An automated homecage system for multiwhisker detection and discrimination learning in mice

Abstract

Automated, homecage behavioral training for rodents has many advantages: it is low stress, requires little interaction with the experimenter, and can be easily manipulated to adapt to different experimental condition. We have developed an inexpensive, Arduino-based, homecage training apparatus for sensory association training in freely-moving mice using multiwhisker air current stimulation coupled to a water reward. Animals learn this task readily, within 1-2 days of training, and performance progressively improves with training. We examined the parameters that regulate task acquisition using different stimulus intensities, directions, and reward valence. Learning was assessed by comparing anticipatory licking for the stimulus compared to the no-stimulus (blank) trials. At high stimulus intensities (>9 psi), animals showed markedly less participation in the task. Conversely, very weak air current intensities (1-2 psi) were not sufficient to generate rapid learning behavior. At intermediate stimulus intensities (5-6 psi), a majority of mice learned that the multiwhisker stimulus predicted the water reward after 24-48 hrs of training. Both exposure to isoflurane and lack of whiskers decreased animals ability to learn the task. Perceptual learning was assessed and following training at an intermediate stimulus intensity, perception was likely heightened as mice were able to transfer learning behavior when exposed to the lower stimulus intensity. Mice learned to discriminate between two directions of stimulation rapidly and accurately, even when the angular distance between the stimuli was <15 degrees. Switching the reward to a more desirable reward, aspartame, had little effect on learning trajectory. Our results show that a tactile association task in an automated homecage environment can be monitored by anticipatory licking to reveal rapid and progressive behavioral change. These Arduino-based, automated mouse cages enable high-throughput training that facilitate analysis of large numbers of genetically modified mice with targeted manipulations of neural activity.