If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Donepezil exerts pro-cognitive effects by nonselectively enhancing acetylcholine (ACh) across multiple brain systems. Two brain systems that mediate pro-cognitive effects of attentional control and cognitive flexibility are the prefrontal cortex and the anterior striatum, which have different pharmacokinetic sensitivities to ACh modulation. We speculated that these area-specific ACh profiles lead to distinct optimal dose ranges for donepezil to enhance the cognitive domains of attention and flexible learning.
Methods
To test for dose-specific effects of donepezil on different cognitive domains, we devised a multitask paradigm for nonhuman primates that assessed attention and cognitive flexibility. The nonhuman primates received either vehicle or variable doses of donepezil before task performance. We measured intracerebral donepezil and its strength in preventing the breakdown of ACh within the prefrontal cortex and anterior striatum using solid phase microextraction neurochemistry.
Results
The highest administered donepezil dose improved attention and made the subjects more robust against distractor interference, but it did not improve flexible learning. In contrast, only a lower dose range of donepezil improved flexible learning and reduced perseveration, but without distractor-dependent attentional improvement. Neurochemical measurements confirmed a dose-dependent increase of extracellular donepezil and decreases in choline within the prefrontal cortex and the striatum.
Conclusions
The donepezil dose for maximally improving attention differed from the dose range that enhanced cognitive flexibility despite the availability of the drug in two major brain systems supporting these functions. These results suggest that in our cohort of adult monkeys, donepezil traded improvements in attention for improvements in cognitive flexibility at a given dose range.
The acetylcholinesterase (AChE) inhibitor donepezil is one of few Food and Drug Administration–approved cognitive enhancers that aims to address a wide range of cognitive deficits in subjects with mild cognitive impairment or dementia (
). Basic research suggests that the cognitive domains that can be enhanced with AChE inhibitors include selective attention, working memory, response inhibition, learning, and long-term memory (
). Consistent with these reports, clinical studies assessing donepezil at one or two doses across larger cohorts of subjects with varying stages of Alzheimer’s disease have found improvements in compound scores of cognitive testing batteries (
Meta-analysis of randomized controlled trials on the efficacy and safety of donepezil, galantamine, rivastigmine, and memantine for the treatment of Alzheimer’s disease.
). It is, however, not clear whether the standard doses of donepezil used in clinical studies improve multiple cognitive domains directly or whether, at a particular effective dose, its major route of action is to enhance arousal, which then provides an indirect, overall cognitive advantage for attention, working memory, learning, and memory processes (
). Assessing whether donepezil affects multiple cognitive domains simultaneously at a given dose is important for evaluating its therapeutic efficiency and to identify cognitive domains that should be targeted in drug discovery efforts for improved future cognitive enhancers.
One potential limitation of donepezil and other AChE inhibitors is that they increase acetylcholine (ACh) concentrations nonselectively across multiple brain systems. Such a nonselective ACh increase has shortcomings when brain systems are differently sensitive to ACh action; the donepezil dose that optimally affects one brain system might over- or understimulate another brain system. In primates, muscarinic ACh subreceptors relevant for attention and memory functions (
), suggesting that the PFC may be more sensitive to modulation by AChE inhibitors than posterior brain areas. Moreover, a comparison of transcription factor (CREB [cAMP-response element binding protein]) activation of the PFC and the striatum to muscarinic modulation by xanomeline has reported a 10-fold higher receptor sensitivity of the striatum (
Striatal, hippocampal, and cortical networks are differentially responsive to the M4- and M1-muscarinic acetylcholine receptor mediated effects of xanomeline [published correction appears in ACS Chem Neurosci 2019; 10:3910].
), consistent with other studies reporting significantly higher muscarinic binding potential and higher AChE activity in the striatum than in other cortical regions (
Effects of acute acetylcholinesterase inhibition on the cerebral cholinergic neuronal system and cognitive function: Functional imaging of the conscious monkey brain using animal PET in combination with microdialysis.
Difficulty of visual search modulates neuronal interactions and response variability in the frontal eye field [published correction appears in J Neurophysiol 2008; 99:1050].
). One consequence of the brain area–specific sensitivity to ACh levels could be that a best dose for enhancing cognitive functions supported by the striatum might not sufficiently stimulate the PFC, and that a best dose for enhancing PFC functions might overstimulate the striatum.
To test for these possible implications of brain region–specific ACh action, we devised a drug testing paradigm for monkeys that assessed the effects of three different doses of donepezil across different domains of arousal, attention, and cognitive flexibility in a single testing session. We evaluated the attention domain with a visual search (VS) task that varied the number and perceptual similarity of distracting objects and quantified the domain of cognitive flexibility with a learning task asking monkeys to flexibly adapt to new feature-reward rules and avoid perseverative responding. This assessment paradigm goes beyond existing nonhuman primate studies of donepezil that so far have found enhanced short-term memory using delayed match-to-sample tasks (
Meta-analysis of randomized controlled trials on the efficacy and safety of donepezil, galantamine, rivastigmine, and memantine for the treatment of Alzheimer’s disease.
Positive allosteric modulator of α7 nicotinic-acetylcholine receptors, PNU-120596 augments the effects of donepezil on learning and memory in aged rodents and non-human primates.
The scopolamine-reversal paradigm in rats and monkeys: The importance of computer-assisted operant-conditioning memory tasks for screening drug candidates.
Effects of acute acetylcholinesterase inhibition on the cerebral cholinergic neuronal system and cognitive function: Functional imaging of the conscious monkey brain using animal PET in combination with microdialysis.
) (surveyed in Table S1) and takes into account that studies in rodents report positive donepezil effects across a wider range of domains, including reversal learning (
A computer-automated touchscreen paired-associates learning (PAL) task for mice: Impairments following administration of scopolamine or dicyclomine and improvements following donepezil.
Dissociable effects of acetylcholinesterase inhibitors and phosphodiesterase type 5 inhibitors on object recognition memory: Acquisition versus consolidation.
Chromaproline and chromaperidine, nicotine agonists, and donepezil, cholinesterase inhibitor, enhance performance of memory tasks in ovariectomized rats.
) (surveyed in Table S2). With our design, we found that donepezil improves interference control over distractors at doses that caused an overall slower response (i.e., reduced speed of processing) and peripheral side effects. In contrast, a lower dose of donepezil caused no clear attentional effect but improved cognitive flexibility. These findings document domain-specific dose-response effects of donepezil for attention and cognitive flexibility.
Methods and Materials
Nonhuman Primate Testing Protocol
Three adult male rhesus macaques were separately given access to a cage-mounted Kiosk Station that provided a touchscreen interface inside the animal’s housing unit to perform cognitive tasks (Figure 1A) (
USE: An integrative suite for temporally-precise psychophysical experiments in virtual environments for human, nonhuman, and artificially intelligent agents.
Figure 1Task design, metastructure, and VS performance as a function of distractor number. (A) (i) Picture of one of the subjects working in the custom-built kiosk, interacting with the touchscreen and receiving fluid reward. (ii) The metastructure of the multitask. Each experimental session consists of three superblocks of VS, FL, and VS, in turn. Each VS block is preceded by 10 familiarization trials, which are identical to a VS trial but without any distractors. Each VS block contains trials with 3, 6, 9, or 12 distractors randomly selected and counterbalanced over the block. In contrast, each FL block contains 0 or 1 irrelevant feature dimensions in addition to the relevant feature dimension (the dimension with the rewarded feature value) counterbalanced over the session. (B) (i) From the grand pool of quaddles, which includes four feature dimensions and a variable number of feature values (9 shapes, 9 patterns, 8 colors, and 11 arms), three feature values from three feature dimensions are chosen. This 3 × 3 pool is then counterbalanced for dimension presentation and feature-reward association and is used for 2 weeks of data collection where all presented quaddles are selected from this 3 × 3 pool. (ii) Example trials. Two example VS trials (top) within the same block with 3 distractors (left) and 9 distractors (right). Each VS block will contain one of five backgrounds, with the VS blocks in the same day never having the same background. All distractors and target objects in VS blocks are three-dimensional objects, and distractors may be duplicated in each trial. Quaddles are spatially randomly presented at the intersections of a 5 × 4 virtual grid pattern on screen. The red box highlights the rewarded target object, which is invariable within the VS block, in these examples. Two example FL trials (bottom) within the same block containing 2D quaddles (1 distracting dimension plus the relevant dimension). The rewarded feature value in this block is the checkered pattern independent of what color feature value it is paired with. Quaddles may be presented in eight possible locations in a circle, each being 17° of visual radius away from the center of the screen. The red box signifies the rewarded target object, which is a variable combination of the rewarded feature value (the checkered pattern in this example) with a random feature value of the distractor dimension (color in this example). (C) The trial structure for both the FL (top) and VS (bottom) blocks of the task are very similar. A trial is initiated by a 0.3 to 0.5 s touch and hold of a blue square (3° visual radius wide), after which the blue square disappears for 0.3 to 0.5 s before task objects, which are 2.5° visual radius wide, are presented on screen. Once the task objects are on screen, the subject is given 5 s to visually explore and select an object via a 0.2 s touch and hold. A failure to make a choice within the allotted 5 s results in an aborted trial and does not count toward the trial count. Brief auditory feedback and visual feedback (a halo around the selected object) are provided on object selection, with any earned fluid reward being provided 0.2 s after object selection and lasting 0.5 s along with the visual feedback. Nonrewarded trials had a different auditory tone and a light blue halo around the selected, nonrewarded object. Rewarded objects had a higher pitch auditory tone, a light yellow halo around the selected rewarded object, and an accompanying fluid (water) reward. (D) Average VS performance by distractor number for the vehicle and all donepezil doses combined, both separated by the first vs. second VS block. VS performance was significantly different for block number (F1,1722 = 22.19, p < .001) and condition (F1,1722 = 19.0, p < .001). The inlay shows individual monkey average VS performance linear fits. (E) Average VS performance by distractor number between vehicle and 0.06, 0.1, and 0.3 mg/kg donepezil doses for the first VS block (F3,896 = 10.77, p < .001). Both the 0.06- and 0.3-mg/kg doses were significantly different from vehicle (Tukey’s, p = .005 and p < .001, respectively). Error bars here reflect standard deviation in this panel. (F) The set size effect of VS performance by distractor number for each condition. The 0.3-mg/kg dose set size effect was significantly shallower than the vehicle set size effect (H3 = 11.46, p = .010; Tukey’s, p = .013). FL, feature-reward learning; n.s., not significant; Num, number; Perf, performance; Prop, proportion; VS, visual search.
We used donepezil (Sigma-Aldrich, catalog number D6821) in three doses, 0.06, 0.1, and 0.3 mg/kg, to operate within the dosing range of previous studies reporting pro-cognitive effects (surveyed in Tables S1 and S2). At this intramuscular (IM) range, plasma concentrations of donepezil are roughly the same when dosing with ∼10x the concentration orally (
). Animals received saline as vehicle control or a dose of donepezil IM injection 30 minutes before starting task performance, taking into account its expected 1-hour half-life (
Estimation of plasma IC50 of donepezil hydrochloride for brain acetylcholinesterase inhibition in monkey using N-[11C] methylpiperidin-4-yl acetate ([11C]MP4A) and PET.
). Administration was double blinded. Drug side effects were assessed 15 minutes after drug administration and after completion of the behavioral performance with a modified Irwin Scale (
) for rating autonomic nervous system (e.g., salivation) and somatomotor system functioning (e.g., posture, unrest). Monkey behavioral status was video-monitored throughout task performance (Figure 1A).
Behavioral Paradigms
In each experimental session, the monkeys performed a VS task to measure attentional performance metrics and a feature-reward learning (FL) task to measure cognitive flexibility metrics (
). Each performance day was made up of an initial set of 100 trials of VS, a set of 21 learning blocks with 35 to 60 trials each of the FL task, and a second set of 100 trials of the VS task (Figure 1Aii). Details of both tasks are provided in the Supplement. The VS task required the monkeys to find and touch a target object among 3, 6, 9, or 12 distracting objects to receive fluid reward (Figure 1B). The target was an object shown in 10 initial trials without distractors. Targets and distractors were multidimensional, 3D-rendered quaddle objects (
USE: An integrative suite for temporally-precise psychophysical experiments in virtual environments for human, nonhuman, and artificially intelligent agents.
) that shared few or many features of different feature dimensions (colors, shapes, arms, body patterns), which rendered the search easier when there were no or few similarities among the features of targets and distractors or more difficult if the target-distractor (T-D) similarity was high (Figure 2A). The FL task required the monkeys to learn through trial and error which object feature was rewarded in blocks of ∼35 to 60 trials (Figure 1C). The rewarded feature changed uncued and switched to a new feature of the same or different feature dimensions, which made the task similar to conceptual set-shifting tasks [e.g., (
)] but different by using a larger set of features that varied within and across sessions to vary task difficulty. In each trial, three objects were shown; they varied either in the features of one feature dimension (e.g., having different colors or body shapes) or in the features of two feature dimensions (e.g., having different colors and body shapes). Choosing the object with the correct feature was rewarded with a probability of 0.8. Blocks in which only one feature dimension varied (1D blocks) were easier because there was lower attentional load than in blocks with two varying feature dimensions (2D blocks).
Figure 2VS task performance and change in difficulty through increasing distractor numbers and T-D similarity. (A) A visual description of the T-D similarity measure in the VS task. For an example target in the red square, three example distractors are presented with 0, 1, and 2 features in common, respectively, from left to right. The cartoon plot below shows the impact of the average T-D similarity of an individual trial on performance. (B) Similar to Figure 1D, but here we plot average VS performance by T-D similarity. There was a significant effect of T-D similarity on performance (F2,627 = 16.17, p < .001) and condition, with both the 0.06 and 0.3 mg/kg donepezil doses being significantly different from vehicle (F3,267 = 7.75, p < .001; Tukey’s p = .034 and p < .001, respectively). (C) The change in the slope of VS performance with 0.06, 0.1, and 0.3 mg/kg donepezil relative to vehicle. The change in slope by distractor number is plotted on the left y-axis (same data as Figure 1F) (H3 = 11.46, p = .010) while the change in slope by T-D similarity is plotted on the right y-axis (H3 = 2.8, n.s.). (D) A visualization of the combined effect of distractor number and T-D similarity on performance. From left to right, each cluster of lines represents increasing distractor numbers while data within each line represents low, medium, and high T-D similarity from left to right. Both distractor number (F3,2615 = 28.85, p < .001) and T-D similarity (F2,2615 = 64.59, p < .001) impact VS performance with no significant interaction (F6,2615 = 0.69, n.s.). dist, distractor; Feat, feature; n.s., not significant; Num, number; Perf, performance; Prop, proportion; sim, similarity; T-D, target-distractor; VS, visual search.
To evaluate the levels of donepezil in brain structures that are necessary for successful attention and learning performance, we measured the ACh metabolite choline and donepezil concentrations in the PFC and the anterior striatum (caudate nucleus) 15 minutes after administering low and high doses of donepezil (0.06 and 0.3 mg/kg, respectively, IM) in separate experiments. Measures of donepezil were made when we observed dose-limiting side effects at the 0.3-mg/kg dose, and the two tested doses were accompanied by pro-cognitive effects in our task (see results). We used microprobes that sampled the local neurochemical milieu with the principles of solid phase microextraction (SPME) (
Data were analyzed with standard nonparametric and parametric tests as described in the Supplement.
Results
Each monkey was assessed in 38 sessions in total, including 17 vehicle days and 7 days with each dose (0.06, 0.1, and 0.3 mg/kg). Drug side effects were noted after IM injections of the 0.3-mg/kg dose in the first 30 minutes after injection as changes in posture, sedation, vasoconstriction, and paleness of skin, but no adverse effects persisted beyond 30 minutes (Table S3). First, we confirmed that the monkeys performed the VS task with a high 84.4% (±0.54%) accuracy (monkeys Ig: 85.2% ± 0.81%; Wo: 88.3% ± 0.94%; Si: 79.8% ± 0.97%) and showed the expected set size effect evident in decreased accuracy and slower reaction times with increasing numbers of distractors (Figure 1D and Figures S1 and S2). When the targets were more similar to distractors (high T-D similarity), VS performance decreased from 92.9% (±0.4%) to 85.5% (±0.3%) and 81.6% (±1.0%) for low, medium, and high T-D similarity, respectively (H[2] = 169.48, p < .001) (Figure 2B). In the feature-reward learning task, the monkeys reached the learning criterion faster in the easier 1D (low distractor load) condition (average trials to ≥80% criterion: 12.5 ± 0.2 SE) than in the 2D (high distractor load) condition (average trials to ≥80% criterion: 15.6 ± 0.2) (Figure 3A and Supplement).
Figure 3Feature-reward learning task learning curves and performance. (A) Average learning curves of each monkey and all monkeys combined for both low and high distractor load conditions. In all instances, monkeys learned faster and with higher plateau performance in low distractor load blocks relative to high distractor load blocks. (B) All-monkey average learning curves for vehicle, 0.06-, 0.1-, and 0.3-mg/kg donepezil doses for both low and high distractor load conditions. (C) Temporal progression of learning speed (LP) for vehicle, 0.06-, 0.1-, and 0.3-mg/kg donepezil doses for the low distractor load condition only. At the 0.06-mg/kg dose, donepezil allows for faster learning in the low attentional load blocks (F3,602 = 3.3, p = .020). Similar to the visual search task, donepezil’s enhancement is only visible early on and relatively close to its intramuscular administration. (D) Average learning speed of vehicle and donepezil doses for low and high distractor load blocks across sessions reveals an interaction between drug condition and distractor load (F3,1052 = 3.59, p = .013). (E) Same as (D) but for choice RTs instead of learning speed. The 0.3-mg/kg donepezil dose slows choice RTs in both low and high distractor load blocks (condition F3,1052 = 12.3, p < .001; Tukey’s, p < .001). (F) Change in the length of perseverative errors from vehicle, where feature values in the distracting dimension were the target of the perseverations. Error bars reflect SEM for intermonkey variability. Donepezil at the 0.06-mg/kg dose significantly reduces perseveration length in the distracting dimension (p = .021); other donepezil doses trends toward this as well. avg, average; LP, learning point; persv, perseverative; Prop, proportion; RT, reaction time.
Dose-Dependent Improvement of VS Accuracy and Slowing of Choice Reaction Times
Donepezil significantly improved the accuracy of the VS task (F1,1722 = 18.95, p < .001) (Figure 1D) but on average slowed search reaction times (F1,1722 = 4.83, p = .028) (Figure S1B). The slower choice reaction times were evident already for the single target object in the 10 target familiarization trials (Figure S1A). These main behavioral drug effects were evident prominently in the first VS block (Figure 1D and Figure S1A). We therefore focused our further analysis on the first search block.
The improved accuracy of VS was dose dependent (F3,896 = 10.77, p < .001). The 0.06-mg/kg dose enhanced performance by 2.5% ± 1.0%, 4.4% ± 1.3%, 6.1% ± 1.4%, and 6.3% ± 1.6% (mean ± SD) for 3, 6, 9, and 12 distractor trials, respectively (Tukey’s, p = .005). The 0.3-mg/kg dose enhanced performance by 2.7% ± 1.0%, 6.3% ± 1.2%, 8.5% ± 1.3%, and 11.0% ± 1.4% (mean ± SD) for 3, 6, 9, and 12 distractor trials, respectively (Tukey’s, p < .001) (Figure 1E). Thus, we found larger improvement when more distractors interfered with the target search. We confirmed this by fitting a regression line across performance at different numbers of distractors, which revealed overall, significantly shallower slopes with donepezil (slopes: −0.013 ± 0.001, −0.009 ± 0.002, −0.015 ± 0.003, and −0.005 ± 0.002 for vehicle, 0.06, 0.1, and 0.3 mg/kg of donepezil, respectively [H3 = 11.46, p = .01]). Pairwise comparison showed that the 0.3-mg/kg drug dose and the vehicle condition showed significantly different slopes (Tukey’s, p = .013) (Figure 1F).
In contrast to improving VS accuracy, donepezil slowed down reaction times across all distractor conditions at the 0.3-mg/kg dose relative to vehicle by on average 100 ± 40 ms, 238 ± 79 ms, 208 ± 99 ms, and 264 ± 102 ms (mean ± SD) for 3, 6, 9, and 12 distractors, respectively (F3,896 = 15.15, p < .001; Tukey’s, p < .001) (Figure S1C). The slope of the regression over different numbers of distractors did not differ between the 0.3-mg/kg dose and vehicle, which showed that reaction time effect was a nonselective effect that was independent of distractors (regression slope on reaction times: 0.061 ± 0.002, 0.065 ± 0.007, 0.067 ± 0.007, and 0.076 ± 0.009 [H(3) = 3.37, not significant (n.s.)] for vehicle, 0.06, 0.1, and 0.3 mg/kg of donepezil, respectively) (Figure S1D).
Across sessions, VS accuracy was correlated with reaction times only for the vehicle (Pearson, ρ: −0.30, p < .001) and 0.1-mg/kg donepezil dose condition (Pearson, ρ: −0.46, p = .034) but not for the 0.06- and 0.3-mg/kg dose conditions in which monkeys showed improved accuracy, which suggests that the accuracy improvement is independent from a slowing of reaction speed (Figure S2A, B).
We next tested whether improved interference control over increasing number of distractor objects was likewise evident when increasing the similarity of distractor and target features (Figure 2A). First, we confirmed that higher T-D similarity overall reduced performance (F2,672 = 16.17, p < .001) (Supplement). Donepezil significantly counteracted this similarity effect and improved performance at the 0.06- and 0.3-mg/kg doses (F3,672 = 7.75, p < .001; Tukey’s, p = .034 and p < .001, respectively). This finding shows that the beneficial effect of donepezil significantly increased when there was higher demand to control perceptual interference from distracting objects (Figure 2B). This was also evident as a statistical trend of a shallower regression slope at 0.06- and 0.3-mg/kg doses of donepezil, which indicates less interference from distracting features when they were similar to the target (Figure 2C) (H3 = 2.79, n.s.; slope changes relative to vehicle for 0.06-, 0.1-, and 0.3-mg/kg doses were 0.0357 ± 0.0236, −0.0289 ± 0.0334, and −0.0656 ± 0.0197, respectively). The improved search performance with donepezil for VS with higher T-D similarity and with a higher number of distractors was evident in significant main effects, but there was no interaction, suggesting that they improved performance independent of each other (F2,2615 = 64.59, p < .001; F3,2615 = 28.85, p < .001; and F6,2615 = 0.69, n.s., respectively) (Figure 2D). This independence was also suggested by the absence of a correlation of the T-D similarity effect and the number-of-distractor effect (Pearson, n.s.) (Figure S3).
Dose-Dependent Improvement of Flexible Learning Performance
Donepezil also improved feature-reward learning performance but only at the 0.06-mg/kg dose (Figure 3B), and it was most pronounced for the first third of the behavioral session (F3,602 = 3.3, p = .020) (Figure 3C). We therefore focused further analysis on the first third of the learning blocks, which revealed that the learning improvement at the 0.06-mg/kg dose was significant for the low distractor load condition (significant interaction effect of drug condition and distractor load [condition × distractor load]: F3,1052 = 3.59, p = .013); for vehicle, 0.06-, 0.1-, and 0.3-mg/kg donepezil doses, the number of trials to criterion were 11.3 ± 0.4, 7.7 ± 0.9, 12.3 ± 1.3, and 11.0 ± 1.2, respectively, with the 0.06-mg/kg dose and the vehicle being significantly different (p = .020, Bonferroni correction) (Figure 3D). There was no change in the learning speed with other doses at low or high distractor load.
Beyond learning speed, we found overall slower choice reaction times at the 0.3 mg/kg donepezil dose (Figure 3E) (main effect of drug condition: F3,1052 = 12.29, p < .001). While reaction times were overall slower at the high distractor load (F1,1052 = 7.18, p = .008), there was no interaction with drug dose (F3,1052 = 0.26, n.s.). After visually inspecting the results, we separately tested the 0.3-mg/kg dose of donepezil and found that it led to significantly slower choice reaction time than the vehicle (Tukey’s, p < .001) (Figure 3E). The changes in choice reaction times did not correlate with changes in learning performance (number of trials to criterion) at any drug condition, indicating that they were independently modulated (Pearson, all n.s.) (Figure S2D).
We predicted that the faster learning at the 0.06 mg/kg donepezil dose could be due to a more efficient exploration of objects during learning, which would be reflected in reduced perseverative choices of unrewarded objects. Overall, perseverative errors (defined as consecutive unrewarded choices of objects with the same feature dimension) made up 20% of all errors. As expected, we found significantly shorter sequences of perseveration of choosing objects within distractor feature dimensions at the 0.06-mg/kg dose of donepezil (Figure 3F). For vehicle, 0.06-, 0.1-, and 0.3-mg/kg doses, the average length of perseverations in the distractor dimension was 2.1 ± 0.1, 1.8 ± 0, 1.9 ± 0.1, and 1.9 ± 0.1 trials, with the difference between the vehicle and the 0.06 dose being significant (p = .021). Perseverative choices in the target feature dimension were not different between conditions (for 0.06-, 0.1-, and 0.3-mg/kg donepezil doses, the average perseveration length in the target dimension was 1.7 ± 0, 1.7 ± 0, 1.6 ± 0, and 1.7 ± 0 trials [n.s.]).
Dissociation of Attention and Learning Improvements, but Slowing Is Correlated
The effects of donepezil on feature-reward learning and VS might be related, but we found that learning speed and search accuracy were not correlated at the doses at which the drug improved learning and search (0.06-mg/kg dose) or improved only VS (0.3-mg/kg dose) (Pearson, all n.s.). A significant correlation was found only for the 0.1-mg/kg dose (Pearson, ρ: −0.54; p = .012) (Figure 4A). Learning at low or high distractor load and the set size (slope) effects in the VS task were uncorrelated (Pearson, all n.s.). However, at the 0.3-mg/kg donepezil dose, the T-D similarity effect (i.e., the search-slope change) in the VS task positively correlated with the difference of learning speed at high versus low distractor load in the learning task (Pearson, ρ: 0.60; p = .008). This effect signifies that better attentional search of a target among similar distractors is associated with poorer flexible learning of new targets when there are multiple object features to search through (high distractor load).
Figure 4The relationship between the VS task and the FL task. (A) Correlation coefficients between FL learning speed and VS performance for vehicle, 0.06-, 0.1-, and 0.3-mg/kg donepezil doses. Only the 0.1-mg/kg donepezil dose had a significant correlation between FL and VS task performance (Pearson, ρ: −0.54; p = .012). No doses showed a significant change in correlation from vehicle. (B) Same as (A) but for FL choice RTs and VS search reaction times. Although vehicle, 0.1, and 0.3 mg/kg donepezil doses had a significant correlation between choice and search reaction times, we found no significant change in correlation relative to vehicle. Corr, correlation; FL, feature-reward learning; n.s., not significant; RT, reaction time; VS, visual search.
In contrast to accuracy, choice reaction times in the learning task and VS were significantly correlated for the 0.1-mg/kg donepezil dose (Pearson, ρ: 0.52; p = .016), the 0.3-mg/kg dose (Pearson, ρ: 0.66; p = .002), and the vehicle control condition (Pearson, ρ: 0.60; p < .001) (Figure 4B).
Determination of Extracellular Donepezil and Choline Levels in the PFC and Anterior Striatum
VS and flexible learning are realized by partly independent brain systems, including the PFC and anterior striatum (
). To determine whether extracellular levels of donepezil were increased to a similar magnitude in the PFC and anterior striatum, we measured its concentration after administering doses of either 0.06 or 0.3 mg/kg donepezil IM in the PFC, assumed to be necessary for efficient interference control during VS (
Difficulty of visual search modulates neuronal interactions and response variability in the frontal eye field [published correction appears in J Neurophysiol 2008; 99:1050].
). We found that donepezil was available in both brain areas, and its extracellular concentration more than doubled after injecting 0.3 mg/kg compared with 0.06 mg/kg in both areas (F1,16 = 9.69, p = .007), with no significant difference between PFC and caudate (F1,16 = 1.44, n.s.) (Figure 5A). Donepezil should cause a depletion of the ACh metabolite choline (
). Using high-performance liquid chromatography–mass spectrometry analysis of the SPME samples, we found that 0.06 and 0.3 mg/kg donepezil reduced choline concentrations by 74.2% ± 14.9% (p = .005) and 85.7% ± 26.9% (p = .007) of their baseline concentrations in the PFC and, by 68.4% ± 13.8% (p = .022) and 81.0% ± 12.9% (p = .009) of respective baseline concentrations in the caudate (Figure 5B). The 11.5% and 12.6% stronger reduction of choline at the 0.3- versus 0.06-mg/kg dose in the PFC or caudate was not significant (n.s.).
Figure 5In vivo extracellular measurements of choline and donepezil, as well as donepezil’s effect on heart rate. (A) Quantified concentration of extracellular unbound donepezil with 0.06- and 0.3-mg/kg donepezil administration in the PFC and CD. We were able to reliably detect higher donepezil concentrations with 0.3 mg/kg dosing relative to 0.06 mg/kg dosing (condition F1,16 = 9.69, p = .007) with solid phase microextraction. We also saw a trend toward higher detectable donepezil in the CD relative to the PFC at the 0.3-mg/kg dose tested; however, we found neither significant group nor interaction effects. (B) We used choline concentrations as a metric for donepezil bioactivity because it deactivates acetylcholinesterase and prevents acetylcholine’s degradation into choline. We extracted average sessionwise change in choline from baseline with 0.06 and 0.3 mg/kg donepezil doses within the PFC and CD. Although we found significant decreases in choline by up to >80% of baseline concentrations, we found no significant effect of dosing in either the PFC or CD. (C) The HR of our fourth monkey was monitored during the neurochemical experiments. This revealed a sharp and transient increase in HR after administration of donepezil at the 0.3-mg/kg dose (Supplement), which led to a higher average beats per minute (bpm). We found that we can significantly distinguish 0.06 and 0.3 donepezil administration via HR (p = .006). CD, caudate; HR, heart rate; n.s., not significant; PFC, prefrontal cortex.
To obtain an independent physiological marker of dose-dependent effects, we quantified how, during actual task performance, donepezil changed the heart rate before versus after drug administration (Supplement). The heart rate showed a transient peak ∼20 minutes after the donepezil injection relative to baseline, which was significant for the 0.3-mg/kg dose (preinjection 102.3 ± 7.1 to postinjection 121.6 ± 2.6; p = .021) but not for the 0.06-mg/kg dose (preinjection: 90.3 ± 4.2 to postinjection: 94.8 ± 5.4; n.s.). The 0.3-mg/kg dose caused a significantly higher heart rate peak than the 0.06-mg/kg dose (p = .006) (Figure 5C).
Discussion
Here, we dissociated donepezil’s improvement of attentional control of interference during VS from improvements of cognitive flexibility during FL. At the highest dose tested, donepezil reduced interference during VS, particularly when there were many distractors and a high similarity of distractors with the target, while concomitantly slowing down overall reaction times and inducing temporary peripheral side effects. In contrast, the lowest dose of donepezil did not affect target detection times during VS but improved adapting to new feature-reward rules and reduced perseverative responding. These findings document a dose-dependent dissociation of the best dose of donepezil for improving attention and cognitive flexibility.
Different Donepezil Dose Ranges for Improving Interference Control and Flexible Learning
Using a behavioral assessment paradigm with two tasks allowed us to discern the differences between the donepezil dose that maximally improved interference control (in the VS task) versus the dose that maximally improved flexible learning (in the reward learning task). In both tasks, donepezil modulated performance early within the session (first of two VS blocks and the first third of the learning blocks) consistent with its short half-life and rapid time to peak concentration with IM delivery (
Estimation of plasma IC50 of donepezil hydrochloride for brain acetylcholinesterase inhibition in monkey using N-[11C] methylpiperidin-4-yl acetate ([11C]MP4A) and PET.
). Our results focused therefore on these early time windows. We do not expect different conclusions if we had altered the task sequence (see Supplemental Discussion). At the 0.06-mg/kg dose, donepezil facilitated flexible learning of a new feature-reward rule and reduced the length of perseverative errors (Figure 3C, F). These behavioral effects are indicators of improved cognitive flexibility across reward learning and set-shifting tasks (
Lesions of the medial striatum in monkeys produce perseverative impairments during reversal learning similar to those produced by lesions of the orbitofrontal cortex.
). At the same 0.06-mg/kg dose, VS response times were unaffected (Figure S1) and VS accuracy was overall improved but independent of the number of distractors, i.e., independent of the degree of interference (Figure 1E, F). In contrast, at the higher donepezil doses, flexible learning behavior was indistinguishable from the no-drug vehicle control condition, showing that improving flexibility required donepezil at a lower dose.
This conclusion is opposite to that of the drug effects on VS performance, which was maximally improved at the 0.3-mg/kg dose. At this dose, subjects not only showed improved resistance to interference when there were more distracting objects (Figure 1E, F) but also improved resistance to distracting objects that were visually similar to the searched-for target (Figure 2B–D). These findings document that donepezil enhances robustness to distraction (
), which critically extends insights from existing primate studies with donepezil that mostly used simpler tasks to infer pro-cognitive effects on working memory or arousal (Table S1). The process of attentional control of interference also goes beyond a short-term memory effect measured with delayed match-to-sample tasks. In the VS paradigm we used, short-term memory of the target object is already necessary for performing the easier trials with 3 or 6 distractors, while an attention-specific effect can be inferred when there is greater improvement in performance with increased attentional demands in trials with 9 or 12 distractors. Thus, our study provides strong evidence that donepezil causes specific attentional improvement at higher doses, which supports the neurogenetic model of cholinergic modulation of attention (
) that recently has received functional support in studies reporting enhanced distractor suppression in nonhuman primates with nicotine receptor–specific ACh modulation (
A computational psychiatry approach identifies how alpha-2A noradrenergic agonist guanfacine affects feature-based reinforcement learning in the macaque.
). We should note, however, that at the high dose, donepezil already caused a nonselective slowing of reaction times indicative of peripheral side effects (see Supplemental Discussion).
The finding that different dose ranges improved flexible learning and VS distractor filtering suggests that these processes have partially independent Yerkes-Dodson style inverted-U dose-response curves (Figure 6). One reason supporting this suggestion is that flexible learning and distractor filtering are supported by partially different brain networks, which likely have differential sensitivity to cholinergic modulation. Lesion studies in nonhuman primates have shown that flexible reward learning is closely associated with the medial and orbitofrontal PFC and the striatum where lesions impair learning visual reward associations (
Lesions of the medial striatum in monkeys produce perseverative impairments during reversal learning similar to those produced by lesions of the orbitofrontal cortex.
). In contrast, VS distractor filtering in primates depends on the dorsolateral PFC (dlPFC) and its connections with posterior parietal cortices, with bilateral dlPFC lesions impairing filtering distraction (
). Brain areas within these partly segregated networks for learning and distractor filtering might be differently sensitive to cholinergic modulation. For example, primate dlPFC has been documented to be uniquely sensitive to neuromodulation by catecholamines and ACh for spatial working memory and switching between distracting features (
). Such inverted-U curves for different receptors are not likely to be homogenous or fully overlapping when taking into consideration variable task demands within a cognitive domain or when considering different cognitive domains entirely (
). Our results may thus reflect different inverted-U curves along a construct of flexible attention shifting, required for optimal performance in our feature-reward learning task, and stable filtering of distractors required for optimal performance in our VS task (Figure 6).
Figure 6Each task will have its own specific demands based on the cognitive domain(s) involved that may be best met with some cholinergic tone, which may be endogenously, or in this case pharmacologically, shifted to reach optimal performance. Here, the feature-reward learning task and visual search task have varying demands in terms of attentional flexibility and therefore different inverted-U curves for optimal performance. These curves may be shifted by changing the attentional flexibility or a given subject may travel along the x-axis due to pharmacological intervention, aging-related changes, and other mechanisms that may change their effective cholinergic tone.
Quantifying Extracellular Levels of Donepezil and Choline in PFC and Striatum
We confirmed the presence of extracellular donepezil in the PFC and anterior striatum at the doses tested (Figure 5A) and that it prevented ACh metabolism as evident in 68% to 86% reduced choline levels (Figure 5B). To our knowledge, this is the first quantification of donepezil action on the breakdown of ACh in two major brain regions in the primate. The observed reduction of choline is higher than reductions of AChE activity (of ∼25%–70%) reported with positron emission tomography or in brain homogenate (
). Previous studies suggest that evaluating blood plasma levels or cerebrospinal concentrations may not predict how effectively AChE drugs influence behavioral outcomes (
) and do not reflect the actual bioactive concentration available in target neural circuits. By confirming that donepezil prevented ACh breakdown in the PFC and striatum, we thus established a direct link between behavioral outcomes and local drug bioavailability in two brain structures that causally contributes to attention and learning (see above) (
Lesions of the medial striatum in monkeys produce perseverative impairments during reversal learning similar to those produced by lesions of the orbitofrontal cortex.
). While our study showed that donepezil has a similar effect on ACh breakdown in both areas, it leaves unanswered whether or how choline concentrations in either brain area relate to finer performance variations across tasks because we measured choline only during one task and with insufficient statistical power to establish such a link at this stage.
The neurochemical measurements of donepezil in the PFC and striatum were achieved with a recently developed microprobe that samples neurochemicals through principles of SPME (
). We believe that leveraging this technique in primate drug studies will be important for clarifying whether systemically administered drugs reach the desired target brain systems in which they are supposed to exert their pro-cognitive effects.
In our study, confirming donepezil action in the PFC and striatum critically constrains the interpretation of the behavioral results, suggesting that different behavioral outcome profiles are not due an uneven drug availability. Rather, the different best doses for VS and flexible learning performance will likely be due to brain area–specific pharmacokinetic profiles of receptor densities, drug clearance profiles, or autoreceptor mechanisms that intrinsically downregulate local drug actions (
). One prediction from the specific distribution and kinetics of nicotinic or muscarinic receptors in the PFC and striatum is that donepezil might, at lower doses, act predominantly in the striatum via activation of muscarinic subreceptors because they have a particularly high binding potential (
Effects of acute acetylcholinesterase inhibition on the cerebral cholinergic neuronal system and cognitive function: Functional imaging of the conscious monkey brain using animal PET in combination with microdialysis.
Striatal, hippocampal, and cortical networks are differentially responsive to the M4- and M1-muscarinic acetylcholine receptor mediated effects of xanomeline [published correction appears in ACS Chem Neurosci 2019; 10:3910].
) (see Supplemental Discussion). However, it is also possible that donepezil recruits nicotinic receptors, which are upregulated with chronic donepezil use (
). It is important to disentangle, in future studies, the role of nicotinic and muscarinic subreceptors in the PFC and striatum to optimize the clinical potential to improve learning and attention functions in conditions with cognitive impairment and particularly in dementia (see Supplemental Discussion).
In summary, our results provide rare quantitative evidence that a prominent Ach-enhancing drug exerts domain-specific cognitive improvements of attentional control and cognitive flexibility at a distinct dose range. A major implication of this finding is that for understanding the strength and limitations of pro-cognitive drug compounds, it will be essential to test their dose-response efficacy at multiple cognitive domains.
Acknowledgments and Disclosures
This study was supported by the National Institute of Mental Health of the National Institutes of Health (Grant No. MH123687 [to TW]).
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
We thank Dr. Carrie K. Jones and Jason Russel for helpful feedback throughout the study and about the manuscript.
Meta-analysis of randomized controlled trials on the efficacy and safety of donepezil, galantamine, rivastigmine, and memantine for the treatment of Alzheimer’s disease.
Striatal, hippocampal, and cortical networks are differentially responsive to the M4- and M1-muscarinic acetylcholine receptor mediated effects of xanomeline [published correction appears in ACS Chem Neurosci 2019; 10:3910].
Effects of acute acetylcholinesterase inhibition on the cerebral cholinergic neuronal system and cognitive function: Functional imaging of the conscious monkey brain using animal PET in combination with microdialysis.
Difficulty of visual search modulates neuronal interactions and response variability in the frontal eye field [published correction appears in J Neurophysiol 2008; 99:1050].
Positive allosteric modulator of α7 nicotinic-acetylcholine receptors, PNU-120596 augments the effects of donepezil on learning and memory in aged rodents and non-human primates.
The scopolamine-reversal paradigm in rats and monkeys: The importance of computer-assisted operant-conditioning memory tasks for screening drug candidates.
A computer-automated touchscreen paired-associates learning (PAL) task for mice: Impairments following administration of scopolamine or dicyclomine and improvements following donepezil.
Dissociable effects of acetylcholinesterase inhibitors and phosphodiesterase type 5 inhibitors on object recognition memory: Acquisition versus consolidation.
Chromaproline and chromaperidine, nicotine agonists, and donepezil, cholinesterase inhibitor, enhance performance of memory tasks in ovariectomized rats.
USE: An integrative suite for temporally-precise psychophysical experiments in virtual environments for human, nonhuman, and artificially intelligent agents.
Estimation of plasma IC50 of donepezil hydrochloride for brain acetylcholinesterase inhibition in monkey using N-[11C] methylpiperidin-4-yl acetate ([11C]MP4A) and PET.
Lesions of the medial striatum in monkeys produce perseverative impairments during reversal learning similar to those produced by lesions of the orbitofrontal cortex.
A computational psychiatry approach identifies how alpha-2A noradrenergic agonist guanfacine affects feature-based reinforcement learning in the macaque.
00160-9&id=gr1.jpg){kind=link}
00160-9&id=gr2.jpg){kind=link}
00160-9&id=gr3.jpg){kind=link}
00160-9&id=gr4.jpg){kind=link}
00160-9&id=gr5.jpg){kind=link}
00160-9&id=gr6.jpg){kind=link}