 gc/ml, UNC vector core). Mice between P21-P30 were anesthetized with 2% isoflurane in oxygen and placed in a stereotaxic frame. Stereotaxic unilateral injections (400 nl; 60 nl/min) were made in the NAc (AP: 1.45 mm ML: ± 0.65 mm DV: - 4.3 mm) and for the initial ex vivo experiments, ChrimsonR was also expressed in the mPFC (AP: 1.7 mm ML: ± 0.3 mm DV: - 2.3 mm). After injection, the pipette was held in place for 10 min prior to retraction to avoid leakage, then removed and exposed cranium was dried and scraped for cannula implantation. Fiber photometry cannulas (400 µm diameter; 5 mm; borosilicate; 0.66 NA) were placed DV 0.1 mm above the previous NAc viral injection coordinates. Before implantation, greater than 50% of light transmission was confirmed for every cannula, the ones that did not meet these criteria were discarded. The cannulas were secured to the skull with dental cement (Meron, Voco). After surgery, animals were allowed to recover from anesthesia with the help of heating pads and returned to the cage once they showed regular breathing and locomotion. Behavioral testing started 2-3 weeks after surgery to allow for viral expression and animal recovery.
Slice preparation and electrophysiology. Experiments were performed on slices from mice (1-3 months, both sexes) unless otherwise indicated. Animals were anaesthetized and prior to decapitation, transcardially perfused with ice-cold protective artificial cerebrospinal fluid (NMDG ACSF) containing (in mM): N-methyl-D-glucamine (NMDG) 93, KCl 2.5, NaH2PO4 1.25, NaHCO3 30, HEPES 20, glucose 25, thiourea 2, Na-ascorbate 5, Na-pyruvate 3, CaCl2 0.5, MgCl2 10. The brain was rapidly removed and placed in ice-cold NMDG ACSF, and slices (300 μm thick) were obtained with a vibratome (Leica Vibratome VT1200S, Germany), maintained at 34 ºC in NMDG ACSF for 10 min and transferred for >1 h incubation at room temperature (RT) in standard ACSF containing (in mM): NaCl 119, KCl 2.5, NaH2PO4 1, MgCl2 1.2, NaHCO3 26, CaCl2 2.5, and glucose 11, and gassed with 95% O2 / 5% CO2 (pH = 7.3).
Slices containing the NAc region were placed into the microscope recording chamber superfused with gassed ACSF and cells were visualized on an Olympus BX50WI microscope (Olympus Optical, Tokyo, Japan) with a 40x water immersion objective. Recordings from NAc medium spiny neurons (MSNs) were obtained using the whole-cell patch-clamp technique. Patch electrodes had resistances of 4-10 MΩ when filled with an internal solution that contained (in mM): KGluconate 135, KCl 10, HEPES 10, MgCl2 1, ATPNa2 2, titrated with KOH to pH 7.3. Recordings were obtained with Multiclamp 700 A/B amplifiers and pClamp software (Molecular Devices). NAc neurons were voltage clamped at -75 mV, unless otherwise noted. For recording of slow inward currents (SICs), the extracellular Mg2+ was equimolarly replaced with Ca2+ and 10 µM glycine was added to optimize NMDAR activation and the recordings were made in presence of 1 µM tetrodotoxin (TTX) to avoid neuronal action potential. SICs were analyzed using the Event Detection protocols in the Clampfit routine of pClamp (the template consisted of a 5 ms baseline, 22 ms rise time, 100 ms decay time). Moreover, to isolate SICs and miniature excitatory synaptic transmission (mEPSCs), ACSF was supplemented with 0.05 mM picrotoxin (GABAA receptor antagonist).
Operant conditioning chamber. Behavioral sessions were conducted in a custom-made operant conditioning chamber, 25 x 30 x 30 cm polypropylene, controlled by an Arduino’s (Arduino Uno) automatized custom-made code system. The 10% sucrose-based solution used as reward was available upon correct performance at two waterholes (1.5 cm diameter) located 2.5 cm above the chamber floor at the same wall of the receptacle. Behind each waterhole, infrared sensors (HW-201) were placed to monitor the animal’s nose-pokes, and reward (10 µl drop dose) was delivered through syringes automatically controlled by mini-pumps (FA0520F; 12V) associated with LED cues based on yellow LEDs (5 mm) placed 4 cm above each waterhole. To track animal movements during behavioral sessions, a camera (Blackfly S BFS-U3-16S2C, Teledyne FLIR LLC) was placed above the chamber.
Behavioral task. Mice from both sexes were housed in 12 hr dark/light cycle with ad libitum access to food and under 60% of water restriction criteria, maintained at no less than 85% of pre-surgical body weight. Behavioral sessions were conducted in presence of white noise as follows: 2 days of habituation, from 0 to 10 days of training (AstroLight labeling on session 8) and 1 test day.
Each training session started with 10 min of pre-training period in which no light or 10% sucrose reward were available, followed by 50 sucrose delivery trials/session or 60 min, performed on a fixed ratio 1 (FR1). In each trial, sucrose was only available in the waterhole signaled by the LED on. To avoid the animal’s prediction, reward location and time between trials were randomized by Arduino custom code; after 15 s fixed interval + 0-15 s variable interval, one of the two LED cues were randomly switched on, signaling reward availability. To avoid an imbalance between both reward locations, the randomized code was set to a maximum of 3 consecutive sucrose deliveries at the same location. Cue-reward association was analyzed across training sessions by monitoring the correct index % (first nose-poke in the LED-on location) and latency (seconds between LED on and reward delivery) parameters. To determine the location of the AstroLight labeling for the reward group (further details in AstroLight labeling and ChR2 optogenetic stimulation sections), the animal’s natural preference towards a specific reward location was monitored across sessions by measuring the % of nose-pokes in each waterhole given by infrared activation during the LED-off pre-training period. After the training period and AstroLight labeling, a final test task was designed to test the behavioral outcomes derived from ChR2 activation in precise astrocytic subsets.
The test session began with a 10-min baseline period, which served as the pre-test period. This was followed by a stimulation phase where we optogenetically activated the astrocytic ChR2 (see section on AstroLight labeling and optogenetic stimulation), without the availability of light or rewards. During this phase, we analyzed the following parameters in each animal compared to their baseline behavior observed during the pre-test period: % nose-poke visits, % time spent in different zones (labeled, unlabeled, and outside areas in the arena), and average velocity. Nose-poke counts were automatically recorded by measuring infrared activation using Arduino, while time in zones and velocity were analyzed by tracking the animal's movements with Ethovision 11.5 software. The ChR2 stimulation was followed by a cue phase where both LEDs and rewards were simultaneously available in FR1 trials (50 sucrose delivery trials per session). The time between trials was randomized using the same criteria as during the training period, with a fixed interval of 15 seconds and a variable interval of 0-15 seconds. To assess the direction of behavior, we measured the preference index by analyzing the % visits towards one location compared to the other, relative to the pre-test period. Representative heatmaps were generated by tracking the mouse's movements using Ethovision software.
AstroLight labeling and ChR2 optogenetic stimulation. To target precise astrocyte ensembles related to specific behaviors, AstroLight labeling was conducted on training session 8 in two different experimental groups, reward group and no-reward group. Reward group labeling was induced by delivering 2.5 s of light (465 nm; 200-500 µW/cm2) during the training session each time the mouse received the sucrose-reward exclusively from the targeted waterhole (left or right). This targeted location was determined by the natural preference of the animal; if mice showed >55% preference towards one waterhole, labeling of the reward-related astrocyte ensemble was automatically conducted in the opposite one. In case of no significant preference, to avoid behavioral artifacts, the animals’ labeling was balanced between both waterhole locations (Fig. S3). No-reward group labeling was induced when mice explored the arena between trials, in absence of LED cues or reward availability. The same approach was used, 2.5 s of light (465 nm; 200-500 µW/cm2) to a final amount of 60 s of labeling light manually triggered at selected time points in which mice were exploring and not directed towards the waterholes.
To activate the ChR2 expressed in astrocytes during the stimulation phase of the test session, we used an optogenetic stimulation protocol (20 s ON, 3 min interval, 5 cycles; 465 nm; 200-500 µW/cm2) automatically triggered by Arduino custom-made code. Both labeling and stimulation light were delivered through the implanted fiber photometry cannula and using the Doric lenses system.
Fiber photometry Ca2+ recordings. To measure calcium activity in AAV5-GFAP-cyto-GCaMP6f injected animals, the Doric GCaMP Fiber Photometry System (FPS_1S_GCaMP, Doric Lenses) was used with a 405 nm LED (5 - 10 µW) as the isosbestic point, and a 465 nm LED (20 - 50 µW) as the excitation-dependent GCaMP fluorescence. Signals were collected interleaved at 100 Hz. For the cohort of animals that underwent training sessions in the operant chamber (Fig. 2D - G), fiber photometry recordings were conducted on alternate days (Day 0, 2, 4, 6, 8, 10) of training sessions to avoid fluorophore exhaustion. To precisely synchronize the fiber photometry recordings with the animal’s performance, digital inputs coming from Doric system, camera (Blackfly S BFS-U3-16S2C, Teledyne FLIR LLC) and Arduino (Arduino Uno) were simultaneously recorded using an Intan RHD USB interface board (Part #C3100, Intan Technologies). Raw signals were processed using Matlab software, including demodulation and a 20 Hz cut-off frequency with 20 dB attenuation, followed by a 1 s moving mean window. Isosbestic signals were subtracted from Ca2+-dependent signals to eliminate motion-related artifacts59,60. The GCaMP6 fluorescence signals were standardized across animals using the equation ΔF = (F-F0)/F0, where F0 was computed by linearly interpolating between local minima of the fluorescence signal within different 45-second time windows to account for photobleaching. To isolate Ca2+ activity related to reward consumption, ΔF signal was aligned to mini-pump digital input signal (recorded by Arduino) using Python software and reward Ca2+ signaling intervals for LED and reward (5 s before LED/reward – 12.5 s after LED/reward) were computed for each trial. Area under curve (AuC) was calculated for each individual interval and the 50 trial/session average was computed for each mouse and fiber photometry recording data.
Fiber photometry Ca2+ recordings were also used for the initial characterization of AstroLight functionality (Fig. 1D and E). Animals were placed in an open field (25 x 30 x 30 cm) and 10-min Ca2+ recordings were performed using the same parameters as the behavioral experiments. After a 5-min baseline recording, mPFC afferents expressing ChrimsonR were stimulated (10 pulses 50 ms at 4 Hz - 4 times, 5 s interval11) followed by another 5 min recording. The raw signal was processed using Matlab software to compute ΔF signal, and the AuC per second (AuC/s) was calculated for the 10-second period before and 40-second period after ChrimsonR optostimulation. This analysis was conducted to assess astrocytic Ca2+ activity during the 40-second temporal window in which we delivered 465 nm light for AstroLight labeling in the analogous group of animals injected with ChrimsonR + AstroLight, following the same experimental approach (Fig. 1F and G).
Three-dimensional spatial quantification. To obtain brain tissue avoiding nonspecific ChR2-YFP expression derived from the stimulation phase protocol, same day after test session was conducted animals were euthanized with pentobarbital and transcardially perfused with ice-cold PBS, followed by 4% paraformaldehyde (PFA), 4% sucrose in PBS. Brains were removed and maintained in the fixative solution overnight (o/n) at 4 ºC. Coronal brain sections 200-300 µm thick were obtained in a VT100S vibratome (Leica) and mounted for imaging in Vectashield antifading mounting medium (Vector Laboratories, Burlingame, CA). Images were acquired with a Leica SP-5 inverted confocal microscope using Leica LAS AF software. Z-stack (10 plane / 10 µm thickness) mosaics were collected for each fluorescence signal: AstroLight-tdTom (561 nm excitation laser) and ChR2-YFP (510 nm excitation laser).
Image analysis was conducted using ImageJ software (public domain software developed at the US NIH). To analyze the entire nucleus, all stacks from the same brain were included. The NAc region was isolated by manually delimiting the region of interest (ROI), and planes were considered at intervals of 20 µm to prevent repeated measurements of the same cells in adjacent 10 µm steps. For each selected plane, images were semi-automatically processed by applying background subtraction and manually adjusting brightness, contrast, and threshold parameters. Subsequently, a binary mask was automatically generated, which included fluorescent particles larger than 50 µm². To avoid false positive detection in the YFP signal, the binary masks for tdTom and YFP signals were compared using the Image Calculator plugin in ImageJ. A final processed YFP mask was obtained, which only included particles that colocalized between the two signals. YFP/tdTom ratio was calculated as # YFP particles / # tdTom particles.
Three-dimensional spatial quantification was performed using the following steps. First, the resulting binary masks for tdTom and processed YFP were registered to a reference NAc mask using the bUnwarpJ plugin in ImageJ. This alignment ensured the same anatomical location across different stack images and animals. Next, X (medial-lateral; M-L) and Z (dorsal-ventral; D-V) coordinates were obtained in the aligned planes by calculating the center of mass for each particle larger than 50 µm². The Y (anterior-posterior; A-P) coordinate was determined based on the 20 µm plane location within the NAc. Using these coordinates, particles were classified in three-dimensional space, and the YFP/tdTom ratio was calculated in 0.05 mm bins for the M-L and D-V axes, and 0.02 mm bins for the A-P axis. Taking into account the anatomy of the NAc61, final quantification (Fig. 3C, G, and J) along the medial-lateral axis was computed as the average YFP/tdTom ratio, including bins from 0 to 0.65 mm for the medial portion and bins from 0.7 to 1.5 mm for the lateral portion. The same approach was applied to quantify the dorsal-ventral axis, where bins from 0 to 0.95 mm were considered dorsal and bins from 1 to 1.95 mm were considered ventral. For the anterior-posterior axis, bins from 1.88 to 1.22 mm were averaged as anterior, and bins from 1.2 to 0.7 mm were averaged as posterior.
In addition to the quantification mentioned above, three-dimensional heat-map gradients of the average expression and location of the YFP/tdTom ratio were generated for the NAc (Fig. 3D, H, and K) using Python software for data visualization. To accomplish this, we applied the PRQ approach (Partition in Regular Quadrants11). First, the binary masks for tdTom signal and processed YFP signal were divided into 50 µm x 50 µm square grids, and the average signal was calculated for each quadrant of the grid. Since the masks contained 8-bit binary data, the average value of each grid ranged from 0 to 255, indicating the spatial probability. A value of 0 indicated no particles in the quadrant, while a value closer to 255 indicated a high number of particles in that specific quadrant. The YFP/tdTom ratio was calculated for each quadrant as the ratio of the processed YFP mask to the tdTom mask. Based on the resulting YFP/tdTom ratio grids, the sections shown in Fig. 3D, H, and K represent 0.1 mm spatial bins containing the average of 20 µm planes from all mice in each experimental group. These sections correspond to the same anterior-posterior (A-P) location.
The same approach was utilized to construct three-dimensional heat-map gradients for analysis of the spatial dispersion of AstroLight-tdTom (Fig. S5C, D, and E) and GFAP-ChR2-mCherry (Fig. S7C) AAV transfections. Binary masks containing tdTom-positive and mCherry-positive particles were divided into 50 µm x 50 µm square grids, and the average signal was calculated for each quadrant, representing the spatial probability of registering particles in that specific area. These resulting grids were then averaged into 0.1 mm spatial bins, encompassing the same anterior-posterior (A-P) 20 µm planes from all mice in each experimental group. This process was repeated for each experimental group. Additionally, the percentage of spatial probability was computed based on the maximum average signal displayed across the resulting 3D maps. Furthermore, to quantify the spatial dispersion maps, the percentage of particles from the total count was calculated for the medial-lateral (M-L), dorsal-ventral (D-V), and anterior-posterior (A-P) portions of the NAc, following the same bin criteria used for the YFP/tdTom ratio quantification.
AAV transfection degree. To measure the degree of AAV transfection for GFAP-cyto-GCaMPf6 (Fig. S2B), AstroLight-tdTom (Fig. S5B), and GFAP-ChR2-mCherry (Fig. S7B), ImageJ software was utilized. A fixed area region of interest (ROI) was employed to measure the fluorescence (a.u.) signal in each NAc slice. Subsequently, the fluorescence (a.u.) values corresponding to the same infection were averaged to determine the final transfection degree.
Immunohistochemistry. Mice injected with AstroLight viral vectors in the NAc were anesthetized using sodium pentobarbital and subsequently transcardially perfused with PBS, followed by ice-cold 4% paraformaldehyde (PFA) and 4% sucrose in PBS. The brain tissue was then postfixed overnight (o/n) in the same fixative solution. Coronal brain slices of 50 µm thickness were obtained using a VT100S vibratome (Leica). The slices were permeabilized with 1% Triton X-100 in PBS, and non-specific binding was blocked for 1 hour at room temperature (RT) using a solution of 0.3% goat serum and 0.1% Triton in PBS. To assess the specificity of AstroLight, the sections were incubated overnight with the corresponding primary antibodies at 4 °C and then with secondary antibodies for 1 hour at RT. The primary antibodies used were rabbit anti-S100 (1:200; ab868, Abcam) and mouse anti-NeuN (1:500; MAB377, Merck). The secondary antibodies used were goat anti-mouse Alexa 647 (1:1000; A21236, Invitrogen) and goat anti-rabbit Alexa 405 (1:1000; A31556, Invitrogen). Following antibody incubation, the sections were washed with 0.1% Triton in PBS and mounted for imaging using Vectashield antifading mounting medium (Vector Laboratories, Burlingame, CA). Fluorescence images were acquired using a Leica SP-5 inverted confocal microscope with Leica LAS AF software. ImageJ software was employed for image analysis.
Drugs and chemicals. Picrotoxin (#1128) and (4R,4aR,5R,7S,9S,10S,10aR,11S,12S)-Octahydro-12-(hydroxymethyl)-2-imino-5,9:7,10a-dimethano-10aH-[1,3]dioxocino[6,5-d]pyrimidine-4,7,10,11,12-pentol citrate (TTX; #1069) were purchased from Tocris.
Statistical analysis. Statistical analysis was conducted using two-tailed unpaired or paired t-tests and repeated measurements (RM) one-way or two-way ANOVA, followed by the Holm-Sidak test for multiple comparisons, unless otherwise stated. Statistical significance was defined as p < 0.05. For normalized data (basal = 100), significant percentage changes from the pre-test were determined using a one-sample t-test. GraphPad Prism 7 software was used for all statistical calculations; the data are presented as mean ± SEM.