Visualizing Learning-Related Plasticity in Neuronal Subtypes of the Barrel Cortex

 Sensory learning is accompanied by synaptic, anatomical, and representational changes  in superficial layers of primary sensory cortex and has been extensively investigated in animal  models, including mice. However, it remains unclear how these changes are distributed across  different neuronal subtypes and how they evolve over time, from naïve to expert animals. We  combined in vivo two-photon calcium imaging with homecage training to track L2/3 neuron  activity in the mouse somatosensory (barrel) cortex during a whisker-dependent task. Because  previous studies showed pathway-specific synaptic changes in acute brain slices and fixed  tissue early in training, we hypothesized that sensory-evoked activity might be altered in vivo,  even in the absence of task-related cues and rewards.  

Before and after training, L2/3 pyramidal neurons (L2/3 Pyrs) exhibited sparse activation.  A modest, transient increase in stimulus-evoked activity was observed in a subset of neurons at  training onset, absent in the pseudotraining paradigm, where stimulus and reward were  uncoupled. The evoked response was negatively correlated with animal performance,  suggesting that learning sparsens sensory representations in barrel cortex.  

The transient potentiation in pyramidal neuron activity may be triggered by the removal  of inhibition, particularly from somatostatin (SST) neurons. We hypothesized that SST neurons  would exhibit a temporarily reduced evoked response at training onset. Surprisingly, in  vivo imaging revealed a persistent, learning-induced reduction in stimulus-evoked activity in the  overall SST population outside the training context, which was mainly driven by a calbindin2 expressing subset. The unsynchronized changes between Pyrs and SST neurons suggests that  Pyrs could be crucial for the initial stages of encoding new sensory information, whereas the  sustained suppression of SST activity might be involved in refining the association between  sensory stimulus and behavioral outcomes. Conversely, pseudotraining enhanced SST neural  activity. These bidirectional changes indicate that SST neurons are highly responsive to  stimulus-reward contingencies.   

Despite the learning-dependent changes in stimulus-evoked responses of L2/3 Pyr and  SST neurons observed in the barrel cortex, its causal role in establishing sensory-reward  associations has been controversial. Lesion and chemogenetic suppression of pyramidal  neuron activity did not prevent sensory-association learning. Overall, the results suggest that  the activity changes in the barrel cortex may reflect a modulatory role, rather than a direct  contribution, in establishing the sensory-reward associations.