GoKawiilAnimals
Wild-type C57BL/6J female and male mice (2–6 months old; Jackson Laboratory) were used as experimental animals. Mice were housed under a reversed 12 h–12 h light–dark cycle (lights off at 07:00 and on at 19:00) in a temperature- and humidity-controlled environment (18–23 °C, 40–60% humidity), with ad libitum access to food and water. Training began when mice were approximately 2–3 months of age. All pairs were age-matched at the start of training. During training and experiments, mice were water restricted but maintained at 80–90% of their initial body weights. Experiments were conducted during the dark cycle, and each training session lasted for about 1 h, during which mice received 0.5–1.5 ml of water from the task. Animals received supplemental water as necessary to maintain their body weights. All procedures complied with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and were approved by the Icahn School of Medicine at Mount Sinai Institutional Animal Care and Use Committee.
Behavioural apparatus
The training arena was an 18 inch × 18 inch white-acrylic square chamber. Four reward zones were positioned at the centre of each arena wall, each with two adjacent water delivery ports. Reward ports were 3D-printed using white and transparent resins, each incorporating an infrared beam for nose-poke detection and a white LED to indicate its availability. Water was delivered through a stainless-steel tube within each port, controlled by a solenoid valve (Lee, LHDB0533418H). An initiation detector, fitted with an infrared LED and a reflective sensor, was positioned at the centre of the arena to detect trial initiation when approached by mice. Analogue signals from all ports were digitized through the Arduino Nano and acquired through a DAQ system (National Instruments, USB-6001). Behavioural control was implemented in LabVIEW (National Instruments, 2014) to manage trial timing, LED signalling, water delivery and data logging. Mouse behaviour was recorded at 30 fps using an overhead camera (Teledyne FLIR, BFS-U3-16S2M-CS).
Video tracking
We used the machine-learning-based tools SLEAP (v1.3.3)55 and Ensemble Kalman Smoother (EKS) (v0.0.0)56 to annotate the nose, neck and torso positions of both mice in each video frame, enabling precise tracking of their movement and posture. To identify individual mice, we shaved a small patch of back fur on one mouse in each pair. Each mouse was annotated with three key points: nose (tip of the nose), neck (centre of the neck) and torso (centre of mass), which were connected to form a skeleton. Approximately 4,200 video frames from 10 mouse pairs were manually annotated for training a SLEAP model, which was then used to automatically estimate the key points of other pairs. To optimize tracking accuracy, we additionally applied EKS—a post-processing method that refines pose estimation outputs by smoothing several model predictions, resulting in more robust tracking. Custom MATLAB scripts (MathWorks, 2024b) were used to align the behaviour data from LabVIEW with the video frames based on the LED onset and offset.
Behavioural definitions
Mouse position was defined using its neck position, which was also used to calculate speed and acceleration. The reward zone was defined as a semicircle with a 10 cm radius centred at the midpoint between each pair of reward ports (Extended Data Fig. 1h). Arrival was defined as the first frame in which a mouse’s neck position entered a reward zone. Reaction time was defined as the time from trial onset to arrival at the chosen reward zone.
Cooperative foraging task training
Before training, mice were water restricted to 1 ml per day for at least 3 days and habituated to the experimenter and the behavioural set-up. Mice underwent one training session per day, progressing through four main stages.
... continue reading