ABSTRACT
Background The nucleus accumbens (NAc) controls multiple facets of impulsivity, but is a heterogeneous brain region with diverse microcircuitry. Prior literature links impulsive behavior in rodents to gamma aminobutyric acid (GABA) signaling in the NAc. Here, we studied the regulation of impulsive behavior by fast-spiking interneurons (FSIs), a strong source of GABAergic synaptic inhibition in the NAc.
Methods Male and female transgenic mice expressing Cre recombinase in FSIs allowed us identify these sparsely distributed cells in the NAc. We used the 5-choice serial reaction time (5-CSRT) task to measure both impulsive action and sustained attention. During the 5-CSRT task, we monitored FSI activity with fiber photometry calcium imaging, and manipulated FSI activity with chemogenetics. We used electrophysiology, optogenetics, and fluorescent in situ hybridization to confirm these methods were robust and specific to FSIs.
Results In mice performing the 5-CSRT task, NAc FSIs showed sustained activity on trials ending with correct responses. In contrast, FSI activity declined over time on trials ending with premature responses, as well as trials in which responses were omitted. Chemogenetic inhibition of NAc FSIs did not change response latencies or general locomotor activity, but significantly increased the number of premature responses, while the number of omission trials either remained constant or decreased.
Conclusions These experiments provide strong evidence that NAc FSIs constrain impulsive actions, most likely by GABA-mediated synaptic inhibition of medium spiny neuron output. Our findings may provide insight into the pathophysiology of impulse control disorders, and inform the development of circuit-based therapeutic interventions.
INTRODUCTION
Impulsivity is defined as the tendency to act prematurely or without foresight, and involves a failure of cognitive control that characterizes various neuropsychiatric diseases (1, 2). Notably, impulsivity represents a vulnerability marker for substance use disorders (3), attention deficit/hyperactivity disorder (ADHD) (4, 5), and suicidality (6). The nucleus accumbens (NAc), a major hub in limbic circuits that regulate reward-seeking behavior, plays a critical role in controlling multiple facets of impulsivity (7). Evidence for this role comes from human brain imaging studies (8–11), as well as local pharmacological manipulations and brain lesion studies in rodents (12–17). These approaches have identified key neuromodulatory systems and provided critical insight into the aggregate function of the NAc, but we still have a limited understanding of how impulsive behavior is regulated by the activity of specific cell types embedded within NAc microcircuitry.
Gamma aminobutyric acid (GABA) has been implicated as a modulator of impulsivity in the NAc (18, 19) and other brain regions (20). A major source of local GABAergic signaling in these brain regions are fast-spiking interneurons (FSIs), a cell type frequently defined by parvalbumin (PV) expression (21). FSIs provide robust feed-forward inhibition that constrains and sculpts the output of projection neurons in the NAc (22–24), dorsal striatum (25–27), and other brain regions (28–30). FSIs in the prefrontal cortex regulate complex cognitive processes like attention (31), while FSIs in the dorsal striatum have been implicated in procedural learning (32) and habit formation (33). Manipulations of FSIs in the NAc alter behavioral responses to addictive drugs (34, 35), but the specific contribution of these cells to impulsivity has yet to be explored.
In this study, we investigated the function of NAc FSIs in mice performing the 5-choice serial reaction time (5-CSRT) task, a classic behavioral assay that measures sustained attention and impulse control (36). We focused on FSIs in the NAc core subregion, where GABAergic signaling has specifically been linked to impulsive behavior in this task (18). To monitor the activity of FSIs in behaving mice, we used a viral approach to express a genetically-encoded calcium indicator, and monitored fluorescent signals using in vivo fiber photometry. These calcium imaging experiments reveal that sustained activity of FSIs is associated with successful control of impulsive action. We then used chemogenetic methods to inhibit the activity of FSIs in the NAc core, and found this manipulation increased impulsivity. Together, our data suggest that FSIs in the NAc play an important role in constraining impulsive actions.
METHODS AND MATERIALS
Subjects
Male and female mice C57/B6J mice were housed with same-sex littermates on a 12 hr light/dark cycle, and used for experiments at 8-20 weeks of age. PV-2A-Cre transgenic mice (37) were obtained from The Jackson Laboratory (JAX stock #012358). All procedures conformed to the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals, and were approved by the University of Minnesota Institutional Animal Care and Use Committee
5-Choice Serial Reaction Time (5-CSRT) Task
As previously described (38), mice were food-restricted to ~85% of free-feeding body weight, and preexposed to 14 mg purified dustless precision pellets (Bio-Serv) in the home cage. Behavioral training took place in standard mouse operant chambers (Med Associates), and began with two consecutive days of magazine training (30 food pellets delivered randomly over 30 minutes). We then began the first of eight training stages (Table 1), using a protocol adapted from previous studies of rats (39) and mice (40, 41). Each trial was signaled by the offset of a house light, followed by illumination of cue lights located inside five nose-poke apertures on the wall across from the food magazine. In Stage 1 of training, cues were immediately and continuously presented in all five apertures, and a response in any aperture was rewarded. In Stage 2 of training, a cue light was immediately and continuously presented in only one aperture. Correct responses in this aperture were rewarded, while incorrect responses in any other aperture were punished with illumination of the house light during a time-out period (five seconds). Mice moved on to the next stage after receiving 50 rewards before the end of each 60-minute session.
In Stage 3 and subsequent stages, each trial began with an inter-trial interval (ITI) prior to cue presentation. Mice were required to withhold responses during the ITI, and premature responses during this period were punished with a time-out. The duration of cue presentation was also limited, and failures to respond (“omissions”) were punished with a time-out. Correct responses to the cued aperture were rewarded, and additional nose-pokes that followed a correct response but preceded retrieval of the reward were recorded as perseverative responses. Mice progressed to the next stage after reaching 50% accuracy, defined as the number of correct trials over the total number of all trial types. At various stages of training in different experiments, we also conducted tests of impulsivity and attention, in which either the ITI length or cue duration varied randomly from trial to trial (2.5-20 sec) within a single session.
Plasmids and Adeno-Associated Viral Vectors
pAAV-Syn-FLEX-rc[ChrimsonR-tdTomato] was a gift from Edward Boyden (Addgene plasmid #62723). pGP-AAV-CAG-FLEX-jGCaMP7s-WPRE was a gift from Douglas Kim (Addgene plasmid #104495). A Cre-inactivated (DO) version (42) of jGCaMP7s with a CaMKII promoter was cloned University of Minnesota Viral Vector and Cloning Core, which also packaged ChrimsonR and jGCaMP7s constructs in the AAVdj serotype (43). Other viral vectors were obtained from the Stanford University Gene Vector and Virus Core (AAVdj-EF1a-FLEX-eYFP) and Addgene (AAV8-hSyn-FLEX-hM4Di-mCherry; viral prep #44362-AAV8). All viral vectors were used at a concentration of 5 × 1012 particles/mL.
Stereotaxic Surgery
Intracranial virus injection and fiber-optic implantation were performed as previously described (38), with the following modifications. Mice were anesthetized with a ketamine:xylazine cocktail (100:10 mg/kg), and a small hole was drilled above target coordinates for the NAc core (AP +1.50, ML ±1.10; DV −4.40). A 33-gauge Hamilton syringe containing viral solution was slowly lowered to these target coordinates, followed by injection of 0.75 μL at a rate of 0.1 μL/min. The syringe tip was left in place at the injection site for 5 minutes, and then slowly retracted over the course of five minutes. For fiber photometry experiments, a 400 μm fiberoptic cannula (Doric Lenses: MFC_400/430-0.48_6mm_MF1.25_FLT) was implanted just above the site of virus injection (+0.02 mm), and fixed to the scull using a dual-cure resin (Patterson Dental, Inc.). After surgery, mice were given 500 μL saline and 5 mg/kg carprofen (s.c.) daily for three days, and recovered a minimum of 7 days before any behavioral training.
Fluorescent in situ Hybridization
PV-2A-Cre mice expressing AAVdj-hSyn-FLEX-ChrimsonR-tdTomato in the NAc were anesthetized with isoflurane and decapitated. Brains were rapidly frozen on dry ice in OCT embedding medium, and coronal sections (10 μm) were collected on positively charged glass slides and stored at −80°C until processing. Fluorescent in situ hybridization was performed using the RNAscope Multiplex Fluorescent Assay (Advanced Cell Diagnostics) according to manufacturer protocol. Briefly, the day before running the assay, slides were postfixed in 4% PFA for 15 min at 4°C, followed by dehydration in increasing concentrations of ethanol (50%>70%>100%) for 5 min each at room temperature (RT). After the final 100% ethanol wash, slides were stored at −20°C in 100% ethanol overnight. The following day, slides were left to air dry and a hydrophobic barrier was painted around sections to be analyzed. Protease IV was applied to sections of interest for 30 min at RT, followed by two brief washes in 1x PBS. Hybridization probes for Pthlh (456521), Pvalb (421931-C2), and tdTomato (317041-C3) were applied to sections and incubated at 40°C for 2 hours, then washed twice in 1x RNAscope buffer for 2 minutes. Amplification was performed using AMP1-4 detection reagents, which were applied sequentially and incubated for 15-30 min each at 40°C, washing with 1X RNAscope wash buffer between each reagent. After the final wash, slides were shaken dry and mounting medium (Invitrogen Prolong Gold antifade reagent with DAPI) was applied before coverslips were placed on the slides. Images were taken using a Keyence BZX fluorescence microscope under a 40X objective.
Electrophysiology
PV-2A-Cre mice expressing AAV8-hSyn-FLEX-hM4Di-mCherry in the NAc were anesthetized with isoflurane and decapitated. Brains were quickly removed and placed in ice-cold cutting solution containing (in mM): 228 sucrose, 26 NaHCO3, 11 glucose, 2.5 KCl, 1 NaH2PO4-H2O, 7 MgSO4-7H2O, 0.5 CaCl2-2H2O. Coronal slices (240 μm thick) containing the NAc were collected using a vibratome (Leica VT1000S) and allowed to recover in a submerged holding chamber with artificial cerebrospinal fluid (aCSF) containing (in mM): 119 NaCl, 26.2 NaHCO3, 2.5 KCl, 1 NaH2PO4-H2O, 11 glucose, 1.3 MgSO4-7H2O, 2.5 CaCl2-2H2O. Slices recovered in warm ACSF (33°C) for 10-15 minutes and then equilibrated to room temperature for at least one hour before use. Slices were transferred to a submerged recording chamber and continuously perfused with aCSF at a rate of 2 mL/min at room temperature. All solutions were continuously oxygenated (95% O2/5% CO2).
Whole-cell current-clamp recordings from FSIs or MSNs were obtained under visual control using IR-DIC optics. FSIs were distinguished from putative MSNs by the presence or absence of mCherry fluorescence, respectively, using an Olympus BX51W1 microscope. Borosilicate glass electrodes (3-5 MΩ) were filled with (in mM): 120 K-Gluconate, 20 KCl, 10 HEPES, 0.2 EGTA, 2 MgCl2, 4 ATP-Mg, 0.3 GTP-Na (pH 7.2-7.3). Cells were injected with a series of current steps (1 s duration) from −100 to +250 pA. For chemogenetic experiments, maximum firing rate and resting membrane potential were analyzed for each cell, before and after bath application of clozapine-N-oxide (CNO, 3 μM). Membrane potential measurements were corrected for a liquid junction potential of ~8 mV. Maximum firing rate was calculated as the average maximum firing rate over the 1 s step that could be sustained without inducing a depolarization block. Action potential (AP) half-width was measured at the half-way point between threshold and the AP peak. Recordings were performed using a MultiClamp 700B (Molecular Devices), filtered at 2 kHz, and digitized at 10 kHz. Data acquisition and analysis were performed online using Axograph software. Series resistance was monitored continuously and experiments were discarded if resistance changed by >20%.
Fiber Photometry
For fiber photometry experiments, separate groups of PV-2A-Cre mice received unilateral injection of the following virus combinations into the NAc:
AAVdj-EF1a-FLEX-eYFP (negative control group, n = 3: 1 male, 2 female)
AAVdj-hSyn-FLEX-ChrimsonR-tdTomato & AAVdj-CAG-FLEX-jGCaMP7s (n = 7: 4 male, 3 female)
AAVdj-hSyn-FLEX-ChrimsonR-tdTomato & AAVdj-CaMKII-DO-jGCaMP7s (n = 4: 2 male, 2 female)
Fiber photometry recordings were conducted a minimum of two weeks after surgery, to allow for sufficient viral expression. Recordings from mice in the 5-CSRT task were conducted after a single session of Stage 8 training. Real-time data were acquired using a RZ5P fiber photometry workstation (Tucker Davis Technologies). As previously described (44), 470 nm and 405 nm LEDs (ThorLabs, M470F3 and M405F1) were modulated at distinct carrier frequencies (531 Hz and 211 Hz, respectively), and passed through a fluorescence mini cube (Doric Lenses) coupled into a patch cord (400 μm, 0.48 NA). For experiments involving red-shifted optogenetic stimulation, a 565nm LED (ThorLabs, M595F2) was connected to the fluorescence mini cube and coupled to the same patch cord, and controlled using a pulse generator (Master-8, A.M.P.I.) The distal end of the patch cord was connected to the implanted fiberoptic cannula with a ceramic sleeve. Fluorescence was back-projected through the mini cube and focused onto a photoreceiver (Newport Model 2151, FC adapter). Signals were sampled at 6.1 kHz, demodulated in real-time, and saved for offline analysis. For the 5-CSRT task, behaviorally relevant events (including trial onset, cue onset/offset, nose poke, and magazine entry) were transmitted to the fiber photometry workstation as TTL signals from the operant chambers.
The change in fluorescent signal (dF/F) was calculated within each session as described (44). Each channel was low-passed filtered (<2Hz) and a linear least-squared model fit the isosbestic control signal (405 nm) to the calcium-dependent signal (470 nm). Change in fluorescence was calculated as ([470nm signal - fitted 405nm signal]/[fitted 405nm signal]). In most experiments, we also subtracted the eighth percentile value of dF/F over a rolling 20 second window (45), which effectively corrected for photobleaching over time. However, this correction was not applied in fiber photometry experiments validating chemogenetic inhibition of FSIs (Supplementary Figure S3), in order to detect slow changes in the fluorescent signal following CNO injection. For event-related analyses, normalized dF/F was computed using the mean value of a two-second window before trial onset as baseline.
Chemogenetic Manipulations of Behavior
PV-2A-Cre mice (n = 14: 10 male, 4 female) and wild-type littermates (n = 14: 7 male, 7 female) received bilateral injection of AAV8-hSyn-FLEX-hM4Di-mCherry into the NAc. After recovery, these mice were trained on the 5-CSRT through Stage 8, which provided at least four weeks of viral expression. After one day of training in Stage 8, mice received i.p. injections of either saline or 2 mg/kg CNO (Hello Bio) 30 minutes prior to testing on Stage 8. Saline and CNO sessions were conducted on separate days in counterbalanced order. After these tests, mice completed attention and impulsivity tests, again preceded by counterbalanced i.p. injections of saline or CNO (2 mg/kg).
We tested open-field locomotor activity in this same group of animals, using an arena (ENV-510, Med Associates) housed within a sound-attenuating chamber. The location of the mouse within the arena was tracked in two dimensions by arrays of infrared beams connected to a computer running Activity Monitor software (Med Associates). All mice were acclimated to the arena for one session on the day prior to testing, and then received i. p. injection of saline or CNO (2 mg/kg) on separate days in counterbalanced order, 30 minutes prior to a 60 minute test session.
Immunohistochemistry
To confirm viral injection location, mice were deeply anaesthetized using Beuthanasia (200 mg/kg, i.p.) and transcardially perfused with PBS followed by ice-cold 4% paraformaldehyde in PBS. Brains were fixed overnight in 4% paraformaldehyde in PBS then sliced at 40 μm thickness using a vibratome (Leica VT1000S). Free-floating coronal sections containing the NAc were incubated for 3hrs in blocking solution containing 0.2% Triton-X, 2% normal horse serum, and 0.05% Tween20. Sections were then incubated for 72 hrs with primary antibodies (mouse anti-mCherry 1:1000, Abcam; mouse anti-eYFP 1:1000, Abcam; rabbit anti-RFP 1:1000, Rockland), washed five times, and incubated overnight with secondary antibodies (donkey anti-rabbit Alexa Fluor 647 IgG 1:1000, Life Technologies; goat anti-mouse Alexa Fluor 488 or 647 IgG 1:1000, Abcam). Slides were washed and counter-stained for 20 minutes with 4',6-diamidino-2-phenylindole (DAPI) (1:50,000, Life Technologies), and mounted. Fluorescent images were acquired on a Keyence microscope using a 20x objective and viewed using ImageJ software.
Statistics
Analysis of variance (ANOVA) was conducted in IBM SPSS Statistics v24, using either one-way or factorial models as appropriate, and with repeated measures on within-subject factors. For main effects or interactions involving repeated measures, the Huynh-Feldt correction was applied to control for potential violations of the sphericity assumption. This correction reduces the degrees of freedom, resulting in non-integer values. No significant main effects or interactions involving sex were observed, so data from male and female mice have been pooled for presentation. The Type I error rate was set to α = 0.05 (two-tailed) for all comparisons. Effect sizes for significant results are expressed as partial eta-squared (ŋp2) values. All data are displayed as mean +/- SEM.
RESULTS
Behavioral Measures of Impulsivity and Attention in Mice
The 5-CSRT task requires mice to withhold a nose-poke response until it can be directed to one of five locations indicated by a brief visual cue (FIGURE 1A). Correct responses to the illuminated nose-poke aperture were rewarded with a food pellet, while all other trial outcomes were punished with a time-out period. These other outcomes included premature responses during the inter-trial interval (ITI) before cue presentation, incorrect responses to the wrong location, and omission trials in which the mouse fails to make a timely response. Training consisted of eight sequential stages of increasing ITI and decreasing cue durations (Table 1). To examine behavioral performance under varying degrees of task difficulty, we systematically varied the duration of either the visual cue (to test attention) or the ITI (to test impulsivity), as previously described (41). In a pilot study, we conducted these attention and impulsivity tests both early and late in training (FIGURE 1B). In the attention test, omission responses increased monotonically as the cue duration decreased (F1.46,10.24 = 13.11, p = 0.002, ŋp2 = 0.65), with little effect of training (F1,7 < 1; Figure 1C). In the impulsivity test, premature responses increased monotonically as the ITI grew longer (F3,21 = 67.62, p < 0.001, ŋp2 = 0.91), while the overall number of premature responses decreased with training (F1,7 = 73.40, p < 0.001, ŋp2 = 0.91;Figure 1D). We used this training protocol in subsequent experiments to investigate how FSIs in the NAc contribute to 5-CSRT task performance.
Characterization of FSIs in Mouse Nucleus Accumbens Core
To specifically target FSIs in the NAc core, we used a mouse line expressing Cre recombinase in a bicistronic fashion from the Pvalb locus (PV-2A-Cre). In this mouse line, Cre is active in cells with both high and low Pvalb expression (37), providing robust labeling of FSIs in both the dorsal striatum (32, 33) and NAc shell (22, 34). To characterize the properties of cells expressing Cre in the NAc core, we stereotaxically injected an adeno-associated virus (AAV) expressing red fluorescent protein in a Cre-dependent fashion (FIGURE 2A). We first carried out fluorescent in situ hybridization to co-label virally-expressed tdTomato with mRNA transcripts for parvalbumin (Pvalb) and parathyroid hormone like hormone (Pthlh), both of which are highly expressed in striatal FSIs (46, 47) (FIGURE 2B). The majority of virally-labelled (tdTomato+) cells expressed Pvalb and Pthlh, although the level of expression varied from cell to cell (FIGURE 2C). We occasionally observed virally-labelled (tdTomato+) cells with no detectable Pvalb or Pthlh expression, as previously reported in this mouse line (34, 48). To confirm that these cells were indeed FSIs, we performed whole-cell patch-clamp recordings from virally-labelled cells in acute NAc brain slices. In current-clamp recordings, these cells showed characteristic properties of FSIs (FIGURE 2D), including a high maximum firing rate, narrow action potential half-width, and short-duration afterhyperpolarization (49). These electrophysiological properties could be easily distinguished from those of medium spiny projection neurons (MSNs), the most abundant type of neuron in the NAc (FIGURE 2E). These data confirm that our viral targeting approach specifically labels FSIs in the NAc core.
Optogenetically-Evoked Calcium Signal in FSIs and MSNs
Our next objective was to characterize endogenous patterns of FSI activity in the NAc of freely behaving mice. Due to the limited number and sparse distribution of FSIs in the NAc, we used fiber photometry to collect bulk calcium signal from FSI ensembles within a large tissue volume (50, 51). PV-2A-Cre mice were stereotaxically co-injected with two AAVs expressing Cre-dependent constructs: the genetically-encoded calcium indicator jGCaMP7s (52), and the red-shifted excitatory opsin ChrimsonR (53). A chronic fiber-optic implant was placed in the NAc just above the site of virus injection, and was used to both deliver excitation light and collect emitted fluorescence (FIGURE 3A). Immunohistochemical examination of infected brain slices revealed sparse and overlapping viral co-expression, as well as proximal placement of fiber optic implant (FIGURE 3B).
To confirm that reliable fiber photometry signals could be recorded from sparsely distributed FSIs in the NAc, we tethered freely moving mice to the fiber photometry system and delivered amber light (594 nm) through the fiber-optic implant to activate ChrimsonR (FIGURE 3C). We detected robust jGCaMP7s signals that varied in a dose-dependent fashion with stimulation frequency (F1.61,9.64 = 8.13, p = 0.011, ŋp2 = 0.58; Figure 3D), pulse number (F2.72,16.35 = 5.72, p = 0.008, ŋp2 = 0.49; Figure 3E), and light power (F1.82,10.89 = 12.84, p = 0.002; ŋp2 = 0.68; Figure 3F). In a separate group of mice, we also used fiber photometry to evaluate the effect of FSI activation on neighboring MSNs. PV-2A-Cre mice were co-injected with AAVs encoding Cre-dependent ChrimsonR (expressed in FSIs), as well as Cre-inactivated jGCaMP7s with a CaMKII promoter to drive expression in MSNs (46) (Figure 3G,H). This strategy allowed for red-shifted optogenetic excitation of FSIs with concurrent fiber photometry monitoring of MSN responses. In mice receiving amber light stimulation at 20 Hz, we observed a significant reduction in calcium signal (F1,3 = 15.00, p = 0.030, ŋp2 = 0.83; Figure 3I), suggesting an inhibitory action of FSIs onto neighboring MSNs within the NAc core.
Fiber Photometry Monitoring of FSIs during the 5-CSRT Task
Having validated the specificity and reliability of the fiber photometry signal from FSIs in the NAc, we proceeded to record this signal while mice performed the 5-CSRT task. Experimental mice included those with validated jGCaMP7s expression from the preceding optogenetic experiments (see Figure 3), as well as a control cohort of PV-2A-Cre mice infected with a Cre-dependent eYFP fluorophore. Upon reaching the final stage of testing, we collected fiber photometry data across five consecutive sessions, with simultaneous excitation of jGCaMP7s at a calcium-dependent wavelength (470 nm), as well as an isosbestic wavelength (405 nm) used to correct for bleaching and movement artifact (FIGURE 4A). Behavioral responses in the 5-CSRT task were time-locked to the fiber photometry signal using TTL outputs from the operant chamber control software (FIGURE 4B).
We first examined fiber photometry signals aligned to trial onset, which coincided with the beginning of the ITI period (Figure 4C, left). FSI activity was sustained throughout the ITI on correct trials, whereas it declined throughout the ITI on both omission and premature trials. These response profiles were absent in control mice expressing the eYFP fluorophore (Figure 4C, right), indicating that motion or other artifacts did not contribute to jGCaMP7s signal. Calcium signal averaged within experimental animal indicated a significant difference in FSI activity following the onset of correct versus premature trials in jGCaMP7s-expressing mice (F1,5 = 16.18, p = 0.010, ŋp2 = 0.76), but the difference between correct and omission trials was not statistically significant (F1,5 = 2.84, p = 0.15, ŋp2 = 0.36; Figure 4D). In mice expressing jGCaMP7s, elevated FSI activity on correct trials was sustained until cue presentation (Figure 4E, left). Signal alignment to nose poke response showed elevated FSI activity before correct versus premature responses (Figure 4E, middle). There was no evidence that this signal varied as a function of spatial location relative to implanted hemisphere (Supplementary Figure S1). Lastly, signal alignment to magazine entry after correct responses revealed an increase in FSI activity associated with food pellet retrieval (Figure 4E, right). After averaging calcium signals within animal, we observed significant increases in FSI activity before cue presentation on correct versus omission trials (F1,5 = 7.46, p = 0.041, ŋp2 = 0.60), as well as before the nose poke response on correct versus premature trials (F1,5 = 7.68, p = 0.039, ŋp2 = 0.61; Figure 4F). Incorrect responses were not analyzed, as they represented a negligible proportion of all trials (1.6 and 1.8% from jGCaMP7s and eYFP mice, respectively). Altogether, these data suggest that FSI activity shapes the selection and timing of behavioral responses in the 5-CSRT task, and reductions of FSI activity predict poor behavioral performance in the form of omissions or premature responses.
Chemogenetic Inhibition of FSIs Increases Impulsivity
Given the relationship between FSI activity and behavioral performance in the 5-CSRT task, we sought to directly manipulate FSI activity levels in vivo using a chemogenetic strategy, and determine the behavioral consequences. Prior to the start of 5-CSRT task training, PV-2A-Cre mice and wild-type littermates received bilateral injections of an AAV expressing Cre-dependent hM4Di in the NAc core (FIGURE 5A). This Gi-DREADD (Designer Receptor Exclusively Activated by Designer Drugs) can be activated by clozapine-N-oxide (CNO), initiating intracellular signaling cascades that reduce neuronal excitability (54). To validate the effect of this manipulation on FSIs in the NAc core, we performed current-clamp recordings from virally-labeled FSIs in acute brain slices, and measured electrophysiological properties before and after bath application of CNO (3 μM; Figure 5B). CNO caused a significant reduction in maximum firing rate (F1,14 = 5.73, p = 0.031, ŋp2 = 0.29), as well as hyperpolarization of the resting membrane potential (F1,14 = 5.09, p = 0.041, ŋp2 = 0.27; Figure 5C, Supplementary Figure S2). These effects in acute brain slices were further corroborated using fiber photometry to verify chemogenetic inhibition of FSIs in vivo (Supplementary Figure S3).
After bilateral injection of AAV-FLEX-hM4Di virus into the NAc core, PV-2A-Cre mice and wild-type littermates were trained on the 5-CSRT task. Upon reaching the last stage of training, we injected either saline or CNO (2 mg/kg) thirty minutes before testing on separate days. In PV-2A-Cre mice, injection of CNO caused a significant increase in the percentage of premature trials, relative to saline (F1,13 = 9.15, p = 0.010, ŋp2 = 0.41; Figure 5D, left). This effect was not observed in wild-type littermates (F1,13 < 1; Figure 5D, right), confirming that the effects of CNO were only observed following viral expression of hM4Di in FSIs. This elevation of premature responding was also robust when comparing difference scores (CNO-saline) between genotypes (F1,26 = 7.08, p = 0.013, ŋp2 = 0.21; Figure 5E), but there were no significant changes in the percentage of other trials types. There was also no change in the number of perseverative responses following injection of CNO versus saline in either genotype (PV-2A-Cre: 1.71±1.76; wild-type: 2.86±1.13; F1,26 < 1). Importantly, chemogenetic modulation of FSIs did not affect latencies to respond or retrieve reward (F1,26 < 1; Figure 5F), nor did it alter locomotor activity in an open field (Supplementary Figure S4).
After determining the effect of chemogenetic FSI inhibition in the final training stage, this same cohort of mice was run through the attention and impulsivity tests described in Figure 1, with injection of saline or CNO (2 mg/kg) thirty minutes before each test session. In the attention test, there was an increase in correct trials following CNO injection in PV-2A-Cre mice (F1,13 = 5.68, p = 0.033, ŋp2 = 0.30; Figure 6A). Although there was a numerical trend for these mice to have fewer omissions after CNO (F1,13 = 4.04, p = 0.066), this was not significantly different from wild-type mice (genotype: F1,13 = 2.71, p = 0.11; Figure 6B; Supplementary Figure S5). Strikingly, in the impulsivity test, administration of CNO increased the percentage of premature trials in PV-2A-Cre mice when the ITI was long (Treatment x ITI interaction: F2.16,28.10 = 9.76, p < 0.001, ŋp2 = 0.43; Figure 6C). These effects of CNO were dependent on hM4Di expression in NAc FSIs (Genotype x ITI interaction: F2.34,60.89 = 3.58, p = 0.028, ŋp2 = 0.12; Figure 6D), as there were no behavioral differences between CNO and saline injection in the wild-type littermate control group (Supplementary Figure S6). Collectively, these data suggest FSIs in the NAc core normally constrain premature responses in the 5-CSRT task, and that disruption of FSI function can lead to impulsive behavior.
DISCUSSION
Decades of research clearly implicate the NAc in controlling multiple facets of impulsivity, but the contribution of specific cell types within this heterogeneous brain region has remained unclear. In this study, we focused on the contribution of FSIs, which are few in number but exert powerful inhibitory control over MSN output from the NAc (22–24). We used PV-2A-Cre transgenic mice to monitor and manipulate the activity of FSIs in the NAc core. Our data suggest the sustained activity of these cells positively correlates with successful control of impulsive behavior in the 5-CSRT task. Furthermore, using chemogenetic methods to directly inhibit FSIs in the NAc core, we observed enhanced impulsivity across a range of task conditions. These findings indicate FSIs in the NAc core play a key role in impulse control.
The sparse distribution of FSIs in the NAc has historically made these cells difficult to study. FSIs represent fewer than 1% of all neurons in the dorsal striatum of rodents (55), with immunohistochemical and electrophysiological evidence suggesting FSIs are even less abundant in the NAc (56, 57). Several aspects of our experimental design facilitated analysis of these scarce interneurons, including bulk calcium imaging of NAc tissue with fiber photometry. We sought to maximize the fluorescent signal from FSIs by virally expressing jGCaMP7s, a recently developed and highly sensitive genetically encoded calcium indicator (52). Using red-shifted optogenetic excitation with ChrimsonR, we confirmed the sensitivity of this method for measuring FSI activation in the NAc core. It should be noted that the kinetics of jGCaMP7s activation and bulk nature of fiber photometry cannot resolve specific spiking patterns of individual FSIs, but instead provide an aggregate view of FSI activation.
This fiber photometry approach enabled real-time monitoring of FSI activity in the NAc core as mice performed the 5-CSRT task. On correct trials, FSI activity was sustained during the ITI, then declined after cue presentation and nose poke in the appropriate location. In contrast, FSI activity declined during the ITI on omission and premature trials, and remained low at the time of premature response (premature trials) or cue presentation (omission trials). This finding raised two possibilities: FSI activity prevents premature responses and/or sustains attention prior to cue presentation. To differentiate these possibilities and move beyond correlations between FSI activity and behavior, we used a chemogenetic approach to directly and specifically inhibit the activity of FSIs in the NAc core. This manipulation increased the frequency of premature responses, especially with long ITI periods, strongly linking FSI activity with successful impulse control. Importantly, CNO injection had no effect in wild-type control animals, demonstrating the behavioral effects of this designer drug depended on hM4Di DREADD expression in NAc FSIs.
Compared to effects on impulsivity, chemogenetic inhibition of FSIs in the NAc had no significant effects on omission responses late in training or in the attention test. However, this manipulation improved performance in the impulsivity test when the ITI was short, due to a significant decrease in omissions. This contrasts with the behavioral function of FSIs in the prefrontal cortex, which are also active during the ITI and appear to sustain attention in the 5-CSRT task (31). It is thus striking that the same cell type in different brain regions has seemingly distinct behavioral functions, especially given that connections between the prefrontal cortex and NAc are involved in 5-CSRT task performance (13). This surprising result makes sense if FSIs in the prefrontal cortex inhibit the activity of pyramidal cells that project to the NAc, thereby reducing excitatory drive to FSIs (58–60). The control of attention by FSIs in the prefrontal cortex (31) may also involve regulation of pyramidal cells that project to other brain regions, such as the dorsomedial striatum (61) or the claustrum (62).
Our ability to monitor and manipulate FSIs in the NAc core was facilitated by use of the PV-2A-Cre mouse line, which labels cells with both high and low expression of Pvalb (37). The high penetrance of Cre expression in this mouse line appears to be critical for labeling the majority of FSIs in the NAc (22) and dorsal striatum (32). To confirm this high penetrance did not reduce the specificity of Cre expression in FSIs, we performed electrophysiological characterization of virally-infected cells in acute brain slices. All of the cells we recorded exhibited characteristic properties of FSIs, consistent with another recent study in the NAc (22). However, in situ hybridization revealed variable expression of Pvalb transcript, including a small number of cells that lacked detectable Pvalb expression. Similar findings with this mouse line have previously been reported in both NAc (34) and hippocampus (48), but electrophysiological data nevertheless suggests these cells are FSIs. The level of Pvalb expression in these cells may be below the threshold of detection with in situ hybridization, while still providing a small amount of Cre expression necessary to activate Cre-dependent viral constructs. FSIs with little Pvalb may be more readily labelled by other markers, such as Pthlh (47) or cannabinoid receptor type 1 (63).
FSIs form strong inhibitory synaptic connections onto MSNs in the NAc (22, 24, 63). Using fiber photometry, we found that optogenetic stimulation of FSIs in the NAc core decreased the calcium signal detected from surrounding MSNs, consistent with a net inhibitory influence of NAc FSIs on MSN activity in vivo. The increased impulsivity observed after chemogenetic FSI inhibition is thus likely to involve disinhibition of NAc output from MSNs. We speculate that the amount of GABA released by FSIs onto MSNs may be a key factor controlling impulsivity, since highly impulsive rats have decreased NAc GABA levels (19), but no change in NAc GABA receptor binding (64). Highly impulsive rats also have lower expression of glutamate decarboxylase in the NAc core (18), and this key GABA synthetic enzyme is encoded by two genes (Gad1 and Gad2) expressed at high levels by striatal FSIs (46). Reduced GABA synthesis by FSIs may thus contribute to increased impulsivity observed after local knockdown of glutamate decarboxylase expression in the NAc core (18).
The synaptic and molecular mechanisms by which NAc FSIs regulate impulsivity will be an important topic for future research, since FSI dysfunction is linked to numerous neuropsychiatric disorders (65, 66). Tourette’s syndrome, a disorder associated with loss of motoric inhibitory control, presents with loss of striatal FSIs (67). Experimental disruption of FSIs in the dorsal striatum can also produce abnormal behaviors in animal models, including stereotypies and dyskinesias (68–71). In the NAc, FSI manipulation alter behavioral responses to addictive drugs (34, 35). Thus, the present findings have broad implications for understanding addiction (8, 72) and other neuropsychiatric disorders (2, 11) epitomized by deficits in impulse control and associated dysfunction of limbic neurocircuitry.
DISCLOSURES
The authors declare no competing financial interests.
ACKNOWLEDGEMENTS
Research reported in this publication was supported by the University of Minnesota’s MnDRIVE (Minnesota’s Discovery, Research and Innovation Economy) initiative (to MTP and PER), as well as grants from the National Institutes of Health (F32MH118794 to MTP and R00DA037279 to PER) and the Whitehall Foundation, a NARSAD Young Investigator Grant from the Brain & Behavior Research Foundation, and an MQ Fellows Award (all to PER). Some of the viral vectors used in this study were generated by the University of Minnesota Viral Vector and Cloning Core, as well as the Stanford University Gene Virus and Vector Core. We thank Justin Lines for fiber photometry analysis troubleshooting, Eshaan Iyer for early piloting of the 5-CSRT task, Sowmya Narayan and Sumyuktha Vijay for assistance with behavioral experiments, David Leipold for maintenance of mouse colonies, and Carlee Toddes and Dieter Brandner for helpful discussion of the manuscript.