Cognition-enhancing and cognition-impairing doses of psychostimulants exert opposing actions on frontostriatal neural coding of delay in working memory

Abstract

The prefrontal cortex (PFC) and extended frontostriatal circuitry play a critical role in executive cognitive processes that guide goal-directed behavior. Dysregulation of frontostriatal-dependent cognition is implicated in a variety of cognitive/behavioral disorders, including addiction and attention deficit hyperactivity disorder (ADHD). Psychostimulants exert dose-dependent and opposing actions on frontostriatal cognitive function. Specifically, low and clinically-relevant doses improve, while higher doses associated with abuse and addiction impair, frontostriatal-dependent cognitive function. Frontostriatal cognition is supported by the coordinated activity of neurons across this circuit. To date, the neural coding mechanisms that support the diverse cognitive actions of psychostimulants are unclear. This represents a significant deficit in our understanding of the neurobiology of frontostriatal cognition and limits the development of novel treatments for frontostriatal cognitive impairment. The current studies examined the effects of cognition-enhancing and cognition-impairing doses of methylphenidate (MPH) on the spiking activity of dorsomedial PFC (dmPFC) and dorsomedial striatal (dmSTR) neurons in 17 male rats engaged in a working memory task. Across this frontostriatal circuit, we observed opposing actions of low- and high-dose MPH on the population-based representation of delay: low-dose strengthened, while high-dose weakened, representation of this event. MPH elicited a more complex pattern of actions on reward-related signaling, that were highly dose-, region- and neuron-dependent. These observations provide novel insight into the neurophysiological mechanisms that support the cognitive actions of psychostimulants.

Introduction

The prefrontal cortex (PFC) supports cognitive processes necessary for goal-directed behavior, including working memory, sustained attention, and cognitive flexibility. Dysregulation of PFC-dependent cognitive function occurs in multiple psychiatric conditions, including attention deficit hyperactivity disorder (ADHD) and addiction [1, 2]. At low and clinically relevant doses, psychostimulants, such as methylphenidate (MPH; Ritalin), are highly effective in treating the symptoms of ADHD. At these doses, psychostimulants improve PFC-dependent cognition in ADHD patients and healthy human and animal subjects [3–8]. In contrast, at higher and behaviorally-activating doses associated with abuse and addiction, these drugs impair PFC-dependent cognition [2, 9–11]. Currently, the neural mechanisms supporting the opposing cognitive actions of psychostimulants are poorly understood.

The PFC extends topographically organized projections to the striatum (STR), forming functionally distinct frontostriatal circuits. Evidence indicates that ADHD and addiction are associated with dysregulation of extended frontostriatal circuitry [1, 12]. In rodents, the dorsomedial PFC (dmPFC), encompassing the dorsal anterior cingulate (dAcg) and dorsal prelimbic PFC, plays a pivotal role in higher cognitive function [13–16]. This region extends a prominent projection to the dorsomedial striatum (dmSTR; [17–19]). Consistent with this, evidence demonstrates the dmSTR plays a critical role in PFC-dependent cognition [13, 14, 20].

Psychostimulants act to elevate catecholamine signaling. At high doses, these drugs robustly elevate (800–1000%) extracellular NE and DA throughout the brain [21]. The motor-activating, reinforcing and arousing actions of psychostimulants involve dopamine (DA) and norepinephrine (NE) signaling in a diversity of cortical and subcortical regions, including the PFC and STR [3, 22–25]. In contrast, clinically relevant and cognition enhancing doses of MPH preferentially elevate extracellular DA and NE moderately within the PFC (100–250%), while eliciting notably smaller increases outside this region, including within the STR (~25–50%; [3, 14]). Consistent with these observations, MPH acts directly in the dmPFC, but not the dmSTR, to improve PFC-dependent cognition [17]. Nonetheless, when administered systemically, simultaneous elevations in catecholamine signaling across frontostriatal circuitry could contribute to the procognitive/therapeutic actions of psychostimulants.

Frontostriatal cognitive function is dependent on the coordinated firing patterns of PFC and STR neurons. For example, in both the dmPFC and dmSTR, subpopulations of neurons encode distinct aspects of working memory testing, including for delay and reward [26–30]. Delay-related activity in the PFC is strongly associated with accurate working memory performance [28]. Extensive evidence indicates that catecholamines exert non-monotonic modulation of working memory related PFC neuronal activity. Thus, moderate rates of NE and DA receptor signaling within the PFC strengthen the signal-to-noise representation of the delay interval, while both low and high levels of catecholamine signaling in the PFC degrade delay-related signaling [31]. To date, the modulatory actions of catecholamines on reward-related signaling within dorsomedial frontostriatal circuitry is unknown. While psychostimulants are expected to dose-dependently impact frontostriatal neural coding of cognition, the nature of these actions is currently unknown. This represents a significant gap in our understanding of the neurobiology of frontostriatal-dependent cognition and limits the development of novel treatments for PFC-dependent cognitive dysfunction. To address this issue, the current studies examined the effects of cognition-enhancing vs. cognition-impairing doses of MPH on delay- and reward-related spiking activity of individual dmPFC and dmSTR neurons in rats tested in a delayed-response task of spatial working memory.

Methods and materials

Subjects

Male Sprague-Dawley rats (n = 17; Charles River, Wilmington DE; 260–280 g) were housed individually with enrichment (Nylabone® chews) on a 13/11 h light/dark cycle (lights on 0600 h) with ad libitum water access. During training and testing, access to food was restricted to maintain motivation (15–17 g of standard chow). All facilities and procedures were in accordance with animal use and care guidelines of the National Institutes of Health of the United States and approved by the Institutional Animal Care and Use Committee of the University of Wisconsin-Madison.

Working memory training

Pre-surgery, animals were trained in a T-maze as described previously [27]. The maze resided in an approximately 6’x6’ sound attenuated chamber, with spatial cues masked or placed symmetrically in the room. Masking white noise (60 dB) was played through a speaker 1-meter directly above the T-intersection. Animals received a 45 mg sucrose pellet (Bio-Serv, Flemington, NJ) when they entered the arm of the maze not chosen on the previous trial. Training consisted of 10 trials/session/day. On error trials, the animal was picked up and returned to the start box in the absence of reward. Subsequent responses to the same side are also scored as errors until the animal chooses the opposite side. After entering an arm, tones provided feedback regarding choice accuracy (Incorrect, 5 kHz; Correct, 10 kHz), triggered by breaking an IR beam 2/3 of the way towards the end of the arm (see Fig. 1). Between trials, animals were placed in the start box and prevented from exiting by a gate for a 0-second delay. Following training, animals were surgically implanted with recording electrodes and recovered for 5-days with ad libitum feeding. Following recovery, food-restriction was reinstated, and training continued until performance fell within treatment specific ranges (see below).

Fig. 1. Frontostriatal recordings in working memory-tested rats treated with varying doses of MPH.

A T-maze schematic illustrating key task events, including delay, two distinct auditory tones serving as outcome-related signals on correct versus error trials, and receipt of reward (sugar) and/or pickup. B Bar graph depicts effects of vehicle, 0.5 mg/kg MPH (Low MPH) and 8.0 mg/kg MPH (High MPH) on working memory performance as measured as percent change from baseline (mean ± SEM). Low-dose MPH significantly improved, while high-dose MPH significantly impaired, working memory. C Representative photomicrograph with arrows indicating ventral extent of electrode placement in (left) layer 5 of dmPFC, and (right) dmSTR. D Left, Action potential waveforms of 4 simultaneously discriminated dmSTR MS neurons. Right, Waveforms from these units exhibit separable clusters in 3D-principal component space. **p < 0.01 vs vehicle; fa, anterior forceps of the corpus callosum; dAcg, dorsal anterior cingulate; PL, prelimbic; CC, corpus callosum; LV, lateral ventricle.

Surgery

50 μM stainless steel, Teflon-coated electrodes were implanted into the dmPFC and dmSTR (NB Labs, Dennison, TX; Fig. 1). For the dmPFC, an 8-wire linear array (250 μM separation) was used to target layer V within the dorsal anterior cingulate or dorsal prelimbic PFC (A + 3.0; L ± 0.8; V-2.2 to 2.7 mm). For the dmSTR, a 2x3x3 matrix (250 μm between wires) was implanted ipsilateral to the dmPFC electrodes (centered at A + 0.45; L ± 3.5; V-3.3 at 11° from midline). Electrodes were counterbalanced between hemispheres (10 left; 7 right) and secured with stainless steel screws (MX-0080-16B-C, Small Parts, Inc., Logansport, IN) and dental acrylic (Plastics One, Roanoke, Virginia).

Drugs

MPH (Sigma-Aldrich, St. Louis, MO) was dissolved in 0.9% saline at a concentration of either 0.5 or 8.0 mg/ml (pH=7.4) and administered subcutaneously (s.c.) at 1 ml/kg.

Testing procedures

Animals were transported in their home cage 2-hours prior to a first recording session and placed on top of the T-maze. The headstage, attached to a 32-channel commutator and connected to a neural acquisition unit (OmniPlex®; Plexon, Dallas, TX), was connected to the animal and single units discriminated. For testing, animals ran 30-trials during a baseline session. Initially, delay duration was set at 5-seconds. Different baseline performance ranges were used for low- vs. high-dose MPH to avoid ceiling and floor effects, respectively. Specifically, the following baseline performance ranges were used: low-dose MPH, 63–80% (mean=71 ± 5%); high-dose MPH, 80–97% (mean=89 ± 6%); vehicle 73–90% (mean=84 ± 5%). Given performance improves over multiple testing sessions, delay duration was increased over time to maintain performance within the target range, resulting in delays ranging from 5–40-seconds. Following the first recording session, animals were returned to their home cage for 2-hours while remaining connected to the recording tether, allowing for within-neuron analyses across baseline and treatment conditions. Animals received vehicle, 0.5 mg/kg, or 8.0 mg/kg MPH s.c. 45-minutes prior to a second 30-trial testing session. To better identify neural correlates of MPH-induced changes in cognition and given challenges associated with s.c. injections to the nape in tethered animals, analyses were limited to only sessions with a > 10% change in performance following MPH administration.

Electrophysiology

Single unit recordings

Putative single ‘units’ of the dmPFC and dmSTR were discriminated in real-time using online template matching algorithms to preliminarily discriminate action potentials exhibiting an ~3:1 signal-to-noise ratio. Following unit sorting, animals remained tethered to the recording hardware and the quality of the discrimination was monitored throughout the testing session. Timestamps of task events were captured using infrared (IR) beams and high-resolution video tracking (Cineplex, Plexon, Dallas, Texas). Receipt of reward was time-stamped from video tracking to identify the moment the animal bit down on the reward. Multiple measures were used to ensure sorted waveforms arose from single units, as previously described [27].

Neuron identification

For the dmPFC, wide-spiking (WS), putative glutamatergic pyramidal vs. narrow-spiking, putative inhibitory interneuron cell types were differentiated based on the trough to peak latency of the action potential waveform (WS > 200 μs; NS, 100–200 μs), as previously described [27, 32]. Few narrow-spiking neurons were observed, and thus, subsequent analyses focused on WS neurons. For the dmSTR, medium spiny neurons (MSNs) and fast-spiking (FS) interneurons were identified using the trough to peak duration (MSN > 250 μs; FS < 200 μs) [26, 33]. Few fast-spiking interneurons were identified, and thus subsequent analyses focused on MSNs.

Under baseline conditions, distinct subpopulations of dmPFC and dmSTR neurons displayed excitatory responses to specific task events, including delay and reward (Figs. 2, 3). Due to the infrequency of error trials during the baseline session, error-related signaling was not analyzed. Within both regions, there were relatively few neurons that responded in a brief and time-locked manner to reward receipt per se. Instead, larger populations were observed that responded to the period between the tone signaling reward and when the animal was picked up (and promptly placed back in the start box following each trial), which we termed “reward proximity”. This resulted in an average 2.99 ± 1.4 second epoch. Tuning of individual neurons (i.e. an excitatory response during either the delay or reward proximity interval) was determined by calculating the z-score of a neuron’s firing rate during the relevant baseline testing session interval (delay or reward proximity) compared to the mean activity in the 14-seconds surrounding, but not including, the interval. Tuned neurons were identified by a baseline z-score of >0.07 for delay, and >0.2 for reward. The longer duration of the delay interval warranted a smaller z-score. For delay, trials with a baseline spiking rate of less than 0.5 Hz were removed. Outliers were identified as trials with a percent change from baseline greater than 3x the 0.1 quantile range, resulting in the exclusion of less than 1% of trials. This approach identified neurons with task-related response properties similar to those described by others (Figs. 2, 3; [28, 34–36]). Units not meeting these criteria were considered ‘untuned’ to task events.

Fig. 2. Effects of MPH on delay-related activity.

Left, Exemplar rasters/PETHs for dmPFC (A) and dmSTR (B) neurons strongly tuned to delay. In raster displays, fiducial markers indicate start of delay (left; lighter) and end of delay (right; darker). Line under raster indicates duration of delay. Bar graphs depict changes in firing rate (mean ± SEM) during the delay interval as measured as percent change from baseline activity following vehicle (VEH), 0.5 mg/kg MPH (Low MPH) and 8.0 mg/kg MPH (High MPH) for neurons strongly tuned (Tuned) and not tuned (Untuned) to delay. A Low-dose MPH had no significant effects on neurons tuned to delay, while significantly suppressing the activity of dmPFC neurons not tuned to delay. In contrast, high-dose MPH significantly suppressed the activity of neurons strongly tuned to delay. B Within the dmSTR, neither vehicle nor low-dose MPH significantly affected delay tuned neurons. While vehicle elicited a slight suppression of activity in neurons not tuned to delay, low-dose MPH further suppressed delay-related activity of untuned neurons. In contrast, high-dose MPH robustly increased the delay-related firing activity of neurons regardless of tuning. ++p < 0.01, +++p < 0.001 vs. baseline. *p < 0.05, ***p < 0.001 vs. vehicle.

Fig. 3. Effects of MPH on reward-related activity.

Left, Exemplar rasters/PETHs for dmPFC (A) and dmSTR (B) neurons strongly tuned to reward. In raster displays, fiducial markers indicate start of reward (left; lighter) and end of reward (right; darker). Line under raster indicates duration of reward proximity. Bar graphs depict changes in firing rate (mean ± SEM) during the reward proximity interval as measured by percent change from baseline activity following vehicle (VEH), 0.5 mg/kg MPH (Low MPH) and 8.0 mg/kg MPH (High MPH) for neurons strongly tuned (Tuned) and not tuned (Untuned) to delay. A Following treatment with low-dose MPH there was a trend (p = 0.07) for suppression of activity of dmPFC reward proximity-tuned neurons relative to vehicle, while increasing activity in untuned neurons. High-dose MPH also showed a trend for suppressed activity in strongly tuned neurons relative to vehicle (p = 0.18). B Within the dmSTR, low-dose MPH did not affect the activity of neurons during the reward interval, while high-dose MPH robustly increased firing among both reward-tuned and untuned neurons. +p < 0.05, +++p < 0.001 vs. baseline. ***p < 0.001 vs. vehicle.

Histology

Animals were deeply anesthetized while cathodal current (15μA) was passed through each electrode (referenced to ground) for 10-seconds. Animals were perfused with cold saline followed by 3.7% formaldehyde, after which brains were removed, and immersed in formaldehyde for at least 24-hours. Coronal sections (40 μm) were collected, mounted on slides and counterstained with neutral red.

Statistical analyses

Two-way linear mixed-effects models (JMP® Pro Version 15.2.1, SAS Institute Inc.) were used to analyze behavioral and electrophysiological measures, with session (baseline vs. treatment), treatment (vehicle, low-dose MPH, high-dose MPH) and their interaction as fixed effects factors. Individual neurons were treated as random effects. For behavioral performance, we analyzed the change in % correct trials between baseline and treatment sessions. For single unit activity, the mean spiking rate during each task interval was determined on a trial-by-trial basis using peri-event time histogram (PETH) analysis, with 8 independent analyses done for each combination of region (dmPFC or dmSTR), interval (delay or reward proximity), and neuronal tuning (tuned or untuned). Planned comparisons using Welch’s T-tests were used to compare the difference between baseline and treatment effect in vehicle to that seen with low-dose and high-dose MPH.

Results

MPH effects on working memory performance

The three treatment groups displayed the following baseline performance levels: vehicle, 84.7 ± 1.6% (n = 12); low-dose MPH (n = 8), 71.3 ± 2.0%; high-dose MPH (n = 11), 88.8 ± 1.70%. As shown in Fig. 1, there was a main effect of treatment (F_2,55.2 = 8.73, p < 0.001) with low-dose MPH significantly improving performance relative to baseline (26.2 ± 6.0%, p < 0.0001) and vehicle (p < 0.01), high-dose significantly impairing performance relative to baseline (−37.3 ± 4.0%, p < 0.0001) and vehicle (p < 0.0001), while vehicle did not significantly affect performance relative to baseline (2.6 ± 4.0%, p = 0.52).

Response properties of dmPFC and dmSTR neurons under baseline conditions

Across all animals tested under baseline conditions, seventy-four (22.4%) WS dmPFC neurons were classified as strongly tuned to the delay interval (Fig. 2) and 67 (20.4%) were identified as strongly tuned to the impending delivery of food reward (reward proximity, Fig. 3). The remaining neurons were classified as being insensitive or ‘untuned’ to these task events (delay, n = 256; reward proximity, n = 261). No significant difference was seen in the baseline firing rate of dmPFC neurons tuned to delay or reward proximity between treatment groups (delay, F_2,71.1 = 0.55, p = 0.58; reward proximity, F_2,59.6 = 1.71, p = 0.19; Supplemental Materials Table S1).

For the dmSTR, 68 (25.8%) neurons were classified as strongly tuned to the delay interval (Fig. 2) and an additional 55 (20.6%) were identified as strongly tuned to reward proximity (Fig. 3). The remaining neurons were classified as being insensitive or ‘untuned’ to these task events (delay, n = 196, reward proximity n = 212). No significant difference was seen in the baseline firing rate of dmSTR neurons tuned to delay or reward proximity between treatment groups (delay, F_2,45 = 0.11, p = 0.89; reward proximity, F_2,51.7 = 0.73, p = 0.48; Supplemental Materials Table S1).

Frontostriatal Neural Coding Actions of a Cognition-Improving Dose of MPH

Delay-related activity

As shown in Fig. 2 and Table 1, within the dmPFC and relative to baseline, vehicle treatment had no significant effect on the activity of neurons strongly tuned to delay (−2.29 ± 2.89%, p = 0.43; n = 25) or on the activity of neurons not tuned to delay (1.4 ± 1.52%, p = 0.36; n = 93). At a low and working memory improving dose (0.5 mg/kg), MPH did not significantly affect the activity of dmPFC neurons strongly tuned to delay relative to baseline (0.69 ± 2.27%, p = 0.76; n = 30) or vehicle (p = 0.21). In contrast, this dose significantly suppressed the activity of the larger population of dmPFC neurons not tuned to delay, relative to both baseline (−7.74 ± 1.31%, p < 0.001; n = 81) and vehicle (p < 0.001). Combined, these actions increase the signal-to-noise representation of delay across the broader population of dmPFC neurons.

Table 1.

Summary of Statistical Analyses.

A.
Interval	Region	Tuning	Treatment	Session	Treatment X Session
Delay	PFC	Tuned	F(2,71) = 0.43, p = 0.65	F(1,3595) = 10.15, p < 0.01	F(2,3595) = 7.07, p < 0.001
		Untuned	F(2,253) = 0.88, p = 0.41	F(1,15084) = 3.52, p = 0.06	F(2,15084) = 18.07, p < 0.001
	dmSTR	Tuned	F(2,61) = 1.52, p = 0.23	F(1,3025) = 1.94, p = 0.16	F(2,3025) = 6.64, p < 0.01
		Untuned	F(2,193) = 0.56, p = 0.57	F(1,11477) = 43.06, p < 0.001	F(2,11477) = 188.6, p < 0.001
Reward	PFC	Tuned	F(2,64) = 1.38, p = 0.26	F(1,3135) = 18.96, p < 0.001	F(2,3135) = 3.14, p = 0.04
		Untuned	F(2,255) = 0.13, p = 0.87	F(1,11871) = 2.09, p = 0.15	F(2,11870) = 6.24, p < 0.01
	dmSTR	Tuned	F(2,52) = 0.42, p = 0.66	F(1,2587) = 26.40, p < 0.001	F(2,2587) = 16.20, p < 0.001
		Untuned	F(2,201) = 1.53, p = 0.22	F(1,9653) = 38.18, p < 0.001	F(2,9653) = 60.74, p < 0.001

B.
Interval	Region	Tuning	Comparison	VEH	Low MPH	High MPH
Delay	PFC	Tuned	Baseline	t3595 = 0.79, p = 0.427	t3595 = 0.3, p = 0.761	t3596 = 4.4, p < 0.001
			Vehicle	N/A	t2764 = 0.82, p = 0.205	t2092 = 1.94, p = 0.026
		Untuned	Baseline	t15084 = 0.91, p = 0.363	t15084 = 5.88, p < 0.001	t15084 = 1.88, p = 0.060
			Vehicle	N/A	t6806 = 4.53, p < 0.001	t6871 = 0.58, p = 0.280
	dmSTR	Tuned	Baseline	t3024 = 0.91, p = 0.908	t3025 = 1.34, p = 0.180	t3026 = 3.53, p < 0.001
			Vehicle	N/A	t2183 = 0.48, p = 0.316	t1851 = 2.02, p = 0.022
		Untuned	Baseline	t11477 = 2.86, p = 0.004	t11477 = 6.35, p < 0.001	t11478 = 18.65, p < 0.001
			Vehicle	N/A	t6806 = 1.70, p = 0.045	t6871 = 16.09, p < 0.001
Reward	PFC	Tuned	Baseline	t3135 = 1.02, p = 0.310	t3135 = 4.18, p < 0.001	t3136 = 2.31, p = 0.021
			Vehicle	N/A	t2823 = 1.63, p = 0.051	t2634 = 0.72, p = 0.237
		Untuned	Baseline	t11866 = 1.22, p = 0.224	t11866 = 3.62, p < 0.001	t11877 = 0.13, p = 0.893
			Vehicle	N/A	t11710 = 3.37, p < 0.001	t10870 = 0.88, p = 0.188
	dmSTR	Tuned	Baseline	t2587 = 0.57, p = 0.570	t2587 = 1.61, p = 0.108	t2587 = 6.39, p < 0.001
			Vehicle	N/A	t2519 = 1.52, p = 0.065	t2069 = 7.84, p < 0.001
		Untuned	Baseline	t9652 = 1.77, p = 0.077	9654 = 0.95, p = 0.343	t9655 = 11.58, p < 0.001
			Vehicle	N/A	t8882 = 0.89, p = 0.188	t8062 = 10.54, p < 0.001

A. Analyses using two-way linear mixed-effects models with session (baseline vs. treatment) and treatment (VEH, low-dose MPH, high-dose MPH) as fixed effects factors. B. Planned comparisons using Welch’s t-tests to compare the change in firing rate from baseline to manipulation of vehicle to that seen with low- and high-dose MPH.

Within the dmSTR, a similar pattern was observed. Specifically, neither vehicle treatment (−0.67 ± 5.83%, p = 0.91 vs. baseline; n = 21) nor low-dose MPH had a significant effect on the activity of neurons strongly tuned to delay (−3.50 ± 2.61%, p = 0.18 vs. baseline; p = 0.32 vs. vehicle; n = 26). For neurons not tuned to delay, vehicle treatment was associated with a modest suppression of activity during the delay interval relative to baseline (−4.27 ± 1.56%, p < 0.01; n = 74). However, low-dose MPH elicited a significantly larger suppression of delay-related activity of untuned neurons relative to vehicle (−7.62 ± 1.20%, p < 0.001 vs. baseline; p < 0.05 vs. vehicle; n = 68). Thus, low-dose MPH strengthens the signal-to-noise representation of delay within the dmSTR, similar to that seen in the dmPFC.

Reward-related activity

As shown in Fig. 3 and Table 1, vehicle treatment had no significant effect on the activity of dmPFC neurons either tuned (−4.32 ± 4.44%, p = 0.31; n = 27) or not tuned to reward proximity (−2.77 ± 2.34%, p = 0.22; n = 90), relative to baseline. Low-dose MPH significantly decreased the activity of neurons strongly tuned to reward proximity relative to baseline (−13.19 ± 3.16%, p < 0.001; n = 23), with a trend towards suppression relative to vehicle (p = 0.052). In contrast, this dose increased the activity of dmPFC neurons not tuned to reward proximity relative to baseline (8.28 ± 2.28%, p < 0.001; n = 87) and vehicle (p < 0.001). These actions result in a decreased population-level representation of this event.

Within the dmSTR, vehicle had no effect on the activity of neurons strongly tuned to reward proximity relative to baseline (−1.86 ± 3.33%, p = 0.570; n = 23). Low-dose MPH had no significant effect on the activity of neurons tuned to reward proximity relative to baseline (5.29 ± 3.29%, p < =0.108; n = 20), but did show a trend towards activation relative to vehicle (p = 0.07). Neither vehicle nor low-dose MPH significantly affected neurons not tuned to reward proximity (vehicle, −5.29 ± 3.16%, p = 0.08; n = 73; low-dose MPH, −1.96 ± 2.06%, p = 0.34 vs. baseline, p = 0.19 vs. vehicle; n = 77).

Frontostriatal neural coding actions of a cognition-impairing dose of MPH

Delay-related activity

As shown in Fig. 2 and Table 1, a high and cognition-impairing dose (8.0 mg/kg) of MPH significantly suppressed the firing rate of dmPFC neurons strongly tuned to delay relative to baseline (−9.75 ± 2.22%, p < 0.001, n = 19) and vehicle (p = 0.026), while having no significant effect on neurons not tuned to delay (2.61 ± 1.38%, p = 0.28 vs. baseline, p = 0.06 vs. vehicle; n = 82). Within the dmSTR, although high-dose MPH significantly increased the activity of MSNs strongly tuned to delay (13.47 ± 3.81%, p = 0.02 vs. vehicle; p < 0.001 vs. baseline; n = 21), this dose elicited an even greater increase in the activity of neurons not tuned to this event (34.75 ± 1.86%, p < 0.001 vs. vehicle and baseline; n = 54). Collectively, these actions decrease the representation of delay-related information across this dorsomedial frontostriatal pathway.

Reward-related activity

As shown in Fig. 3 and Table 1, although high-dose MPH significantly decreased the activity of dmPFC WS neurons tuned to reward proximity relative to baseline (−8.44 ± 3.6%, p = 0.23; n = 17), this was not significant relative to vehicle (p = 0.237). This dose had no effect on neurons not tuned to reward proximity relative to baseline (0.36 ± 2.64%, p = 0.893; n = 84) or vehicle (p = 0.479). Within the dmSTR, high-dose MPH robustly increased the activity of MSNs tuned to reward proximity (71.1 ± 11.1%, p < 0.001 vs. baseline and vehicle; n = 12). However, a largely similar excitation was observed for neurons not tuned to this event (50.1 ± 4.3%, p < 0.001 vs. baseline and vehicle; n = 62).

Analyses of all animals

The above analyses were limited to MPH-treated animals displaying a > 10% change in performance to avoid significant overlap with vehicle-related changes in performance. Nonetheless, additional analyses were conducted for all animals receiving low vs. high dose MPH in the absence of a performance cut-off. As shown in Supplemental Materials, Figure S1 and Table S2, when analyzed in this manner a largely similar pattern of results was observed, but with a reduced effect size.

Discussion

Psychostimulants exert dose-dependent actions on frontostriatal-dependent cognition; low-doses improve and high doses impair [14]. To better understand the neurophysiological actions underlying these divergent cognitive actions, the current studies compared the effects of cognition-improving and cognition-impairing doses of MPH on delay- and reward-related activity of dmPFC and dmSTR neurons in animals engaged in a delayed response task of spatial working memory. These tasks require maintenance of information over short delays to guide subsequent goal-directed behavior and have been used extensively to study the neurobiological basis of PFC-dependent cognitive function [28, 31, 37]. Importantly, performance in these tasks is improved by all approved ADHD medications and impaired by acute stress in healthy human and animal subjects [14], demonstrating their translational relevance. Delay-related activity of PFC neurons is closely associated with accurate working memory and has been posited to protect the maintenance of task-relevant information from interference from distractors [28, 37–40]. As in earlier studies [26, 27], we observed a robust population of neurons tuned to delay in the dmPFC and dmSTR. At a cognition-improving dose, MPH significantly decreased the activity of the larger population of frontostriatal neurons not tuned to delay, while leaving unaffected the activity of neurons strongly tuned to delay. Collectively, this strengthens the signal-to-noise representation of delay-related information across the broader neuronal population within this circuit and could contribute to the procognitive actions of low-dose MPH. The small, yet significant, decrease in delay-related activity of untuned dmSTR neurons in vehicle-treated animals could contribute to the small (but nonsignificant) improvement in performance in these animals. However, the additional suppression of this activity in low-dose MPH-treated animals, relative to vehicle, was not proportional to the change in performance. These observations may indicate that the procognitive actions of low-dose psychostimulants are more dependent on dmPFC coding of delay or on the combined actions of MPH on delay-related coding across this circuit. At a cognition-impairing dose, MPH suppressed the activity of dmPFC neurons strongly tuned to delay while leaving unaffected neurons not tuned to this event. While this higher dose increased the activity of delay-tuned dmSTR neurons, it elicited an even greater increase in the activity of the larger population of dmSTR neurons not tuned to this event. Combined, these actions likely degrade the representation of delay across dorsomedial frontostriatal circuitry.

In contrast to that seen with delay, low-dose MPH reduced the activity of dmPFC neurons strongly tuned to reward proximity, while increasing the activity of neurons not tuned to this event. While speculative, these actions could reduce the salience of potential rewards and thus facilitate resistance to distraction and/or promote cognitive control [41]. The fact that low-dose MPH does not affect reward-related coding in the dmSTR indicates the procognitive actions of MPH are not dependent on alterations in reward-related activity in this region. High-dose MPH tended to decrease the activity of dmPFC neurons tuned to reward, similar to that seen with low-dose MPH. In contrast, within the dmSTR, high-dose MPH elicited a robust increase in activity of neurons tuned (+71%) and untuned (+50%) to reward proximity. Given the larger population of untuned neurons, high-dose MPH may degrade the representation of this event at the population level. Alternatively, the robust excitatory actions of high-dose MPH on reward-tuned dmSTR neurons could strengthen reinforcement learning, increasing attention to rewards, and shift frontostriatal control from a PFC-driven and flexible mode to a less flexible, bottom-up form of reward seeking, similar to that seen in addiction [42]. These changes in reward-related processing may transition the animal from an optimal strategy requiring working memory to an easier, and suboptimal win-stay, lose-shift strategy that results in impaired accuracy in this task.

In both ADHD patients and healthy human subjects, imaging studies have yielded conflicting results on the effects of stimulants on PFC activity during a working memory task. Whereas some studies observed improved working memory was associated with an increase in PFC activity [43–45], others observed a decrease in PFC activity [46–48]. Aligning these results to the current observations is difficult given imaging studies have typically not examined drug effects across distinct stages of a task (i.e. delay vs outcome) nor do they segregate by neuronal subtype (e.g. only glutamatergic neurons in the PFC).

Catecholamine mechanisms underlying MPH modulation of frontostriatal neural coding

Catecholamines exert non-monotonic modulation of PFC-dependent cognitive function as well as delay-related activity of PFC neurons [49]. Under optimal conditions associated with moderate levels of catecholamine release, high-affinity postsynaptic noradrenergic α2A receptors increase the activity of PFC neurons tuned to delay, while dopaminergic D1 receptors suppress delay-related activity to task-irrelevant stimuli (i.e. ‘noise’; [50]). Consistent with these observations, the working memory improving actions of MPH are dependent on PFC α2 and D1 receptors [51, 52]. The ability of low-dose MPH to decrease the activity of dmPFC neurons not tuned to delay observed in the current study is consistent with moderate rates of D1 receptor activation. The fact that this dose did not increase the activity of delay-tuned neurons in the PFC could indicate that α2A receptors are maximally activated under baseline conditions. Under higher rates of catecholamine release in the PFC, activation of lower affinity noradrenergic α1 receptors and high rates of D1 receptor activation suppress the activity of delay-tuned neurons [53]. The ability of high-dose MPH to suppress delay-related firing of PFC neurons is thus consistent with involvement of both receptors. The modulatory actions of catecholamines on reward-related signaling within the PFC have not been extensively studied. However, the current observations suggest that at moderate levels, PFC catecholamines differentially impact reward- vs. delay-related signaling. The differential actions of low-dose MPH across task-related events indicates the cognition-improving effects of low-dose MPH do not involve a simple change in excitatory or inhibitory tone within the PFC. In contrast to that seen with the higher dose of MPH in the current study, a locomotor activating dose of amphetamine was observed to increase reward-related activity of PFC neurons [54]. However, unlike MPH, higher doses of amphetamine also elevate extracellular serotonin and thus may have different neuromodulatory actions than MPH [21]. Moreover, the 8 mg/kg dose of MPH used in these studies did not lead to robust locomotor activation.

Relative to the higher dose, low-dose MPH had limited effects on task related activity of dmSTR neurons. This is consistent with the modest effects of this dose on extracellular striatal DA [3, 14], particularly in contrast to that seen with higher doses of psychostimulants [3, 21]. The relatively large excitatory effects of high-dose MPH on delay- and reward-related firing of dmSTR neurons are consistent with the well-documented excitatory actions of DA on striatal neurons [55–58].

Lastly, it should be noted that high-dose psychostimulants affect neuromodulators other than catecholamines, including glucocorticoids [59]. Glucocorticoids are known to act within the PFC to impair cognitive function [60]. Currently, the role of non-catecholamine neuromodulators in the cognitive actions of high-dose psychostimulants remain to be determined.

Author contributions

RCS and CWB conceived the experiments. RCS and AJM executed the experiments. RCS and AJM, analyzed the data with assistance from DMD. RCS and CWB wrote the manuscript.

Funding

This work was supported by the National Institutes of Health MH09863 (RCS); MH80813, MH116526 (CWB).

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41386-023-01738-6.