A cerebello-olivary signal for negative prediction error is sufficient to cause extinction of associative motor learning

Abstract

The brain generates negative prediction error (NPE) signals to trigger extinction, a type of inhibitory learning that is responsible for suppressing learned behaviors when they are no longer useful. Neurons encoding NPE have been reported in multiple brain regions. Here, we use an optogenetic approach to demonstrate that GABAergic cerebello-olivary neurons can generate a powerful NPE signal, capable of causing extinction of conditioned motor responses on its own.

Diverse inhibitory projections from the cerebellar interposed nucleus

“…Inhibitory projections from cerebellum to sensory areas would seem to be ideally situated to modulate the sensory gain of predicted sensory consequences of movement ( Brooks et al, 2015 ; Shadmehr, 2020 ). Moreover, negative sensory prediction error could be used to actively cancel predicted sensory reafference ( Kim et al, 2020 ; Requarth and Sawtell, 2014 ; Shadmehr, 2020 ; Conner et al, 2021 ), raising implications for a combined role of negative sensory prediction error in guiding learning both through modulation of climbing fiber signaling in IO and through modulation of sensory signals reaching the cerebellum upon which associative learning is built. Second, GABAergic projections to the pontine nuclei, which are themselves a major source of mossy fiber inputs to the cerebellum, suggest a regulatory feedback pathway that could operate as a homeostat akin to the feedback loops through the IO ( Medina et al, 2002 ).…”

Widespread inhibitory projections from the cerebellar interposed nucleus

“…Inhibitory projections from cerebellum to sensory areas would seem to be ideally situated to modulate sensory gain of predicted sensory consequences of movement (Brooks et al, 2015; Shadmehr, 2020). Moreover, negative sensory prediction error could be used to actively cancel predicted sensory reafference (Kim et al, 2020; Requarth & Sawtell, 2014; Shadmehr, 2020; Conner et al, 2021), raising implications for a combined role of negative sensory prediction error in guiding learning both through modulation of climbing fiber signaling in IO and through modulation of sensory signals reaching the cerebellum upon which associative learning is built. Second, GABAergic projections to the pontine nuclei, which are themselves a major source of mossy fiber inputs to the cerebellum, suggests a regulatory feedback pathway that could operate as a homeostat akin to the feedback loops through the IO (Medina et al, 2002).…”

“…Dichotomous roles for different cell types have been most clearly hypothesized in delay eyelid conditioning models, where glutamatergic neurons are proposed to produce the conditioned response while inhibitory neurons regulate the learning 'setpoint' via projections to the IO, the source of climbing fibers (Bengtsson et al, 2007;Garcia & Mauk, 1998;Kim et al, 1998;Kim et al, 2020;McCormick & Thompson, 1984;Medina et al, 2001;Medina et al, 2002;Ten Brinke et al, 2017;Thompson & Steinmetz, 2009). These studies assume that premotor and nucleo-olivary neurons respond in roughly equivalent ways during behavior (Shadmehr, 2020).…”

A recurrent circuit links antagonistic cerebellar modules during associative motor learning

“…We recorded a total of 138 Purkinje cells (61 from wildtype mice trained with the 220-ms ISI, 17 cells from a wildtype mouse trained with the 370-ms ISI, 40 cells from PCP2-ChR2 mice, 11 cells from nNOS-ChR2 mice, and 9 cells from EAAT4-GFP mice). The -25-spike sorting and the electrophysiological identification of Purkinje cells was performed as described previously (Ohmae and Medina, 2015;Kim et al, 2020).…”

“…The following two methods to suppress rAIN while recording Purkinje cells in parallel were less successful: First, we tried to infuse muscimol or lidocaine into rAIN (pump or iontophoresis), but the infusion site was so close to the Purkinje cells being recorded (about 0.6 mm) that the direct diffusion of muscimol/lidocaine to the Purkinje cells affected their spontaneous firing rates even when the infusion volume was strictly controlled to the minimum required to suppress CR. Second, we attempted Arch expression in rAIN with AAV5-hSyn-Arch but abandoned it because the toxicity of AAV5-hSyn-Arch could interfere with learning (Kim et al, 2020) and most of the rAIN cells died in the pilot experiment.…”