Basal forebrain parvalbumin neurons modulate vigilant attention and rescue deficits produced by sleep deprivation

Abstract

Attention is impaired in many neuropsychiatric disorders, as well as by sleep disruption, leading to decreased workplace productivity and increased risk of accidents. Thus, understanding the neural substrates is important. Here we test the hypothesis that basal forebrain neurons which contain the calcium-binding protein parvalbumin modulate vigilant attention in mice. Furthermore, we test whether increasing the activity of basal forebrain parvalbumin neurons can rescue the deleterious effects of sleep deprivation on vigilance. A lever release version of the rodent psychomotor vigilance test was used to assess vigilant attention. Brief and continuous low power optogenetic excitation (1 s, 473 nm @ 5 mW) or inhibition (1 s, 530 nm @ 10 mW) of basal forebrain parvalbumin neurons was used to test the effect on attention, as measured by reaction time, under control conditions and following 8 hr of sleep deprivation by gentle handling. Optogenetic excitation of basal forebrain parvalbumin neurons that preceded the cue light signal by 0.5 s improved vigilant attention as indicated by quicker reaction times. By contrast, both sleep deprivation and optogenetic inhibition slowed reaction times. Importantly, basal forebrain parvalbumin excitation rescued the reaction time deficits in sleep deprived mice. Control experiments using a progressive ratio operant task confirmed that optogenetic manipulation of basal forebrain parvalbumin neurons did not alter motivation. These findings reveal for the first time a role for basal forebrain parvalbumin neurons in attention and show that increasing their activity can compensate for disruptive effects of sleep deprivation.

INTRODUCTION

Attention is impaired in many neuropsychiatric disorders (Gilmour et al., 2019) and by sleep disruption (Goel et al., 2009; Lim and Dinges, 2008). Sleep loss or disruption result in an increased homeostatic drive to sleep (i.e. sleepiness) and impairs multiple aspects of cognition, including attention, in both human and non-human subjects (Doran et al., 2001; McCoy & Strecker, 2011). Elevated sleepiness diminishes extradimensional shift discrimination in an attentional set-shifting task (McCoy et al., 2007), increases impulsivity in the Go/No-Go task (Chuah et al., 2006), and increases response latencies and omissions in both the continuous performance test (Gruber et al., 2007) and the 5-choice serial reaction time task (Córdova et al., 2006). Additionally, sleep disruption impairs performance in both the rodent and human psychomotor vigilance task (PVT; Christie et al., 2008b; Doran et al., 2001). Vigilant attention (Robertson & O’Connell, 2010) is altered by the state of arousal and is the measure of cognition that is most sensitive to and reliably impaired by sleep loss (Lim & Dinges, 2008). These effects of sleep disruption on attention have important real-world consequences since they lead to decreased workplace productivity and increased risk of accidents (Goel et al., 2009; Lim and Dinges 2008).

Given the important outcomes of sleep disruption on attention, understanding the underlying neural substrates is important for developing treatments. The basal forebrain (BF) is a brain region which degenerates in dementia (Ray et al., 2018) and is implicated in the negative effects of sleep disruption on vigilance and cognition (Christie et al., 2008a; McCoy & Strecker, 2011). Indeed, the BF comprises a major ascending arousal pathway important in both wakefulness and attention (Aguilar & McNally, 2022; Maness et al., 2022; Sarter & Bruno, 2002). Disturbances of sleep result in elevation of the inhibitory neuromodulator adenosine in the BF, leading to inhibition of multiple subtypes of cortically-projecting BF neurons (McKenna et al., 2007; Porkka-Heiskanen et al., 1997; Yang et al., 2013). While BF cholinergic neurons are no doubt important for attention (Burk et al., 2018; Everitt & Robbins, 1997; Sarter & Lustig, 2019), excitotoxic BF lesions which preferentially affect non-cholinergic neurons have effects on attention which are generally more profound than selective lesions of cholinergic neurons (Burk & Sarter, 2001; McGaughy et al., 2002; Muir et al., 1994). However, which of the many non-cholinergic BF neurons (Yang et al., 2017) corresponds to these attention-related BF neurons has not been determined. Recent work has shown that one population of BF GABAergic neurons containing the calcium-binding protein parvalbumin (PV) controls cortical fast oscillations and arousals from sleep (Hwang et al., 2019; Kim et al., 2015; McKenna et al., 2020; McNally et al., 2021), but their role in awake behavior is unclear. Thus, here we test the hypothesis that BF-PV neurons modulate vigilant attention in mice.

Several different tasks have been used in rodents to examine attention and the effects of sleep loss (McCoy & Strecker, 2011). Vigilant attention is assessed here in food-restricted mice using a lever release version of the rodent PVT (rPVT, Figure 1A) which was adapted from previous studies in rats (Amalric & Koob, 1987; Christie et al., 2008b). The rPVT is a simple signal detection reaction time test that requires monitoring of a stimulus location (lightbulb) for a brief and unpredictable cue signal (light flash), which models the human PVT widely used to detect drowsiness in human vigilance studies (Lim & Dinges, 2008; McCoy & Strecker, 2011). The “lever release rPVT” introduced herein is a self-paced version of the rPVT that assures task engagement by requiring mice to depress a lever to initiate a trial and maintain the lever press until presentation of the cue signal. Importantly, since reaction time is registered as soon as mice release the lever, the measure is not contaminated by movement speed or by locomotion to engage the operandum as in other tests of vigilance in mice (Christie et al. 2008a,b; MacQueen et al., 2018; Young et al., 2009). As illustrated in Figure 1A, mice must hold the lever down for a random delay to then report detection of the cue light signal by releasing the lever within the response window to receive reward. Premature responses are defined as lever releases before cue light onset and provide a measure of impulsivity. Trials where mice fail to report signal detection within the response window are scored as omissions and model “attentional lapses” in human vigilance studies (Lim & Dinges, 2008). This task can be readily learned by mice although it does not have some of the features (e.g. correct rejections) of more sophisticated tasks measuring sustained attention in mice (St. Peters et al., 2011; Young et al., 2009). Using the lever release rPVT, combined with brief low power optogenetic excitation or inhibition, we examined the role of BF-PV neurons in vigilant attention for the first time.

Figure 1. Optogenetic excitation of basal forebrain (BF) parvalbumin (PV) neurons enhances vigilant attention and rescues deficits induced by sleep deprivation (SD).

A: Design of the lever release version of the rodent psychomotor vigilance task (rPVT) used to assess vigilant attention. Food-restricted mice are required to press a lever down to start each trial. Once the lever is depressed a random delay (0.7– 6.0 s) is followed by delivery of a 200 ms cue light signal. Mice report cue signal detection by releasing the lever within 1 s of signal onset (i.e., a 1 s response window) to receive a 20 mg sucrose pellet reward for their correct response. Reaction time, the primary measure of vigilant attention, is the time between cue light onset and lever release. Continuous bilateral optogenetic excitation of BF-PV neurons began 500 ms before cue light onset and continued for 500 ms thereafter. B: Schematic showing the BF target area for optogenetic excitation. C: Optogenetic excitation of BF-PV neurons quickens reaction times (N=6; within-subjects). Sorting correct trial reaction times from fastest to slowest illustrates the consistent pattern of quickened reaction times during the BF-PV excitation session. D: 8 hr of sleep deprivation slows reaction times throughout the session and in a separate session optogenetic excitation of BF-PV neurons in sleep deprived mice returns reaction times to baseline levels, indicating rescued performance. E: Summary of mean session reaction times in the four experimental conditions with individual data points shown as open circles. Group mean data are shown as horizontal bars ± SEM. Asterisk denotes p < 0.05.

METHODS

Subjects and experimental design

Adult (4–6 months) homozygous PV-Cre mice (B6.129P2-Pvalb^tm1(cre)Arbr/J) were purchased from Jackson Laboratory (Stock#017320; Bar Harbor, Maine) and bred in house. Mice were housed at 21°C with a 12-h light/dark cycle (7AM–7PM light phase), and food and water ad libitum until entering behavioral experiment protocols. This study used a within-subjects design with N=6 mice in the BF-PV excitation experiments, N=5 mice in the BF-PV inhibition experiments, and N=3 GFP control mice. A within-subjects design was used because it increases statistical power by reducing variability and allows the use of fewer animals to reach statistical significance (Charness et al., 2012). A power analysis based on preliminary data from a mouse lever release rPVT pilot study performed by our group indicated that sample sizes of at least 5 provide 90% power to detect large effect sizes (d>0.8) using a within-subjects design. Following initial baseline measures, the order of treatments was randomized for each animal mitigating the concern of possible order effects. Although the experimenter was not blind to treatment conditions, the behavioral tests were run in automated operant cages reducing the risk of experimenter bias. All procedures were performed in accordance with the National Institutes of Health guidelines and in compliance with the animal protocol approved by the VA Boston Healthcare System Institutional Animal Care and Use Committee.

Stereotaxic viral vector injections and implantation of optical fibers and local field potential/electromyography electrodes

For optical excitation of BF-PV neurons, we used adeno-associated viral vectors serotype 5 (AAV5) with Cre recombinase-dependent expression of a fusion protein consisting of channelrhodopsin-2 (ChR2) and enhanced yellow fluorescent protein (EYFP) (AAV5-DIO-ChR2-EYFP; 7 × 10¹² viral particles/ml estimated by DotBlot, University of North Carolina Vector Core; Chapel Hill, NC). For inhibition of BF-PV neurons, AAV5-CAG-flex-reverse-ArchT-GFP, bearing the archaerhodopsin (ArchT) strain Halorubrum sp. TP009 (8.5 × 10¹² viral particles/ml estimated by DotBlot), was purchased from the University of North Carolina vector core. Non-opsin fluorophore-only control experiments used Cre-dependent AAV1/2-EGFP vectors (Genedetect, Auckland, NZ) with virion concentration estimated to be 2 × 10¹² particles/ml by DotBlot. The transduction specificity and penetrance of these vectors for targeting of BF-PV neurons was validated in our previous study (McKenna et al., 2020). Viral vectors (500 nL per side) were bilaterally injected into the BF (AP +0.4 mm, ML ± 1.6 mm, DV −5.4 mm) of PV-Cre mice under isoflurane anesthesia (induction, 5%; maintenance, 1% – 2%) using a Hamilton syringe (50 nL/minute) driven by a high-precision injector pump (model 250; KD Scientific). Optical fibers (200 μm, 0.37 NA, Doric Lenses; Quebec City, Quebec, CA) were implanted above BF injection sites (AP +0.4 mm, ML ± 1.6 mm, DV −5.2 mm) and each mouse was instrumented with an insulated stainless-steel local field potential (LFP) electrode in medial prefrontal cortex (mPFC; AP +1.8 mm, ML +0.45 mm, DV −2.0 mm). Bilateral optical fibers targeted intermediate/caudal BF where cortically projecting BF-PV neurons are concentrated (McKenna et al., 2013). Reference and ground electroencephalogram (EEG) screw electrodes were placed into the skull above the cerebellum on opposite sides of the midline. Electromyography (EMG) electrodes were placed in the nuchal muscle. Electrode leads were connected to EEG/EMG headmounts (Pinnacle Technology Inc., part # 8402-SS, KS, USA) and the installation was secured to the skull using dental cement. The scalp incision was sutured closed, and mice were allowed to recover.

Optogenetic Excitation and Inhibition

Bilateral optical illumination was delivered using fiber-coupled 473 nm solid state laser diode (Cat # CL473–050-O; CrystaLaser; Reno, Nevada) or 530 nm laser diode (Cat # CL530–050-O; CrystaLaser; Reno, Nevada) and a patch cord (Doric Lenses) attached to implanted optical fibers with zirconium sleeves. Each sleeve was fitted with black shrink-wrap and the headmount was wrapped with black tape to shield light escape. Software-generated transistor–transistor logic (TTL) pulses (WinWCP; Strathclyde Institute of Pharmacy and Biomedical Sciences, or Spike2; Cambridge Electronic Design) were used to drive optical illumination. After a minimum of four weeks following stereotaxic surgery, mice injected with ChR2-encoding virus were subjected to a 40 Hz optogenetic BF-PV excitation protocol to verify efficacy of the manipulation in vivo (Kim et al., 2015) with 10 ms pulses delivered at 40 Hz (1 s ON 2 s OFF) for 5 min (100 trials). LFP signals from mPFC were amplified and conditioned (16 Channel Amplifier; AM System) and recorded with WinWCP at a sampling rate of 2 kHz (bandpass filtered at 1–200 Hz). Custom MATLAB scripts assessed evoked 40 Hz (±5 Hz) gamma power. ChR2 animals used for behavioral experiments exhibited >2000 μV²/Hz evoked 40 Hz (±5 Hz) gamma power during photoexcitation. For BF-PV optogenetic manipulations during rPVT performance, 1 s of continuous laser illumination (5 mW 473 nm blue light for ChR2-mediated BF-PV excitation; 10 mW 530 nm green light for ArchT-mediated BF-PV inhibition) was delivered on every trial beginning 500 ms prior to cue light onset. Animals injected with non-opsin encoding fluorophore control virus (GFP) received delivery of blue and green laser in separate sessions. Effective optogenetic excitation and inhibition of BF-PV neurons using these vectors was previously confirmed in our laboratory using in vitro electrophysiological recordings (Kim et al., 2015).

Behavioral procedures for rodent psychomotor vigilance task and progressive ratio task

Mice were subjected to one behavioral session per day during the light phase (7am-7pm). During the post-surgical recovery period (at least one week), mice were individually housed and provided with ad libitum access to water and rodent chow. Mice were then food-restricted and fed 2–3 g of chow per day to maintain 90% of their ad libitum weight throughout experiments. To acclimate to the food rewards (20 mg sucrose pellets, TestDiets 5TUT), mice were given 10 pellets/day in their home cages for three days prior to the onset of behavioral training. Both behavioral procedures (rPVT and progressive ratio task (PR)) consisted of transferring mice to software-controlled operant chambers (GraphicState4.0; Coulbourn Instruments LLC, PA, USA) equipped with two retractable levers on either side of a sucrose pellet dispenser and panel lights located directly above each lever. In both experiments, mice were first trained to retrieve rewards from the dispenser through a pseudorandom delivery of pellets (30 pellets over a 60 min session). In the next session, mice were trained to press the active lever to receive a single pellet per press (fixed ratio; FR1) until 30 pellets were earned. The active lever was randomly assigned across mice and the inactive lever was never rewarded. The procedures for the two experiments then diverged.

The lever release version of the rPVT has several advantages over previous rodent vigilant attention tasks. Task designs that successfully isolate and reduce variance of the primary measure of an investigated construct, such as reaction time for vigilant attention, are most useful when examining the neural substrates. The lever release rPVT provides considerably more accurate measures of reaction time, and attentional lapses are not contaminated by movement time or task disengagement as in previous rodent sustained attention tasks (Christie et al., 2008a,b; Young et al. 2009, MacQueen 2018). Requiring mice to press and hold the lever down throughout the fore period assures that the distance between the mouse, the lever, cue light, and the reward port remain relatively fixed. As soon as mice react to the cue light by darting towards the food cup, the lever is released, and reaction time is recorded. Because mice choose to initiate a trial by engaging with the lever, relatively few omissions are seen, and those that are recorded are not due to task disengagement caused by grooming or exploration. Consequently, these omissions more closely resemble the “attentional lapses” described in human vigilance studies (Lim & Dinges, 2008).

In the lever release rPVT experiment, the next six sessions trained mice to press and hold the active lever down for an increasing amount of time (hold duration) to receive pellet rewards (30 pellets per session). The hold duration criterion began at 200 ms for each of the 30 rewards in the first session. In the next session, hold criterion remained at 200 ms for the first five rewards, increased to 260 ms for the next ten rewards, then increased to 300 ms for the last 15 rewards. Incremental training continued in this manner across the next four sessions until a final hold duration of 700 ms was achieved. The next phase of the rPVT experiment trained mice to release the lever in response to brief illumination of the panel light cue (200 ms) located directly above the active lever. First, mice were introduced to the pseudorandom presentation of the light signal after a minimum hold of 500 ms. Then, mice were required to release the lever within 1 s of cue light onset (response window) to receive reward (correct trial) while premature responses (lever release before cue onset) were punished by retraction of both levers for 3 s (time-out). If mice failed to release the lever within 1 s after light onset, the trial was scored as an omission (lapse in attention), no reward was delivered, and the levers remained extended (no time-out).

The time between cue onset and lever release on correct trials is the primary measure of vigilant attention (reaction time (RT)). As described by Amalric and Koob (1987) and Young et al. (2009), performance accuracy was assessed according to the following formula: % correct = # correct responses/(# correct responses + # of omissions). Measures of impulsivity (premature responses and inactive lever presses) were analyzed separately. Training continued until performance with a 0.7–6 s random delay between lever press and cue light presentation was stable across 3 sessions (<10% Δ in mean session RT) and the mean reaction time of these 3 sessions was used as a baseline for each mouse. It typically took 4 to 6 weeks of training for mice to reach the stable baseline criterion. The order of treatments was counterbalanced for each animal to minimize the risk of carryover and practice effects of the within-subjects design (Charness et al., 2012).

In this study, each rPVT session terminated after 30 rewards were received to prevent decreases in motivation from impacting reaction time. Here, rPVT session duration is typically between 10–20 mins. In human studies, PVT sessions as short as 3 min can be used to reveal effects of sleep deprivation on reaction time (Grant et al., 2017) and 10 min PVT sessions are widely used (Lim & Dinges, 2008).

Mice were awake throughout all sessions including those preceded by sleep deprivation. Sleep deprivation was achieved by gentle handling for 8 hr prior to rPVT sessions. Sleep deprivation by gentle handling was conducted using previously validated and widely used procedures such as adding new objects into the cages and a gentle touch of the animals by a brush when mice were attempting to sleep. The effectiveness of these procedures in producing near continuous wakefulness was previously validated using concurrent EEG/EMG recordings (Yang et al., 2018).

In the progressive ratio experiment, the FR1 session was followed by two sessions of FR3 (three presses per pellet). Then, mice were trained on a progressive ratio schedule according to the following formula: [Response Requirement = (5e^{0.2 × Pellet Number}) – 5] as detailed by Richardson and Roberts (1996) and Schiffino et al. (2019) until the number of pellets earned was stable across three sessions (<10% change). Optogenetic manipulations of BF-PV activity used the same parameters as above and delivered laser for 1 s every 30 s throughout the 1 hr test session. The measure of motivation is the number of pellets earned and inactive lever presses provide a measure of impulsivity.

Perfusion/brain extraction

Mice were anesthetized with sodium pentobarbital (50 mg/mL), exsanguinated with ice-cold PBS, and perfused transcardially with 10% formalin (Cat # HT5011; Sigma-Aldrich; St. Louis, MO). Brains were post-fixed overnight in 10% formalin then transferred to 30% sucrose solution. 40 μm-thick coronal slices were collected and stored at 4°C.

Tissue mounting/microscopy and photography

Coronal sections were mounted onto gel-alum subbed slides and coverslipped using Vectashield Hard Set mounting medium (Cat #H-1400; Vector Laboratories; Burlingame, CA). Fluorescent microscopy and photography were performed using a Zeiss Image2 microscope, with a Hamamatsu Orca R2 camera (C10600). The tracks of optogenetic fiber optic cannula were visible in nuclear stained tissue and the locations were mapped onto appropriate schematic templates employing Adobe Illustrator (v.CS5.1).

RESULTS

Excitation of BF-PV Neurons Enhances Vigilant Attention and Rescues Deficits Induced by Sleep Deprivation.

To test the effect of exciting BF-PV neurons on vigilant attention, we bilaterally injected double-floxed viral vectors into the BF of PV-Cre mice to transduce PV neurons with the excitatory opsin ChR2 coupled to EYFP (Figure 1B). To verify ChR2 transfection efficacy in BF-PV neurons in vivo, we recorded LFP in mPFC and used 40 Hz optogenetic BF-PV excitation to evoke 40 Hz gamma oscillations as previously shown (Kim et al., 2015; McKenna et al., 2020). All mice showed robust 40 Hz responses. In our prior experience, animals that exhibit strong 40 Hz responses were found to have >80% of ChR2 expressing cells stained positive for PV (specificity), and >50% of PV positive neurons expressed ChR2 (penetrance) (Kim et al., 2015; McKenna et al., 2020). Postmortem histological analysis confirmed that viral expression and optic fiber placements were localized to the BF (Figures S1 and S2).

Vigilant attention was assessed with the lever release rPVT. After establishing stable baseline performance (<10% change in mean session reaction time across three consecutive sessions), mice (N=6) were tested under each of three conditions: ChR2 (1 s, 5 mW continuous BF-PV excitation beginning 0.5 s prior to cue light onset; Figure 1A), Sleep Deprivation (8 hr), and Rescue (sleep deprivation + BF-PV excitation). Bilateral, brief, and continuous low-wattage BF-PV excitation was used to enhance activity of BF-PV neurons without attempting to drive specific firing rate frequencies. Moreover, brief excitation closely timed to stimulus onset was used since prolonged (>30 s) continuous BF-PV excitation disrupts performance in other tasks (McNally et al., 2021).

Sorting trials by reaction time reveals a consistent quickening of reaction time during the BF-PV excitation sessions (Figures 1C, 1E). By contrast, sleep deprivation led to a consistent slowing of reaction time, a deficit that was rescued when sleep deprived mice received BF-PV excitation 0.5 s prior to the signal light (Figures 1D, E). Repeated-measures ANOVA with Greenhouse-Geisser correction on mean session reaction time (Figure 1E) shows a significant treatment effect (F (3,15) = 37.02, p < 0.001) where mean session reaction time is faster in BF-PV excitation sessions relative to baseline (Baseline: 445 ± 19 ms vs ChR2: 382 ± 18 ms, p < 0.001), slower following 8 hr sleep deprivation (SD: 533 ± 23 ms, p < 0.01), and returns to baseline when sleep-deprived mice receive BF-PV excitation (SD + ChR2: 422 ± 27 ms, p > 0.20) indicating rescued performance.

We did not observe significant treatment effects on premature responses (F (3,15) = 2.20, p > 0.16; Baseline: 33 ± 3; ChR2: 49 ± 8; SD: 29 ± 7; SD + ChR2: 45 ± 8) or inactive lever presses (F (3,15) = 0.68, p > 0.49; Baseline: 0.3 ± 0.2; ChR2: 1 ± 0.4; SD: 0.3 ± 0.2; SD + ChR2: 1 ± 0.7), suggesting that quickened reaction times during BF-PV excitation are not attributable to increased impulsivity. Moreover, we found no evidence of a speed-accuracy tradeoff during BF-PV excitation as performance accuracy (% correct) did not differ (F (3,15) = 4.06, p > 0.09; Baseline: 93 ± 1%; ChR2: 96 ± 1%; SD: 79 ± 7%; SD + ChR2: 98 ± 1%). Lastly, we did not observe significant treatment effects on omissions (F (3,15) = 3.45, p > 0.12; Baseline: 2.3 ± 0.6; ChR2: 1.2 ± 0.4; SD: 9.0 ± 3.8; SD + ChR2: 0.7 ± 0.3) or session length (F (3,15) = 1.84, p > 0.21; Baseline: 12 ± 1min; ChR2: 13 ± 1min; SD: 10 ± 1min; SD + ChR2: 11 ± 1min). Additional analyses (Figure S3A) confirmed that mice responded to the cue light and did not use laser onset to time their response.

Inhibition of BF-PV Neurons Impairs Vigilant Attention and Mimics the Effects of Sleep Deprivation.

To determine the effect of inhibiting BF-PV neurons on vigilant attention, we again used the lever release rPVT task (Figure 2A). A viral vector expressing ArchT/green fluorescent protein (GFP) in the presence of Cre recombinase was bilaterally injected into the BF of PV-Cre mice and bilateral optical fibers were installed in the BF (Figure 2B).

Figure 2. Optogenetic inhibition of BF-PV neurons impairs vigilant attention.

A: Continuous bilateral optogenetic inhibition of BF-PV neurons began 500 ms before cue light onset and continued for 500 ms thereafter. B: Schematic showing the BF target area for optogenetic inhibition. C: Sorting correct trial reaction times from fastest to slowest illustrates that BF-PV inhibition (N=5; within-subjects) produces a consistent slowing of reaction times throughout the session, an effect that closely resembles that of sleep deprivation (see Figure 1). D: Mean session reaction time is slower in the BF-PV inhibition session relative to baseline. Group mean data are shown as horizontal bars ± SEM and individual data points are shown as open circles. Asterisk denotes p < 0.05.

After establishing stable baseline rPVT performance, mice (N=5) received 1 s of continuous 10 mW green light for ArchT-mediated BF-PV inhibition beginning 0.5 s before cue onset. Sorting trials by reaction time shows that BF-PV inhibition produced a consistent slowing of reaction time (Figure 2C). Mean session reaction time (Figure 2D) was significantly slower in BF-PV inhibition sessions relative to baseline sessions (Baseline: 426 ± 21 ms vs ArchT: 531 ± 11 ms, p < 0.02). Although the ArchT mice were not exposed to sleep deprivation, slowing of reaction time by BF-PV inhibition closely resembled reaction time deficits produced by sleep deprivation in the ChR2 mice (c.f., Figure 1D, E & Figure 2C, D; SD: 20 ± 4% slower vs ArchT: 26 ± 7% slower). We observed a small but significant decrease in inactive lever presses in BF-PV inhibition sessions (Baseline: 2.7 ± 0.8 vs ArchT: 0.2 ± 0.2, p < 0.03). Finally, BF-PV inhibition increased omissions (Baseline: 2 ± 1 vs ArchT: 14 ± 4, p < 0.05) with a concomitant decrease in performance accuracy (Baseline: 93 ± 2% vs ArchT: 71 ± 7%, p < 0.05), but did not alter premature responses (Baseline: 39 ± 8 vs ArchT: 31 ± 8, p > 0.58) or session length (Baseline: 13 ± 1 min vs ArchT: 18 ± 4 min, p > 0.21).

Control Experiments

A group of control mice injected with a non-opsin fluorophore-only encoding control virus (GFP; N=3; Figure S4) exhibited no differences in mean session reaction time relative to baseline when given either the blue or green laser light conditions - to control for ChR2 excitation and ArchT inhibition, respectively (F (2,4) = 0.19, p > 0.98). These control data indicate that altered reaction times during BF-PV excitation/inhibition sessions were not due to non-specific effects of laser illumination. An additional control experiment used the progressive ratio operant task (a method commonly used to assess changes in motivation; e.g. drive to procure food; Schiffino et al., 2019) to demonstrate that neither BF-PV excitation nor BF-PV inhibition altered motivation for food (Figure S5).

DISCUSSION

The BF has long been implicated in the regulation of attention, a component of cognition which is impaired by sleep disruption and affected in many severe neurological and psychiatric disorders. Here, we directly tested the role of PV-expressing GABAergic neurons of the BF in vigilant attention using optogenetic techniques and a rodent signal detection task that closely mimics human tasks widely used to assess the effects of sleep disruption on attention (Lim & Dinges, 2008). Optogenetic excitation and inhibition began shortly before delivery of the cue, since unit recording studies showed that a population of fast-firing BF non-cholinergic neurons ramp up their activity after trial start in anticipation of uncertain events (Hangya et al., 2015; Zhang et al., 2019) in a manner that predicts subsequent reaction time (Hangya et al., 2015). In our experiments here, brief and precisely-timed optogenetic excitation of BF-PV neurons led to consistently faster reaction times in the rPVT whereas optogenetic inhibition consistently slowed reaction times, providing the first evidence using gain- and loss-of function approaches of a role for BF-PV neurons in attention. The present results, along with our previous work (Kim et al., 2015; Hwang et al., 2019; McNally et al., 2021; McKenna et al., 2022), indicate that the activity of BF-PV projection neurons can bidirectionally modulate cortical gamma activity (30–80 Hz, typically 40 Hz) and behavior. It can be hypothesized that balanced and well-timed gamma oscillations are important for vigilant attention and cognition while excessive gamma activity is detrimental. This hypothesis is consistent with a known relationship between arousal and cognitive performance which follows an inverted-U function whereby optimal attention and learning are associated with intermediate levels of arousal, and performance is impaired by both low and excessive levels of arousal (Maness et al., 2022; McCoy & Strecker, 2011; Yerkes & Dodson, 1908).

The BF is one of the most important regions involved in the detrimental effects of sleep loss on cognition. As with the effects of BF lesions on attention, the role of the BF in sleepiness-induced cognitive deficits has generally been ascribed to inhibition of cholinergic neurons (Christie et al., 2008a). Several lines of evidence suggested that BF-PV neurons might also regulate transient fluctuations in alertness and therefore vigilant attention. BF-PV neurons are large or medium sized neurons which project to cortex and thalamic reticular nucleus and discharge at high rates (20–60 Hz) during wakefulness (Duque et al., 2000; Kim et al., 2015; McKenna et al., 2013; Xu et al., 2015), similar to the discharge rates of a subset of non-cholinergic attention-related BF neurons identified in macaques and mice (Hangya et al., 2015; Zhang et al., 2019). Optogenetic excitation of BF-PV neurons rapidly and briefly wakes animals from sleep (McKenna et al., 2020) and potently enhances fast cortical gamma oscillations (Hwang et al., 2019; Kim et al., 2015; McNally et al., 2021). Furthermore, appropriately timed excitation of BF-PV neurons modulates the cortical topography of sensory stimuli-induced gamma oscillations in a manner reminiscent of the effects of attention (Hwang et al., 2019), and optogenetic inhibition of BF-PV neurons impairs gamma oscillations evoked by auditory stimuli (Kim et al., 2015). Previous studies have linked cortical gamma oscillations to successful performance in attention tasks in mice (Kim et al., 2016). Thus, although we have not recorded cortical electrical activity during behavior in this study, the previous literature suggests that modulation of cortical gamma oscillations is a plausible physiological mechanism which could account for our results.

In the present experiment, optogenetic inhibition of BF-PV neurons impaired reaction time in the rPVT to a similar extent as sleep deprivation. Moreover, we found that optogenetic excitation of BF-PV neurons rescued the effects of sleep deprivation on reaction time, suggesting that direct excitation of these neurons can override the inhibitory effects of adenosine on glutamatergic inputs to BF-PV neurons (Yang et al., 2013) which normally reduces their activity during sleep deprivation. Finally, since it is “universally agreed that vigilance is the component of cognition that is most consistently and drastically affected by periods without sleep” (Lim & Dinges, 2008), our observation that sleep deprivation significantly slowed reaction time validates the lever release rPVT as an assessment of vigilant attention in mice.

The goal of this study was to assess the role of BF-PV neurons in vigilant attention as measured by reaction time. Although we cannot rule out the impact of BF-PV manipulations on motor readiness, our observation of similar levels of premature responses across conditions as well as a paucity of exceedingly quick reaction times (<2% of RTs <175ms, Figure S3A) argues against the notion that changes to impulsivity altered reaction times in the rPVT. Another potential confound of our conclusion that BF-PV neurons modulate vigilant attention is that excitation or inhibition of BF-PV neurons might affect motivation and therefore alter performance in attention tasks. Indeed, unit recording studies in rodents and macaques have suggested that BF neurons are important for motivationally-guided attention (Zhang et al., 2019). However, control experiments here using a progressive ratio operant task suggested that our optogenetic manipulations of BF-PV activity did not affect motivation to procure food. These control experiments are also consistent with findings that optogenetic excitation of BF-PV neurons does not alter food intake (Zhu et al., 2017). Lastly, although short periods of experimental sleep deprivation elevate endocrine measures of stress, studies using adrenalectomized rodents and exercise control groups suggest that these neuroendocrine changes do not mediate the vigilance and cognitive impairments produced by sleep loss (Tartar et al., 2009; Meerlo et al., 2008; McCoy & Strecker, 2011). Indeed, the cognitive impairments following 6 h of sleep deprivation produced by gentle sensory stimulation were recently shown to be independent of the stress-induced glucocorticoid response (Raven et al., 2020). Thus, it is unlikely that the ability of BF-PV excitation to rescue vigilance impairments produced by 8h of sleep loss was mediated by alterations in stress hormones since the experimental and control conditions were exposed to identical sleep deprivation procedures.

In a previous study we showed that BF-PV neurons fulfil the criteria for a system which responds to danger signals during sleep by transiently enhancing alertness to support appropriate responses to threat (McKenna et al., 2020). Our results here show that during wake, BF-PV neurons are important for vigilant attention in a food-motivated signal detection task. These new findings support a general role for BF-PV neurons in enhancing alertness, regardless of whether this enhancement is motivated by aversive or appetitive situations. Indeed, it is plausible that BF-PV neurons rapidly activate the cortex in anticipation of, or in response to, motivationally significant events to mobilize cognition and behavior. Thus, therapeutic interventions which enhance the activity of BF-PV neurons may be useful in improving vigilance and therefore attention and cognition in sleep disorders and neurodegenerative diseases.