Parvalbumin-Positive Interneurons Regulate Cortical Sensory Plasticity in Adulthood and Development Through Shared Mechanisms.

Parvalbumin-positive neurons are the largest class of GABAergic, inhibitory neurons in the central nervous system. In the cortex, these fast-spiking cells provide feedforward and feedback synaptic inhibition onto a diverse set of cell types, including pyramidal cells, other inhibitory interneurons, and themselves. Cortical inhibitory networks broadly, and cortical parvalbumin-expressing interneurons (cPVins) specifically, are crucial for regulating sensory plasticity during both development and adulthood. Here we review the functional properties of cPVins that enable plasticity in the cortex of adult mammals and the influence of cPVins on sensory activity at four spatiotemporal scales. First, cPVins regulate developmental critical periods and adult plasticity through molecular and structural interactions with the extracellular matrix. Second, they activate in precise sequence following feedforward excitation to enforce strict temporal limits in response to the presentation of sensory stimuli. Third, they implement gain control to normalize sensory inputs and compress the dynamic range of output. Fourth, they synchronize broad network activity patterns in response to behavioral events and state changes. Much of the evidence for the contribution of cPVins to plasticity comes from classic models that rely on sensory deprivation methods to probe experience-dependent changes in the brain. We support investigating naturally occurring, adaptive cortical plasticity to study cPVin circuits in an ethologically relevant framework, and discuss recent insights from our work on maternal experience-induced auditory cortical plasticity.

Introduction

The cerebral cortex has a remarkable capacity to remodel its synaptic structures and refine its neuronal activity in response to changing conditions throughout life. Such cortical plasticity interacts with a wide range of fundamental biological processes including aging (Freitas et al., 2011), stress (Takatsuru and Koibuchi, 2015; McGirr et al., 2020; Muehlhan et al., 2020), injury and recovery (Grasso et al., 2020; Moreno-López and Hollis, 2021), and sensory learning (LeMessurier and Feldman, 2018). A substantial body of evidence supports the crucial role of GABAergic inhibition in sensory cortical plasticity both during development and adulthood.

The neocortex is built around a “canonical circuit,” a set of vertically connected circuit motifs composed of diverse excitatory and inhibitory cell types that establish common hierarchical neuronal computations (Pluta et al., 2015). This arrangement amplifies thalamic input and integrates that input with projections from other regions to accomplish the complex tasks of sensory processing, learning, and memory. Within the cortex, there is a wide disparity between the numbers of excitatory and inhibitory neurons; glutamatergic excitatory neurons compose the overall majority, whereas GABAergic inhibitory interneurons account for only ∼10–20% of neurons (Markram et al., 2004; Petilla Interneuron Nomenclature Group et al., 2008; DeFelipe et al., 2013; Chu and Anderson, 2015). Despite this mismatch in relative densities, cortical inhibitory drive tends to match excitatory drive (Shadlen and Newsome, 1994; Rubenstein and Merzenich, 2003). This suggests that changes to the synaptic connectivity, morphology, intrinsic properties, and firing patterns of inhibitory neurons play a disproportionate role in controlling the timing and specificity of cortical responses (Hensch, 2004; Lu et al., 2007; Trachtenberg, 2015).

In the last two decades, the contributions of specific inhibitory interneuron subtypes to sensory learning, memory, and plasticity have been studied in the visual, auditory, somatosensory, and olfactory cortices. Here we review some of that literature, focusing on cortical parvalbumin-expressing interneurons (cPVins), which are the most abundant inhibitory cell type in the brain. We discuss how the physiological and molecular features of cPVins equip sensory networks for plasticity. We argue that cPVins across sensory cortices share core intrinsic features and mechanisms that enable periods of experiential learning. We further propose that approaches exploiting ethologically relevant behaviors are important for understanding how cPVin directed plasticity is engaged naturally in the adult brain. While there are many varieties and mechanisms of “plasticity,” our use of the term here denotes adjustments to intrinsic and extrinsic connectivity in sensory cortices that optimize representation of stimuli. We particularly focus on the mechanisms that open and close episodes of heightened sensitivity of cortex to stimuli in response to developmental programs or novel sensory experience.

Inhibitory Control of Cortical Plasticity

Many of the fundamental features of cortical plasticity were first identified by studies examining the visual cortex during developmental critical periods (e.g., Wiesel and Hubel, 1963; Levay et al., 1978; Gordon and Stryker, 1996; Hensch et al., 1998; Huang et al., 1999; Sawtell et al., 2003). Much of this work observed the consequences of restricting sensory experience, as occurs in monocular deprivation when vision is restricted in one eye. When performed during a specific period early in life, this manipulation weakens responses corresponding to the deprived eye and strengthens responses to the remaining eye, a phenomenon known as ocular dominance plasticity (Hubel and Wiesel, 1963; Fagiolini et al., 1994; Gordon and Stryker, 1996).

The onset and offset of developmental critical periods are defined by changes in cortical GABAergic transmission (Fagiolini et al., 2004; Maffei et al., 2006, 2010). Premature maturation of GABAergic inhibitory networks terminates the V1 developmental critical period precociously (Huang et al., 1999). Mice that are deficient in glutamate decarboxylase (GAD), the rate limiting enzyme involved in synthesis of GABA, have altered neural circuitry and are insensitive to monocular deprivation (Hensch et al., 1998; Fagiolini and Hensch, 2000; Chattopadhyaya et al., 2007). Further, pharmacological disruption of GABA transmission is sufficient to impair typical critical period opening (Jiang et al., 2005; Deidda et al., 2015). On the other hand, complementary approaches that facilitate the development of inhibitory networks accelerate the onset of the critical period (Hanover et al., 1999; Fagiolini and Hensch, 2000; Di Cristo et al., 2007; Sugiyama et al., 2008).

Plasticity continues to be mediated by GABAergic transmission in adulthood (Nahmani and Turrigiano, 2014). The mouse visual cortex retains some plasticity in response to monocular deprivation following critical period closure (Sawtell et al., 2003; Hofer et al., 2006; Fischer et al., 2007; Sato and Stryker, 2008). In the adult V1, cortical potentials corresponding to stimulation of the non-deprived eye are strengthened with experience (Sawtell et al., 2003; Sato and Stryker, 2008). This is associated with enhanced visual acuity in the spared eye (Iny et al., 2006). Manipulations of inhibitory signaling in the adult visual cortex can reactivate experience-dependent plasticity mechanisms. For example, restoring inhibitory function in GAD mutants rescues the capacity for plasticity in response to visual deprivation (Hensch et al., 1998; Fagiolini and Hensch, 2000; Iwai et al., 2003; Chattopadhyaya et al., 2007). Interestingly, transplant of inhibitory precursor cells enhances adult cortical plasticity (Southwell et al., 2010; Tang et al., 2014; Davis et al., 2015). On the other hand, Harauzov et al. (2010) reported that reducing intracortical inhibition reactivated visual plasticity, possibly due to a reduction in chondroitin sulfate proteoglycan structures in the extracellular matrix (ECM) called perineuronal nets (PNNs).

Several pieces of early circumstantial evidence implicated cPVins, among all cortical GABAergic populations, as having a central role in regulating visual cortical critical periods. Mutant mice that exhibit an early critical period also show early cPVin maturation (Hanover et al., 1999; Krishnan et al., 2015). Plasticity can be restored in GAD mutants by manipulating GABA receptors, which are highly and specifically enriched on cPVins (Fagiolini et al., 2004). More directly, dissolution of cPVin-specific extracellular structures (perineuronal nets- PNNs) reinstates plasticity in adult mice (Pizzorusso et al., 2002). Evidence from Kuhlman et al. (2011, 2013) established that cPVins play a permissive role in critical period plasticity by disinhibiting the excitatory network. However, attempts to reproduce this effect by chemogenetically and optogenetically suppressing cPVin inhibition following monocular deprivation in older mice has yielded mixed results (Kuhlman et al., 2013; Saiepour et al., 2015; Kaplan et al., 2016). This may be due to methodological differences in the parameters used for manipulating PV neurons such as the depth of optical penetration, magnitude, duration, and frequency (El-Boustani and Sur, 2014) or the timing of administration of chemogenetic tools. Interestingly, recent work has shown that monocular deprivation can drastically change the number of synaptic inputs onto V1 cPVins (Ribic, 2020). This suggests that the observed fluctuations in inhibitory drive may be a compensatory mechanism downstream of aberrant synaptic pruning of excitatory inputs onto cPVins (Severin et al., 2021).

Molecular and Structural Plasticity of Cortical Parvalbumin-Expressing Enterneurons

At the finest spatial scale, cPVin control of cortical plasticity is evident at the level of subcellular interactions with perineuronal structures. The extracellular matrix (ECM) is a complex structure composed of elastin, fibronectin, integrin, glycoprotein, and polysaccharide macromolecules that surrounds cells, and which accounts for 10–20% of the brain’s volume (Cragg, 1979). The ECM supports a wide range of physiological processes including remyelination (You and Gupta, 2018), stem cell storage (Roll and Faissner, 2014), protection against reactive oxygen species (Cabungcal et al., 2013), synapse development (Chan et al., 2020), and stabilization of mature synapses (Frischknecht and Gundelfinger, 2012).

Extracellular matrix components are not uniformly distributed across brain regions or cell types (Dauth et al., 2016). For example, perineuronal nets (PNNs) are chondroitin sulfate glycosaminoglycan (CS-GAG)-based, net-like structures found within ECMs that preferentially surround cPVins and their proximal dendrites (Härtig et al., 1999; Wegner et al., 2003). A hyaluronan backbone acts as a scaffold for assembling CS-GAG, link proteins, and other anchoring proteins into these condensed mesh structures which attach to cell membranes (Brückner et al., 1993; Day and Prestwich, 2002). PNNs contain a number of specific CS-GAGs (e.g., aggrecan, neurocan, brevican) which represent diverse sidechains that attach to lectican protein cores (Giamanco et al., 2010; Frischknecht and Gundelfinger, 2012). How the specific combination of these molecules influences PNN properties is an active area of investigation (Galtrey et al., 2008; Miyata et al., 2018).

Perineuronal nets envelop cPVins and their synapses as developmental windows of plasticity close (Pizzorusso et al., 2002; Berardi et al., 2003; McGee et al., 2005; Nowicka et al., 2009; Carulli et al., 2010; Beurdeley et al., 2012; Ye and Miao, 2013; Happel et al., 2014; Krishnan et al., 2015; Balmer, 2016). This deposition occurs in an activity dependent manner (Reimers et al., 2007; Favuzzi et al., 2017); in fact, cPVin activity is necessary for PNN assembly (Cisneros-Franco and de Villers-Sidani, 2019). The pervasiveness of PNNs fluctuates over the course of an animal’s lifetime in response to sensory experience. These fluctuations are enabled by endogenous metalloproteinase (MMP)— enzymes that degrade the backbone structures of the PNNs (Rossier et al., 2015), and which are themselves regulated by Tissue Inhibitory of Matrix Metalloproteinase (TIMP)-1 (Magnowska et al., 2016). The expression of MMPs declines with the closure of developmental critical periods, but retention of MMPs in the adult brain allows for sensory plasticity through PNN degradation (Murase et al., 2017).

Expression of the parvalbumin (PV) protein itself also fluctuates according to cPVin activity (Kamphuis et al., 1989; Patz et al., 2004). Downregulation of PV has broad consequences for the physiological and synaptic properties of cPVins (Caillard et al., 2000; Schwaller, 2012), as well as for behavior (Wöhr et al., 2015). These effects are likely related to PV’s role in rapidly buffering excess calcium (Aponte et al., 2008). Downregulation of PV expression may also reflect shifts in regional excitatory-inhibitory balance resulting from pathology, for example, as broadly observed in animals models and post-mortem tissue from patients with Autism Spectrum Disorders (Schwaller, 2012; Filice et al., 2016). Moreover, there is evidence of strong mutual influence between the composition and maturity of PNNs and the excitability and activity of cPVins (Balmer, 2016; Favuzzi et al., 2017; Chu et al., 2018; Gottschling et al., 2019; Devienne et al., 2021). The complex relationship between PV expression, cPVin physiology, and PNN assembly are matters of ongoing study, and will further inform our understanding how cPVins regulate cortical plasticity.

Well-developed PNNs appear to be a necessary component for termination of developmental critical periods. Manipulations that disrupt critical period closure (e.g., dark-rearing) delay disinhibition and slow the formation of PNNs (Lander et al., 1997; Pizzorusso et al., 2002). Moreover, genetic knockout of PNN components (Rowland et al., 2021), enzymes required for the synthesis of those components (Hou et al., 2017), or link proteins necessary to establish PNN architecture (Carulli et al., 2010; Ribic, 2020) also prevent critical period closure. Meanwhile, pharmacological inhibition of endogenous MMP activity disrupts ocular dominance plasticity in the adult visual cortex, i.e., a shift in V1 activity to preference the spared eye (Pielecka-Fortuna et al., 2015; Akol et al., 2022).

Conversely, plasticity can be reinstated by pharmacological dissolution of PNNs in the adult cortex, which allows for enhanced synaptic plasticity and physical restructuring of mature PV+ synapses (Magnowska et al., 2016; Ferrer-Ferrer and Dityatev, 2018). For example, in the adult visual cortex, removal of PNNs reinstates most aspects of ocular dominance plasticity, although it is “incomplete” compared to that instated during developmental critical periods (Pizzorusso et al., 2002). This includes disinhibition seen as a drop in V1 cPVin firing rates (Lensjø et al., 2017), and remobilization of dendritic spines, allowing them to lengthen and potentially change the architecture of established circuitry (de Vivo et al., 2013; Faini et al., 2021). Likewise, the effects of monocular deprivation during development on V1 can be partially reversed during adulthood via environmental enrichment, potentially due to a drop in PNNs and reduced inhibitory drive (Sale et al., 2007).

Perineuronal net-fluctuations as a marker of plasticity can also be observed outside of the context of monocular deprivation. For example, reintroduction of light in dark-exposed animals triggers V1 plasticity through the endogenous degradation of the ECM by MMP-9, and resulting disinhibition of fast-spiking interneurons (Murase et al., 2017). In the barrel cortex, whisker trimming leads to downregulation of PV-PNNs (McRae et al., 2007). Finally, evidence from our lab and others also suggests that aberrantly high or low numbers of PNNs during critical periods of sensory learning may impair A1 sound processing (Krishnan et al., 2017; Wen et al., 2018; Pirbhoy et al., 2020).

Physiological Properties of Cortical Parvalbumin-Expressing Enterneurons

cPVins represent an estimated 5–20% of cells in the brain, and up 40–50% of the inhibitory population, based on developmental progenitor and tracing studies (Tamamaki et al., 2003; Butt et al., 2005; Rudy et al., 2011; Rodarie et al., 2021). Despite similar genetic profiles (Tasic, 2018), and gross numbers of afferent thalamic inputs (Pouchelon et al., 2021) across cortical IN subtypes, individual cPVins have unique cellular physiological characteristics that enable them to locally shape sensory responses.

Temporal Control of Sensory Responses

One such characteristic is their high spiking rates they exhibited spontaneously in vivo or in response to somatic current injection (Connors and Gutnick, 1990), leading to their original label of “fast spiking” cells (Kawaguchi and Hama, 1987; Kawaguchi and Kubota, 1993; Kawaguchi, 1995). “Fast spiking” cells are now widely acknowledged to correspond to PV expressing GABAergic basket cells and to a lesser extent, chandelier cells (Kawaguchi et al., 2019). The fast spiking attributes of cPVins result from their expression of rapidly repolarizing Kv3 potassium channels (Rudy and McBain, 2001) and Ca2+-permeable AMPA receptors which have rapid gating kinetics (Jonas et al., 1994; Geiger et al., 1997).

The brisk spiking output and high reliability of cPVin firing enables them to exert tight temporal control over cortical excitation. They convey precisely timed, feedforward inhibition that lags just behind feedforward excitatory input from the thalamus, thereby imposing strict temporal limits (Wehr and Zador, 2003; Zhang et al., 2003; Mittmann et al., 2005; Wilent and Contreras, 2005; Heiss et al., 2009). The potency of this mechanism is increased by cPVins’ particular sensitivity to onsets and offsets of sensory stimuli (Wehr and Zador, 2003; Weible et al., 2014; Keller et al., 2018; Li et al., 2021). Over broader timescales, cPVins make important contributions to stimulus specific adaptation, which allows the brain to maintain sensitivity to novel or behaviorally relevant stimuli among repetitive distractors (Pérez-González and Malmierca, 2012; Natan et al., 2015).

Precise temporal regulation of cortical activity by cPVins is also relevant for more persistent changes in sensory responses through long-term potentiation (LTP) or depression (LTD). These mechanisms rely on coincidence detection of pre- and post- synaptic neuronal spiking. Spike timing-dependent plasticity (STDP), which modifies synapses based on the precise relative timing of individual pre- and post- synaptic spikes (Bi and Poo, 1998), has been widely observed in the cortex (Bender et al., 2006; Meliza and Dan, 2006; Feldman, 2012) and supported by computational models of cortical cell assembles (Wenisch et al., 2005; Klampfl and Maass, 2013). Only more recently has the importance of STDP at inhibitory output synapses, and at cPVin specific synapses, been appreciated (D’Amour and Froemke, 2015; Vickers et al., 2018). Computational modeling predicts that maturation of cPVins helps stabilize STDP such that the most temporally coherent inputs onto cortical pyramidal neurons are strengthened (Kuhlman et al., 2010). Experimental data indicate that LTD of cPVin output onto principal neurons implements cortical disinhibition to regulate plasticity, and is necessary for remodeling the structure of sensory maps (Vickers et al., 2018).

Selectivity and Gain Control of Sensory Responses

Another important cPVin feature is their broad tuning relative to neighboring pyramidal cells (Runyan et al., 2010; Kuhlman et al., 2011; Li et al., 2012), and other inhibitory subtypes (Kerlin et al., 2010; Ma et al., 2010; Li et al., 2015), however, this observation varies among different sensory brain regions, laminar depth, and among experimentalists (Hofer et al., 2011; Zariwala et al., 2011; Moore and Wehr, 2013).

Many studies have examined the specific operations that cPVins perform on sensory responses through optogenetic activation and/or inactivation of the population. Collectively, these studies have uncovered a panoply of conflicting transformations of sensory representations by cPVins. These include non-selective divisive gain control (Atallah et al., 2012; Wilson et al., 2012; Hamilton et al., 2013), receptive field sharpening (Lee et al., 2012; Kaplan et al., 2016), temporal sharpening (Andermann et al., 2011; Vecchia et al., 2020), non-selective stimulus adaptation (Natan et al., 2015), and spatial constraint of neuronal ensemble activity (Agetsuma et al., 2018).

Broad inhibition of cortical output by cPVins derives from their ability to integrate excitatory activity across a wide area via their extensive dendritic fields (Poo and Isaacson, 2009; Sohal et al., 2009; Packer and Yuste, 2011; Karnani et al., 2014; Hage et al., 2022). This allows cPVins to exert gain control over cortical activity and to maximize the signal-to-noise ratio of information carried to downstream targets. Their intrinsic features also support their synchronized firing (Galarreta and Hestrin, 2001; Bartos et al., 2007; Ferguson and Gao, 2018). First, cPVins are highly sensitive to a large number of high amplitude, phase-locked inputs as a result of their Kv1 channel expression (Goldberg et al., 2008). Second, cPVins are coupled by gap-junctions (Fukuda and Kosaka, 2000; Tamás et al., 2000; Galarreta and Hestrin, 2001; Bartos et al., 2002). Third, cPVins maintain highly branched dendrites to access synaptic input over a sprawling area (Bartholome et al., 2020). In sum, these features facilitate the spread and synchronization of phasic, oscillatory activity across large brain regions.

cPVin Control of Cortical Network Activity and Behavioral State

At the broadest spatiotemporal scale, cPVin networks have a capacity to coordinate cortical output of large areas. On a population level, the rhythmic firing of cPVins establishes gamma oscillations, a high frequency, (30–80 Hz) periodic signal that results from synchrony of local field potentials and synaptic activity (Buzsáki and Draguhn, 2004). Genetic reduction of the number of cPVins attenuates gamma oscillations (Kalemaki et al., 2018), while optogenetically driving cPVin pools triggers gamma activity (Cardin et al., 2009; Sohal et al., 2009; Etter et al., 2019). Computational models suggest that gamma wave strength directly correlates with the expression of PV and GAD67 (Volman et al., 2011).

Gamma oscillations critically depend on NMDA receptor expression by cPVins (Korotkova et al., 2010; Gonzalez-Burgos and Lewis, 2012). In contrast to the fast spiking properties of cPVins, NMDA receptors activate synaptic currents with slow kinetics (Forsythe and Westbrook, 1988). The key contribution of NMDA receptors to the establishment of gamma rhythms appears to be their capacity to stabilize and coordinate recruitment of cPVins into synchronous ensembles (Carlén et al., 2012; Cornford et al., 2019).

The functional roles of gamma oscillations are less well understood. Oscillations reflect cPVin task-dependent activity and recruitment of greater numbers of cPVins during sensory learning (Gotts et al., 2012; Brunet et al., 2014; Ainsworth et al., 2016). Changes in gamma activity are critical for learning rules, such as during operant conditioning (Caroni, 2015; Lintas et al., 2021). Moreover, they are necessary for updating cortical responses when there is a mismatch between the history of stimuli-salience association and the new rules of the rules of sensory learning (Cho et al., 2020).

Further, the presence of gamma oscillations also correlates with the familiarity of sensory stimuli, which suggests they facilitate memory traces in sensory cortices (Womelsdorf et al., 2006; Headley and Weinberger, 2011; Weinberger et al., 2013; Cooke et al., 2015; Kissinger et al., 2018, 2020). In addition these oscillations may contribute to storing memories (Galuske et al., 2019). Gamma waves can also be triggered in response to meaningful behavioral outcomes (Fries et al., 2001; Kim et al., 2015; Ray and Maunsell, 2015; Cardin, 2016; Cho et al., 2020), which could imply their contribution to salience encoding.

Finally, gamma power in sensory cortices correlates with behavioral states, such as attention (Börgers et al., 2008; Sobolewski et al., 2011; Bosman et al., 2012; Kim et al., 2016), locomotion (Niell and Stryker, 2010), and arousal (Kim et al., 2015; Vinck et al., 2015). Evidence to date suggests that gamma rhythms in sensory cortex reflect top-down, modulated cPVin activity that integrates learning and memory contexts.

cPVin Control of Maternal-Experience Induced Auditory Cortical Plasticity

In mice, plasticity in the auditory cortex (ACtx) is triggered by maternal experience with pups. A broad base of evidence supports the conclusion that auditory plasticity in adult females helps to sharpen and amplify responses to behaviorally salient distress vocalizations from the pups. Here we review these data, highlighting work by our lab identifying cPVin-specific physiological mechanisms underlying “maternal,” experience-dependent auditory plasticity. These mechanisms share features of classical models of experience-dependent cortical plasticity, including those observed in developmental critical periods. Importantly, however, this plasticity is activated not by an exogenous experimental manipulation, but simply by social experience.

Primiparous mice learn to recognize and respond to pup ultrasonic vocalizations (USVs), which signal distress when pups are isolated and become hypothermic. This learning is manifested behaviorally in the retrieval of pups back to the nest by the dams (Sewell, 1970; Ehret et al., 1987). However, the same learning is also exhibited by co-housed, virgin females (“surrogates”) outside of the influence of pregnancy-induced hormone fluctuations (Rosenblatt, 1967; Galindo-Leon et al., 2009; Cohen et al., 2011; Cohen and Mizrahi, 2015; Stolzenberg and Champagne, 2016; Krishnan et al., 2017; Lau et al., 2020; Carcea et al., 2021). Importantly, retrieval learning correlates with plasticity of ACtx inhibitory networks (Liu and Schreiner, 2007; Galindo-Leon et al., 2009; Cohen et al., 2011; Lin et al., 2013; Cohen and Mizrahi, 2015; Marlin et al., 2015; Lau et al., 2020). Specifically, ACtx pyramidal cells from pup or ultrasonic vocalization exposed females demonstrate time-locked responses. However, these responses are attenuated in naïve counterparts.

Our mechanistic understanding of maternal experience-induced plasticity has been enhanced by our research with Mecp2 mutant mouse models. While wild-type (WT) maternal surrogates readily learn to perform pup retrieval, adult female mice that are missing one copy of the X chromosome -linked Mecp2 gene (Mecp2) fail to perform this task (Krishnan et al., 2017). We examined potential molecular mechanisms underlying this behavioral phenotype and found that exposure to pups initiated a transient increase in expression of GAD67 in the ACtx of both Mecp2 subjects and wild type littermates (Mecp2) (Krishnan et al., 2017). However, pup experience also leads to overexpression of parvalbumin protein (PV) and cPVin-associated PNNs in the ACtx of Mecp2 subjects, relative to naïve and Mecp2 controls (Krishnan et al., 2017). In the context of the literature discussed above, we hypothesized that the overexpression of PV and PNNs in the brains of Mecp2 surrogates inhibits the ACtx plasticity underlying retrieval learning, paralleling restriction of developmental critical periods. That is, while depression of these markers was not observed in WT subjects, their aberrant elevation in Mecp2 surrogates is sufficient for impairing behavioral responses to USVs. Therefore we surmise that a threshold level of PV and PNN expression, below which ACtx cells of Mecp2 controls maintain expression, allows for normal learning.

Further, we predicted that dissolution of PNNs might reinstate ACtx plasticity, as has been found in other sensory cortices. Indeed, PNN dissolution in the ACtx just prior to birth of the pups ameliorated overexpression of PNNs and facilitated retrieval performance in Mecp2 subjects (Krishnan et al., 2017). Again this suggests a cPVin specific mechanism or disruption that underlies the ability of Mecp2 subjects to respond behaviorally to USV cues. We further linked cPVins to maternal learning by showing that deletion of Mecp2 in cPVins was sufficient to disrupt early retrieval (Krishnan et al., 2017).

We next examined the effects of pup exposure on stimulus-driven activity in the ACtx of awake mice. Consistent with a central role of cPVin inhibitory drive in the pup-induced ACtx plasticity, we found that, in Mecp2 subjects, responses of cPVins to USVs were weaker after experience with pups (Lau et al., 2020). The diminished output of individual cPVins was mirrored by a complementary increase in the magnitude of responses from excitatory, putative pyramidal ACtx neurons (Lau et al., 2020). That is, maternal experience triggers a drop in cPVin responses to USV playback in the ACtx of Mecp2. This is complemented by increased pyramidal neuron responses to USVs.

In contrast, these physiological changes were not observed in the ACtx of Mecp2. More specifically, cPVin activity in response to USVs remained high despite experience with pups. These firing patterns matched the observed elevation in PV protein expression in Mecp2, but not Mecp2 subjects. That is, given the activity-dependent expression of PV, the continued high stimulus-evoked firing rates of cPVins in Mecp2 links the cPVin-specific molecular expression to a functional lack of plasticity.

This pattern of results is also consistent with our understanding of sensory cortical plasticity in other systems, which is often initiated by disinhibition. In sum, inhibitory ACtx plasticity maximizes neural “contrast” and appears to align temporal dynamics of cortical response to USVs, enabling pup retrieval (Liu and Schreiner, 2007; Shepard et al., 2015). These findings are very much in keeping with the physiological properties of cPVin that allow them to rapidly regulate the timing of cortical output.

As others have suggested, studying sensory processing in natural and/or socially salient contexts may be particularly effective at triggering rapid or robust plasticity (Taborsky et al., 2015). Our work uses a maternal experience paradigm to evoke a natural and spontaneous form of cPVin-mediated sensory cortical plasticity in adult animals. However, there are many remaining questions to be answered. What does the population level structure of cPVin activity, including cortical gamma oscillatory activity, in response to USV calls look like, and how does experience with pups change those patterns? Do large scale oscillations reflect, or regulate, synchronous activity that favors synaptic modification in response to behavioral state? How does the activity of cPVins, and other inhibitory cortical populations, differ in freely behaving mice, as opposed to awake, head-fixed mice? What is the contribution, if any, of other cortical inhibitory subpopulations to maternal experience-induced auditory plasticity? How is cPVin-mediated, sensory cortical plasticity modulated by multisensory integration? And how might it be modulated by behavioral states including attention, arousal, and emotional salience or valence?

Here we briefly summarize the shared features of plasticity during visual, developmental critical periods and during adult, auditory learning in natural contexts. First, both are marked by dynamic downregulation of cortical inhibition that results from decreased stimulus-evoked firing by cPVins, and leads to a concomitant increase in excitability of principal cortical neurons. This likely reflects changes to relative synaptic strengths from cPVins that lower the LTP threshold of pyramidal cells (Smith et al., 2009). Second, this cortical disinhibition is permissive for changes in connectivity, including the number of inputs by cPVins onto pyramidal cells. Third, periods of plasticity are delineated by fluctuations in the expression of parvalbumin protein and perineuronal nets, including prominent elevation of these markers as windows of heightened plasticity close. These molecular fluctuations may aid in mobility, or pruning, of synaptic inputs (Huang et al., 2015) that facilitate synaptic plasticity. Finally, in both cases, changes that reflect cortical plasticity do not preclude mechanisms of plasticity upstream of the cortex, e.g., changes to thalamic projections onto cPVins spurred by sensory experiences (Sommeijer et al., 2017). Nevertheless, the similarity in cortical plasticity features provides a general framework through which we can examine cPVin-specific contributions to sensory learning.

Conclusion and Future Directions

Here we have focused on the known contributions of functional and structural cPVin plasticity for supporting sensory learning. However, the current understanding is far from complete. There are many areas of on-going investigation which promise to enhance this understanding including: contributions of cPVin to encoding natural or other complex stimuli (Zhu et al., 2015), real-time dynamics between cPVins and other cell types that make up canonical units (Rikhye et al., 2021), modulation of cPVins by top-down brain-state dynamics (Pakan et al., 2016), better dissection of transcriptomic (Fishell and Kepecs, 2020) and spatial (Large et al., 2018) profiles, and cPVin circuit specificity (Jiang et al., 2021).

We have also emphasized, by highlighting our own work, the need for further dissection of cell type specific plasticity in naturally occurring, socially salient contexts. Technical advancements, such as the development of less invasive approaches, are particularly important for implementing such investigations. For example, extracranial delivery of optogenetic (Chen et al., 2021) and ultrasonic stimulations (Estrada et al., 2021) as well as wireless recording capabilities (Shin et al., 2022) will make observation and manipulation of cell type specific populations during freely moving task engagement less cumbersome.

Advancing our understanding of cPVin plasticity has implications beyond scientific inquiry. Aberrant cPVin circuitry has been implicated in a diverse set of neurocognitive conditions (Huang, 2014; Ferguson and Gao, 2018; Filice et al., 2020). Therefore, a thorough understanding of cell type specific contributions to synaptic plasticity and network-wide oscillatory patterns has wide-spread implications for diagnostic and therapeutic advancement.

Author Contributions

Both authors listed have made a substantial, direct, and intellectual contribution to the work, and approved it for publication.

Conflict of Interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher’s Note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.