Concomitant activation of D₁ dopamine and α_2A adrenergic receptors improves cognition better than methylphenidate in two rodent behavioral tests

Abstract

Rationale

Methylphenidate is commonly prescribed to manage symptoms of attention-deficit/hyperactivity disorder (ADHD), but like other stimulants it has variable effectiveness. Methylphenidate works by increasing the synaptic availability of dopamine and norepinephrine, resulting in stimulation of dopaminergic and adrenergic receptors. One hypothesis is that selective receptor targeting may be more effective clinically and have fewer side effects than non-selective stimulants.

Objectives and methods

To test this hypothesis, we compared methylphenidate with three compounds: the selective D_1/5 dopamine agonist 2-methyldihydrexidine; the selective α_2A adrenergic agonist guanfacine; and the cannabinoid compound cannabigerol, that has α_2A agonist properties and was included given the increasing recreational cannabis use among individuals with ADHD. Acute effects on temporal order memory, cognitive flexibility, and spatial working memory were evaluated using two rodent behavioral tasks.

Results

Co-administration of an α_2A agonist and a D₁ agonist produced greater cognitive improvement than methylphenidate, but only in rats with poor baseline performance.

Conclusions

These findings suggest that potential benefits may emerge from the coadministration of selective agents (e.g., α_2A and D₁ agonists) and should be considered for further study, especially regarding individuals with decrements in cognitive function.

Introduction

Attention-deficit/hyperactivity disorder (ADHD) is one of the most common neurodevelopmental disorders, affecting about 9% of children, that often persists throughout adulthood [1, 2]. This cognitive disorder is categorized by impulsivity and inattention that negatively impacts learning and social behavior [3, 4]. Besides the well-known symptoms of ADHD, such as reduced working memory (WM), struggling to make decisions, and difficulties sustaining attention, individuals who have this disorder tend to have higher instances of substance abuse, anxiety, and depression [2, 5, 6]. Most importantly, there are no known cures for ADHD, and patients are generally required to take medication (e.g., stimulants) indefinitely. Non-pharmacological interventions, such as cognitive behavioral therapy, are generally less effective than stimulants and usually used concomitantly with stimulants [7–9].

ADHD has a complex and heterogeneous etiology involving many unknown factors (genetic, developmental, environmental); this complicates a search for personalized therapy [10, 11]. Thus, the gold standard of care for ADHD has been formulations of stimulants such as methylphenidate (MPH) and amphetamine [12–14]. The exact mechanisms of how MPH and other stimulants affect ADHD are not fully understood [15], but MPH affects numerous receptor systems via the inhibition of catecholamine reuptake (i.e., of dopamine and norepinephrine). This explains why these compounds have undesired side-effects or risk of abuse. Although the complex etiology of ADHD may make precision medicine approaches difficult, improved symptomatic control is still very important to a majority of patients.

MPH is an indirect agonist of adrenergic and dopaminergic receptors like the α_2A adreno receptor (α_2AR) and the D₁ dopamine receptor (D₁R) [16–20]. Activation of the α_2AR is known to enhance prefrontal cortical (PFC) function and network connectivity [21], and dysfunction of this system is common in ADHD [17]. The D₁R also plays an important role in optimizing PFC activity, and aberrant dopamine signaling in the PFC leads to deficits in WM and other cognitive processes [22–24], which are also seen with dysfunctional norepinephrine signaling [25]. Thus, both the α_2AR and D₁R systems play pivotal roles in promoting optimal PFC function; the α_2AR enhances network connectivity, whereas the D₁R finely tunes the signal-to-noise ratio by regulating excitatory and inhibitory balance [21, 26–31]. Increasing either dopamine and norepinephrine signaling results in an inverted-U-shaped dose-response curve; too little or too much causes suboptimal cognitive processes [30, 32–37].

It has been speculated that individuals with ADHD have dysregulated signaling in one or both of these systems [28], suggesting them to be useful therapeutic targets. In fact, guanfacine (GFX), a selective α_2AR full agonist, is approved for clinical use in the treatment of ADHD and is a useful alternative option to stimulants [16, 38–41]. It is, however, noteworthy that a recent study indicated GFX slightly improved WM, though not as effectively as MPH [39]. On the other hand, our lab, as well as others, have reported that selective D₁ agonists improve both spatial WM and temporal order memory in rodents, with the effects in our study surpassing those of MPH [32–36, 42]. Based on this collective evidence, we therefore sought to test the hypothesis that D₁ and α_2A coactivation might produce better cognitive improvements, and thus compared them alone or in combination versus the standard-of-care methylphenidate. We used two well-established behavioral paradigms: temporal-order object recognition (TOR) in the open field and delayed alternation response (DAR) in the T-maze. The TOR task measures object recognition with a focus on the temporal order, while the DAR task measures spatial WM. Both tasks critically rely on PFC function, as lesioning the PFC hindered rodents’ ability to perform these two tasks [42, 43]. More importantly, patients with ADHD experience a reduction in both temporal order memory and spatial WM [44, 45]. We hypothesized that a selective α_2A agonist and a selective D₁ agonist, whether administered alone or together, will more effectively rescue these memory deficits than MPH. Furthermore, we hypothesize that cotreatment of an α_2A agonist and a D₁ agonist will be more efficacious than either compound alone.

Materials and methods

Subjects

A total of 31 male Fischer rats were included in the study, with 24 obtained from Envigo (Frederick, MD) and seven from the National Institute on Aging (Hollister, CA; Raleigh, NC; or Kingston, NY). At the start of the experiments, rats weighed between 295 and 495 g. Depending on their body weight, they were housed individually, in pairs, or in groups of three, under a 12-hour light/dark cycle. Water was available at all times. Animals assigned to the TOR task were given free access to food, whereas those in the DAR task were maintained on a restricted diet of Bio-Serv rat chow to keep their body weight at 90–95% of their free-feeding weight, allowing food to be used for motivation. A palatable chocolate-flavored sucrose pellet (Bio-Serv, Flemington, NJ) was used as the behavioral reward. All animal care and experimental procedures were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals and Penn State Hershey Animal Resources Program, with approval from the Institutional Animal Care and Use Committee (IACUC) of Penn State College of Medicine (protocol # 202001749).

Drug preparation and administration

All drug solutions were freshly prepared on the day of experimentation. Four compounds were used: MPH (methylphenidate, Ritalin^®), 2-methyldihydrexidine (2MDHX), GFX (guanfacine, Intuniv ER^®), and cannabigerol (CBG).

MPH powder was gifted by Novartis. It was dissolved in purified water (10 mg/mL) and then mixed into Nutella^® to produce an oral dose of 1.5 mg/kg, corresponding to the allometric equivalent used in ADHD patients [20, 46, 47]. Nutella alone served as the vehicle control.

The D₁ agonist 2MDHX was synthesized following a modified protocol [33, 48]. Because 2MDHX is prone to oxidation, it was prepared in 0.1% ascorbic acid vehicle and administered subcutaneously due to its limited oral bioavailability. A dose of 3 µg/kg was selected based on prior studies demonstrating its efficacy in enhancing WM in rats [32–35]. 0.1% ascorbic acid (w/v in distilled water) served as the vehicle control.

The α_2A agonist GFX was purchased from Tocris (Bristol, UK). It was dissolved in saline vehicle and administered intraperitoneally at a 0.1 mg/kg dose. This dose was chosen because it improved WM in rats without causing hypotension [49, 50]. Saline served as the vehicle control.

CBG is a cannabinoid that has α_2A agonist activity, but has very low CB1 or CB2 receptor activity [51, 52]. Given the increased recreational use of cannabis, including among individuals with ADHD [53], we decided to include CBG in this study. It was purchased from Cayman Chemical (Ann Arbor, MI) and dissolved in saline based vehicle (1:1:18 DMSO: Tween80:Saline) for intraperitoneal administration at 1 mg/kg. This dose was chosen because it was reported to be functionally equivalent to GFX in terms of cardiovascular effects, but has a lower risk of hypotension in mice [54].

Notably, for compounds not yet approved for clinical use (2MDHX and CBG), doses were selected based on behavioral paradigms within this study (DAR and TOR tasks), prioritizing therapeutically relevant doses that enhanced spatial working memory and temporal object recognition. For clinically approved compounds (MPH and GFX), doses were determined using established human-to-rat allometric conversions. Together, these selections were guided by prior evidence to maximize translational relevance.

Temporal-order object recognition (TOR) test

Apparatus

Open field observational chambers (E63-20, Coulbourn Instruments, Whitehall, PA) were used for the temporal-order object recognition (TOR) experiments. The chambers were 41 × 41 cm squared, and made of opaque walls, with black floors. CCD cameras (30 frames/second, STC-TB33USB-AS, SenTech, Carrollton, TX) were set above to track the free movement in real time, and footage was analyzed in LimeLight software (Actimetrics, Coulbourn Instruments, Whitehall, PA). Several types of objects were used, including opaque plastic bottles and enrichment toys from Bio-Serv (Kong Genius, Kong Toys, Precious Gem, Dumbbells; Bio-Serv, Flemington, NJ). They all were as tall as or no taller than twice the size of the rat [55, 56].

Behavioral task procedure

Rats were habituated to all procedures and tested in a modified classic novel object recognition test (Fig. 1a). There was first a three-minute habituation phase. After habituation, rats performed the acquisition phase that consisted of two sampling trials, followed by the retrieval phase that consisted of one choice trial [57–59]. Each trial lasted five minutes, and rats were returned to their home cages after each of these five-minute explorations in the test chambers. For the first sampling trial, each rat was presented with two identical copies of one object (O1 + O1). The second sampling trial, one hour after the first, was similar except the rat would be presented with two identical copies of a second object (O2 + O2) located where the O1s were during the first sampling trial. After a three-hour delay, the choice trial involved presenting a third copy for each of the two prior objects (O1 + O2). The arena was cleaned with 70% EtOH between each trial. The objects were randomly assigned for each test session, as well as the location that two objects were assigned during the choice trial (i.e., O1 on the left vs. O2 on the left).

Fig. 1.

Comparison of treatment effects on TOR task performance. a: Schematic of TOR task and rats’ baseline performances when vehicle was administered. Each circle represents one vehicle session. Note the large variation. I_TOR=0.7 was used as the criteria to categorize rats into groups having inferior (≤ 0.7) or superior (> 0.7) baseline WM-related performance; I_side = 0.5 was used as the criteria to categorize rats into flexible (< 0.5) or rigid (≥ 0.5) performance groups. b–c: Changes of I_TOR (b) and I_side (c) under various drug conditions. Top row from each panel shows the results from rats whose baseline performances showed poor WM-related temporal-order object recognition (b) or cognitive flexibility (c); bottom is for the rats having better temporal-order recognition (b) or cognitive flexibility (c). Bars represent mean ± SD, and gray shades represent means at the baseline when vehicles were administered. Each circle represents an individual rat. * indicates significant difference from vehicle, and # indicates significant difference between two drug conditions

Experimental design

A total of 22 rats were used for this behavioral paradigm, and were habituated to all procedures before the start of the experiments. To ensure that the observed behavioral changes were induced by drugs, there was always a vehicle test day right before a drug test day to serve as the reference (Fig. 2). The vehicle test day used 0.1% ascorbic acid (sc), saline (ip), Nutella (pos), or their combination. The drug test day had one of the first four drug combinations (Fig. 2) administered. The drug or vehicle administration was right after the three-minute habituation phase and 15 min before the first sampling trial of the TOR test. After a drug test day, there were at least five consecutive days in which no substance was given for “washout” prior to reusing the animal (Fig. 2). The test order of the first four drug combinations was randomized for each rat. Once these four drug combinations were tested for a rat, another two drug combinations (i.e., CBG replacing GFX) were tested in a counter balanced order in all rats (Fig. 2). The entirety of this experiment was completed in approximately four months.

Fig. 2.

Experimental design. A within-subject design was employed. To ensure that the observed behavioral changes were induced by drugs, there was always a vehicle test day right before a drug test day to serve as the reference. After a drug test day, there were at least five consecutive days in which no substance was given for “washout” prior to reusing the animal. A total of six unique drug conditions were examined. The first four drug conditions were prioritized and tested in randomized order, followed by the remaining two based on primary and secondary hypotheses. The test order for the last two also was randomized in a counter balanced order among all rats. “+” indicates that the drug was administered; “−” indicates that it was not. A total of 22 rats were used for TOR task, and an additional nine rats were used for DAR task. However, due to the prolonged duration of the experiment, some rats did not complete all tests because of health concerns unrelated to the study. The exact number of rats per test is shown in the two far-right columns

Delayed alternation response (DAR) test

Apparatus

A standard T-maze was used, which had one start runway (56 cm long, 10 cm wide, 18 cm high) and two finish arms (41 cm long, 10 cm wide, 18 cm high). The lower portion of the start arm served as the start box, which could be cordoned off by a solid gate. At the intersection of the maze, the runways to the two finish arms, left or right, also can be cordoned by solid gates. A CCD camera (30 frames/second, STC-TB33USB-AS, SenTech, Carrollton, TX) was mounted directly above the maze to capture animal movement. Video data were collected using the Limelight video recording system (Actimetrics, Coulbourn Instruments, Whitehall, PA). Pre-defined zones and grids were used to quantify behavioral parameters such as decision latency (time spent in the choice zone) and arm selection (grid crossings). The Limelight software automatically computed the duration spent in a zone and recorded the time point at which the animal crossed a grid.

Behavioral task procedure

This is a discrete T-maze alternation task where each trial was composed of two runs (Fig. 3a). Each run began when the rat was released from the start box after the tester raised the gate. The first run was the sample phase in which the runway to one of the two (left or right) finish arms was randomly cordoned by a solid gate. The rat would enter one of the finish arms during this sample phase, and then gently picked up and placed back in the start box for a predetermined five seconds delay (a general temporal scale of WM tasks [60]). After the delay time the rat was released from the start box again, and the second run, the choice phase, began. During the choice phase, neither left or right finish arms was cordoned, and the rat was free to choose either finish arm. There was no visual, odor, or audio cue for the choice. The rat intrinsically tends to explore novel places and therefore should visit the arm that was cordoned in the first run and thereafter not explored during the sample phase. This was reinforced with a hand-fed food reward after the rat had made the turn. On the contrary, the wrong/incorrect choice (i.e., the rat continuously visited the same arm that was not cordoned and thereafter had been explored during the sample phase) led to no reward. The completion of the choice phase was when the rat was gently picked up from one of the finish arms and placed back in the start box. The second trial then began. This was repeated for 10 trials for one test session to be completed. For each run during each trial, rats needed to run from the start box to the intersection and then turn to one of the finish arms in less than two minutes, otherwise the trial was aborted, and the rat was gently picked up and placed back in the start box to restart. The maze was cleaned with 70% EtOH between runs and trials.

Fig. 3.

Comparison of treatment effects on DAR task performance. a: Schematic of DAR task in the T-maze. Note that the sample run cordons off one of the choice arms, while the choice run has both arms accessible. Black arrow indicates correct choice; gray arrow represents incorrect choice. b: Behavioral index used to quantify cognitive flexibility (I_side). Each circle represents one vehicle session. Note that the majority of rats showed no side bias (I_side < 0.5) at the baseline when vehicles were administered. c: Changes of I_side under 2MDHX treatment compared to its vehicle. Each line connects the data from the same individual rat. Data from the only one rat having rigid side bias (I_side ≥ 0.5) was not included

Experimental design

A total of nine rats were used. They were trained once per day and given rest on weekends. To ensure acclimation to experimental handling, they underwent mock sessions in which all procedures were simulated without actual drug administration (e.g., needle insertion without injection). Once they were well trained, the test session was implemented on the next day. First was the vehicle session where 0.1% ascorbic acid (sc), saline (ip), Nutella (os), or their combination was administered; the drug session was then implemented on the next day where one of the first four drug combinations was administered (Fig. 2). The test order of these four drug combinations was randomized for each rat. CBG was not evaluated in this behavioral test because of prolonged training. All drug and vehicle treatments were administered 20 min prior to the task. Following each drug testing session, a minimum five-day “washout” period was implemented before the same animal was tested again. During this interval, rats continued their routine behavioral training sessions without receiving any drug or vehicle treatments. This experiment was finished for each rat in about 1–3 months.

Data analysis

TOR task performance relies on multiple cognitive processes. Among them, WM-related temporal order object recognition was quantified by the index I_TOR=O1/(O1 + O2), where O1 and O2 represent the time a rat interacts with the objects during the choice trial in the retrieval phase. Here, O1 is the earlier-in-sequence object that has been presented during the first sampling trial in the acquisition phase, and O2 is the later-in-sequence object that has been presented during the second sampling trial in the acquisition phase [36]. If neither O1 nor O2 was explored by a rat during the retrieval phase, resulting in an undefined I_TOR value, data from that session were then excluded from the analysis. Besides temporal order recognition, rats exhibited preference to explore objects located on either the left or right side during two sampling trials when WM was less needed. This side bias was quantified by the index , where O_l and O_r represent the average time a rat interacts with the object located on its left or right side, respectively. If a rat did not explore any object, resulting in an undefined I_side value, data from that session were then excluded from the analysis.

Spatial WM-related T-maze alternation was quantified by the percentage of alternated choices a rat made over all completed trials. Besides alternation, rats exhibited preference to explore the left or right arm during the sample phase when rats can only explore one arm without choice and WM was not needed. This side bias was quantified, similarly as for the TOR task, by the index , where T_L and T_R represent the average time a rat takes to turn into the left or right arm in the T-maze, respectively.

No data were excluded from the analyses except undefined values described above; however, due to the prolonged duration of the experiment, some rats did not complete all tests because of health concerns unrelated to this study, resulting in missing data. To account for the unequal data structure, both the Linear Mixed Model and the paired t-test were used to determine whether behavioral performance differed across drug conditions or from the vehicle. A p-value of < 0.05 was considered statistically significant, with multiple comparison corrected by Bonferroni. All analyses were done using SPSS 30 and GraphPad Prism 10. Normality was checked by Kolmogorov-Smirnov test and Shapiro-Wilk test. Data are presented as mean ± standard error (SE) if not mentioned otherwise.

Results

Effects on TOR task

Overall, there was no significant difference when comparing WM-related temporal order object recognition (i.e., I_TOR) amongst different drug conditions (p = 0.825, F_6,216=0.477). It is noteworthy, however, that the spread of I_TOR under vehicle condition was wide, ranging from 0–1, with coefficient of variation as high as 78% (Fig. 1a), indicating substantial but expected variability consistent with prior literature [59]. Lower I_TOR is interpreted as reflecting poor WM-related temporal order recognition. We then sought to determine if our compounds effectively rescued poor performance on rats who had lower I_TOR. Based on the vehicle session, rats were split into either a “superior” (I_TOR> 0.7) or “inferior” (I_TOR ≤ 0.7) performing group. This a priori cutoff is consistent with the literature and our previous work, indicative of superior vs. inferior task performance, well above chance [36, 61]. A k-means clustering analysis further revealed two performance clusters with centroids on either side of I_TOR =0.7. 2MDHX cotreatment with CBG significantly increased I_TOR of rats with poor baseline performance (Figs. 1b and 2MDHX + CBG vs. vehicle = 0.526 ± 0.103 vs. 0.256 ± 0.024; p = 0.036, t₁₄₂= 3.193; d = 1.004, 95% CI [0.383, 1.626]). Interestingly, despite the rescue effect from 2MDHX and CBG coadministration, neither compound significantly increased I_TOR when administered alone (Table 1). On the other hand, for rats with superior baseline performances, 2MDHX cotreatment with CBG significantly decreased I_TOR (Figs. 1b and 2MDHX + CBG vs. vehicle = 0.340 ± 0.310 vs. 0.898 ± 0.014; p = 0.027, t₅₅=3.395; d = − 2.047, 95% CI [− 3.256, − 0.839]); all the other drug conditions did not cause significant changes (Table 2). Together, these differential effects in inferior versus superior performers suggest that treatment outcomes may be dependent on basal performance. Consistent with this finding, significant correlations were observed between baseline performance and 2MDHX- or 2MDHX + GFX-induced changes (Table 3), but not for other drug conditions.

Table 1.

Comparison of treatment effects on rats with lower WM-related TOR task performance (i.e., low I_TOR) at the baseline

	Mean	±	SEM	(n)	p-values
					MPH	GFX + 2MDHX	CBG + 2MDHX	GFX	CBG	2MDHX
MPH	0.500	±	0.114	(11)		> 0.999	> 0.999	> 0.999	> 0.999	> 0.999
GFX+2MDHX	0.471	±	0.089	(15)	> 0.999		> 0.999	> 0.999	> 0.999	> 0.999
CBG+2MDHX	0.526	±	0.103	(12)	> 0.999	> 0.999		> 0.999	> 0.999	> 0.999
GFX	0.489	±	0.110	(12)	> 0.999	> 0.999	> 0.999		> 0.999	> 0.999
CBG	0.524	±	0.150	(9)	> 0.999	> 0.999	> 0.999	> 0.999		> 0.999
2MDHX	0.337	±	0.108	(12)	> 0.999	> 0.999	> 0.999	> 0.999	> 0.999
Vehicle	0.256	±	0.024	(78)	0.206	0.179	0.036	0.191	0.074	1.000

Overall Linear Mixed Model result: p < 0.001, F_6,142 = 4.432. p values of pairwise comparisons are reported in the columns next the mean ± SEM. Statistically significant results are highlighted in bold

Table 2.

Comparison of treatment effects on rats with higher WM-related TOR task performance (i.e., high I_TOR) at the baseline

	Mean	±	SEM	(n)	p-values
					MPH	GFX + 2MDHX	CBG + 2MDHX	GFX	CBG	2MDHX
MPH	0.557	±	0.149	(6)		> 0.999	> 0.999	> 0.999	> 0.999	> 0.999
GFX+2MDHX	0.750	±	0.130	(6)	> 0.999		0.643	> 0.999	> 0.999	> 0.999
CBG+2MDHX	0.340	±	0.310	(3)	> 0.999	0.643		> 0.999	> 0.999	> 0.999
GFX	0.630	±	0.255	(3)	> 0.999	> 0.999	> 0.999		> 0.999	> 0.999
CBG	0.614	±	0.213	(5)	> 0.999	> 0.999	> 0.999	> 0.999		> 0.999
2MDHX	0.578	±	0.162	(6)	> 0.999	> 0.999	> 0.999	> 0.999	> 0.999
Vehicle	0.898	±	0.014	(33)	0.133	> 0.999	0.027	> 0.999	0.617	0.203

Overall Linear Mixed Model result: p = 0.002, F_6,55 = 3.951. p values of pairwise comparisons are reported in the columns next the mean ± SEM. Statistically significant results are highlighted in bold

Table 3.

Correlation of drug effects with baseline performance

	ITOR			Iside
	r S	p	n	r S	p	n
MPH	0.162	0.535	17	0.571	0.007	21
GFX+2MDHX	0.505	0.020	21	-0.023	0.920	21
CBG+2MDHX	− 0.255	0.360	15	0.297	0.218	19
GFX	0.131	0.642	15	0.247	0.268	22
CBG	− 0.053	0.858	14	0.782	< 0.001	18
2MDHX	0.545	0.019	18	− 0.181	0.434	21

Spearman’s correlation coefficients (r_S) and corresponding p values are reported for I_TOR and I_side, respectively, for each drug condition. Statistically significant results are highlighted in bold

Besides temporal order recognition, rats exhibited preference (i.e., I_side) to explore objects located on either the left or right side during two sampling trials when WM was less needed. High I_side indicates strong side bias, implying less cognitive flexibility [20, 62–64], and was observed on many rats in this study (Fig. 1a). We then examined if our compounds effectively improved the flexibility on rats who had higher I_side. Based on the vehicle session, rats were split into either a “flexible” (I_side < 0.5) or “rigid” (I_side ≥ 0.5) performing groups. This cutoff is supported by k-means clustering, which revealed two performance clusters with centroids on either side of I_side =0.5. 2MHDX alone, as well as its coadministration with GFX, significantly decreased I_side of rats who had rigid baseline performance (Figs. 1c and 2MDHX vs. vehicle = 0.541 ± 0.088 vs. 0.800 ± 0.017, p = 0.036, t₁₂₇=3.206, d = − 1.164, 95% CI [− 1.882, − 0.445]; 2MDHX + GFX = 0.549 ± 0.125, p = 0.045, t₁₂₇= 3.117, d = − 1.130, 95% CI [− 1.847, − 0.413]). All the other drug conditions did not cause significant changes (Table 4). Interestingly, for rats with flexible baseline performance, MPH and CBG did not change their I_side (Fig. 1c; Table 5), whereas other drug conditions, i.e., 2MDHX (2MDHX vs. vehicle = 0.577 ± 0.082 vs. 0.231 ± 0.018; p = 0.006, t₁₀₆=3.719; d = 1.658, 95% CI [0.774, 2.543]), GFX (GFX = 0.654 ± 0.060; p < 0.0001, t₁₀₆=7.418; d = 2.039, 95% CI [1.494, 2.584]), and cotreatment of 2MDHX with GFX (2MDHX + GFX = 0.528 ± 0.093; p = 0.024, t₁₀₆=3.275; d = 1.431, 95% CI [0.565, 2.297]) or CBG (2MDHX + CBG = 0.580 ± 0.091; p = 0.009, t₁₀₆=3.592; d = 1.649, 95% CI [0.739, 2.560]), all significantly increased I_side. The I_side after GFX administration was significantly different from the I_side under MPH treatment (MPH = 0.342 ± 0.083; p = 0.011, t₁₀₆=3.534; d = 1.489, 95% CI [0.198, 2.779]). Overall, the divergent responses of rigid versus flexible performers suggest that baseline performance may influence treatment effects. Consistent with this finding, significant correlations were observed between baseline performance and MPH- or CBG-induced changes (Table 3), but not for other drug conditions.

Table 4.

Comparison of treatment effects on rats with lower cognitive flexibility (i.e., high I_side) at the baseline during the TOR task performance

	Mean		±		SEM	(n)	p-values
							MPH	GFX + 2MDHX	CBG + 2MDHX	GFX	CBG	2MDHX
MPH	0.581		±		0.094	(12)		> 0.999	> 0.999	> 0.999	> 0.999	> 0.999
GFX+2MDHX	0.549		±		0.125	(9)	> 0.999		> 0.999	> 0.999	> 0.999	> 0.999
CBG+2MDHX	0.688		±		0.082	(13)	> 0.999	> 0.999		> 0.999	> 0.999	> 0.999
GFX	0.691		±		0.104	(10)	> 0.999	> 0.999	> 0.999		> 0.999	> 0.999
CBG	0.636		±		0.101	(14)	> 0.999	> 0.999	> 0.999	> 0.999		1.000
2MDHX	0.541		±		0.088	(9)	> 0.999	> 0.999	> 0.999	> 0.999	> 0.999
Vehicle	0.800		±		0.017	(67)	0.067	0.045	> 0.999	> 0.999	0.452	0.036

Overall Linear Mixed Model result: p < 0.001, F_6,127 = 4.255. p values of pairwise comparisons are reported in the columns next the mean ± SEM. Statistically significant results are highlighted in bold

Table 5.

Comparison of treatment effects on rats with higher cognitive flexibility (i.e., low I_side) at the baseline during the TOR task performance

	Mean	±	SEM	(n)	p-values
					MPH	GFX + 2MDHX	CBG + 2MDHX	GFX	CBG	2MDHX
MPH	0.342	±	0.083	(9)		0.747	0.809	0.011	> 0.999	0.685
GFX+2MDHX	0.528	±	0.093	(12)	0.747		> 0.999	> 0.999	> 0.999	> 0.999
CBG+2MDHX	0.580	±	0.091	(6)	0.809	> 0.999		> 0.999	0.809	> 0.999
GFX	0.654	±	0.060	(12)	0.011	> 0.999	> 0.999		0.747	> 0.999
CBG	0.293	±	0.185	(4)	> 0.999	> 0.999	0.809	0.747		0.689
2MDHX	0.577	±	0.082	(12)	0.685	> 0.999	> 0.999	> 0.999	0.689
Vehicle	0.231	±	0.018	(58)	0.997	0.024	0.009	< 0.001	> 0.999	0.006

Overall Linear Mixed Model result: p < 0.001, F_6,106 = 15.420. p values of pairwise comparisons are reported in the columns next the mean ± SEM. Statistically significant results are highlighted in bold

Finally, we examined if our compounds affected total exploration time, an indicator of motor function, and found that there was no significant difference amongst different drug conditions (p = 0.823, F_6,247=0.480).

Effects on DAR test

Overall, the alternation rate was not significantly changed under any drug condition (Table 6). For preference to explore the left or right arm during the sample phase when WM was not needed, only one rat showed strong side bias (I_side ≥ 0.5), whereas others were relatively flexible (I_side < 0.5; Fig. 3b). We then examined if our compounds impaired rats’ cognitive flexibility. Only 2MDHX alone, when compared with its vehicle, tended to increase I_side (Fig. 3c and 2MDHX vs. vehicle = 0.221 ± 0.056 vs. 0.100 ± 0.020; p = 0.040 before Bonferroni correction, t₆ = 2.604; d = 0.984, 95% CI [0.059, 1.909]). All other drug conditions did not change I_side (Table 7; overall Linear Mixed Model result: p = 0.409, F_4,45=1.016; p > 0.05 for all pairwise comparisons).

Table 6.

Comparison of treatment effects on spatial WM-related DAR task performance (i.e., alternation rate)

	Mean	±	SEM	(n)	p-value
					MPH	GFX+2MDHX	GFX	2MDHX
MPH	0.595	±	0.028	(6)		> 0.999	> 0.999	> 0.999
GFX+2MDHX	0.569	±	0.080	(7)	> 0.999		> 0.999	> 0.999
GFX	0.629	±	0.044	(8)	> 0.999	> 0.999		> 0.999
2MDHX	0.553	±	0.046	(8)	> 0.999	> 0.999	> 0.999
Vehicle	0.647	±	0.020	(29)	> 0.999	> 0.999	> 0.999	0.542

Overall Linear Mixed Model result: p = 0.257, F_4,53 = 1.368. p values of pairwise comparisons are reported in the columns next the mean ± SEM

Table 7.

Comparison of treatment effects on side bias during the DAR task performance

	Mean	±	SEM	(n)	p-value
					MPH	GFX+2MDHX	GFX	2MDHX
MPH	0.172	±	0.084	(5)		> 0.999	> 0.999	> 0.999
GFX+2MDHX	0.217	±	0.044	(6)	> 0.999		> 0.999	> 0.999
GFX	0.172	±	0.046	(7)	> 0.999	> 0.999		> 0.999
2MDHX	0.250	±	0.057	(7)	> 0.999	> 0.999	> 0.999
Vehicle	0.203	±	0.036	(25)	> 0.999	> 0.999	> 0.999	> 0.999

Overall Linear Mixed Model result: p = 0.409, F_4,45 = 1.016. p values of pairwise comparisons are reported in the columns next the mean ± SEM. Data from the only one rat having rigid side bias (I_side ≥ 0.5) was not included

Discussion

This study sought to determine whether co-administration of an α_2A agonist and a D₁ agonist would improve spatial WM and/or temporal order memory in rats. In addition, we also investigated if selective α_2A agonists and D₁ agonists, either administered alone or together, would be more effective than the non-selective stimulant, MPH.

We found that when the selective full D₁ agonist 2MDHX was administered concomitantly with a selective α_2A agonist (either CBG or GFX), either temporal order memory or cognitive flexibility was improved in rats with deficits. MPH, on the other hand, did not improve temporal order memory, spatial WM, or cognitive flexibility. A trend toward lowering rigidity in cognitive flexibility was observed with MPH, but the effect was not statistically significant. These findings are in general agreement with existing literature that suggests that the effects of MPH on memory and cognitive functions could be task-dependent [65–68]. Our previous study [35] reported similar results, showing that MPH improved spatial WM in some rats performing a continuous T-maze alternation task, but impairing performance in others. The current findings underscore the limitations of using stimulants, such as MPH, to treat ADHD, mirroring observations in patients [69, 70]. While it is reasonable to suggest using a behavioral paradigm where MPH has shown more consistent efficacy, our objective was to use tasks that probe specific cognitive domains often impaired in ADHD but less reliably improved by MPH. This approach allowed us to evaluate whether selective α_2A agonists and D₁ agonists can better address these particular cognitive domains. Our data provide a basis for future clinical studies to evaluate cotreatment of selective agents, such as an α_2A agonist with a D₁ agonist, as an alternate, potentially superior therapeutic option. Future studies will extend this work across a broader range of behavioral paradigms to further characterize the relative strengths and limitations of these treatment strategies.

The benefits of co-treatment, however, were tempered by the detrimental effects on rats that were performing well already. Co-administration of the selective D₁ agonist 2MDHX with the selective α_2A agonist CBG or GFX, not only significantly lowered cognitive flexibility, but also impaired temporal order memory. Similar detrimental effects were observed with monotherapies as well. When either GFX or 2MDHX was administered alone, there was significantly decreased cognitive flexibility. One limitation of these studies was that the drug dose chosen might have been on the descending phase of the inverted-U dose response curves of these rats with superior baseline performance [34, 71]. Additionally, although the compounds used are receptor-selective, off-target interactions cannot be entirely ruled out and could potentially lead to adverse effects [48–52, 72, 73]. However, given the low dose administered and the fact that these adverse effects were observed only in the good-performing group, off-target effects are unlikely to be the primary mechanism. It is interesting to note, however, that when compared to these selective agents, the non-selective drug, MPH, was the only treatment that did not negatively affect cognitive flexibility or temporal order memory in rats with superior baseline performance. These findings support the hypothesis that when brain activities, such as those in the PFC, are operating normally, they are more susceptible to disruption by selective agents, such as D₁ agonists and α_2A agonists [27, 74]. The greater risk of perturbation is likely to lead to a more significant imbalance in signaling, ultimately impairing cognitive functions. This hypothesis merits further investigation.

Our results also suggest that D₁Rs may play a more critical role than α_2ARs in regulating cognitive flexibility. The selective full D₁ agonist 2MDHX alone was capable of lowering rigidity in cognitive flexibility, whereas neither of the α_2A agonists GFX or CBG demonstrated a similar benefit. On the other hand, for the rats who had already performed well, 2MDHX alone was capable of causing impairments on cognitive flexibility, whereas the detrimental effects of the α_2A agonists were more mixed. CBG alone did not affect cognitive flexibility, whereas GFX alone significantly reduced it. One resulting hypothesis is that α_2ARs require optimal D₁Rs function, that is, co-administration of a D₁ agonist with an α_2A agonist can significantly affect temporal order memory, whereas neither an α_2A agonist alone, nor a D₁ agonist alone, produced the same effects. The exact mechanisms underlying these findings require further investigation. One potential explanation is that D₁Rs function to balance the excitatory-inhibitory neural activities in the PFC, thereby fine-tuning the signal-to-noise ratio to achieve this potentially superior effect with α_2ARs [26, 37].

Our results also revealed differences between CBG and GFX. CBG had more profound effects on temporal order memory. Its concomitant administration with a D₁ agonist significantly rescued poor temporal memory, but also caused impairment in rats who had already performed well. These effects were not observed with GFX. Instead, GFX was more effective in modulating cognitive flexibility. Administration of GFX, either alone or in combination with a D₁ agonist, either significantly improved cognitive flexibility in rats with rigidity or caused impairment in those rats who already had good cognitive flexibility. CBG, by contrast, did not produce the same effects, nor were they as pronounced as those observed with GFX. Both CBG and GFX have α_2A agonist activity [51, 52, 72], but CBG also is a 5HT_1A antagonist [52]. Indeed, its pharmacological properties of CBG are still not completely defined, and actions at other receptors mechanisms may help explain the different effects observed of CBG and GFX.

Conclusions

In summary, we found that coadministration of selective agents that are agonists at different receptors, in this case the α_2A adrenoreceptor and the D₁ dopamine receptor, may have the potential to rescue cognitive function. Compared to non-selective stimulants, this approach may offer a potential therapeutic alternative for impairments in PFC-dependent temporal order memory and cognitive flexibility, functions often disrupted in ADHD; although further validation in direct ADHD models is warranted. One caution, however, is that selective agents may pose a greater risk of disrupting brain activities, potentially causing cognitive impairment in subjects who are already well-functioning. Before this idea is translated into human trials, a significant amount of additional research is needed. The complex nature of these experiments made it difficult to do the desired full dose-response studies. Thus, some of the detrimental effects on the measured behaviors might result from overactivation of the relevant circuits. It is also important to note that subgroup sizes in these experiments are relatively small, limiting statistical power and the ability to detect subtle effects, though the within-subject design helps mitigate variability between rats, supporting the robustness of the observed results.

Finally, there is now a centrally active D₁ agonist on the clinical horizon. Tavapadon is a high intrinsic activity partial agonist that is functionally selective with bias for canonical versus β-arrestin signaling that recently completed four Phase-3 trials for Parkinson’s disease. Recent publications of the TEMPO-1 and TEMPO-3 studies reported positive findings for both primary and secondary endpoints [75, 76]. Press releases also have reported positive findings for TEMPO-2. The year-long TEMPO-4 study is not yet peer-reviewed, but recent presentations at scientific meetings have also reported positive results [77]. To our knowledge, the US FDA has never failed to approve a drug candidate that had positive primary and secondary indications and an acceptable side effect profile. Of note, the D₁ agonists that have been used in prior human clinical trials are injectable only agents with relatively short duration. Tavapadon, as a one pill day oral medication, will be a very useful tool for doing more well-controlled human studies (i.e., over longer time-periods) in cognition. Moreover, tavapadon is a partial agonist with significant functional selectivity (i.e., highly biased toward cAMP signaling vs. β-arrestin) [78], thus differing from the prior injectable compounds [33, 79]. We predict it will be very useful step in determining if and how a D₁ agonist affects cognition in both CNS disorders and aging.

Author contributions

Conceptualization: Mi Zhou, Richard Mailman, Yang Yang. Formal analysis and investigation: Luke Bransom, Yang Yang. Writing—original draft preparation: Luke Bransom Writing—review and editing: Luke Bransom, Ava Bassett, Mi Zhou, Richard Mailman, Yang Yang. Funding acquisition: Ava Bassett, Mi Zhou, Richard Mailman, Yang Yang. Supervision: Yang Yang.

Funding

This work was supported by the National Institutes of Health (R01 AG071675, TL1 TR002016, and UL1 TR002014), Children’s Miracle Network Research Grant (2022–2025) and Trainee Grant (2023–2025), Penn State Center for Biodevices Seed Grant (2025–2027), and the Penn State Translational Brain Research Center.

Data availability

All the data supporting the findings of this study are contained within the paper. The raw data will be made available by the authors, without undue reservation.

Declarations

Ethics approval and consent to participate

The animal study was approved by Penn State Institutional Animal Care and Use Committee and conducted in accordance with the local legislation and institutional requirements.

Consent for publication

All authors consent to publish.

Competing interests

RBM has intellectual property related to dopamine D1 agonists and their use that are not actively under development but could be constructed as a conflict-of-interest that has been managed by the Penn State College of Medicine and the University of Virginia. The remaining authors declare no competing financial interests.