Dopamine effects on memory load and distraction during visuospatial working memory in cognitively normal Parkinson’s disease

Abstract

Visuospatial working memory (WM) impairments in Parkinson’s disease (PD) are more prominent and evolve earlier than verbal WM deficits, suggesting some differences in underlying pathology. WM is regulated by dopaminergic neurotransmission in the prefrontal cortex, but the effect of dopamine on specific processes supporting visuospatial WM are not well understood. Dopamine therapeutic effects on different WM processes may also differ given the heterogeneity of cognitive changes in PD. The present study examined the effect of dopamine therapy on memory load and distraction during visuospatial WM. Exploratory analyses evaluated whether individual differences in medication effects were associated with a gene, catecho-O-methyltransferase (COMT), which regulates prefrontal cortex dopamine levels. Cognitively normal PD participants (n = 28) and controls (n = 25) performed a visuospatial WM task, which manipulated memory load and the presence/absence of distractors. PD participants performed the task on and off medication. PD COMT groups were comprised of Met homozygote (lower COMT activity) and heterozygote and Val homozygote carriers (higher COMT activity, Het/Val). The results showed that handling higher memory loads and suppressing distraction were impaired in PD off but not on medication. Medication improved distraction resistance in Met, but not Het/Val group. COMT did not modulate medication effects on memory load. These findings demonstrate that dopaminergic therapy restores visuospatial WM processes in patients without cognitive impairment and suggest that COMT variants may partly explain the mixed effects of medication on specific processes governed by distinct brain systems. Future investigations into gene-modulated effects of medication could lead to individualized strategies for treating cognitive decline.

Introduction

Working memory (WM) is impaired in Parkinson’s disease (PD) and associated with the development of PD dementia (Pal et al., 2018; Siegert et al., 2008). However, the mechanisms for WM impairments are not well understood, despite its importance in cognition, and daily activities including walking, balance, and driving (Le Bras et al., 1999; Cowan, 2014; Ranchet et al., 2011; Stolwyk et al., 2015; Xu et al., 2014). WM decline correlates with motor symptom severity in PD (Domellöf et al., 2011), although the association depends on the motor phenotype and modality of WM. These findings suggest that the pathology underlying different modalities of WM may differ, consistent with some modality-specific WM patterns of cortical activation in healthy adults (O. Gruber & von Cramon, 2001; Oliver Gruber & von Cramon, 2003). Visuospatial WM impairments are more pronounced and occur earlier in PD than verbal WM deficits (Siegert et al., 2008), suggesting that modality-specific pathologies underlying visuospatial WM may evolve earlier in the disease. Thus, elucidating specific mechanisms for visuospatial WM impairments in PD is of considerable importance.

While WM deficits in PD are attributed to dysfunctional dopaminergic neurotransmission (Ekman et al., 2012), the effects of dopaminergic therapy on different facets of WM are not well understood. Dopaminergic medications improve verbal WM in PD (Lewis et al., 2005), but only for higher WM loads (Poston et al., 2016). Dopaminergic medications also improved performances on visuospatial n-back tasks (Beato et al., 2008; Costa et al., 2003), but not on letter or face n-back tasks (Beato et al., 2008), digit span (Grogan et al., 2018) or a spatial WM task (Mollion et al., 2003). Conflicting findings may be due to differences in the demands placed on component processes that support WM, including the ability to resist distraction and handle higher memory loads. In this regard, one study found reduced resistance to distraction in PD patients while off than on medication (Fallon et al., 2017). Yet another study of PD off patients found impaired distraction resistance only in individuals with low memory capacity (Lee et al., 2010). The discrepant findings may be due to differences in the cognitive status of participants, which was not rigorously evaluated through neuropsychological testing of different cognitive domains. This is critical because medication therapy may not have the same effects on WM in patients with greater cognitive decline and presumably more significant neuropathology. A related issue is that the effects of dopaminergic therapy on both distraction resistance and handling higher memory loads during visuospatial WM (Vogel et al., 2005) need to be studied in the same PD cohort to better unravel the causes for WM decline.

There is also indirect evidence that individual differences in the effects of dopaminergic therapy on cognition may be partly related to genes that regulate dopamine levels, such as catechol-O-methyltransferase (COMT val¹⁵⁸met), which influences prefrontal cortex dopamine levels (Chen et al., 2004). Met homozygotes have lower COMT activity and higher active prefrontal dopamine levels than Val homozygotes, and heterozygotes have intermediate levels (Chen et al., 2004; Wu et al., 2012). In healthy adults, Met homozygotes generally exhibit better executive functioning than Val homozygotes (Egan et al., 2001; Goldberg & Weinberger, 2004). In levodopa-treated PD patients, however, Met homozygotes typically show worse executive functioning (e.g, attention, planning, set-shifting) than Val homozygotes, ostensibly due to overstimulation of prefrontal cortex (Foltynie et al., 2004; Williams-Gray et al., 2008, 2007). These findings are compatible with the proposed inverted U-shaped relationship between dopamine levels and prefrontal cortex functioning, wherein frontal-executive impairments are a result of a hyperdopaminergic state in the prefrontal cortex early in the course of the disease when dopamine loss is less severe (Fallon et al., 2013; Williams-Gray et al., 2008, 2007). Still, other studies report that levodopa-treated PD Met homozygotes have better executive functioning than Val homozygotes (Fang et al., 2019; Morley et al., 2012), whereas one study found no effect of COMT variants on executive functioning in PD on patients (Hoogland et al., 2010). A recent study of verbal WM in PD on patients found a linear effect of COMT activity, with Met homozygotes (n = 6) showing the best performance, followed by heterozygotes (n = 15) and then Val homozygotes (n = 33), who showed the poorest digit-span backward performance (Fang et al., 2019). While the effects of COMT activity on executive and WM tasks remain debated, discrepant findings may be due to studying tasks, rather than manipulating specific processes that govern WM performance within a task. This is important because WM is composed of multiple component process that are supported by different brain systems (Christophel et al., 2017), which may be differentially affected a loss in dopamine and other neurotransmitters. Importantly, on/off medication study designs are essential to determine if therapeutic effects of dopamine on cognition depend on COMT activity, which could have translational implications for managing treatment. The potential significance of COMT activity levels for understanding WM proficiency was convincingly demonstrated in healthy adults using placebo-controlled design that tested the effects of a COMT inhibitor, which slows down levodopa metabolism (Farrell et al., 2012). When given a placebo, Met homozygotes performed better than Val homozygotes on a 2-back digit task, whereas a COMT inhibitor reversed this effect by worsening WM in Met subjects and enhancing WM in Val subjects. However, studies of the effect of COMT on WM in PD are limited, and interactions of COMT activity levels with dopaminergic therapy on different visuospatial WM processes have not been studied in PD using on/off medication designs.

The present study builds upon previous research by investigating the effects of dopaminergic therapy on different facets of visuospatial WM, namely distraction resistance and memory load, using an on/off medication design. To control for generalized effects of cognitive decline on WM processes, a cognitively normal PD cohort without mild cognitive impairment (MCI) was studied. In exploratory analyses we also examined whether the effects of dopaminergic therapy on WM processes were associated with COMT activity. We hypothesized that distraction resistance and memory capacity would both improve when patients were on relative to off medication, and that the benefits of medication therapy on WM proficiency would differ between patients with higher and lower COMT activities. Specifically, we predicted that Met homozygotes would perform worse on than off medication if dopaminergic therapy produced a hyperdopaminergic state. Dopaminergic therapy was expected to improve WM performances in both heterozygote and Val homozygote carriers.

Materials and methods

Participants

Twenty-eight PD patients meeting the UK PD Brain Bank criteria for idiopathic PD (Hughes et al., 1992) and 25 controls were studied. Exclusion criteria were neurological diagnoses other than PD, psychiatric diagnoses including anxiety and depressive disorders, subjective cognitive decline, history of alcohol or substance abuse, and use of cognition enhancing medications (e.g., Donepezil). All participants had Mini Mental State Examination (MMSE) scores ≥ 26. No participants met the Movement Disorders Society Level I criteria for MCI based on neuropsychological testing (Litvan et al., 2012). Two patients were taking only dopamine agonists, four were on levodopa monotherapy, and 22 were on levodopa combination therapy. Patients were in early to mid-stages of the disease on the Hoehn and Yahr scale [Stages 1 (n = 2), 2 (n = 21), and 3 (n = 5)] (Hoehn & Yahr, 1967). Motor symptoms were evaluated on (within an hour after medication) and off (> 14 hours after last dose) medication using the Unified Parkinson’s Disease Rating Scale Part III (UPDRS III). The study was approved by the local Institutional Review Board and conducted in accordance with the Declaration of Helsinki. Participants provided written informed consent before participation.

All participants underwent neuropsychological testing. Patients were tested during their regular medication regimens. The Wechsler Adult Intelligence Scale Version 3 (WAISIII) Digit Span and Spatial Span subtests tested verbal and visuospatial WM span. The Trail Making Test Part A (TMT-A) tested attention. The Trail Making Test Part B (TMT-B) and the Delis Kaplan Executive Function System (D-KEFS)-Letter Fluency subtest tested executive functioning. Judgment of Line Orientation (JoLO) tested visuospatial functioning. The California Verbal Learning Test Version 2 (CVLT-II) tested verbal memory. The D-KEFSCategory Fluency tested language.

Blood samples were collected to evaluate the COMT val158met polymorphism (rs4680) using an allelic TaqMan discrimination assay. COMT data were missing for three PD patients. Of the 25 PD patients with COMT genotyping, 11 were Met homozygotes (lower COMT activity), 8 were Met/Val heterozygotes (intermediate COMT activity), and 6 were Val homozygotes (higher COMT activity). Owing to the small sample of Val homozygotes, Val homozygote and heterozygote carriers were combined into one Het/Val group (n = 14), which was compared to the Met group (n = 11) for the main statistical analyses. Follow-up analyses compared heterozygote carriers with each homozygote group to ensure that effects of COMT were not due to combining the high and intermediate activity groups.

Visuospatial working memory paradigm

The task was a modified version of similar paradigms (Fukuda & Vogel, 2009), and manipulated distraction and WM load (array size) (Figure 1). On each trial, a cue was presented (rectangle or square) signaling the shape to attend to. After a delay, one of the three array types was presented: 3 shapes no distractors (control condition), 3 shapes 3 distractors (distraction condition), and 5 shapes no distractors (memory load condition). The array was followed by a brief delay, after which a target was presented and remained visible until the participant pressed a button with their right index or middle finger to indicate if the target was the same or different color than the shape presented in the same location of the array.

Figure 1.

Illustration of visuospatial working memory paradigm. On each trial, a cue was presented (rectangle or square) signaling the shape to attend to. After a 500 ms delay, one of the three array types were presented. Figure 1a, 3 shapes no distractors; Figure 1b, 3 shapes 3 distractors; and Figure 1c, 5 shapes no distractors. The array was followed by a 900 ms delay, then a target was presented and remained visible until the participant responded. The participant pressed a button with their right index or middle finger to indicate if the target was the same or different color than the shape presented in the same location of the array.

No two arrays were identical. Half of the targets were the same and half were different from the shape in an array. Each of the three array types were randomly presented 24 times across two blocks of 36 trials (72 total trials). To test for the effects of array size, 3 shapes without distractors and 5 shapes without distractors were compared. To test for the effects of distraction, 3 shapes without distractors were compared to 3 shapes with 3 distractors. Thus, in both analyses the control condition was 3 shapes without distraction. The outcome measures included 1) d prime, a measure of sensitivity that adjusts for response biases (d’ = inverse of the standard normal cumulative distribution of hits – inverse of the standard normal cumulative distribution of false alarms) and 2) reaction time (RT) for correct trials, which was the time from the onset of the target to the button press. PD patients performed the WM paradigm in two separate sessions, one while on and the other while off medication. The order of on/off testing was counterbalanced. To control for practice effects in the PD group, controls also performed the task twice in different sessions. For the data analyses, first session data were used in the analyses for half of the control participants. For the remaining control participants, second session data were used in the analyses. Selection of first versus second session data was randomized for each control participant.

Statistical analyses

The within-subject effects of medication (on versus off), array size (3 versus 5 shapes), distraction (3 shapes with and without distractors) and their interactions were analyzed using repeated measures ANOVAs. PD on and off performances were compared separately to the controls using mixed-model ANOVAs, in which group (control versus PD on; control versus PD off) was the between-subject effect and array size and distraction were within-subject effects. In the PD group, mixed-model ANOVAs were used to test the between-subject effect of COMT (Met versus Het/Val) and its interactions with medication, separately for tests of array size and distraction.

Results

ANOVAs and Chi square statistics tested for group differences in demographics and neuropsychological test scores (Table 1). The PD group performed worse than controls on Letter Fluency. Post-hoc Pearson correlations found no relationships between Letter Fluency scores and visuospatial WM (d’, RT) in any of the conditions (false discovery rate corrected for multiple comparisons q >.20). No other group differences were found in neuropsychological scores or demographics. Medication significantly reduced motor symptoms (UPDRS III: F(1,27) = 36.9, p < .001, η_p² = .58).

Table 1.

Demographics and cognitive scores of Parkinson’s and Control participants.

	Parkinson’s (n = 28)	Controls (n = 25)	p	ηp2
Age	65.9 (8.3)	64.7 (8.3)	.61	.005
Sex, female %	53%	40%	.32
Education (years)	16.3 (2.3)	17.5 (2.7)	.09	.056
Disease duration (years)	6.1 (4.8)
LDE (Tomlinson et al., 2010)	778.4 (440.7)
UPDRS III on ‡	25.8 (10.0)
UPDRS III off	34.1 (12.0)
MMSE	28.9 (1.2)	29.2 (0.9)	.44	.012
WAIS III: Digit Span Total	18.1 (3.4)	18.3 (3.0)	.78	.002
WAIS III: Spatial Span Total	16.4 (2.8)	15.6 (4.3)	.43	.012
Trail Making Test Part A	36.6 (13.3)	31.2 (10.5)	.11	.050
Trail Making Test Part B	78.1 (41.4)	71.3 (29.7)	.50	.009
DKEFS Letter Fluency	40.7 (11.4)	48.2 (12.3)	.03	.094
Judgment of Line Orientation	25.4 (3.9)	26.8 (3.7)	.16	.039
CVLT Short Delay Free Recall	7.6 (1.6)	7.9 (1.2)	.42	.013
CVLT Long Delay Free Recall	7.3 (1.7)	7.6 (1.3)	.46	.011
DKEFS Category Fluency	13.0 (3.3)	13.9 (3.6)	.32	.019

All variables are reported as mean (standard deviation) or percentages. ANOVAs and Chi square statistics (sex) tested for group differences in demographics and neuropsychological test scores.

^‡Total symptoms on the UPDRS III were significantly greater off than on medication [F(1,27) = 36.9, p <.001, η_p² =.58].

CVLT: California Verbal Learning Test Version 2; DKEFS: Delis-Kaplan Executive Function System; LDE: Levodopa dose equivalency; MMSE: Mini Mental State Examination; UPDRS III: Unified Parkinson’s Disease Rating Scale Part III; WAIS III: Wechsler Adult Intelligence Scale Version 3.

Distraction and memory load effects on visuospatial working memory performance

Figure 2 displays the d’ data for each task condition in the control and PD on/off groups. The RT data for are shown in Figure S1 (Supplemental Materials) for these groups. First, the effects of array size and distraction on WM performance measures was tested separately in the control and the PD on group to demonstrate that the experimental manipulations had the expected effects in each group. For both groups, distraction and array size had large effects on WM performance. RTs were longer [Control: F(1,24) = 18.49, p < .001, η_p² = .44; PD on: F(1,27) = 7.04, p < .013, η_p² = .21] and d’was lower [Control: F(1,24) = 6.24, p < .016, η_p² = .22; PD on: F(1,27) = 4.26, p < .05, η_p² = .13] in the distraction than the no distraction condition. As for array size, RTs were longer [Control: F(1,24) = 11.12, p < .003, η_p² = .32; PD on: F(1,27) = 11.54, p < .002, η_p² = .30] and d’ was lower [Control: F(1,24) = 29.83, p < .001, η_p² = .55; PD on: F(1,27) = 17.78, p < .001, η_p² = .40] for the large than small array.

Figure 2.

Working memory performances in the PD and control groups. Significant group differences in sensitivity (d’) are designated by brackets for each task condition. Medication significantly improved d’ during distraction (3 shapes, 3 distractors) and the large array size (5 shapes, no distractors). Controls showed significantly higher d’ than the PD off, but not on group in the distraction and large array size conditions.

Medication effects

Next, the PD on and off conditions were compared to test the effects of medication on WM performance. Follow-up analyses then compared the PD on and off conditions with the control group to determine if WM performance was impaired. For the d’ analyses of array size, Figure 2 displays the significant medication by array size interaction [F(1,27) = 13.5, p < .001, η_p² = .33]. Simple effects analyses showed that medication improved d’ for the large arrays [F(1,27) = 14.1, p < .001, η_p² = .34], but had no effect on d’ for the small arrays (p = .74). Compared to controls, d’ was lower in the PD off group, regardless of array size [F (1,51) = 8.68, p < .005, η_p² = .15]. Comparisons between the controls and the PD on group revealed a medication by group interaction for d’ [F(1,51) = 4.20, p < .046, η_p² = .08], indicating that sensitivity was slightly lower in the PD on group for small [F(1,51) = 5.28, p < .03, η_p² = .09], but not large arrays [F(1,51) = 0.00, p = .93]. For the RT analyses of array size, the interaction of medication by array size [F(1,27) = 1.89, p > .18) and the main effect of medication [F(1,27) = 1.05, p > .31) were nonsignificant (Figure S1). Compared to the controls, RTs were slower in the PD on group [F(1,51) = 4.50, p < .05, η_p² = .08], regardless of array size, but not in the PD off group [F(1,51) = 2.4, p > .13].

For the d’ analyses of distraction, Figure 2 displays the significant medication by distraction interaction [F(1,27) = 4.34, p < .047, η_p² = .14]. Simple effects analyses demonstrated that medication improved d’ during distraction [F(1,27) = 4.47, p < .04, η_p² = .14], but had no effect on d’ in the no distraction condition (p = .74). Compared to controls, d’ was lower [F(1,51) = 7.60, p < .008, η_p² = .13] in the PD off group regardless of distraction. No significant differences were found between the controls and PD on group in d’ (p = .08). For the RT analyses of distraction, the interaction of medication by distraction [F(1,27) = 1.61, p > .21] and the main effect of medication [F(1,27) = 0.78, p > .39] were nonsignificant, indicating that response speed did not differ when PD patients were on or off their medications (Figure S1). However, when compared to the control group, RTs were slightly slower in the PD on group [F(1,51) = 4.37, p < .05, η_p² = .08], regardless of distraction, but not in the PD off group [F(1,51) = 2.91, p > .09, η_p² = .05].

COMT Met homozygotes versus Het/Val carriers

Table S1 (supplemental materials) details the demographic, clinical, and neuropsychological test performance data for the Met and Het/Val PD groups. Demographics (age, education, sex; p > .07), disease duration (p > .71), levodopa equivalent daily dose (p > .92) (Tomlinson et al., 2010), and neuropsychological test scores (p > .11) did not differ between the PD COMT groups. For the UPDRS III, the main effect of COMT [F (1,23) = 0.12, p > .73] and the COMT by medication interaction [F(1,23) = 2.1, p > .16] were nonsignificant. This finding was not related to differences between the COMT groups in the severity of motor symptoms on the left or right side, which did not differ on (p > .60) or off (p > .82) medication. Thus, medication reduced motor symptoms to the same extent in both COMT groups.

As for the visuospatial WM results, Figure 3 displays d’ for each task condition as a function of medication, separately for the Met and Het/Val groups. Figure S2 displays the RT data for these two COMT groups. For the analyses of array size, no significant main effects of COMT [d’: F(1,23) = 0.06, p > .80; RT: F(1,23) = 1.27, p > .27] or COMT interactions with medication [d’: F(1,23) = 0.92, p > .35; RT: F(1,23) = 0.18, p > .67] were found for d’ and RT.

Figure 3.

Effects of dopaminergic therapy on d’ for each task condition in the PD Met and Het/Val groups. Brackets designate significant differences between the PD on and off conditions for the analyses of distraction and array size. In the analyses of distraction effects, medication significantly improved d’ in Met homozygote carriers (n = 11) during distraction (p <.01, η_p² = .51), but had no effect on d’ in Het/Val carriers (n = 14). COMT group differences in d’ were nonsignificant off medication. On medication, the main effect of COMT was nonsignificant (p =.059, η_p² =.15), although the large effect size suggested that d’ tended to be higher in the Met than the Het/Val group. In the analyses of array size, medication significantly improved d’ for large arrays in both Met homozygotes and Het/Val carriers.

For the analyses of distraction, very large effect sizes (η_p² > .20) were obtained for the COMT by medication interaction [F(1,23) = 6.17, p = .021, η_p² = .21]. Follow-up analyses demonstrated that medication improved d’ in the Met [F(1,10) = 6.10, p = .03, η_p² = .38], but not the Het/Val group (F(1,13) = 0.27, p > .61). In addition, no differences were found between the Met and Het/Val groups off medication (d’: p = .69). On medication, the difference between the COMT groups was nonsignificant [F(1,23) = 3.93, p = .059, η_p² = .15], although the large effect size indicated that d’ tended to be higher in the Met than the Het/Val group. To better specify the locus of the medication effect we tested for the COMT and the COMT by medication interaction separately for the no distraction and distraction conditions. In the no distraction condition, the COMT [F(1,23) = 0.50, p > .48] and the COMT by medication interaction [F(1,23) = 1.98, p > .17] were nonsignificant. In the distraction condition, there was a significant COMT by medication interaction [F (1.23)= 5.25, p < .03, η_P² = .19] (Figure 3). Follow-up simple-effect analyses of this interaction showed that during distraction, medication had a large effect on improving d’ in Met homozygotes [F(1,10) = 10.58, p < .01, η_P² = .51], but had no effect on d’ in the Het/Val carriers [F(1,13) = 0.23, p > .63]. For the distraction analyses of RT (Figure S2), the main effect of COMT [F(1,23) = 1.87, p > .18] and the COMT by medication interaction [F (1.23) = 0.01, p > .90] were nonsignificant.

COMT heterozygotes versus Val and Met homozygotes

To determine if the above results for d’ were influenced by combining the Val (high COMT activity) and Het (intermediate COMT activity) groups, follow-up analyses compared heterozygotes with each homozygote group. These analyses demonstrated that distraction, array size, and medication effects on WM were not related to combining Val and Het carriers. Summaries of these data and the statistical analyses are detailed in Tables S2 and S3.

For analyses of array size, comparisons between heterozygotes and Val homozygotes showed that the COMT by medication (p > .48) and COMT by array size (p > .09) interactions were nonsignificant. However, tests for the main effect of COMT indicated that heterozygotes showed overall higher sensitivity [F(1,12) = 5.1, p < .05, η_p² = .30; Mean (SD): d’ = 1.6 (0.19)] than Val homozygotes [Mean (SD):d’ = 0.9 (0.22)]. For comparisons between the heterozygote and Met groups, tests for the COMT by medication (p > .60), COMT by array size (p > .96), and the COMT main effect (p > .34) were nonsignificant.

For analyses of distraction, comparisons between Het and Val homozygotes showed that the COMT by medication (p > .75) and COMT by distraction (p > .20) interactions were nonsignificant. Medication also did not significantly affect WM performance in either the Val homozygotes (p > .93) or the heterozygotes (p > .53). Despite the nonsignificant main effect of COMT [F(1,12) = 3.87, p = .073, η_p² = .25], the large effect size suggested that heterozygotes tended to show better overall WM proficiency [Mean (SD): d’ = 1.6 (0.20)] than Val homozygotes [Mean (SD): d’ = 0.9 (0.26)]. For comparisons between the heterozygote and Met homozygotes, the medication by COMT interaction [F(1,17) = 4.89, p < .04, η_p² = .22] showed that medication improved WM in Met homozygotes [F(1,10) = 6.10, p < .03, η_p² = .38], but not heterozygotes (p > .53). The COMT by distraction interaction (p > .54) and the COMT main effect (p > .70) were nonsignificant, indicating that overall WM proficiency was similar between Met homozygotes [Mean (SD): d’ = 1.5 (0.17)] and heterozygotes [Mean (SD): d’ = 1.6 (0.20)].

Discussion

In a cognitively normal PD cohort without MCI, we found that memory load capacity and distraction resistance were impaired while off medication, whereas dopaminergic therapy restored the ability to handle higher memory loads and suppress distraction, although the latter effect was modulated by COMT expression. These results were not driven by generalized cognitive decline, which has not been rigorously evaluated in past studies. Rather, neuropsychological test scores across five cognitive domains did not differ between the PD on group and controls. An exception was lower Letter Fluency scores, which did not correlate with visuospatial WM performance on or off medication. In exploratory analyses, we also revealed for the first time that medication effects on distraction resistance, but not on handling higher memory loads, were associated with COMT activity levels, suggesting the possibility that these cognitive functions may be supported in part by different neural circuitry. Dopamine therapy specifically improved distraction resistance only in Met homozygotes (lower COMT activity). Notably, the benefit of medication on motor symptoms was not associated with COMT activity levels, consistent with COMT regulation of prefrontal cortex dopamine levels.

Our results build upon findings from a limited number of studies reporting mixed effects of dopaminergic therapy on various visuospatial WM tasks (Beato et al., 2008; Costa et al., 2003; Fallon et al., 2017; Mollion et al., 2003). In the same PD cohort, we demonstrated that medication not only restored distraction resistance as previously shown (Fallon et al., 2017), but also normalized the ability to manage higher WM loads. Though past studies found mixed effects of dopamine therapy on WM (Beato et al., 2008; Costa et al., 2003; Mollion et al., 2003), the reasons for discrepant results are difficult to unravel since medication effects on specific processes that govern WM performance were not studied. This is important because WM tasks differ in the extent to which distraction, memory load, or other key processes are at play and therefore, the extent to which their underlying neural mechanisms are vulnerable to PD.

Emerging research suggests that overlapping and unique neurocircuitry supports different WM processes. In PD patients off medication, a loss of striatal-prefrontal coupling was observed during high visuospatial WM memory loads and distraction (Harrington et al., 2020). Yet the engagement and disengagement of other neurocircuitry differed depending on specific WM processing demands. For example, PD off patients upregulated frontal and parietal connectivity during distraction. In contrast, connectivity of the dorsolateral prefrontal cortex (DLPFC) with the parietal cortex was lost when retrieving items from high memory loads. However, the neural systems that underlie medication effects on different visuospatial WM processes have not been investigated in PD, and in on/off medication studies of verbal WM in PD, discrepant findings have been reported with dopaminergic therapy increasing, decreasing, or having no effect on frontal and/or striatal activation (Poston et al., 2016; Simioni et al., 2017). Future neuroimaging studies are therefore needed to unravel the systems by which dopaminergic therapy exerts beneficial, neutral, or potentially detrimental effects on different visuospatial WM processes in cognitively normal PD participants.

The results from our exploratory analyses also suggested that COMT may modulate the positive effect of dopaminergic therapy for specific processes that are engaged during visuospatial WM and by inference, neurobiological systems that support these processes. Dopaminergic therapy specifically improved the ability to suppress distraction, but only in Met homozygotes. This finding may be linked to upregulation of prefrontal dopamine levels in patients with lower COMT activity (Wu et al., 2012). One speculation is that in Het/ Val carriers, dopaminergic therapies may not upregulate prefrontal cortical dopamine to levels that are effective for suppressing attentional responses to distractions. Distraction resistance also tended to be better in Met homozygotes relative to Het/Val carriers when patients were on, but not off medication, despite the absence of COMT group differences in neuropsychological test performances. In comparisons between Het and Val homozygote carriers, we found that WM tended to be better in Het than Val carriers, although medication did not alter performance in either group. Although larger samples are need to validate these initial results, this linear association between COMT activity in PD on patients is consistent with a study of verbal WM (digit span backward) reporting that PD on Met homozygote carriers (n = 6) performed better than Val homozygote carriers (n = 33), with heterozygote carriers (n = 15) showing intermediate levels of performance (Fang et al., 2019). However, another study found that COMT activity levels showed an inverted U-shaped relationship with spatial WM performance in PD on patients (Fallon et al., 2015), with heterozygote carriers performing better than both Met and Val homozygotes whose performance did not differ. The reason for this discrepant finding is unclear, but the results may be partly related to using a complex WM search task, which likely engages a variety of cognitive processes. The importance of this factor cannot be overstated because COMT activity in PD on patients is known to have different relationships with attention and executive functions, which can be more or less engaged depending on WM task difficulty. For example, in PD on patients COMT activity has a linear relationship (Met > Het > Val) with attention (Morley et al., 2012) and set-shifting (Fallon et al., 2015; Fang et al., 2019), although not always for other attention and executive tasks (Hoogland et al., 2010; Mata et al., 2017). In contrast, planning (Tower of London) can be better in Val homozygote than Met homozygote carriers (Foltynie et al., 2004; Williams-Gray et al., 2008, 2007). The results also depend on disease duration, as worse planning in Met homozygotes was found only in early disease duration patients (1.6 years), possibly due to overstimulation of prefrontal dopamine systems by dopamine therapy (Williams-Gray et al., 2009). In our study, however, average disease duration was about 6 years, suggesting that prefrontal cortex dopamine levels were likely reduced in patients, thereby mitigating the potential for dopaminergic therapy-induced hyperactivity, which would be expected to negatively impact WM. Thus, relationships between cognition and the COMT genotype appear to follow an inverted U-shaped function whereby optimal levels of prefrontal cortex dopamine depend both on the cognitive processing demands and disease duration (Fallon et al., 2015; Goldman-Rakic, 2000). This proposal needs to be more fully considered in studies that evaluate medication effects on different cognitive processes and their regulation by COMT activity levels in PD.

Interestingly, COMT had no effect on medication benefits for storing larger memory arrays, despite dopamine therapy-induced improvement in d’ for high memory loads. This finding may suggest that in addition to dopamine, interdependent brain systems that are linked to non-dopaminergic neurotransmission are also at play. For example, lower gamma-aminobutyric acid (GABA) in the DLPFC is associated with greater performance declines with higher memory loads but has no influence on distraction resistance in healthy adults (Yoon et al., 2016). This illustrates the specificity of GABA, which has opposing effects on dopamine levels, for processing high memory loads. In addition, the parietal cortex, which is linked to non-dopaminergic neurotransmission (Gratwicke et al., 2015), specifically exhibits load-dependent activity that is associated with WM precision (Galeano Weber et al., 2016). Moreover, in healthy adults, transcranial direct current stimulation of the parietal cortex, but not the prefrontal cortex enhances visuospatial WM capacity when WM loads exceed capacity limits (Wang et al., 2019). Altogether, these findings suggest that non-dopaminergic systems are also recruited when the burden on WM storage capacity is high, which could diminish the influence of COMT activity levels on performances. In contrast, medication therapy boosted selective attention and distraction resistance only in Met homozygote carriers, presumably due to increasing prefrontal cortex dopamine to optimal levels needed for these cognitive functions (Toepper et al., 2010). Medication failed to improve distraction resistance in both heterozygote and Val homozygote carriers, suggesting that it had little or no effect on prefrontal cortex dopamine levels in patients with higher COMT enzymatic activity. Neuroimaging studies are needed to test these proposals by characterizing neural systems that respond to dopaminergic therapy during WM-related processes and their modulation by COMT variants.

Despite the very large effect sizes obtained in our exploratory analyses of COMT activity levels, future studies containing larger PD samples for each COMT variant are needed to validate these findings. However, our initial results showing a linear effect of COMT activity on visuospatial WM in PD on patients was largely consistent with other studies of PD on patients for verbal WM (Fang et al., 2019), attention (Morley et al., 2012) and executive functions (Fallon et al., 2015; Fang et al., 2019; Williams-Gray et al., 2009). Another potential limitation concerns the heterogeneity of cognitive changes in PD, which can dilute tests for medication and genetic effects on visuospatial WM. Although this is a limitation of most studies to date, we do not believe that it was a major factor in our study since unlike past studies, participants were screened for MCI and group differences in five cognitive domains were generally nonsignificant. We were also unable to control for the possibility that levodopa and dopamine agonists have different effects on WM in PD. WM is associated with D1 dopamine receptors, which can be affected by levodopa, but not dopamine agonists. Levodopa exerts effects through both D1 and D2 dopamine receptors, whereas commonly used dopamine agonists have a higher affinity for D2 receptors (Moustafa et al., 2013). The limited number of studies that have compared dopamine agonists and levodopa in early stage PD patients suggest that dopamine agonists can both worsen (Brusa et al., 2003; Costa et al., 2003) and improve WM (Costa et al., 2009). Nonetheless, potential differences in medication therapies may indeed interact with COMT variants. Studies containing larger samples sizes than our study are needed to evaluate this prospect as the results could guide decision making for the selection of optimal dopaminergic treatment for PD patients with WM impairments. Lastly, we did not conduct COMT genotyping for our control participants owing to the focus on testing whether dopaminergic replacement effects on WM in PD were modulated by COMT activity levels. Tests of whether dopaminergic activity restored WM performance to normal levels would be best tested through comparisons between PD and control groups with the same COMT genotypes. A related issue is that distraction effects on d’ were found in our control group, but not in PD Met homozygotes on medication. This finding might suggest that medication boosted Met homozygotes’ ability to resist distraction beyond that of the control group, whereas distraction suppression was not optimal in controls owing to the decline in dopaminergic functioning in normal aging (Bäckman et al., 2000; Erixon-Lindroth et al., 2005). To test this proposal, however, COMT genotyping in the control group would be required.

In conclusion, dopaminergic therapy in PD restores the capacity to handle higher memory loads and suppress distractions during visuospatial WM, which should enhance everyday activities that amplify cognitive workloads often in the face of distractions. Importantly, the benefits of medication on WM can be present before significant cognitive impairment. Our results also underscore the potential relevance of genetic variants in explaining the heterogeneous effects of medication on specific facets of cognition in PD. These initial findings should encourage investigations into dopamine regulating genes and their potential for moderating dopaminergic therapy effects on different cognitive domains and underlying brain systems. There are no preventive or disease-modifying approaches available for cognitive decline in PD and there is an urgent need for efficient treatments given the substantial impact of cognitive decline in quality of life for PD participants and their families. Outcomes from this line of research could lead to individualized strategies for optimizing the therapeutic benefit of dopamine medications on cognitive decline in PD patients with specific genetic profiles.