M1-selective muscarinic allosteric modulation enhances cognitive flexibility and effective salience in nonhuman primates

Abstract

Acetylcholine (ACh) in cortical neural circuits mediates how selective attention is sustained in the presence of distractors and how flexible cognition adjusts to changing task demands. The cognitive domains of attention and cognitive flexibility might be differentially supported by the M₁ muscarinic acetylcholine receptor (mAChR) subtype. Understanding how M₁ mAChR mechanisms support these cognitive subdomains is of highest importance for advancing novel drug treatments for conditions with altered attention and reduced cognitive control including Alzheimer's disease or schizophrenia. Here, we tested this question by assessing how the subtype-selective M₁ mAChR positive allosteric modulator (PAM) VU0453595 affects visual search and flexible reward learning in nonhuman primates. We found that allosteric potentiation of M₁ mAChRs enhanced flexible learning performance by improving extradimensional set shifting, reducing latent inhibition from previously experienced distractors and reducing response perseveration in the absence of adverse side effects. These procognitive effects occurred in the absence of apparent changes of attentional performance during visual search. In contrast, nonselective ACh modulation using the acetylcholinesterase inhibitor (AChEI) donepezil improved attention during visual search at doses that did not alter cognitive flexibility and that already triggered gastrointestinal cholinergic side effects. These findings illustrate that M₁ mAChR positive allosteric modulation enhances cognitive flexibility without affecting attentional filtering of distraction, consistent with M₁ activity boosting the effective salience of relevant over irrelevant objects specifically during learning. These results suggest that M₁ PAMs are versatile compounds for enhancing cognitive flexibility in disorders spanning schizophrenia and Alzheimer's diseases.

Keywords: acetylcholine; attention; cognitive control; donepezil; learning.

Fig. 1.

Task Design and Feature-Reward Learning Task Performance Enhancement by VU0453595 (A) Images of cage-mounted kiosk and monkey Ba utilizing the touch screen to perform the feature-reward learning (FRL) task taken via the video monitoring system. Both images are taken from the same time point from different angles. (B) Trial progression of the FRL task (Top) and the VS task (Bottom). The example FRL task here is a block with “high distractor load” where objects vary in both color and pattern. Each object contains only 1 feature from each feature dimension. Although the red checkered object was correctly chosen in this trial, the animal would need to learn through trial and error if the red or checkered feature was correct in order to optimally acquire reward from future objects in this block. The example VS block here shows a trial with three distractors and a target object that is defined by three features: blue, striped, and straight conical arms. The red distractor has zero features in common with the target, the yellow distractor has one feature in common with the target (striped pattern) and the blue distractor has two features in common with the target (blue color and straight conical arms). Trials in either task were initiated by a 0.3 to 0.5 s touch and hold of a central blue square (3° visual radius wide) after which the square disappears (for 0.3 to 0.5 s) and task objects (2.5° visual radius wide) are presented on screen. For the VS task, a structured background scene (here: a lawn) was used to distinguish the VS from the FRL task (which had a grey uniform background). For each visual search block, we drew a different random background scene from a set of five backgrounds independent of the target search feature. In either task, subjects have 5 s to select one of the objects with a 0.2 s touch and hold. Failure to choose an object resulted in an aborted trial which was ignored. Feedback for choice selection was provided 0.2 s after object selection for 0.5 s via both a visual halo around the chosen object as well as a auditory cue alongside any earned fluid. Both the frequency of the audio feedback and color of the feedback halo differed based on outcome. (C) Block-wise average learning curves for the low distractor load blocks of the FRL task aligned to block start for vehicle, 0.3, 1, and 3 mg/kg VU0453595, smoothed after the first three trials with a sliding window (shaded area: SE). Dotted horizontal lines signify 0.33 and 0.66 probabilities. (D) The same as C but for the high distractor load blocks. (E) Median trials-to-criterion, calculated as the first trial in a block that led to >70% performance over 10 subsequent trials, for the low and high distractor load blocks of the FRL task. For the low distractor load blocks, trials-to-criterion were 11.03 (SE: 0.38), 8.94 (SE: 0.75), 7.93 (SE: 0.81) and 10.88 (SE: 0.94) for vehicle, 0.3, 1, and 3 mg/kg doses of VU0453595, respectively. Only the 1 mg/kg dose was significantly different from vehicle [F(3,691) = 3.54, P = 0.01; η² = .015; Tukey’s, P = 0.028; Cohen’s d = −0.352). For the high distractor load blocks, trials-to-criterion were 12.85 (SE: 0.43), 13.56 (SE: 1.03), 11.65 (SE: 1.00), and 12.92 (SE: 0.98) for vehicle, 0.3, 1, and 3 mg/kg doses of VU0453595, respectively, with no significant effect [F(3,565) = .40, n.s.].

Fig. 2.

Feature-Reward Learning Task Efficiency and cognitive flexibility improvements with VU0453595 (A) The average RT curve of each session (correct trials only) aligned to block start for the low distractor load blocks of the FRL task for vehicle, 0.3, 1 and 3 mg/kg doses of VU0453595 (shaded area: SE) (B) The same as A but for high distractor load blocks of the FRL task. (C) Block-wise averages of the traces plotted in A and B visualized to compare RTs between distractor load conditions. Low distractor load blocks had RTs of 960 ms (SE: 11), 923 ms (SE: 24), 870 ms (SE: 23) and 974 ms (SE: 23) for vehicle, 0.3, 1, and 3 mg/kg doses of VU0453595, respectively. High distractor load blocks had RTs of 984 ms (SE: 11), 965 ms (SE: 26), 960 ms (SE: 23), and 937 ms (SE: 22). Only the 1 mg/kg dose of VU0453595 was significantly different from vehicle [F(3,1672) = 2.97, P = 0.03; η² = 0.005; Tukey’s, P = 0.04; Cohen’s d = −0.350] (D) Trials-to-inflection for RTs in the low distractor load blocks defined as the first trial per block (excluding trial 2) that RTs become faster (error bars: SE). Trials-to-inflection was 8.7 (SE: 0.3), 8.0 (SE: 0.6), 6.5 (SE: 0.5) and 8.5 (SE: 0.5) for vehicle, 0.3, 1, and 3 mg/kg doses of VU0453595, respectively. Only the 1 mg/kg dose was significantly different from vehicle [F(3,193) = 2.68, P < 0.05; η² = .040; Tukey’s, P = 0.017; Cohen’s d = −0.674]. (E) Block-wise average trials-to-criterion after extradimensional shifts were 12.2 (SE:1.0), 8.9 (SE: 2.4), 4.0 (SE: 0.7) and 9.3 (SE: 1.4) for vehicle, 0.3, 1, and 3 mg/kg doses of VU0453595, respectively. Only the 1 mg/kg dose showed a significant difference from vehicle [F(3,122) = 3.15, P = 0.03; η² = 0.072; Tukey’s, P = 0.02; Cohen’s d = −0.868]. (F) Block-wise average trials-to-criterion after intradimensional shifts were 12.6 (SE: 0.5), 10.0 (SE: 0.7), 9.3 (SE: 0.7) and 12.3 (SE: 1.1) for vehicle, 0.3, 1, and 3 mg/kg doses of VU0453595, respectively. Only the 1 mg/kg dose showed a significant difference from vehicle [F(3,518) = 3.26, P = 0.02; η² = .019; Tukey’s, P = 0.04; Cohen’s d = −0.349]. (G) The proportion of errors that were perseverative in nature with the feature that was perseverated being from the same feature dimension as the target feature. The proportion of perseverative errors from the target feature dimension were 10.7% (SE: 0.2), 11.5% (SE: 0.7), 8.5% (SE: 0.6) and 11.0% (SE: 0.7) for vehicle, 0.3, 1, and 3 mg/kg doses of VU0453595, respectively, with only the 1 mg/kg dose being significantly different from vehicle [F(3,1679) = 3.74, P = 0.01; η² = 0.007; Tukey’s, P = 0.01; Cohen’s d = −0.243]. (H) The same as G but with the feature that was perseverated being from the distracting feature dimension (different from the target feature dimension). Proportions of perseverative errors from the distracting feature dimension were 17.3% (SE: 0.3), 19.6% (SE: 0.8), 15.6% (SE: 1.0) and 15.6% (SE: 1.0) for vehicle, 0.3, 1, and 3 mg/kg doses of VU0453595, respectively. There was a nonsignificant trend for a main effect of experimental condition [F(3,844) = 2.36, P = 0.07; η² = 0.008].

Fig. 3

Distractor Effect and Interference Control Are Not Consistently Impacted by VU0453595 (A) Target detection durations (reaction times) as a function of distractor number for the second VS block. There was a significant main effect of experimental condition with a significant different between the 3 mg/kg dose of VU0453595 compared with vehicle [F(3,944) = 3.67, P = 0.01; η² = .008; Tukey’s, P < 0.05]. The 3 mg/kg dose improved search times from 1.16 s (SE: 0.02), 1.37 s (SE: 0.02), 1.54 s (SE: 0.03) and 1.72 (SE: 0.03) with vehicle to 1.11 s (SE: 0.04), 1.30 s (SE: 0.04), 1.48 s (SE: 0.05) and 1.58 s (SE: 0.05) for 3, 6, 9, and 12 distractors, respectively. There was no significant change in the first VS block (data not shown). (B) The set size effect, operationalized as the slope of the linear fit of search times as a function of distractor numbers for the second VS block (0.057 (SE: 0.003), 0.060 (SE: 0.005), 0.058 (SE: 0.005) and 0.049 (SE: 0.005) for vehicle, 0.3, 1, and 3 mg/kg doses of VU0453595) was not significant [F(3,236) = 0.67, n.s.]. There was also no significant set size effect in the first VS block (data not shown). (C) VS task performance as a function of distractor number for the first VS block. There was a significant main effect of experimental condition with a significant different between the 3 mg/kg dose of VU0453595 compared with vehicle [F(3,944) = 3.80, P = 0.01; η² = .010; Tukey’s, P = 0.04]. The 3 mg/kg dose reduced performance from 95.8% (SE: 0.4), 91.8% (SE: 0.7), 88.3% (SE: 0.8) and 84.1% (SE: 1.0) with vehicle to 94.5% (SE: 1.2), 89.4% (SE: 2.0), 83.1% (SE: 2.2) and 81.8% (SE: 2.7) for 3, 6, 9 and 12 distractors, respectively. There was no significant change in the second VS block (data not shown). (D) The set size effect, operationalized as the slope of the linear fit of performance as a function of distractor numbers for the first VS block [−0.013 (SE: 0.001), −0.011 (SE: 0.003), −0.014 (SE: 0.002) and −0.015 (SE: 0.002) for vehicle, 0.3, 1, and 3 mg/kg doses of VU0453595] was not significant [F(3,236) = 0.60, n.s.]. There was also no significant set size effect in the second VS block (data not shown). (E) VS task search times as a function of target-distractor similarity for the second VS block. There was a significant main effect of experimental condition with a significant different between the 1 mg/kg dose of VU0453595 compared with vehicle [F(3,708) = 4.67, P = 0.003; η² = 0.018; Tukey’s, P = 0.02] but no significant set size effect [F(3,236) = 1.81, n.s.]. Search times were faster from 1.29 s (SE: 0.02), 1.48 s (SE: 0.02), and 1.49 (SE: 0.03) with vehicle to 1.18 s (SE: 0.04), 1.42 s (SE: 0.05) and 1.33 s (0.05) for low, medium, and high average target-distractor similarity, respectively. (F) VS task performance as a function of target-distractor similarity for the second VS block. There was a significant main effect of experimental condition but no significant post hoc comparison was found [F(3,708) = 2.84, P = 0.04; η² = 0.011; Tukey’s, n.s.]. We also failed to find a significant set size effect [F(3,236) = 0.53, n.s.].

Fig. 4.

Results Summary and Consistency across Monkeys. Key results from the feature-reward learning task (Top) and the visual search task (Bottom). Measures and the respective figures showing each result (if applicable) are stated on the y-axis; asterisks indicate a significant effect for all monkeys combined. Values for each monkey represent the average change at the 1 mg/kg VU0453595 dose relative to vehicle scaled arbitrarily for each measure. The scaling for each measure is indicated on the right for 1 arbitrary unit along the x-axis. At the 1 mg/kg dose, VU0453595 enhances virtually all measures of the feature-reward learning task to some degree while no reliable changes in the visual search task were observed.