The development of an association between a low-salience cue with a subsequent aversive stimulus is a well-studied phenomenon and is the basis of classical conditioning paradigms. Less often considered are the effects of other neutral cues that are incidental to the conditioned stimulus — these too can influence behaviour indirectly by a process called mediated learning. The mechanisms underlying mediated learning are not well understood, but in this study, Busquets-Garcia et al. show that cannabinoid receptor 1 (CB1R) expressed on hippocampal interneurons plays a crucial role.

Credit: filadendron/Getty

The authors developed a modified classical conditioning paradigm that could separately measure direct and mediated learning: mice were first preconditioned by repeated exposure to pairs of low-salience sensory stimuli (an odour and a taste) to form an incidental association between them. Then, on subsequent days, mice were conditioned to associate one of the sensory stimuli (but not the other) with an aversive stimulus. They found that both the conditioned stimulus but also the incidental preconditioned stimulus produced aversion behaviour, indicating the acquisition of both direct aversion learning and mediated aversion learning, respectively.

Hippocampal CB1Rs have been implicated in a variety of direct conditioning responses, and the authors found that mice lacking the CB1R throughout the brain (CB1R-KO) showed impaired mediated learning but intact direct aversion learning. Moreover, administration of rimonabant (a CB1R antagonist) to wild-type mice before the preconditioning phase reduced mediated learning for both taste and odour modalities, but rimonabant administration before the test phase had no effect, further supporting a role for the CB1R in incidental associations.

The hippocampus has previously been implicated in mediated learning. The authors found that in mice lacking CB1Rs selectively in hippocampal GABAergic interneurons, mediated learning was compromised but direct learning was unaffected. In addition, they found that hippocampal CB1R expression was increased following preconditioning, and attenuated by pretreatment with the protein synthesis inhibitor anisomycin, suggesting mediated learning might involve synthesis of new CB1Rs in hippocampal interneurons.

Stimulation of CB1Rs expressed by hippocampal interneurons has the effect of reducing their activity. To test whether the absence of a CB1R-mediated reduction in GABAergic inhibition might underlie the reduced mediated learning found in GABA-CB1R-KO mice, the authors mimicked CB1R signalling in hippocampal interneurons of GABA-CB1R-KO mice using an inhibitory designer receptor exclusively activated by designer drugs (DREADD). This treatment rescued mediated learning in these mice. Further support for this hypothesis was provided by artificial enhancement of GABAergic transmission in the hippocampus of wild-type mice using an excitatory DREADD expressed in hippocampal GABAergic neurons, which impaired mediated learning.

Of the two main subpopulations of hippocampal interneurons — cholecystokinin (CCK)-expressing and parvalbumin (PV)-expressing — CB1Rs are expressed only in the former. As CCK is expressed in pyramidal cells, selective targeting is technically challenging, so the authors chose to rule out the involvement of PV-positive interneurons using an activating DREADD. Selective activation of PV-expressing interneurons during preconditioning did not affect mediated learning, indicating, by elimination, the likely role for CCK-expressing interneurons in this effect.

Overall, these findings suggest that CB1Rs expressed on specific subpopulations of hippocampal GABAergic interneurons play a key role in the formation of incidental associations between two sensory stimuli, which involves protein synthesis of new CB1Rs and can in turn contribute to mediated learning.