The Role of Parvalbumin Interneurons in Mouse Primary Visual Cortex during Visual Discrimination and Perceptual Learning

Abstract

Parvalbumin-expressing (PV) cells are the most common class of inhibitory interneurons in the rodent cortex and play important roles in balancing cortical circuit activity and learning (Xu et al., 2010). PV cells are suggested to mediate feedforward inhibition and control gain modulation (Markram et al., 2004; Tremblay et al., 2016). Although increasing their activity may improve discrimination (Lee et al., 2012), findings are inconsistent and the role of PV cells has been under debate (Atallah et al., 2012; El-Boustani et al., 2014; Wilson et al., 2012). The primary visual cortex (V1) has an early and a late phase of activity, likely reflecting bottom-up (feedforward) and integration of top-down (feedback) sensory processing (Lamme and Roelfsema, 2000; Wyatte et al., 2014). The feedforward sweep alone is sufficient to enable discriminations in mice (Resulaj et al., 2018), however, it is not known how PV cells contribute to this early processing and whether both easy and difficult discriminations are equally impacted by changes in inhibition. Perceptual learning increases the selectivity of V1 pyramidal (Pyr) and PV cells (Poort et al., 2015), and the emergence of PV-Pyr ensembles has been reported (Khan et al., 2018). Thus, PV cells may modulate Pyr cell activity differently depending on the stage of learning. However, such interactions have only been examined in a naïve V1 network and have not yet been studied after extensive behavioural training on a visual task. Plasticity following visual perceptual learning is thought to manifest in lower-level areas like V1 due to the high ‘specificity’ for the trained features (Fiorentini and Berardi, 1980; Schoups et al., 1995). Generalization and transfer of learning with simpler stimuli (McGovern et al., 2012), changes in higher-level representations, and influences from attentional feedback (Szpiro and Carrasco, 2015), decision-making areas (Diaz et al., 2017), and the oculomotor system (Kwon et al., 2013), challenges this view. Indeed, mice are extremely flexible in generalizing the task rule and adapt their behaviour over days to successfully discriminate previously inexperienced stimuli using a categorization strategy (Goltstein et al., 2021; Reinert et al., 2021). However, due to the complexity of perceptual learning likely affecting several brain areas and multiple levels of processing, it is important to have a full understanding of the flexibility of V1 representations and how the network adapts to reflect these changes over shorter timescales. Therefore, the thesis aims to investigate: 1) How does manipulating the strength and timing of PV cell activation in V1 affect visual discriminations of varying difficulty? 2) How does the influence of PV cells on the local V1 network change with experience? 3) How does V1 rapidly represent previously untrained orientations given the flexibility of mice to generalize their task knowledge? To address these questions, I used a simple go/no-go orientation discrimination task and a combination of optogenetics to activate PV cells and two-photon imaging to record the activity of V1 cell populations. First, I found that performance improved for easy discriminations but only when PV cells were stimulated during the early phase of V1 activity. This was supported by changes in V1 neuron activity, where selectivity to the rewarded stimulus increased following PV cell activation. Second, I identified that PV cells exert global effects on the V1 population through divisive inhibition, both before and after learning. However, when taking into consideration stimulus relevance, the inhibition exerted by PV cells after learning was orientation-dependent. Third, I found that V1 representations are remarkably flexible, adapting the same processes that affect trained orientations to novel, but perceptually similar, orientations thus enabling mice to perform accurate discriminations. Taken together, these findings increase our understanding of the role of PV cells and V1 in visual discrimination and perceptual learning.