Prefrontal Interneurons: Populations, Pathways, and Plasticity Supporting Typical and Disordered Cognition in Rodent Models

Abstract

Prefrontal cortex (PFC) inhibitory microcircuits regulate the gain and timing of pyramidal neuron firing, coordinate neural ensemble interactions, and gate local and long-range neural communication to support adaptive cognition and contextually tuned behavior. Accordingly, perturbations of PFC inhibitory microcircuits are thought to underlie dysregulated cognition and behavior in numerous psychiatric diseases and relevant animal models. This review, based on a Mini-Symposium presented at the 2022 Society for Neuroscience Meeting, highlights recent studies providing novel insights into: (1) discrete medial PFC (mPFC) interneuron populations in the mouse brain; (2) mPFC interneuron connections with, and regulation of, long-range mPFC afferents; and (3) circuit-specific plasticity of mPFC interneurons. The contributions of such populations, pathways, and plasticity to rodent cognition are discussed in the context of stress, reward, motivational conflict, and genetic mutations relevant to psychiatric disease.

Introduction

The PFC coordinates neural communication across expansive brain networks to facilitate high-order cognitive functions. Complex inhibitory microcircuits within PFC, comprised of its many GABAergic interneurons and their interconnections, dynamically gate and integrate neural input from distal brain regions to support contextually tuned behaviors (Miller and Cohen, 2001). Unsurprisingly, disruptions to these microcircuits have been implicated in a wide array of neuropsychiatric disorders, including schizophrenia, autism spectrum disorder, and depression (Dienel and Lewis, 2019; Fogaça and Duman, 2019; Yan and Rein, 2022). Therefore, characterization of the rich diversity of the interneurons that comprise PFC microcircuits, their synaptic connectivity, and the plastic nature of this connectivity stand to inform the neural basis of typical and disordered cognition. Here, we briefly review recent advances in the parsing of rodent mPFC interneurons into molecularly, anatomically, and functionally defined subpopulations. We also describe newly uncovered complexity in the synaptic connections between mPFC interneurons and their distal inputs, and novel mechanisms of plasticity and neuromodulation that regulate this long-range synaptic connectivity. How these distinct cell types, circuits, and circuit adaptations guide rodent cognitive functions is discussed in relation to stressful and rewarding experiences, motivational conflict, and disease-relevant genetic insults.

Prefrontal interneuron populations

The rodent mPFC harbors a network of inhibitory interneurons interspersed among glutamatergic pyramidal neurons. Despite their vast heterogeneity stemming from differences in developmental origin, genetic profile, morphology, connectivity, and functional properties (DeFelipe et al., 2013; Kepecs and Fishell, 2014), nearly all mPFC interneurons can be broadly (albeit imperfectly) classified based on their expression of parvalbumin (PV), somatostatin (SST), or the 5HT3a serotonin receptor (Rudy et al., 2011; Tremblay et al., 2016). PV-INs, the most abundant of these interneuron subtypes, exert potent inhibitory control over pyramidal neurons by synapsing onto their somatic and perisomatic compartments (Kepecs and Fishell, 2014). SST-INs are less abundant and tend to innervate the distal dendrites of pyramidal neurons (Kepecs and Fishell, 2014), positioning them to gate the influence of local and long-range inputs. Recent work indicates that mPFC SST-INs can also disinhibit pyramidal neurons through their monosynaptic inhibition of neighboring PV-INs (Xu et al., 2019; Cummings and Clem, 2020; Jiang et al., 2021). Interneurons that express vasoactive intestinal polypeptide (VIP-INs) represent the largest group of 5HT3a-positive interneurons (Rudy et al., 2011; Tremblay et al., 2016). VIP-INs are the canonical “disinhibitors”: in mPFC and neocortex broadly, VIP-INs inhibit SST-INs and to a lesser extent PV-INs to promote pyramidal neuron activity (Pi et al., 2013). Importantly, the foundational studies examining interneuron specializations were conducted in sensory cortices (Rudy et al., 2011; Tremblay et al., 2016). While recent work indicates that interneuron classes are broadly conserved across neocortex (e.g., Tasic et al., 2018), specific classifications and properties of mPFC interneurons are likely to differ from those of sensory regions (e.g., Whissell et al., 2015; Y. Kim et al., 2017).

Novel molecularly and functionally defined prefrontal interneuron subpopulations

Within these heterogeneous populations are interneuron subclasses, each with unique anatomical connectivity, physiological properties, and behavioral contributions, that are defined by expression of additional proteins and neuropeptides, such as, calretinin, (Saffari et al., 2019), cholecystokinin (CCK) (Nguyen et al., 2020), and corticotropin-releasing factor (P. Chen et al., 2020; see also Tasic et al., 2018; Yao et al., 2021). Among these is a novel subclass of interneurons expressing prodynorphin (PDyn)-derived peptides, including dynorphins (Dyn). In mPFC, PDyn is expressed in pyramidal neurons and a subset of SST-INs (Sohn et al., 2014; ACNP 60th Annual Meeting, 2021). Tejeda and colleagues have recently demonstrated that Dyn⁺ SST-INs, which comprise ∼10% of SST-INs, are localized to deeper mPFC layers relative to Dyn-lacking SST-INs (Fig. 1A) (ACNP 60th Annual Meeting, 2021). Using intersectional viral and genetic approaches, Tejeda and colleagues further demonstrated that Dyn⁺ SST-INs are heavily activated by footshocks and footshock-predictive cues (ACNP 60th Annual Meeting, 2021). Interestingly, Dyn⁺ SST-INs immediately adapted their activity on the first omission of the shock during threat extinction procedures by switching to inhibitory responses during the shock-predictive cue. In contrast, Dyn-lacking SST-INs were activated by footshocks but showed little response to footshock-predictive cues during associative learning or inhibitory responses during threat extinction, consistent with the notion that Dyn⁺ SST-INs represent a distinct subtype of SST-IN.

Figure 1.

A, Schematic represents distinct and overlapping populations of SST-INs in mPFC (ACNP 60th Annual Meeting, 2021; Cummings et al., 2022). B, Schematic represents simplified connectivity motifs between superficial (Calb⁺) and deep (Calb-lacking) pyramidal neurons in vHPC with downstream mPFC neuron types (Sánchez-Bellot et al., 2022), and between long-range GABAergic neurons in mPFC with VIP-INs in dHPC (Malik et al., 2022). SST-INs, somatostatin-positive interneurons; dHPC, dorsal hippocampus; mPFC, medial prefrontal cortex; PN, pyramidal neurons; FSI, fast-spiking interneurons; ISI, irregular-spiking interneurons; RSI, regular-spiking interneurons. Created with www.BioRender.com.

Supplementing these and other genetic approaches to dissecting novel interneuron subclasses and their behavioral contributions (Ma et al., 2006; He et al., 2016) is new work that seeks to characterize experientially and behaviorally defined interneuron populations (Cummings et al., 2021). In a recent study, Cummings et al. (2022) developed a novel intersectional viral, transgenic, and activity-dependent tagging strategy to gain genetic access to mPFC SST-INs that are activated in response to an experimental manipulation. With this approach, the authors were able to tag, manipulate, and functionally interrogate populations of mPFC SST-INs activated by auditory fear conditioning or morphine administration (Fig. 1A). Doing so, they found that fear-activated SST-INs, which represented ∼30% of all prelimbic mPFC SST-INs, were selectively reactivated during memory retrieval and were necessary and sufficient for the expression of cued fear. Fear-tagged SST-INs also exhibited unique circuit properties, including greater inhibitory drive onto PV-INs and non–fear-tagged pyramidal neurons. Moreover, morphine treatment recruited a non-overlapping population of SST-INs that promoted motivational reward behaviors and opposed fear memory expression. Furthermore, optogenetic activation of fear- and morphine-responsive SST-INs recruited distributed brain networks related to fear and reward processing, respectively (Cummings et al., 2022).

Long-range GABAergic prefrontal projection neurons

The vast majority of mPFC GABAergic neurons project locally and play crucial roles in local microcircuit computations and input-output transformations. In contrast, a small fraction of mPFC GABAergic neurons send axons to remote cortical and subcortical brain regions, thus forming long-range GABAergic projections (Lee et al., 2014; Tomioka et al., 2015; Malik et al., 2022). Although the existence of sparse long-range GABAergic projection neurons in hippocampus (Sik et al., 1994; Jinno, 2009) and sensorimotor cortices (Tamamaki and Tomioka, 2010; Melzer and Monyer, 2020) has been known for over a decade, we are only now beginning to understand the organization and function of prefrontal long-range GABAergic projections.

The first evidence of long-range GABAergic projection neurons in the mPFC came from Lee et al. (2014), who described the properties and function of mPFC GABAergic projections to the nucleus accumbens, a key subcortical node for reward and aversion circuits in the brain. The authors demonstrated that optogenetic activation of this molecularly heterogeneous population of accumbens-projecting long-range GABAergic neurons enhanced aversion behavior. This and subsequent studies have revealed axon collaterals of mPFC long-range GABAergic neurons in multiple subcortical structures, such as the BLA, claustrum, striatum, and VTA (Tomioka et al., 2015).

The recent discovery of long-range GABAergic neurons from the mPFC to the dorsal hippocampus (dHPC) by Malik et al. (2022) has disrupted long-held assumptions about the neural pathways supporting mPFC-HPC communication and the nature of this top-down information flow. Indeed, it was widely assumed that the mPFC transmits information to the dHPC near exclusively via indirect excitatory connections with the thalamic nucleus reuniens (Vertes et al., 2007). Enriching this understanding, Malik et al. (2022) demonstrated that monosynaptic projections from a molecularly and physiologically mixed population of GABAergic neurons in the mPFC preferentially inhibit dHPC VIP-INs (Fig. 1B). As in mPFC and cortex generally, VIP-INs disinhibit hippocampal microcircuits (Acsády et al., 1996); thus, stimulating mPFC-dHPC long-range GABAergic projections increased dHPC feedforward inhibition (a mechanism capable of enhancing the “signal-to-noise ratio” of select excitatory neural pathways/ensembles) (Buzsáki, 1984) and enhanced object-related dHPC spatial encoding of objects in the environment (Malik et al., 2022). Accordingly, activating or inhibiting long-range GABAergic projections enhanced or suppressed object exploration, respectively.

Projection-specific targeting of prefrontal interneurons

mPFC receives long-range excitatory inputs from across the brain, including contralateral mPFC, thalamus, BLA, and ventral hippocampus (vHPC) (Anastasiades and Carter, 2021). Each of these projections is proposed to support unique functional roles during behavior (Sierra-Mercado et al., 2011). For example, BLA inputs are thought to support the learning, expression, and updating of affective associations (e.g., aversive threat memory) (Sotres-Bayon et al., 2012; Janak and Tye, 2015), whereas vHPC inputs are thought to convey context, or state information on which associations can be formed (Gershman et al., 2010; Maren et al., 2013; Marek et al., 2018).

Differential targeting of prefrontal interneurons by inputs from widespread brain regions

Although much evidence for the function of long-range inputs into mPFC has focused on their interactions with pyramidal neurons, a key means by which these long-range inputs influence mPFC circuitry is via dense and specific connectivity with local interneurons (Anastasiades and Carter, 2021). Anatomical studies using anterograde and retrograde tracing techniques suggest that all interneuron types in mPFC receive glutamatergic inputs by long-range sources (Ährlund-Richter et al., 2019; Q. Sun et al., 2019). Accordingly, electrophysiological recordings have shown that long-range inputs form excitatory synaptic connections with mPFC PV-, SST-, and VIP-INs, albeit in different layers that correspond to the biased laminar positioning of each interneuron type (Delevich et al., 2015; McGarry and Carter, 2016; Marek et al., 2018; Lee et al., 2019; Anastasiades et al., 2021). However, researchers have identified notable preferential targeting of interneuron types by different long-range inputs. For example, contralateral mPFC inputs preferentially target PV⁺ chandelier cells in layer 2/3 (Lu et al., 2017), whereas vHPC inputs show robust targeting of layer 5 CCK-INs (Liu et al., 2020) and layer 2/3 VIP-INs (Lee et al., 2019). Similarly, inputs from mediodorsal thalamus target a specific class of layer 1 VIP-INs, whereas inputs from ventromedial thalamus show biased targeting of apical tuft-targeting, neuron-derived neurotrophic factor-expressing interneurons (Collins et al., 2018; Anastasiades et al., 2021). This network of generalized and specific targeting of different interneuron types provides the foundation needed for contextually tuned mPFC computations.

Differential targeting of prefrontal interneurons by intermingled long-range inputs

The complex innervation of mPFC interneuron types is accompanied by marked heterogeneity of afferent projection neurons. Indeed, regions that provide mPFC afferents, such as the hippocampus (Cembrowski et al., 2016, 2018a, 2018b; Gergues et al., 2020), thalamus (C. Gao et al., 2020), contralateral mPFC (Murugan et al., 2017), and BLA (J. Kim et al., 2016), are composed of intermingled populations of genetically distinct neurons that often have opposing function during behavior.

Recent work characterized important structural and functional heterogeneity within vHPC inputs to the mPFC (Sánchez-Bellot et al., 2022). Consistent with a previous study showing at least two molecularly distinct populations of mPFC-projecting vHPC neurons (Cembrowski et al., 2018a), Sánchez-Bellot et al. (2022) found two populations of mPFC-projecting neurons in vHPC that were differentiated by their expression of Calbindin1 (Calb1), their position along the radial axis of the vHPC pyramidal neuron layer (deep and superficial), and their biased targeting of mPFC cell types. Calb1⁺ vHPC input preferentially targeted adapting mPFC interneurons (corresponding to CCK-INs and SST-INs), whereas Calb1-lacking vHPC input preferentially targeted pyramidal neurons and fast-spiking interneurons (corresponding to PV⁺ basket cells; Fig. 1B). Thus, these parallel vHPC-mPFC pathways are well placed to control the balance of feedforward inhibition onto pyramidal neuron dendrites and somas. Sánchez-Bellot et al. (2022) further showed that the parallel inputs from vHPC had distinct activity during, and control over, exploratory behavior. Calb1⁺ input was preferentially active on entry to the open arms of the elevated plus maze, whereas Calb1-lacking input was active on entry to the closed arms. Consistent with this opposing activity in the two pathways and known behavioral effects of directly manipulating the mPFC populations they target (Soumier and Sibille, 2014; Canetta et al., 2016; Berg et al., 2019), activation of Calb1⁺ input promoted exploration of the open arms, whereas activation of Calb1-lacking input reduced open arm exploration.

Prefrontal interneuron plasticity

Plastic changes of interneuron structure and function, whether through disease-relevant genetic insult, exogenous neuronal activation, receptor modulation, or experience-induced alterations, are poised to remodel computations within mPFC, and the routing of circuit-specific information through mPFC, to sculpt behavior. Numerous interneuron adaptations have been reported following each of these types of manipulations, manifesting as changes to mesoscopic and microscopic interneuron structure (e.g., Boksa et al., 2016; Al-Absi et al., 2020; Gildawie et al., 2020; Bueno-Fernandez et al., 2021), protein expression (e.g., Stedehouder et al., 2018; Mukherjee et al., 2019; Reichelt et al., 2021), and intrinsic physiology (e.g., Campanac and Hoffman, 2013; Dao et al., 2020; Zorrilla de San Martin et al., 2020). Here, we highlight some recent advances in our understanding of how plasticity in the synaptic and circuit connectivity of mPFC interneurons can reshape typical and disordered mPFC function and associated behavior (see also Yang et al., 2021).

Genetic insult-induced prefrontal interneuron plasticity

Rodent models bearing genetic mutations relevant to psychiatric diseases, such as schizophrenia and autism spectrum disorder, have deepened our understanding of how specific disease-related genes regulate prefrontal interneuron structure and function. Indeed, a wealth of data now links the pathophysiology of these neurodevelopmental disorders with impaired local and long-range synaptic connections with mPFC inhibitory microcircuits (e.g., Cho et al., 2015; Vogt et al., 2015; Selimbeyoglu et al., 2017; Delevich et al., 2020).

Animal models for the study of autism spectrum disorder present particularly robust links between specific gene disruptions and mPFC interneuron synaptic dysfunction. For example, a recent study showed that mice lacking one copy of the Pogz gene, which is involved in chromatin regulation and strongly linked with autism (Stessman et al., 2016), exhibit abnormal anxiety-related avoidance, impaired oscillatory synchrony between the vHPC and mPFC, and deficits in hippocampal excitatory input to fast-spiking (putative PV⁺) mPFC interneurons (Cunniff et al., 2020). Further support for autism-related microcircuit adaptations come from mice modeling the 16p11.2 duplication syndrome (Weiss et al., 2008), which show deficient GABAergic synaptic transmission and concurrent hyperexcitability in mPFC pyramidal neurons, as well as social and cognitive deficits. All of these phenotypes were rescued by restoring expression of the GABA synapse regulator, Npas4 (Rein et al., 2021). Likewise, a loss-of-function mutation in the autism-associated Shank3 gene (Durand et al., 2007) was recently shown to reduce dendritic inhibition onto mPFC pyramidal neurons via decreased NMDAR currents in, and reduced firing of, SST-INs (Ali et al., 2021). Strikingly, selective deletion of Shank3 from only BLA-projecting mPFC pyramidal neurons resulted in reduced inhibitory transmission onto these cells and reduced sociability (S. Kim et al., 2022). Together, these and many other studies are helping to inform the causal links between disease-relevant genetic insults, microcircuit connectivity, and behavior.

Activity-induced interneuron plasticity

Numerous studies have described long-term synaptic plasticity at long-range connections with the rodent mPFC following brain stimulation (e.g., Laroche et al., 1990; Takita et al., 1999; Maroun and Richter-Levin, 2003). In particular, connectivity between vHPC and mPFC has been demonstrated to be highly plastic, and this plasticity has been linked with cognitive function and disease-relevant dysfunction (e.g., Jay et al., 2004; Baudin et al., 2012; Tripathi et al., 2020; Park et al., 2021). Although mPFC interneurons have been implicated in gating some forms of activity-induced vHPC-mPFC plasticity (Caballero et al., 2014; Alvarez et al., 2020), plastic changes at vHPC and other long-range inputs to mPFC interneurons themselves are vastly unexplored (Lu et al., 2007; Sarihi et al., 2008; H. X. Chen et al., 2009). Further, given roles for mPFC interneurons in cognition-relevant functional connectivity between vHPC and mPFC (Abbas et al., 2018; Lee et al., 2019), and vHPC-mPFC dysconnectivity in models relevant to psychiatric disease (Mukai et al., 2015; Tamura et al., 2016; Song et al., 2022), it is important to understand how vHPC inputs interact with mPFC inhibitory microcircuits, whether these interactions are disrupted in disease-relevant models, and whether these interactions are plastic, offering a potential path to correcting circuit dysconnectivity (Kupferschmidt and Gordon, 2022).

To these ends, Clarity and colleagues used an all-optical approach to characterize in vivo dynamics and activity-induced plasticity of discrete mPFC interneuron population responses to vHPC input stimulation (Kupferschmidt et al., 2022). In wildtype and Df(16)A^+/– mice that model the schizophrenia-predisposing 22q11.2 deletion syndrome, vHPC inputs were optogenetically stimulated and postsynaptic Ca²⁺ responses in mPFC SST-, VIP-, and PV-INs were monitored with fiber photometry. SST-IN responses to vHPC terminal stimulation were weak at baseline in wildtype and Df(16)A^+/– mice but progressively increased over 50 d of minimal, periodic optogenetic stimulation (Fig. 2). The potentiation of SST-IN Ca²⁺ responses was blunted in Df(16)A^+/– relative to wildtype mice, and partially recovered with additional high-frequency optical vHPC input stimulation. In contrast, VIP- and PV-IN responses to vHPC input stimulation were initially strong but rapidly suppressed in wildtype and Df(16)A^+/– mice that received additional high-frequency optical stimulation. By reshaping the recruitment of mPFC interneurons by long-range inputs, these forms of plasticity and others like them stand to bias the routing of pathway-specific information through mPFC and may be leveraged to influence cognition-relevant circuit function and dysfunction.

Figure 2.

Schematic represents simplified long-range connectivity changes following activity-induced plasticity or stress/fear-associated experience. Left, Cell type-specific changes in in vivo functional connectivity between vHPC inputs and mPFC interneuron population types following repeated optical stimulation of vHPC inputs to mPFC (Kupferschmidt et al., 2022). Right, Bidirectional changes to synaptic connectivity between BLA inputs and mPFC SST-INs versus PV-INs following exposure to restraint stress (Joffe et al., 2022) or fear conditioning (Cummings and Clem, 2020). vHPC, ventral hippocampus; mPFC. medial prefrontal cortex; SST, somatostatin; VIP, vasoactive intestinal polypeptide; PV, parvalbumin; PN, pyramidal neuron. Created with www.BioRender.com.

Modulation-induced interneuron plasticity

The mPFC is rich with neuromodulators capable of regulating the synaptic input, output, and intrinsic excitability of mPFC interneurons. Dopamine (e.g., W. J. Gao and Goldman-Rakic, 2003; Floresco and Tse, 2007; Tierney et al., 2008; Anastasiades et al., 2019), acetylcholine (e.g., Komal et al., 2015; Tikhonova et al., 2018; Maksymetz et al., 2019), serotonin (e.g., Puig et al., 2010; Zhong and Yan, 2011), norepinephrine (e.g., Wang et al., 2013; Luo et al., 2015), endocannabinoids (e.g., Younts and Castillo, 2014; Liu et al., 2020), and glutamate (e.g., H. Sun and Neugebauer, 2011; Maksymetz et al., 2021), as well as various neuropeptides (e.g., Nakajima et al., 2014; Aracri et al., 2015; Vollmer et al., 2016; Birdsong et al., 2019; Casello et al., 2022) are among the many agents that exert complex and interacting receptor-, cell-, and circuit-specific modulation of mPFC interneurons.

Dyn, through its actions on κ opioid receptors (KORs), is emerging as a potent modulator of mPFC interneuron circuit function and stress-related motivated behavior. Recent work by Tejeda and colleagues revealed that Dyn inhibits glutamate release from various KOR-expressing mPFC afferents (e.g., BLA, paraventricular nucleus of the thalamus, contralateral mPFC) onto pyramidal neurons (ACNP 60th Annual Meeting, 2021). Further, Dyn/KOR signaling differentially regulates BLA inputs onto mPFC interneurons (ACNP 60th Annual Meeting, 2021). Specifically, direct excitation of SST-INs by BLA inputs to the mPFC was inhibited by Dyn, an effect that was absent at BLA synapses innervating mPFC PV interneurons. These results suggest that Dyn/KOR signaling can filter excitatory inputs onto mPFC interneurons in a synapse-specific manner and reduce SST-IN-mediated feedforward inhibition of pyramidal neurons. Moreover, by inhibiting local GABA release from KOR-expressing SST- and PV-IN terminals, Dyn potently inhibited polysynaptic inhibition driven by incoming glutamatergic inputs, regardless of whether the inputs themselves express KORs. Together, these findings demonstrate that Dyn/KOR signaling is poised to directly suppress KOR-expressing excitatory inputs while concurrently amplifying mPFC engagement by KOR-lacking inputs via disinhibition (ACNP 60th Annual Meeting, 2021). This complex synaptic modulation by Dyn appears to have implications for behavior under threatening environmental conditions, as evidenced by in vivo Dyn release within mPFC in response to environmental threat, and impaired toggling between active and passive defense strategies following PDyn knockdown in the mPFC (ACNP 60th Annual Meeting, 2021).

Experience-induced interneuron plasticity

By engaging mechanisms of activity- and modulation-induced neural plasticity, salient experiences can trigger robust and persistent synaptic changes in mPFC. Although experience-induced plastic changes in synaptic physiology of mPFC pyramidal cells have been the subject of significant research, we are just beginning to appreciate how salient events and environmental factors shape synaptic function in mPFC interneurons (e.g., Canetta et al., 2016; Skorput and Yeh, 2016; Slaker et al., 2018).

Recent findings reveal that stressful experiences readily facilitate synaptic adaptations in mPFC interneurons. In separate studies, restraint stress (Joffe et al., 2022) and footshock conditioning (Cummings and Clem, 2020; ACNP 60th Annual Meeting, 2021) were shown to each enhance Ca²⁺ responses within SST-INs and potentiate excitatory transmission from the BLA onto mPFC SST-INs (Fig. 2). The potentiation induced by these aversive experiences appears selective to SST-INs, as excitatory transmission in PV-INs was unaltered by restraint stress (Joffe et al., 2022) and seemingly reduced following footshock conditioning (Perova et al., 2015; Cummings and Clem, 2020) (Fig. 2). Initial efforts to parse the mechanisms mediating this stress-induced potentiation of excitatory drive onto SST-INs have implicated postsynaptic mGlu₅ metabotropic glutamate receptor signaling. Indeed, mice lacking mGlu₅ receptors in SST-INs showed no mGlu_1/5 agonist-induced LTP, no stress-induced increases in excitatory drive onto SST-INs (or corresponding increases in pyramidal cell inhibition), resilience to stress-induced deficits in spatial working memory task performance, and impaired cue-associated fear learning (Joffe et al., 2022).

Although stressful experiences can alter both synaptic transmission in mPFC inhibitory microcircuits and behavior, whether these synaptic alterations promote the encoding of experience-induced learning and behavioral adaptations is less clear. Early support for this more causal role comes from Cummings and Clem (2020), who showed that excitatory drive onto SST-INs was potentiated in mice that formed associative fear memories through paired footshock-tone presentations, but not in mice that received unpaired footshocks and tones. These data provide compelling evidence that persistent synaptic changes in mPFC SST-INs are not an unavoidable consequence of a stressful experience; rather, interneuron plasticity appears to instruct the formation of persistent memories (i.e., CS-US association) and future behaviors (i.e., conditioned freezing).

Further evidence for a causal link between experience-induced synaptic and behavioral adaptations comes from studies of PV-IN plasticity following stress and drug exposure. Perova et al. (2015) showed that male mice displaying a phenotype of “helplessness” following repeated footshocks (i.e., fewer escapes and longer escape latencies) showed reduced excitatory synaptic strength in mPFC PV-INs. In contrast, PV-IN synaptic strength was unaltered in “resilient” mice, despite undergoing an identical shock experience. A similar link was established by Ferranti et al. (2022) in a study of the neural adaptations encoding alcohol reward. By manipulating the timing of the same intoxicating dose of alcohol, the researchers conditioned mice to express either a place preference or aversion to the drug. Despite both groups of mice receiving identical alcohol exposure, only those that formed a rewarding alcohol association exhibited enhanced excitatory drive in PV-INs. Thus, bidirectional synaptic adaptations in mPFC interneurons may help encode specific behavioral adaptations to salient experience, rather than simply reflect a history of such experience.

Conclusions

In conclusion, we have highlighted some recent advances in our understanding of the rich diversity of mPFC interneuron subpopulations, the neural pathways they are embedded within and regulate, and the dynamic changes they undergo to remodel mPFC computations and distal network interactions to shape cognitive functions. That these interneurons and their connections are engaged and dysregulated by stress and psychoactive drug exposure, fear learning, motivational conflict, and disease-relevant genetic insults suggests their privileged contributions to various forms of disordered cognition. Importantly, while this mini-review did not segregate findings based on mPFC subregions, considerable regional variations in the anatomical, molecular, and functional properties of mPFC interneurons and their embedded circuits further complicate the study of interneuron contributions to typical and disordered cognition (Heidbreder and Groenewegen, 2003; Euston et al., 2012; Laubach et al., 2018; Anastasiades and Carter, 2021). Through continued innovation and democratization of transcriptomic sequencing, elaboration of intersectional genetic strategies, and expansion of the toolset to tag populations based on dynamic cellular processes, the next decade of neuroscience research will see remarkable advances in the parsing of these important cells into increasingly refined subclasses (Bugeon et al., 2022; Zeng, 2022). By pairing these advances with more sophisticated tools to monitor and manipulate interneuron populations, we stand to build a detailed guide to their interconnections, plasticity, and behavioral contributions. As tools emerge that enable synapse-specific manipulations (e.g., altered connectivity between defined presynaptic and postsynaptic elements) (Ransey et al., 2021; Prakash et al., 2022), we will similarly need innovation in behavioral analyses sensitive to the subtle effects these targeted manipulations may yield. Last, alongside efforts to identify and target molecules and synaptic processes unique to select cell types and circuits in the rodent brain, we must advance similar efforts in nonhuman primates with due consideration of the anatomical, functional, and behavioral homology across species.