Nigrostriatal dopaminergic integrity in relation to prefrontal cortex activity and gait performance during dual-task walking in older adults

Abstract

Background

Walking automaticity facilitates maintenance of gait speed without prefrontal cortex (PFC) resources. Brain aging may cause shifts to attentional gait with greater engagement of the PFC. Nigrostriatal dopaminergic integrity likely facilitates walking automaticity and gait speed during attentional dual-tasks.

Methods

Older adults (n = 201; age = 74.9; 63.2% women, gait speed = 1.10 m/s) completed a dual-task protocol of saying every other letter of the alphabet while walking. PFC activation was measured by functional near-infrared spectroscopy. Four groups were defined based on PFC activation (increased during dual-task, PFC+, or not, PFC−) and gait speed performance (maintained during dual-task, gait+, or slowed, gait−). To compare nigrostriatal dopaminergic integrity across groups, we assessed binding of the type-2 vesicular monoamine transporter density in the sensorimotor and associative striatum using [¹¹C]-(+)-a-dihydrotetrabenazine (DTBZ) positron emission tomography. Multinomial regression estimated adjusted associations of DTBZ binding with group membership. We hypothesized that DTBZ binding was highest for those who maintained gait speed without additional PFC activation when switching from single- to dual-task (PFC−/gait+; ie, highest gait automaticity).

Results

In bivariate analyses, the PFC−/gait+ group had the highest DTBZ binding in the sensorimotor striatum (P = .05); binding in the associative striatum was similar across groups (P = .1). Results were similar in adjusted regression analyses; DTBZ binding in sensorimotor striatum was associated with lower likelihood to be in the PFC−/gait− (odds ratio [OR] = 0.28; 95% CI, 0.08-0.94) or the PFC+/gait+ (OR = 0.29; 95% CI, 0.10-0.84) groups compared to PFC−/gait+ reference group.

Conclusion

These results provide support for dopaminergic involvement in sustaining gait automaticity at older ages.

Introduction

Walking has traditionally been viewed as primarily modulated in the brain by the motor cortex and subcortical structures. A body of recent evidence has established the prefrontal cortex (PFC) as a critical component of gait control, particularly in older adults and during complex walking conditions.^1,2 Motor automaticity is characterized by lower frontal and increased posterior-subcortical engagement when performing overlearned motor tasks. However, aging-related declines in the neural integrity of motor-control regions lead to reduced walking automaticity and an increased reliance on PFC activity during walking.^3,4 This increased reliance on PFC activity can be compensatory, preserving walking performance despite declines in the sensorimotor regions at the expense of using PFC resources that may otherwise be needed for competing attentional tasks^2,3 which can result in increased fall risk.⁵ However, increased PFC activity can also be found in the presence of performance deterioration when it is interpreted as neural inefficiency and may represent noisy signaling in the PFC.6 Therefore, interpretation of PFC activation during walking tasks in older adults is challenging in the absence of information pertaining to performance changes⁷ or other neural resources controlling movement.

Nigrostriatal dopaminergic integrity of striatal sensorimotor regions contributes to motor learning and motor automaticity.⁸ This has been primarily studied in individuals with Parkinson’s disease where marked declines in nigrostriatal dopaminergic integrity occur. In individuals with Parkinson’s, reduced motor automaticity is evident early in the disease process, leading to increased activity in the PFC when performing presumably automatic movements.⁸ Therefore, lower nigrostriatal dopaminergic integrity in older adults without Parkinson’s may contribute to loss of walking automaticity, particularly during challenging conditions, such as dual-task walking. The importance of nigrostriatal dopaminergic pathways in walking among older adults without Parkinson’s disease is supported by prior research demonstrating that lower dopamine transporter (DAT) expression in striatum, as measured by [¹¹C]β-CFT positron emission tomography (PET), is related to slower gait speed and greater likelihood of nonspecific parkinsonian signs.^9,10

Further, the nigrostriatal associative regions have strong connectivity to the PFC and a key role in modulating activity of the PFC. Therefore, dopaminergic integrity of the associative regions could support compensatory PFC activation, facilitating maintenance of gait speed under challenging conditions in the presence of reduced walking automaticity. Executive control function, which is mediated via dopaminergic signaling in the associative regions of the striatum, is strongly related to gait speed and motor control in older adults, particularly during dual-task walking.¹ However, prior studies with measures of nigrostriatal dopaminergic integrity have focused on gait speed under unchallenged conditions and have not explored adaptation to walking conditions with increased attentional demands, such as dual-task walking. In addition, there are few studies using multimodal methods to assess both PFC function during walking tasks and aspects of brain health related to walking automaticity, and none currently determining the role of dopaminergic integrity by PET imaging.

We tested whether the level of nigrostriatal dopaminergic integrity measured via type-2 vesicular monoamine transporter (VMAT2) binding from dihydrotetrabenazine ([¹¹C]DTBZ) PET imaging¹¹ of the sensorimotor and associative regions of the striatum was associated with the PFC activation and gait speed response when moving from single- to dual-task walking among older adults without Parkinson’s disease. We hypothesized that greater dopaminergic integrity of the somatosensory network supports more automatic walking behaviors, requiring less PFC engagement. This would result in gait speed being maintained without increasing PFC activation. We further hypothesized that greater dopaminergic integrity of the associative network would support greater compensatory engagement of the PFC as needed to maintain gait speed when switching from single- to dual-task gait speed. Therefore, we hypothesized that the greatest DTBZ binding for sensorimotor striatum would be seen in the group that maintained gait speed from single- to dual-task with no additional activity in the PFC and also that the greatest DTBZ binding for the associative striatum would be in the group that maintained gait speed from single- to dual-task with additional PFC activity. These results will improve understanding of automatic motor control during walking and decline during aging with implications for intervention development and future research.

Methods

Participants

Participants came from 2 studies with harmonized protocols for magnetic resonance imaging (MRI), PET, and functional near-infrared spectroscopy (fNIRS) data collection. A subset of participants from the Monongahela-Youghiogheny Healthy Aging Team (MYHAT) study were included in Move MYHAT (MMH), a substudy of dopaminergic contributions to mobility resilience. The parent study is a community-based study of risk factors for mild cognitive impairment. Recruitment of community-dwelling adults aged 65 and older in small towns in Southwest Pennsylvania occurred between 2006-2008 and 2016-2019 (n = 2685), and annual follow-up continues.^12,13 Of 386 eligible and referred, MMH recruited participants (n = 87) who were still in the study in 2019-2021 and could walk unassisted, were eligible for and agreed to participate in the MR and PET protocols, and were free of dementia or other neurodegenerative conditions, including Parkinson’s disease.

The second study was an ancillary of the Study of Muscle, Mobility and Aging (SOMMA; n = 879)¹⁴ to assess dopaminergic contributions to fatigability in older adults. SOMMA recruited at the University of Pittsburgh and Wake Forest University School of Medicine between April 2019 and December 2021. Individuals were eligible if they were aged 70 and older with gait speeds ≥0.6 m/s, who were able to complete a 400 m walk, were free of neurological diagnoses including Parkinson’s disease and dementia, had a body mass index (BMI) < 40 kg/m², and were eligible for MRI and muscle biopsy. The SOMMA Brain ancillary further recruited individuals from the 439 participants enrolled at the Pittsburgh site between 2020 and 2022. Of the 285 potentially eligible participants whose muscle biopsy for the parent study was less than 12 months prior, a total sample of n = 150 was recruited for the ancillary and completed neuroimaging. Data for SOMMA came from the November 2024 data release. Participants from both studies (n = 85 from MMH and n = 116 from SOMMA) were included in these analyses if they had usable data on all relevant measures.

All neuroimaging data (fNIRS and PET) were collected on the same instruments, and the dual-task walking protocols were completed in the same lab space with the same data collection team for both studies. All protocols were approved by participating Institutional Review Boards, and all participants gave informed consent prior to any data collection.

Dual-task walking

Participants walked a 15-meter straight path at their comfortable pace with and without a concurrent cognitive task (single- and dual-task walking).¹⁵ Single- and dual-task walking were each repeated 4 times, and the order of presentation was randomized. The dual-task consisted of reciting every other letter of the alphabet starting with “B” while walking.¹⁶ The alphabet task was also completed for 20 seconds while standing. Twenty-second quiet standing periods were included before each task as a baseline condition for the fNIRS signal. All tasks were completed while wearing a wireless fNIRS system on the forehead (Artinis OctaMon). Participants were not given explicit instructions with regard to task prioritization. Start and end times of each task were recorded to mark events within the fNIRS time series. Gait speed was measured by stopwatch and calculated in meters/second (m/s) for each walking task. Change in gait speed from single- to dual-task was calculated with maintenance of gait speed (ie, less slowing) from single- to dual-task considered to be the healthier response. The rate of correct letters generated was calculated as the number of correct letters/s.

fNIRS data collection and processing

We followed consensus recommendations in our fNIRS data processing and reporting.¹⁷ Participants wore an 8-channel continuous wave fNIRS system across the forehead to measure changes in activity in the right and left PFC regions. The fNIRS instrument measured oxygenated hemoglobin (HbO) and deoxygenated hemoglobin concentrations at 850 and 760 nm, respectively, with one detector and 4 sources on each side of the head. The probe approximately covered bilateral Brodmann areas BA9, BA44, BA45, and BA46.18 Optical data were collected at 10 Hz and stored using OxySoft software (Artinis Medical Systems).

Raw signals were exported to MATLAB (MATLAB and Statistics Toolbox Release 2021b) for processing with the NIRS Brain AnalyzIR Toolbox.¹⁹ Flat channels due to saturation or equipment error were removed from analysis. This occurred for 8% of total recorded channels across the 4 trials for the 201 participants. Visual checks were performed to confirm no time overlap across tasks and for data quality. Trial observations were excluded where protocol deviations (eg, participant walked during a standing task) occurred. Because all tasks were completed 4 times, exclusion of some observations did not impact overall data availability or quality. We included data from a participant as long as there were valid fNIRS data for at least 2 of the 4 trials. At least one trial was removed from analyses in 67 total participants, and 11 participants were removed from analyses due to missing fNIRS data.

The fNIRS light intensity signals were converted to optical density and resampled to 4 Hz to reduce computational burden. The modified Beer-Lambert law was applied to generate the concentration of HbO (in µM) from optical density data.²⁰ In first-level modeling, a canonical model with autoregressive prewhitening approach using iteratively reweighted least-squares was implemented to model changes in HbO for each task relative to a global baseline, which in our data is the first and last 2 seconds of the data recording.²¹ Previous work demonstrated that this iterative regression model, which uses a robust statistical estimator to downweigh outlier data points, is less prone to bias from motion artifacts in the fNIRS data, which result from slippage of the head cap.²¹ Channel-wise student’s t tests compared the level of HbO change for each task from the quiet standing baseline state. These coefficients from the linear regression were then used to combine the results of the 4 trials using a mixed-effects model with fixed effect of task. The mixed effects model used robust weighted least square regression with the estimated noise covariance matrix derived from the first-level regression model. The channel-specific t-statistic values were averaged over the combined right and left hemispheres of the PFC for all subsequent analyses.

PET/MR imaging

VMAT2 binding by DTBZ provides a measure of the degree of dopamine innervation in the striatal regions of interest and more accurately reflects nerve terminal density compared to other radioligands such as DAT where the signal also reflects a component of compensatory adaptation. [¹¹C]-(+)-a-DTBZ was synthesized with high molar activity (59.9 ± 26.3 MBq/nmol at time-of-injection) using a modification of the methodology reported by Kilbourn et al.²² The major modifications included the use of [¹¹C]methyl triflate as the labelling synthon and the purification of the desired [¹¹C]DTBZ using a semipreparative HPLC column. Radiochemical and chemical purities were >95% as determined by analytical HPLC.

Simultaneous acquisition of [¹¹C]DTBZ PET and 3 Tesla MR images was performed on a Siemens Biograph mMR scanner (Siemens Medical Solutions USA).²³ PET emission data were collected in list-mode over 60-minute postinjection and were reconstructed using filtered back projection with Fourier rebinning into a dynamic series of 20 frames ranging in duration from 15 seconds to 300 seconds. Attenuation correction of PET emission data was performed analytically using previously described model-based methods.^24,25 The [¹¹C]DTBZ distribution volume ratio was estimated from indices of [¹¹C]DTBZ binding derived using the noninvasive Logan graphical analysis method²⁶ using the occipital cortex as a reference tissue. Volumes of interest for sampling of PET image data were defined on the magnetization-prepared rapid gradient echo (MPRAGE) MR series using an automated atlas-based method (CIC Atlas, Clinical Imaging Center, Imperial College).²⁷

We included 2 a priori defined regions of interest: the sensorimotor striatum (posterior putamen) and the associative striatum (anterior putamen, precommissural dorsal caudate, postcommissural dorsal caudate). The sensorimotor striatum is primarily connected to sensorimotor regions of the cortex, including the primary and supplementary motor cortices, which are involved in automatic motor control. The associative striatum is involved in executive functions and has strong connections to the associative cortex, including the PFC. VMAT2 was chosen as the target as it is responsible for the translocation of monoamine neurotransmitters from the cytoplasm into synaptic vesicles. The use of VMAT2 as an index of nigrostriatal dopaminergic integrity is predicated on the fact that ∼95% of monoaminergic synapses in the striatum are dopaminergic.^28,29

Covariates

Age at MRI, sex, race, and education were self-reported. Due to low numbers of non-White participants, we categorized race as either White or non-White, which included 17 Black, 1 Asian, and 1 Native American participant, and 1 participant with unreported race. Education was recorded as completed less than high school, high school or General Educational Development (GED), or greater than high school; this was recoded as less than or completed high school compared to those with more than high school education. Cognitive status was harmonized across the 2 datasets as either unimpaired or impaired. In MMH, cognitive impairment was determined in the parent study by Clinical Dementia Rating³⁰ of 0.5. In SOMMA, cognitive impairment was derived from age, sex, race, and education adjusted percentile scores on the Montreal Cognitive Assessment (MoCA)³¹ in the SPRINT trial³²; the 10th percentile was used to identify both mild cognitive impairment and dementia. Body mass index was calculated from measured height and weight. Systolic blood pressure was recorded by a single measurement in MMH and by the average of 2 measurements in SOMMA. Health conditions in each study are coded as the presence of any of self-reported cancer (excluding nonmelanoma skin cancer), cardiac arrhythmia, chronic kidney disease, chronic obstructive pulmonary disease, coronary artery disease (myocardial infarction), congestive heart failure, diabetes, and stroke. Structural brain measures were obtained by structural 3T MR imaging concurrent with PET imaging. 3T MR images were acquired by T1-weighted MPRAGE and Dixon sequences.³³ Total white matter hyperintensities (WMHs) were assessed using the lesion segmentation tool³⁴ for the MRI analysis toolbox SPM-12. White matter hyperintensity volumes were normalized to the total cerebral white matter volume. Brain atrophy was defined as the ratio of total gray matter volume to intracranial volume.

Statistical analysis

All analyses were conducted in SAS 9.4 using the combined dataset for the 2 studies and with statistical significance at α ≤ 0.05. Variables are presented as either means with standard deviations (SDs) or as n and percent for categorical variables for the combined sample and for each study individually.

We defined 4 groups based on PFC and gait speed change when transitioning from single- to dual-task walking (Figure 1). PFC change was defined by the t-statistic from the fNIRS signal as either a negative or 0 change (PFC−) or a positive change (PFC+) from single- to dual-task. Gait speed change was defined using a threshold of 0.1 m/s decline; those having less than this change were considered to maintain speed from single- to dual-task (gait+), and those with more than 0.1 m/s decline were considered to have slowed (gait−). This gait speed change cut-point was chosen because prior work has demonstrated that differences of 0.1 m/s during dual-task walking are related to greater risk for subsequent onset of mobility disability or cognitive decline,^35,36 and it was approximately equivalent to a 1 SD difference for our sample. Four groups were then generated: (1) PFC−/gait+ who did not have an increase in PFC activation and maintained or increased gait speed from single- to dual-task walking (ie, automatic gait), (2) PFC−/gait− who did not have an increase in PFC activation and had a decline in gait speed from single- to dual-task walking (ie, loss of automaticity without compensation), (3) PFC+/gait+ who had an increase in PFC activation and maintained or increased gait speed from single- to dual-task walking (ie, compensatory PFC signaling), and (4) PFC+/gait- who had an increase in PFC activation with a concurrent decline in gait speed from single- to dual-task walking (ie, neural inefficiency).

Figure 1.

Change from single- to dual-task walking in prefrontal cortex (PFC) activation by functional near-infrared spectroscopy and gait speed (m/s) in older adults (n = 201). Analytic groupings are shown along with the prevalence of each.

The relation of demographic, health, performance, and neuroimaging variables with the PFC/gait groups was assessed by Chi-square tests for categorical variables, except for race and cognitive impairment, where Fisher’s exact test was used to account for small sample sizes within cells. Continuous variables were compared across PFC/gait groups using ANOVA. All bivariate comparisons are presented for overall group differences. Multinomial logistic regression was used to assess the relation of DTBZ binding levels as the independent variable with the PFC/gait groups as the dependent variable with adjustment for covariates. Group 1 (PFC−/gait+) served as the reference group for all analyses. Separate models were built with DTBZ in either the sensorimotor striatum or in the associative striatum as the independent variable. Due to the limited sample size, covariates of study, age, sex, race, education, BMI, blood pressure, presence of health conditions, WMH and brain atrophy were assessed in the model one at a time. Covariates that were significant at P < .05 in these models were then retained in the final adjusted model. Although study, age, and sex were not significant in the individual models, they were felt to be important confounders and were retained in the final adjusted model, along with BMI and the brain atrophy index, which were significantly associated. Sensitivity analyses assessed whether gait performance alone or PFC activation alone was related to DTBZ binding and also excluded those with cognitive impairment.

Results

The analytic sample combined across the 2 cohorts (n = 201, Table 1) was 75 years old on average and predominantly women (63%), White (92%), and with a greater than high school education (83%). In the total sample, 50% had at least one health condition, and 7% were cognitively impaired. Mean single-task gait speed was 1.10 (SD = 0.20) m/s, and mean change in gait speed from single- to dual-task was −0.13 (SD = 0.12) m/s. Alphabet performance did not differ between standing and walking conditions (mean change = 0.03, SD = 0.08). On average, PFC activation decreased from the standing alphabet task to the walking alphabet task (mean change = −1.14, SD = 3.66).

Table 1.

Demographic, health, neuroimaging, and dual-task walking characteristics by group based on prefrontal cortex (PFC) activation and gait performance from single- to dual-task walking in older adults.

		Did not increase PFC (PFC−)		Increased PFC (PFC+)
	Total	Maintained gait (gait+)	Slowed down (gait−)	Maintained gait (gait+)	Slowed down (gait−)
	N = 201	N = 42, 21%	N = 30, 15%	N = 61, 30%	N = 68, 34%
	Mean (SD) or n (%)	Mean (SD) or n (%)	Mean (SD) or n (%)	Mean (SD) or n (%)	Mean (SD) or n (%)	P-valuea
Age (years)	74.9 (4.6)	74.1 (4.2)	74.4 (4.7)	75.1 (4.6)	75.3 (4.8)	.5
Women	127 (63.2%)	26 (61.9%)	19 (63.3%)	38 (62.3%)	44 (64.7%)	.9
White race	185 (92.0%)	40 (95.2%)	27 (90.0%)	56 (91.8%)	62 (91.2%)	.9b
>HS education	165 (82.5%)	32 (76.2%)	26 (86.7%)	52 (86.7%)	55 (80.9%)	.5
BMI	28.4 (5.5)	30.3 (6.2)	29.1 (5.7)	28.5 (4.7)	26.9 (5.3)	.02
Systolic blood pressure (mmHg; n = 173)	130.8 (15.0)	131.0 (13.3)	134.9 (15.1)	130.5 (15.7)	128.8 (15.4)	.4
Any health condition	100 (50.0%)	22 (53.7%)	13 (43.3%)	27 (44.3%)	38 (55.9)	.5
Cognitively unimpaired	186 (93.0%)	39 (92.9%)	27 (90.0%)	57 (95.0%)	63 (92.7%)	.8b
Study (SOMMA)	116 (57.7%)	21 (50.0%)	16 (53.3%)	39 (63.9%)	40 (58.8%)	.5
WMH (mm2; n = 197)	0.014 (0.02)	0.021 (0.046)	0.011 (0.010)	0.014 (0.012)	0.012 (0.018)	.3
Brain atrophy	0.34 (0.03)	0.34 (0.03)	0.34 (0.03)	0.33 (0.03)	0.33 (0.03)	.08
Dual task
Gait speed, single task (m/s)	1.10 (0.20)	1.08 (0.21)	1.07 (0.20)	1.09 (0.16)	1.14 (0.23)	.3
Gait speed change single- to dual-task (m/s)c	−0.13 (0.12)	−0.04 (0.06)	−0.18 (0.05)	−0.06 (0.04)	−0.22 (0.14)	<.001
Rate of correct letters, single task (letters/s)	0.56 (0.16)	0.59 (0.17)	0.54 (0.14)	0.56 (0.17)	0.55 (0.15)	.5
Rate of correct letters change single- to dual-task (letters/s)	0.03 (0.08)	0.03 (0.08)	0.03 (0.08)	0.05 (0.09)	0.02 (0.07)	.09
HbO change single- to dual-task walking (t-statistic)c	0.78 (2.21)	−1.45 (1.08)	−1.56 (1.42)	2.02 (1.55)	2.08 (1.45)	<.001
HbO change single- to dual-task ABC (t-statistic)	−1.14 (3.66)	−2.26 (3.75)	−1.25 (3.62)	−1.14 (3.66)	−0.40 (3.51)	.08
DTBZ binding
Sensorimotor striatum (BPND)	2.44 (0.43)	2.58 (0.48)	2.38 (0.45)	2.35 (0.42)	2.46 (0.38)	.05
Associative striatum (BPND)	2.09 (0.34)	2.17 (0.39)	2.05 (0.40)	2.02 (0.33)	2.12 (0.28)	.1

Abbreviations: BMI, body mass index; BP_ND, binding potential, nondisplaceable; DTBZ, dihydrotetrabenazine; HbO, oxygenated hemoglobin; HS, high school; m/s, meters/second; SOMMA, Study of Muscle, Mobility and Aging; WMH, white matter hyperintensities.

^a P-values are for overall group effect.

^bFisher’s exact test.

^cThese variables only differ across levels that define the groups.

There were few notable differences across the 2 cohorts, but a greater proportion of participants in SOMMA had a greater than high school education (92%) compared with MMH (69%). Mean values and SDs for neuroimaging variables (MRI, PET, and fNIRS) were comparable across the 2 cohorts (Supplementary Table 1). Performance during dual-task walking for both gait speed and rate of correct letters generated was also highly comparable across the cohorts (Supplementary Table 1).

The sample was fairly evenly split among the 4 PFC/gait groups, though PFC+/gait− was the most common (34% of the sample), whereas PFC−/gait− was the least common (15% of the sample; Figure 1). Most demographic and health characteristics did not differ by PFC/gait group (Table 1). BMI was highest for the PFC−/gait+ group (mean = 30.3) and lowest for the PFC+/gait− group (mean = 26.9, P = .02). Change in HbO signal and gait speed from single- to dual-task significantly differed by group (P < .001 for each) as these variables were used to define the groups. Gait speed during single-task walking and rate of correct letters generated on the cognitive task alone did not differ among groups (P = .3 for each). DTBZ binding in the sensorimotor striatum was significantly different between PFC/gait groups in bivariate analyses; PFC−/gait+ had the highest levels of DTBZ binding (mean = 2.58), whereas PFC+/gait+ had the lowest (mean = 2.35; P = .05). There were no significant differences in DTBZ binding in the associative striatum across groups in bivariate analyses (P = .1).

In adjusted multinomial regression analyses (Table 3), those with greater DTBZ binding in the sensorimotor striatum were less likely to be in the PFC−/gait− (odds ratio [OR] = 0.28; 95% CI, 0.08-0.95) or the PFC+/gait+ (OR = 0.29; 95% CI, 0.10-0.84) groups compared to the PFC−/gait+ reference group. There was no difference in DTBZ binding in the sensorimotor striatum for those in the PFC+/gait− compared to the PFC−/gait+ reference group. In adjusted multinomial regression analyses for DTBZ binding in the associative striatum, there was a significant difference for those in the PFC+/gait+ group compared to the PFC−/gait+ reference group (OR = 0.23; 95% CI, 0.06-0.85). No other significant differences were observed for DTBZ binding in the associative striatum, although ORs were comparable to those for the sensorimotor striatum.

Table 2.

Adjusted multinomial logistic regression models of dihydrotetrabenazine (DTBZ) binding for dopaminergic neurotransmission by region of interest with groups based on prefrontal cortex (PFC) activation and gait performance from single- to dual-task walking in older adults (n = 201).

	Did not increase PFC (PFC−)		Increased PFC (PFC+)
	Maintained gait (gait+)	Slowed down (gait−)	Maintained gait (gait+)	Slowed down (gait−)
	N = 42	N = 30	N = 61	N = 68
	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Sensorimotor striatum
Unadjusted	REF	0.31 (0.10, 0.97)	0.28 (0.10, 0.76)	0.48 (0.18, 1.24)
Adjusteda	REF	0.28 (0.08, 0.95)	0.29 (0.10, 0.84)	0.53 (0.18, 1.54)
Adjusteda CN onlyb	REF	0.33 (0.09, 1.18)	0.27 (0.09, 0.83)	0.47 (0.15, 1.44)
Associative striatum
Unadjusted	REF	0.34 (0.08, 1.41)	0.27 (0.08, 0.92)	0.63 (0.20, 2.04)
Adjusteda	REF	0.27 (0.06, 1.23)	0.23 (0.06, 0.85)	0.49 (0.13, 1.82)
Adjusteda CN onlyb	REF	0.34 (0.07, 1.63)	0.21 (0.05, 0.80)	0.47 (0.12, 1.82)

Abbreviations: CI, confidence interval; CN, cognitively normal; OR, odds ratio; REF, reference.

^aAdjusted for study, age, sex, body mass index, and brain atrophy.

^b n = 186.

In sensitivity analyses that defined groups based solely on PFC activation or on gait performance (rather than on the combination of the 2), there were no associations of DTBZ binding with either PFC activation alone or gait performance alone (Table 3). In additional sensitivity analyses excluding those with cognitive impairment (Table 2), results were largely unchanged with the exception that the association of DTBZ binding at the sensorimotor striatum for PFC−/gait− compared to PFC−/gait+ was no longer significant, though the OR was of similar magnitude to analyses including those with cognitive impairment.

Table 3.

Adjusted logistic regression models of dihydrotetrabenazine (DTBZ) binding for dopaminergic neurotransmission by region of interest with groups based on either prefrontal cortex (PFC) activation or change in gait performance from single- to dual-task walking in older adults (n = 201).

	Did not increase PFC (PFC−)	Increased PFC (PFC+)	Maintained gait (gait+)	Slowed down (gait−)
	N = 72	N = 129	N = 103	N = 98
	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Sensorimotor striatum
Unadjusted	REF	0.63 (0.31, 1.25)	REF	0.90 (0.47, 1.72)
Adjusteda	REF	0.70 (0.33, 1.49)	REF	0.92 (0.45, 1.87)
Adjusteda CN onlyb	REF	0.59 (0.27, 1.30)	REF	0.94 (0.45, 1.97)
Associative striatum
Unadjusted	REF	0.67 (0.29, 1.60)	REF	1.13 (0.50, 2.57)
Adjusteda	REF	0.60 (0.23, 1.53)	REF	1.00 (0.42, 2.43)
Adjusteda CN onlyb	REF	0.49 (0.19, 1.32)	REF	1.13 (0.46, 2.80)

Abbreviations: CI, confidence interval; CN, cognitively normal; OR, odds ratio; REF, reference.

^aAdjusted for study, age, sex, body mass index, and brain atrophy.

^b n = 186.

Discussion

In this analysis of community-dwelling older adults, we found that participants with higher dopaminergic integrity in the sensorimotor and associative striatum were more likely to maintain their gait speed during dual-task walking without increased PFC activation (PFC−/gait+) than to do so with increased PFC activation (PFC+/gait+). The highest striatal DTBZ binding was consistently found in the group that maintained gait performance without increasing PFC activation (PFC−/gait+). This is presumably the group with the most automatic control of gait³ who can effectively adapt to the demands of dual-task walking without additional engagement of the PFC, which may be supported by greater nigrostriatal dopaminergic integrity. The PFC+/gait+ group may represent those with reduced automaticity and increased PFC engagement to maintain performance. Although this compensatory engagement of the PFC during walking is important for maintenance of performance under challenging conditions such as dual-tasking, it may come at the expense of availability of PFC resources for dealing with other environmental threats that could lead to falls.⁵ Of note, this group may also be at greater risk for falls based on their lower dopamine levels,³⁷ potentially compounding the risk from lower gait automaticity.

We further found that greater dopaminergic integrity in only the sensorimotor striatum was related to lower likelihood of being in the group that declined in gait speed during dual-task without additional PFC activation (PFC−/gait−) compared to the group that maintained gait speed without increasing PFC activation (PFC−/gait+), though these findings were less robust. This may suggest a loss of automaticity in this group without compensatory PFC activation, possibly due to neural pathologies that we did not assess here. There were no statistical differences in dopaminergic integrity for either region with being in the group that declined in gait performance despite increased PFC activation (PFC+/gait−). It may be that the effects of loss of dopaminergic integrity differ across individuals based on aspects of brain health, facilitating maintenance of dopaminergic neurotransmission and preservation of function in the presence of presynaptic losses in nigrostriatal dopamine neurons. This is supported by the fact that putaminal VMAT2 reductions may be >60% before clinical motor manifestations of Parkinson’s disease are detected. We assessed some other aspects of neurologic health (volume of WMH³⁸ and brain atrophy) as confounders in our analyses, but there are other potential aspects of brain health that could interact with dopaminergic function, such as cholinergic function^39,40 and functional connectivity,^41,42 to determine performance.

Notably, no differences by dopaminergic integrity were observed when groupings were made using either PFC activation alone or gait performance alone, suggesting that neither component alone is driving the observed associations. This finding emphasizes the importance of interpreting cortical fNIRS signals in the context of performance changes.⁷ In addition, findings were generally less strong for the associative striatum compared to the sensorimotor striatum, which may suggest that sensorimotor striatal dopaminergic integrity as measured by DTBZ binding is more indicative of motor automaticity than the ability of the PFC to provide compensatory signaling.

Prior research has suggested that polymorphisms of the catecholamine-O-methyl-transferase (COMT) gene, which is related to greater dopaminergic signaling in the PFC, are related to higher gait speed both cross-sectionally and longitudinally.^43–45 In addition, older adults with mild parkinsonian signs, including slow gait speed, had lower functional PFC-striatal connectivity.⁴² These prior results suggest that dopaminergic integrity in older adults may play a critical role in how the PFC is involved in the control of mobility. However, there are limited data assessing how nigrostriatal dopaminergic integrity is related to mobility and motor control in older adults without Parkinson’s disease, and no prior studies have assessed the relation of dopamine and PFC-mediated motor control during walking.

There are several limitations of this study to note. We assume that declines in dopaminergic integrity precede reduced automaticity, but we cannot establish this temporal relation with our cross-sectional data. Our sample had limited generalizability and primarily represents White individuals with higher educational attainment. Our interpretation of DTBZ-specific binding in sensorimotor and associative striatum as an index of the density of presynaptic terminals of the nigrostriatal dopamine projection is predicated on previous preclinical evidence that VMAT2 expression is linearly related to the integrity of nigrostriatal dopamine neurons and that VMAT2 expression is insensitive to compensatory regulation attributed to lesion or dopamine-enhancing drugs.^29,46 Further, DTBZ binding based on a single resting state acquisition is a predominant static measure of dopaminergic integrity and does not necessarily reflect the dynamic processes needed for adaptation to dual-tasking. Potential speech-related artifacts in the fNIRS signal cannot be ruled out when comparing single-task walking to our verbal dual-task paradigm. However, we observed a lower PFC signal during the standing and talking task compared to standing alone, suggesting that the observed signal during walking and talking is not due to speech-related movement. There are no established cutoffs for defining meaningful change from single- to dual-task gait speed. We based our cutpoint on existing evidence, but our results may be sensitive to this threshold choice. Further, we only assessed gait speed and not other aspects of gait quality which are sensitive to dual tasking⁴⁷ and may be relevant to clinically meaningful outcomes.^48,49 Finally, we had a limited sample size within our PFC/gait groups, which may have limited our power to detect true differences in some cases. However, this is the first sample to our knowledge to assess both PFC activation during complex walking tasks using fNIRS and dopaminergic neurotransmission in the striatum using PET. In addition, we considered the combination of fNIRS signal and gait performance in order to disentangle adaptive versus nonadaptive PFC activation.

In conclusion, we found that higher nigrostriatal dopaminergic integrity, particularly in the sensorimotor region, was related to maintenance of gait speed during dual-task walking without increased PFC activation compared to those who maintained gait speed with increased PFC activation. This suggests that nigrostriatal dopaminergic integrity supports gait automaticity in older adults without Parkinson’s disease and that dopaminergic therapy may prove useful in this population. However, a better understanding of how higher dopaminergic neurotransmission is related to compensatory compared to inefficient PFC signaling in older adults is needed.

Funding

The MYHAT study was supported by the National Institute on Aging (R01AG023651). The Study of Muscle, Mobility, and Aging is supported by funding from the National Institute on Aging (R01 AG 059416). Study infrastructure support was funded in part by NIA Claude D. Pepper Older American Independence Centers, at the University of Pittsburgh (P30 AG024827) and Wake Forest University (P30 AG021332), and by the Clinical and Translational Science Institutes, funded by the National Center for Advancing Translational Science, at Wake Forest University (UL1 TR001420). This analysis was supported by NIA awards (R01AG075025 and U01AG061393).

Conflicts of interest

None declared.

Data availability

Data are available through the parent cohort studies by request.