Basal forebrain activation improves working memory in senescent monkeys

Abstract

Brain aging contributes to cognitive decline and risk of dementia. Degeneration of the basal forebrain cholinergic system parallels these changes in aging, Alzheimer’s dementia, Parkinson’s dementia, and Lewy body dementia, and thus is a common element linked to executive function across the lifespan and in disease states. Here, we tested the potential of one-hour daily intermittent basal forebrain stimulation to improve cognition in senescent monkeys, and its mechanisms of action. Stimulation in five animals improved working memory duration in 8–12 weeks across all animals, with peak improvements observed in the first four weeks. In an ensuing three month period without stimulation, improvements were retained. With additional stimulation, performance remained above baseline throughout the 15 months of the study. Studies with a cholinesterase inhibitor produced inconsistent improvements in behavior. One of five animals improved significantly. Manipulating the stimulation pattern demonstrated selectivity for both stimulation and recovery period duration. Brain stimulation led to acute increases in cerebrospinal levels of tissue plasminogen activator, which is an activating element for two brain neurotrophins, Nerve Growth Factor (NGF) and Brain-Derived Growth Factor (BDNF). Stimulation also led to improved glucose utilization in stimulated hemispheres relative to contralateral. Glucose utilization also consistently declines with aging and some dementias. Together, these findings suggest that intermittent stimulation of the nucleus basalis of Meynert improves executive function and reverses some aspects of brain aging.

Introduction

The cerebral cortex of mammals receives innervation from forebrain cholinergic neurons(1), which facilitate neurotransmission and maintain blood flow and neurotrophic tone in the neuropil(2, 3). The neurodegeneration in this pathway parallels declines in cognition that occur with aging and dementia, and in early phase degeneration in Alzheimer’s disease(4–8). The cholinergic basal forebrain region in the inferior aspect of the globus pallidus is the nucleus basalis of Meynert in primates, and it supplies cholinergic innervation to the cortical mantle(9).

The aged monkey represents a reasonable model to investigate the potential of deep brain stimulation of nucleus basalis of Meynert to halt or reverse cognitive decline. The Rhesus macaque ages at roughly three times the rate of humans(10). The aged primate has declines in cognition analogous to those in humans(11). Beta-amyloid plaques form with age although the presence of neurofibrillary tangles composed of tau is more rare and appears shifted to a later age(12). We investigated whether deep brain stimulation of the nucleus basalis of Meynert in senescent monkeys could halt or reverse cognitive decline.

Results

Sixteen Rhesus monkeys (Macaca mulatta) over the age of 25 (analogous to age >75 in humans) underwent stimulation and/or pharmacological administration in these experiments. We evaluated the behavioral effects of stimulation primarily in five Rhesus monkeys (2 male, 3 female) that were implanted bilaterally with stimulating electrodes and had stable behavioral performance. These monkeys completed sufficient numbers of daily trials before and after stimulation to quantify stable delay period thresholds in an adaptive tracking paradigm (Figure 1B) version of a working memory task (Figure 1A). Prior to implantation, animals learned the task requiring them to remember a cue stimulus presented at the center of a screen. After a variable delay period, the monkey selected between two test stimuli, one of which matched the cue. The behavior featured a two-alternative forced choice procedure and used adaptive tracking to adjust the duration of the delay period. If the monkey chose correctly on three consecutive trials, the trial delay grew roughly 50% longer. Each error resulted in the delay being shortened by 33%. This 3-up, 1-down adaptive tracking can be analyzed by averaging an even number of tracking reversals to converge on the 79% correct delay threshold(13) and allowed monkeys to perform using the same task parameters daily. Our primary behavioral metric was the mean of adjacent pairs of adaptive tracking reversals, one upward facing, and one downward facing, which has the expected value of the 79% correct threshold.

Figure 1.

A. Monkey task. The monkey touches the cue on the touchscreen. The squares darken for the delay period. Then, a match and a distractor appear in the lower squares for the choice. B. Adaptive tracks for monkey performance for one animal before stimulation onset (left), and in the last four days of stimulation in the first 12 weeks (right). Each line traces the delays used in one training session. C. Target electrode placement from CT superimposed on a coronal MRI. The target was the mediolateral center of the globus pallidus, in the floor of the globus pallidus, 2 mm posterior to the crossing of the pallidus by the anterior commissure. MRI section was taken from the NMT template V2 (20). Green outline depicts the activating function radius of the implanted electrode. D. Single animal adaptive tracking threshold estimates in the control period and the ensuing first 8–12 weeks of stimulation. E. Normalized individual changes in threshold each four weeks, and group average (thick gray line), for the same five animals as in D. F. Threshold changes compared across the last four weeks of stimulation for four of the animals in D compared to the next 12 weeks after ceasing stimulation. G. Changes caused by donepezil treatment for 12 weeks in five animals not receiving stimulation.

The implantation target was the floor of the globus pallidus, in the medial-lateral center of the pallidal floor, and 2 mm posterior to the crossing of the globus pallidus by the anterior commissure. Placement was verified with CT imaging, shown in Figure 1C and Figure S1, and/or histological lesions, postmortem. After implantation and recovery, animals performed the working memory task daily, and received unilateral deep brain stimulation consisting of a 60 Hz train of pulses for 20 seconds interspersed with 40 second periods without pulses, for one hour daily. A pulse amplitude of 0.5 mA was chosen to produce an estimated activating function radius of 700 μm. We were particularly interested in the usefulness of deep brain stimulation as a means of improving cognitive function even when not applied concurrently with task execution. Therefore, we delivered stimulation during the task on half the sessions, the during condition, and delivered the stimulation after the task on half the sessions, the after condition. All data in Figures 1 and 2 were recorded in the after condition.

Figure 2.

Long-term changes in the five animals from Figure 1D. The three month time period includes the three month average data from Figure 1D, while the other time points include data from 1F and after in which variable stimulation protocols were followed as detailed in Figure S6. The control data are shown lower in the figure with error bars measuring one standard deviation. The dotted line shows unity.

Cholinergic DBS produces long-term cognitive improvement

Over the three-month period after the initiation of stimulation, the animals’ delay thresholds were higher (Figure 1B, D). In this 12 week period, the monkeys were correct more often, and tracked to lonqger delays. Their changes in threshold are shown in Figure 1D. A generalized linear mixed-effects model (GLMM) in which monkey identity was treated as a random effect and stimulation was a fixed effect estimated a 49% increase in threshold, which was significant (main effect of treatment, Z = 6.83, p=7.3e-12). Post hoc t-tests indicated that each of the five monkeys significantly increased their delay threshold (M1 t₃₃=4.32_, p=0.00013, M2 t₉₉=3.34 p=0.0012, F1 t₃₃=2.42_, p=0.022, F2 t₁₃₅=3.48_, p=0.00068, F3 t₉₈=3.05 p=0.0029).

To assess the time course of changes, we grouped the monkey threshold estimates in four week periods after the initiation of stimulation. These are plotted after normalization in Figure 1E. A GLMM using each four week period as a fixed effect found that the 0–4 week period was 34% larger than the 4–8 week period (main effect of treatment, Z=3.80, p=7.2e-5), and the 8–12 week period was 19% larger than the 4–8 week period (main effect of treatment, Z=2.30, p=0.026). Individual animal thresholds were all highest in the first four weeks, which indicates we observed the largest changes from stimulation in those first four weeks.

In the three months after the initial stimulation period, four of the monkeys were continued in testing for 12 weeks without further stimulation. All four subjects’ performance remained elevated above baseline in the off period, shown in Figure 1F, which suggests endurance of the long-term benefits of stimulation. A GLMM assessing stimulation stopping as a fixed effect resulted in a 14.4% increase in delay threshold after stimulation ended (main effect of treatment Z=2.385, p=0.017). Three monkeys had mean thresholds slightly higher in the 12 week off period than in the last four weeks of stimulation, and one had a slightly lower mean threshold, and no individual animals reached statistical significance in post hoc testing.

To further quantify effects over longer time periods, the animals that had received deep brain stimulation were given more stimulation, either unilateral or in some cases bilateral (see Figure S6), and their working memory thresholds tracked, for months after onset. All five animals retained above baseline working memory thresholds throughout study, over time periods as long as 15 months after implantation, as shown in Figure 2. A GLMM using stimulation as a fixed effect and comparing the pre-stimulation period to each three month period was significant for each three month period (Z scores 5.271, 6.162, 6.031, 6.669 at 6, 9, 12, and 15 months). The three month data points in Figure 2 are the same depicted in Figure 1D.

Control monkey data, which consisted of animals in the same age range that performed the same task over months of time without stimulation, were compiled for comparison. Working memory performance declines with age in macaques, with animals over age 25 reaching delays that are approximately 50% lower than animals under age 10 (14). Each animal’s threshold was measured over each nonoverlapping 90 day period for which data existed, and then averaged within each animal to create a distribution from these nine animals. The average three month change across these nine control animals was a 3.4% decrease in working memory threshold, with a standard deviation of 16.6%. We projected this three month statistical trend each three months to fifteen months for comparison. At three months, the smallest difference in working memory delay change between the control mean and one of the five stimulated animals was a Z-score of 1.645. Five out of five animals exceeding that Z score has an associated probability of 3.1×10⁻⁷. At the 6, 9, 12, and 15 month intervals, the data were also significant (minimal Z=1.17, n=5, p=2.59×10⁻⁵, Z=1.857, n=5, p=3.18×10⁻⁸, Z=1.865, n=4, p=1.12×10⁻⁶, Z=1.65, n=2, p=0.002). These sample control data are shown below the stimulation animal data in Figure 2.

Cholinesterase inhibitors and off-target stimulation

To assess whether cholinesterase inhibitors could also provide enduring improvement of working memory, monkeys were given the commonly prescribed cholinesterase inhibitor donepezil daily after their working memory tests. The treatment of dementia often applies donepezil, which prolongs the action of acetylcholine by inhibiting its hydrolysis. We have previously shown that it can acutely improve working memory in monkeys if given before testing(15). Monkey performance in the period before and after donepezil administration is shown in Figure 1G, for a cohort of five monkeys that did not receive stimulation during this testing period. A GLMM on the impact of donepezil on delay threshold resulted in a significant 15.9% increase in delay threshold (Z=2.378, p=0.017). Post hoc t-tests found a positive significant impact only in animal F1 (t₂₈=7.29_, p = 6.02e-8). Animals that appear both in Figure 1D and 1G were tested for the impact of donepezil prior to washout and implantation.

A further positive stimulation control treatment consisted of electrodes that were not placed in the subpallidal basal forebrain and were stimulated with the same parameters as the treatment group. As shown in Figure S3, this treatment did not result in a consistent improvement in working memory duration. Electrode misplacement by more than 2 mm from the desired targets occurred in either the mediolateral or anterior posterior plane. A GLMM using stimulation as a fixed effect on these three animals resulted in a significant 17.0% decrease in threshold caused by stimulation (main effect of treatment, Z=2.46, p=0.014).

Timing and duration of stimulation

Prior work has shown that basal forebrain stimulation could acutely impair or improve working memory depending on the stimulation pattern in young monkeys(15, 16), in addition to producing long-term improvements. We examined this effect in elderly monkeys by comparing behavior in the during condition to the after condition, shown in Figure 3A. A GLMM using acute stimulation as a fixed effect, as opposed to interleaved behavioral days in the after condition, revealed that thresholds were 12.9% lower during stimulation (main effect of treatment, Z=2.784, p=0.005). Post hoc testing reached significance for lower thresholds in the during condition only for monkey M1 (t₂₅₃=4.38 p=0.00001). Four of the five animals had lower mean and median threshold estimates in the during condition, while one had a lower mean and median condition in the after condition. This effect contrasts with our observations in young monkeys using the same stimulation pattern, 20 seconds on and 40 seconds off, for which all animals performed better in the during condition. We next examined whether these elderly monkeys would benefit from different stimulation patterns.

Figure 3.

A. Acute effects of stimulation. Plotted in purple are estimates of threshold taken during stimulation, and plotted ingreen are estimates of threshold on stimulation days in the after condition. Green data points represent the same data shown Figure 1D. B. Adaptive tracking delay plots from five sample behavioral sessions from monkey F2 during the prestudy period (left), during bilateral 20–40 stimulation (middle), and during 10–40 stimulation (right). C. Threshold estimate distributions for animals F1 and F2 under different bilateral recovery period conditions. D. Threshold estimate distributions for animals F1 and F2 under different bilateral stimulation period conditions.

The stimulation pattern was altered to use a balance of stimulation and recovery that had less stimulation or more recovery. The prior parameters had impacts on long-term behavior that were found serendipitously, and not through systematic exploration of the parameters. Recent mouse imaging work also suggested that the stimulation period was too long, and the recovery period too short, to achieve maximal acetylcholine bursts in each cycle(17). The two animals most recently implanted in our study (subjects F1 and F2) were stimulated bilaterally, and different parameters were tested.

As shown in Figures 3C–D, S4, and S5, working memory threshold was highest during periods in which the stimulation parameters included a 10 second on period and a 40 second off period (10–40) in both animals. Example adaptive tracking delays from female monkey F2 are shown in Figure 3B for the pre-stimulation period on the left, after applying the 20–40 stimulation pattern in the center, and after applying the 10–40 stimulation pattern on the right. Two different experiments compared the effect of altering the stimulation period by comparing 5–40, 10–40, and 20–40 patterns, and of altering the recovery period by comparing 10–30, 10–40, and 10–50 patterns. In both cases, stimulation occurred at 60 pulses per second, and 60 cycles of the pattern were repeated. A GLMM assessing a fixed effect of stimulation period found a significant 28.4% longer delay threshold in the 10–40 period than in the 20–40 period, which was significant (main effect of treatment, Z=2.815, p=0.005). The difference between 10–40 and 5–40 was not significant, and the 10–40 mean was higher. A GLMM using the recovery period as a fixed effect found a significant 42.8% longer delay threshold in the 10–40 period than in the 10–50 period, which was significant (main effect of treatment, Z=2.815, p=0.006). The difference between the 10–30 and the 10–40 period was not significant, but the 10–40 mean was higher. The entire series of tested stimulation parameters are shown in Figure S4.

Metabolic and neurotrophic mechanisms of long-term improvement

We investigated the underlying mechanisms through which stimulation improved cognitive performance in two ways. In the first, brain glucose utilization was evaluated in three animals by applying Fluorodeoxyglucose F18 (FDG) PET imaging before and after a 7–11-month period that included 2–3 months of unilateral stimulation. Glucose utilization is a marker that consistently tracks progression of aging and Alzheimer’s dementia(18, 19). We anticipated that the effect of time would be to decrease glucose utilization, while perhaps unilateral stimulation would reduce that decrease, or even increase utilization. Animals received one hour of daily intermittent unilateral stimulation, and the effect of stimulation on the standardized uptake value ratio (SUVr) relative to the ipsilateral pons was calculated(20). Both sides decreased in their average SUVr. The control side, contralateral to stimulation, averaged a decrease of 5.84% with a standard deviation of 0.70%. The stimulation side decreased by 3.59% with standard deviation of 2.12%. To look at potential regional changes in glucose uptake, a voxel-wise cluster analysis was performed. Starting with the control side (Figure 4A-left), the statistical maps were thresholded to identify contiguous clusters in which voxels trended down more strongly. These clusters with decreased utilization made up 18% of the hemisphere, and there were no clusters of increased utilization. Then, the stimulated side was analyzed after adjusting for the average changes in the control hemisphere. Three significant patches of relatively increased glucose utilization resulted (all t₈>3.355, p<0.005, see Methods), with no clusters of decreased utilization, shown in Figure 4A-right. Using atlas parcellation from the precise CHARM atlas (AFNI)(21), we also compared glucose utilization cross-hemisphere by region in these three animals, and the regions strongly trended to have a smaller decrease in utilization on the stimulated hemisphere (40 of 51 comparisons positive, p = 5.70×10⁻⁵, binomial test) as shown in Figure 4B. Similar results were obtained when the cerebellum was used for the calculation of the SUVr (Figure S5, 38 of 51 comparisons positive, p=0.0003 binomial test).

Figure 4.

4. A. Statistical maps of lateralized effects on 18-F PET-FDG SUVr. Hemisphere on image left: significant clusters with time-dependent changes in glucose uptake. Hemisphere on image right: significant clusters with stimulation-dependent changes in glucose update, after controlling for effects in the control hemisphere. B. Comparison of change ratios in PET SUVr between stimulated and unstimulated in 17 cortical regions. C. Tissue plasminogen activator was measured in cerebrospinal fluid under neutral conditions or after a one hour stimulation period in four animals.

The second investigation of potential mechanisms underlying effects on working memory assessed whether activation of neurotrophic pathways was occurring in our preparation. Reduced neurotrophin expression has been documented in parallel with brain changes in aging and dementias, and increases in neurotrophins are being tested as treatments for dementias(22). Rodent work has indicated that expression of the receptors for the neurotrophins nerve growth factor (NGF) and brain derived neurotrophic factor (BDNF) are increased by this stimulation(17), which suggests that neurotrophins are being released and/or activated. This release is hypothesized to occur into the interstitial fluid of the cerebral cortex, which slowly mixes with cerebrospinal fluid(23). The fluid turns over in 6–8 hours, and mixes completely with cerebrospinal fluid (CSF) in less than half that time. In four of the monkeys, CSF was sampled before and after one hour of stimulation in the 20–40 condition. The fluid was tested for its concentration of tissue plasminogen activator (tPA), which is an activating element in the pathways for both NGF and BDNF. We found tPA levels in the same range as is documented in humans(24), and that the levels increased significantly and by up to three-fold by stimulation (t₃=3.59,p<0.04), as shown in Figure 4C.

Discussion

Deep brain stimulation has been used extensively in the treatment of movement disorders, and increasingly for the treatment of depression, epilepsy, and obsessive-compulsive disorder(25–27). Multiple studies have attempted stimulation of the nucleus basalis of Meynert in clinical trials(28–31). In most applications, continuous, 24 hour per day high frequency pulse trains are used to normalize the hyperactive circuits that receive the stimulation. Multiple hypotheses exist on the mechanisms that support this normalization(32). In contrast, shorter pulse trains are used in neurobiology to activate cortical circuits(33–35), rather than to suppress them. Limited duration pulse train stimulation of the basal forebrain increases cortical acetylcholine levels(36). Pulse stimulation depolarizes membrane potentials in proportion to the second spatial gradient of the voltage field, often termed the activating function(37). With suprathreshold, repetitive pulse stimulation, axons within the activating function radius will depolarize and generate action potentials(38), recruit closely synaptically coupled neurons(39, 40), and activate the local circuit.

We thus applied short pulse trains, with recovery periods, over one hour daily with intent on increasing the activity of the circuits. Our results establish this as a viable strategy for the long-term cognitive function improvement in aged individuals. Basal forebrain stimulation was superior to donepezil administration, and further, the effect was selective for the targeted subpallidal basal forebrain, as nearby, but off target, stimulation was impairing. The superiority of basal forebrain stimulation was evident in the overall increase in delay threshold, 49% for stimulation versus 15.9% for donepezil, and by the number of animals that showed a statistically significant improvement upon post hoc analysis, all five animals for stimulation and only one out of five animals for donepezil. This response rate for donepezil in aged monkeys is somewhat lower than the rate in Alzheimer’s clinical trials, although a wide range of response rates have been observed(41). The neurodegeneration of the basal forebrain region predicts non-motor executive function deficits in Alzheimer’s, Lewy Body and Parkinson’s dementia(7, 8, 42, 43), so its improved function from intermittent deep brain stimulation is a candidate for improving cognition in these populations.

The main mechanism of long-term improvement did not require stimulation to be delivered concurrently with behavior. The duty cycle of 20 seconds on, 40 seconds off led to impaired acute performance in aged monkeys. Our prior work in young male monkeys found consistent acute improvement with the same parameters(15, 16, 44). Together, these data suggest that the acute impact of intermittent stimulation using parameters defined from young male monkeys result in different neuromodulation in elderly animals. The tuning of the parameters for optimizing long-term improvements in working memory further revealed that half as much stimulation per cycle, 10 seconds instead of 20, resulted in 28.4% longer working memory duration. Elderly animals may require a different balance of intermittent stimulation and recovery to optimize acute and long-term effects.

Longitudinal PET-FDG imaging was used to examine the impacts of stimulation on brain glucose utilization. Generally, advancement in age is associated with decreased glucose utilization(45). The effect can be translated as a 7% loss in glucose utilization per decade of life(46, 47), and these findings apply equally in longitudinal scans of the same individuals. Our findings indicate that glucose utilization was decreased overall as expected in each cortical hemisphere compared to control regions not normally impacted by age or AD (i.e., the pons and cerebellum), but the decrease was lower in the stimulated hemisphere relative to the unstimulated hemisphere.

As intended, the stimulation perfuses the cerebral cortex with acetylcholine, with a recovery period, prior to perfusing the cerebral cortex again. The increased NGF and BDNF receptor expression found from this stimulation in mice(17), along with the increased CSF concentration of tPA observed in the present study, support the premise that this pattern of stimulation increases the forebrain projection-related acetylcholine release to enhance neurotrophic activity. Cholinergic activation has previously been shown(3, 48) to cause the release of neurotrophins and tissue plasminogen activator (tPA), which lead our hypotheses for underlying processes causing the improved behavioral function. The neurotrophins are necessary for multiple forms of synaptic and map plasticity in the cerebral cortex(49, 50), and in the nonhuman primate preparation tPA and neurotrophins may be readily measured by their concentrations in CSF. The release of NGF by stimulation should elaborate the arborization of cholinergic axons into the cerebral cortex, and result in greater acetylcholine release(51). BDNF, in turn, has effects that include increased coupling between parvalbumin interneurons and pyramidal neurons(52).

The potential application of this form of stimulation to improve executive function, especially in dementia, is also beginning to be explored. Three Parkinson’s patients, receiving this pattern of stimulation for two weeks, improved in their sustained attention, compared either to baseline, or to performance after two weeks of continuous pulsetrain stimulation(53). Deep brain stimulation to reduce activity in the stimulated neural circuits has been heavily explored for the past two decades; this work, and the work of others(35), are beginning to explore the potential of brain stimulation to increase activity in the stimulated neural circuits.

Highlights.

• The basal forebrain and its cholinergic projections are the sole source of acetylcholine for the cortical mantle in primates and humans.

• Forebrain function tracks cognitive loss throughout the adult lifespan.

• One hour per day intermittent stimulation of this region improves executive function behaviors and plausibly reverses some aspects of brain aging, a large risk factor in dementias.

• This stimulation exceeds impacts of standard pharmacotherapies, is enduring, recruits brain neurotrophic pathways and improves cortical glucose utilization.

Data and Material Availability:

all raw data will be uploaded and accessible freely upon publication. Implantable pulse generators and accompanying hardware were provided by Boston Scientific through a public/private partnership. Contact dblake@augusta.edu about any other materials.