Effects of repetitive transcranial magnetic stimulation combined with cognitive training on resting-state brain activity in Alzheimer’s disease

Abstract

Objectives

Repetitive transcranial magnetic stimulation (rTMS) is a promising tool to modulate brain plasticity, but the neural basis has been little addressed. The purpose was to investigate the effects of rTMS on resting-state brain activity in patients with Alzheimer’s disease (AD).

Methods

Seventeen patients with mild or moderate AD were enrolled and randomly divided into one of the two intervention groups: (1) real rTMS combined with cognitive training (real group, n = 9); (2) sham rTMS with cognitive training (sham group, n = 8). 10 Hz rTMS was used to stimulate the left dorsolateral prefrontal cortex and then the left lateral temporal lobe for 20 min each day for 4 weeks. Each patient underwent neuropsychological assessment and resting-state functional magnetic resonance imaging (rsfMRI) before and after treatment. The fractional amplitude of low frequency fluctuation (fALFF) of rsfMRI data in real group were: (1) compared to sham; (2) correlated with rTMS-induced cognitive alterations.

Results

Significantly increased fALFF in right cerebellum/declive, left lingual/cuneus and left cingulate gyrus, as well as decreased fALFF in left middle frontal gyrus were found after 10 Hz rTMS, but not after sham stimulation. Using these suprathreshold regions, we found that rTMS increased functional connectivity between the right cerebellum/declive and left precentral/postcentral gyrus. The fALFF increase in left lingual/cuneus and right cerebellum/declive was associated with significant improvement in cognitive function.

Conclusions

rTMS combined with cognitive training induced increased low frequency fluctuation neural oscillations and functional connectivity in brain regions subserving cognition, suggesting a possible neuronal mechanism of the beneficial effects of rTMS.

Introduction

Alzheimer’s disease (AD) is a neurodegenerative disorder characterized by cognitive decline progressively. Currently available drugs for AD treatment are only for symptom improvement. ¹ For this reason, alternative or supplemental therapies to pharmacological treatments have been increasingly utilized in recent years. Applied to the brain transcranially, repetitive transcranial magnetic stimulation (rTMS) is a noninvasive neuromodulation tool that produces favorable outcomes for patients with various neurological and psychiatric disorders. ² Previous studies have demonstrated that rTMS could help improve cognitive function ³ and alleviate behavioral and psychological symptoms of dementia (BPSD) in AD. ⁴ However, among these clinical studies, the neural basis of the effects induced by rTMS has been little addressed.

Over the past decade, various imaging techniques had been employed to improve our understanding on the neural basis of the effects induced by rTMS. Among them, resting-state functional MRI (rsfMRI) provides a versatile tool to endeavor this. Amplitude of low frequency fluctuation (ALFF) is an index to quantitatively measure the total voxel-wise amplitude of low frequency oscillations (LFOs),^5,6 which reflects the local features of brain oscillatory activities at rest. As an advanced ALFF method, fractional ALFF (fALFF) has been utilized in resting-state fMRI research. ⁶ The alterations of fALFF in the AD spectrum were symptom-relevant, which showed gradual disturbances in intrinsic brain activity as the disease progressed.^7,8 fALFF also allows researchers to obtain stable data across scan sessions,^9,10 making it an ideal tool for pre-post treatment assessments. Recently, clinical treatment studies demonstrated that the accurate localization of fALFF alteration was crucial for the evaluation of therapeutic effects in schizophrenia, ¹¹ depression, ¹² and cigarette smokers. ¹³

As a key part of the executive control network and frontoparietal network, the left dorsolateral prefrontal cortex (DLPFC) is one of the most commonly used rTMS targets for treating cognitive impairment. ¹⁴ Stimulation of the temporal lobe (TL) has also been reported in AD patients to improve executive function. ¹⁵ Our previous study has indicated that high frequency rTMS in combination with cognitive training (rTMS-COG) can produce more sustained improvement by changing the metabolic levels of these target sites in AD patients. ¹⁶ However, the effects of rTMS on the resting-state brain activity in these patients was not clear. In this study, we hypothesized that the neural basis may be attributed to an increase of the resting-state brain activity, especially for the most commonly cited low-frequency components in patients with AD. This finding may provide new insights into the underlying mechanism of the neural basis of rTMS treatment for AD.

Materials and methods

Study design and participants

This was a prospective, double-blind, placebo-controlled pilot study carried out from March 2016 to May 2018. All AD patients were randomly divided into real rTMS-COG or sham rTMS-COG with a web-based randomization generator (http://www.randomization.com). Before treatment, all patients underwent magnetic resonance imaging (MRI) scanning and neuropsychological assessment on Saturday (T0). The treatment was scheduled for Monday through Friday. Another MRI scan and neuropsychological assessment (T1) was scheduled for Saturday immediately after all treatments.

This study was approved by the ethics committee of Tongji Hospital (Wuhan, China; Chinese Clinical Trail Registry Registration number: ChiCTR-INR-16009227). Written informed consents from all the subjects or family members were obtained before participation. All AD subjects were recruited from the Department of Neurology in Tongji Hospital. Inclusion criteria are as follows: (1) age more than 50 years old, right-handed; (2) 6 years of education at least; (3) met the criteria of probable AD based on NINCDS-ADRDA ¹⁷ ; (4) Clinical Dementia Rating scale (CDR) memory score of 0.5–2; 5) Hachinski Ischemia Score (HIS) less than 4; (6) stable anti-dementia medication (donepezil or memantine) for at least 3 months. Significant neurological or psychiatric diseases that might result in cognitive dysfunction, unstable systemic condition, and MRI/rTMS contraindications (such as metal implants or claustrophobia) were regarded as disqualified subjects and excluded.

rTMS-COG protocol

The rTMS protocol was applied through a Butterfly focal coil (MCF-B65 coil) guided by marked coordinates via an optical navigation system (Magventure, Denmark). The parameters were as follows: repetition of 10 Hz with each train lasting for 5 s (25 s interstimulus interval), 20 trains were applied which totaled 1000 pulses. We applied 10 Hz rTMS first over the left DLPFC (Talairach coordinates: X = −35, Y = 24, and Z = 48) and then over the left lateral -TL (Talairach coordinates: X = −60, Y = −15, Z = −15). ¹⁸ The treatment lasted for 4 weeks (5 times/week) with no maintenance sessions. The sham stimulation oriented with front edge touching the scalp at 90° (sham conditional coil) at the same scalp position as real rTMS.

The rTMS intensity was set at 100% of the patient’s resting motor threshold to meet safety recommendations. ¹⁹ A screening questionnaire was requested before administering TMS. ²⁰ To ensure that the patient was in good condition, the patient’s blood pressure and heart rate were measured before and after the intervention.

All patients underwent cognitive training up to one hour in conjunction with cortical stimulation by rTMS. The cognitive tasks were selected and downloaded on an iPad tablet (version 9.1, Apple Inc, USA) by a cognitive therapist. Cognitive tasks involved memory tasks, attention tasks, mathematical calculations, agility drills, language tasks and logic thinking tasks. ¹⁶ The memory tasks were selected during stimulation by the cognitive therapist. Other tasks were practiced after the stimulation ceased with no additional maintenance session. All cognitive tasks were guided by the same experienced therapist and completed on the same iPad tablet touch screen.

Neuropsychological assessment

The primary measure is the Alzheimer’s Disease Assessment Scale–Cognitive Subscale (ADAS-cog) score, which is a common tool used clinically to evaluate the reliability and validity of cognitive changes after interventions. ²¹ The maximum score is 70 points, with lower score indicating better cognitive performance. Two versions of the ADAS-cog were used to avoid the learning effect at baseline and after treatment.

The second were the scores on the Addenbrooke’s Cognitive Examination III (ACE-III), ADL, and Neuropsychiatric Inventory (NPI). The higher scores of ACE-III indicates better cognitive function. The lower scores of ADL and NPI indicate better condition. All neuropsychological assessment before and after treatment were assessed by a neuropsychological assessor who was blinded to the allotment.

MRI data acquisition and preprocessing

Both structural MRI and rsfMRI were acquired at 3T MR scanner (Discovery 750,GE healthcare, United States) with a 32-channel head coil. During the scan, a cushion was placed to immobilize the head and reduce motion. For rsfMRI scanning, the subjects passively viewed a fixation cross, not focused on anything and were kept awake. High-resolution T1-weighted structural MRI was acquired with a 3D-T1 brain volume (BRAVO) sequence with repetition time/echo time (TR/TE) = 8.2/3.2 ms, preparation time = 450 ms, FOV = 240×240 mm, matrix = 256×256, NEX = 1, slice thickness = 1 mm. The rsfMRI images were obtained using a gradient-recalled echo-planar imaging (EPI) sequence with the following specifications: TR/TE = 2000/35 ms, FOV = 240×240 mm, matrix = 64×64, slice thickness = 4 mm, time points = 240, and 29 axial slices covering the whole brain.

Data prepossessing of rsfMRI was performed using Data Processing Assistant for Resting State fMRI (DPARSF) V4.3 ²² and Resting-State fMRI Data Analysis Toolkit (REST). ⁹ Functional images, removing the first 10 time points, underwent slice-timing correction and realignment. The maximum translational motion of all subjects was less than 2 mm and the maximum rotation was less than 2°. Functional images registered with the 3D-T1 structural images and then registered into the MNI space at a resolution of 3 × 3 × 3 mm³. They were then smoothed with an isotropic 3D Gaussian kernel with a full width at half maximum of 8 mm. Finally, functional images with linear trend were removed. Several sources of spurious variance (24 head motion parameters, averaged signal from white matter, cerebrospinal fluid and global signal) were regressed out using multiple linear regression analysis.

fALFF analysis

We defined fALFF as the ratio of power spectrum of low frequency (0.01–0.08 Hz) range to that of the entire frequency range computed in (6). Finally, the fALFF value of each voxel became normalized through dividing the global mean fALFF value. A two-sample t-test was performed to examine the possible fALFF difference of the pre-stimulation condition (T0) between the rTMS and sham cohort. Paired t-tests were then used to statistically infer the within-subject pre-(T0) and post- (T1) rTMS or sham fALFF difference. Statistical significance of the analysis results was defined by p < .005 at the voxel level and cluster size>46 at the cluster level (corrected for multiple comparison using the Monte-Carlo simulation-based approach as implemented in AlphaSim). ²³

Functional connectivity (FC) analysis

We focused on FC of any part of the brain to regions with significant rTMS fALFF effects identified in above analysis by using those regions as the seeds. FC was calculated as the correlation coefficient between the mean time-course of the seed region and the assessed voxel. Fisher Z transform was used to convert the correlation coefficient map into a Z map, and the within subject stimulation effects and across-group rTMS versus sham FC change difference were then assessed using paired t-test and two sample t-test like those used for fALFF analysis.

Statistical analysis

Differences in demographic characteristics and baseline neuropsychological measures (T0) between the two groups were analyzed using independent two sample t-test for continuous variables and the Fisher exact test for categorical variables. Paired t-tests were then used to statistically infer the cognitive difference within-subject pre-(T0) and post-(T1) rTMS or sham. A two-sample t-test was also performed to infer the between group (rTMS vs sham) pre-(T0) and post-(T1) stimulation difference of cognitive function.

For correlation analyses, regions with significant fALFF changes during the course of rTMS treatment were extracted as ROIs. Then, the mean fALFF value in these ROIs was then forwarded into the correlation with the alterations in neuropsychological measurements (pre-post) using Spearman correlation analysis. The data analysis was conducted using IBM SPSS software (version 21.0; IBM, Armonk, New York, USA). p < .05 was considered statistically significant.

Results

The study flow chart is shown in Figure 1. Fifty-one AD patients were recruited for the present study. Twenty-nine patients were excluded for no MRI examination. Three patients in rTMS group withdrew after being treated several times for feeling nervous during treatment or unable to hold up when MRI scanning. Two patients in rTMS group were excluded for registration error. Finally, only nine patients in real rTMS group and eight patients in Sham group completed both baseline and follow up neuropsychological tests and MRI scan. The mean frame-wise displacement, computed by averaging the frame-wise displacement at each time point for each subject, was not significant before and after treatment (p > .05). The mean age of the patients in the sham rTMS-COG group (5 women and 3 men) was 66.3 ± 8.1 years and the mean illness duration was 3.6 ± 1.5 years. In the real rTMS-COG group (7 women and 2 men), the mean age was 66.9 ± 7.4 years and the mean illness duration was 3.2 ± 1.5 years. The baseline clinical and demographic variables of the participants are presented in Table 1. At baseline, the two groups did not differ significantly in gender, age, education level, duration of illness, anti-dementia medication (donepezil or memantine), or in the scores for the CDR, ADAS-cog, ACE-III, ADL, and NPI (p > .05).

Figure 1.

Flow diagram of the study. MRI: magnetic resonance imaging; rsfMRI: resting-state fMRI; rTMS: repetitive transcranial magnetic stimulation.

Table 1.

Demographic and clinical characteristics of the study participants.

	sham rTMS-COG (n = 8)	Real rTMS-COG (n = 9)	p Value
Gender (F/M)	5/3	7/2	0.620 a
Age (yrs.)	66.3 ± 8.1	66.9 ± 7.4	0.867 b
Education (years.)	11.5 ± 2.8	12.3 ± 2.0	0.490 b
Duration of illness (years.)	3.6 ± 1.5	3.2 ± 1.5	0.587 b
Medication (donepezil/memantine)	7/1	8/1	1.000 a
CDR (mild/moderate)	5/3	7/2	0.620 a
ADAS-cog	29.3 ± 6.6	25.4 ± 12.0	0.427 b
ACE-III	47 ± 14.0	62.1± 20.9	0.111 b
MMSE	17.6 ± 5.1	19.3 ± 6.9	0.264 b
ADL	31.6 ± 5.2	28.6 ± 6.4	0.298 b
NPI	11.3 ± 5.3	10.1 ± 5.7	0.678 b

rTMS-COG: rTMS combined with cognitive training; CDR: clinical dementia rating; ADAS-cog: Alzheimer’s Disease Assessment Scale–Cognitive Subscale; ACE-III: Addenbrooke’s Cognitive Examination-III; MMSE: Mini-Mental State Examination; ADL: Activities of daily living; NPI: Neuropsychiatric Inventory. Results are expressed as mean±standard deviation.

^aThe p value was obtained using Fisher’s exact test.

^bThe p value was obtained using two-sample two-tail Student’s t-test.

No significant baseline (pre-stimulation condition) fALFF difference was observed between the two groups, regressing out the covariates of age, gender and education years. After 4 weeks treatment, rTMS induced significant fALFF changes but sham didn’t. Figure 2 shows the post-rTMS versus pre-rTMS fALFF comparison results. Significant fALFF increase after 10 Hz left DLPFC/TL rTMS was found in right cerebellum/declive (peak MNI coordinates: x = 9, y = −66, z =−24), left lingual/cuneus (peak MNI coordinates: x = −3, y = −63, z = 6), and left cingulate gyrus (peak MNI coordinates: x = −12, y = −9, z = 39). Significant fALFF decrease after 10 Hz left DLPFC/TL rTMS was found in left middle frontal gyrus (peak MNI coordinates: x = −30, y = −3, z = −69).

Figure 2.

10 Hz rTMS induced significant changes of fALFF in (A) right cerebellum/declive, (B) left lingual/cuneus, (C) left cingulate gyrus, (D) left middle frontal gyrus (Significance level was defined at p < .005, cluster size>46 voxels, AlphaSim corrected). The left side of the image corresponds to the right side of the brain. Color bar represents t values. The warm and cold colors represent higher and lower fALFF after rTMS, respectively. (E) The box plots showed the mean fALFF changes of the four brain regions before and after sham/real rTMS treatment (p < .05). fALFF: fractional amplitude of low-frequency fluctuation; R: right; L: left.

Functional connectivity analysis was based on the right cerebellum/declive seed, the left lingual/cuneus seed, the left cingulate gyrus seed, and the left middle frontal gyrus seed, as defined by the aforementioned suprathreshold post-versus pre-rTMS fALFF difference analysis. Increased right cerebellum/declive-FC after 10 Hz rTMS was found in left precentral/postcentral gyrus (peak MNI coordinates: x = −54, y = −12, z = 36). No significant left lingual/cuneus-FC, left cingulate gyrus-FC, or left middle frontal gyrus-FC was found for comparison of post-rTMS versus pre-rTMS. Sham group didn’t show any significant FC changes.

The changes of ADAS-cog score (p = .028) and NPI score (p = .011) were significant in the real-rTMS group, but not in sham group (p > .05). The ACE-III and ADL scores all had significant changes both in rTMS and sham group (all p < .05). Compared to sham, 10 Hz rTMS-COG yielded greater reduction in NPI score (p = .020) and ADL score (p = .014), indicating better condition of those patients. The fALFF increase in the left lingual/cuneus was associated with significant decrease of ADAS-cog score (r = −0.499, p = .041) (Figure 3(a)) and ADL score (r = −0.701, p = .002) (Figure 3(b)). The fALFF increase in right cerebellum/declive was significantly correlated with the decrease of ADL score (r = −0.500, p = .041) (Figure 3(c)). That is, the more the regional brain activity (i.e. fALFF) in left lingual/cuneus and right cerebellum/declive increased, the more cognitive function improved, accompanying an improvement in the ability of daily living. There was no correlation between the fALFF change in the left cingulate gyrus and cognitive measures, nor in middle frontal gyrus (p > .05).

Figure 3.

The relationship between cognitive changes and fALFF changes in different brain regions. fALFF: fractional amplitude of low-frequency fluctuation; R: right; L: left; ADAS-cog: Alzheimer’s disease assessment scale-cognitive subscale; ADL: activities of daily living.

Discussion

This study used the same population as the published paper from our group, ¹⁶ which used MR spectroscopy to detect the metabolic changes on the target area of rTMS. In the published study, we found a 4-weeks rTMS-COG treatment can improve cognitive function by changing the NAA/Cr ratio in the left DLPFC. However, the effects of rTMS on resting-state brain activity were not clear. In this study, we assessed the effects of left DLPFC/TL rTMS-COG on resting-state brain activity regarding the low-frequency fluctuations and inter-regional FC. Our results showed significantly increased fALFF in right cerebellum/declive, left lingual/cuneus and left cingulate gyrus, as well as decreased fALFF in left middle frontal gyrus after real rTMS but not after sham stimulation. rTMS increased functional connectivity between the right cerebellum/declive and left precentral/postcentral gyrus. After 4 weeks of rTMS-COG treatment, the increase of fALFF in left lingual/cuneus and right cerebellum/declive was associated with significant improvement in cognitive function. These findings provided us new insights into the underlying mechanism of the neural basis of rTMS treatment for AD.

Previous studies have suggested that regional brain activity and distant connectivity in various neural circuits are altered in AD. Specifically, the prefrontal-limbic circuit, including the frontal lobe, cingulate cortex, and parahippocampal gyrus, show aberrant brain activity and connectivity.^8,24-26 It is well known that rTMS can produce changes of cortical activity that outlast the duration of the stimulus train per se(3). The mechanistic basis is now increasingly conceptualized as involving regional brain activity dysfunction and altered connectivity of neural circuits.^3,27 Our findings provide evidence that fALFF alterations in brain regions involving lingual/cuneus, cingulate cortex, and cerebellum, play a key role in the biologic mechanisms underlying the therapeutic effects of rTMS therapy in AD. fALFF represents a method of investigating LFOs specific to a particular region. ⁶ There was evidence demonstrating the abnormalities in fALFF of the lingual/cuneus and cingulate cortex were linked to cognition dysfunction (memory) in early AD.^8,25 The cerebellum is involved in cognitive associative learning and cerebellar fALFF changes has been supposed to correlate with cognitive impairment in Alzheimer Spectrum.^7,24 Patients with mild cognitive impairment (MCI), a less severe clinical state that can evolve in AD, exhibited altered cerebello-thalamo-cortical activations during Stroop tests. ²⁸ We speculated that rTMS stimulation on left DLPFC/TL in this study induced intrinsic brain activity changes through the frontal-limbic and cerebello-thalamo-cortical pathway in those regions. A unique aspect of this study is that we measured both regional brain activity and FC, which may be a more possible explanation for mechanisms of rTMS in AD than using either one alone. FC between right cerebellum/declive and left precentral/postcentral gyrus in part reflects direct and indirect pathways between cerebellum and frontal-parietal cortices. In a word, our study indicates that the effects induced by rTMS can occur at distant interconnected sites in the brain through distinct neural circuits.

To observe the effect of rTMS treatment longitudinally, we assess the relationships between alterations of fALFF and neuropsychological test scores in AD patients before and after rTMS therapy. Our findings indicate that increases of spontaneous brain activity after rTMS, measured by fALFF, were significantly associated with the improvement of cognitive performance, in regions known to subserve memory and cognitive associative learning, such as lingual/cuneus and cerebellum. fALFF values of lingual/cuneus gyrus were significantly correlated with cognitive test scores in cross sectional studies,^7,8 suggesting that these low fluctuation oscillations were linked to clinical symptoms in Alzheimer spectrum. Previous study also found that rTMS-induced hypoconnectivity within default mode network (DMN) was associated with clinical cognitive improvements in patients with MCI, ²⁹ indicating the correlation of resting-state brain activity and rTMS induced effects. However, to the best of our knowledge, none have evaluated both regional intrinsic neural oscillations and distant connectivity after rTMS in patients with mild AD.

This study has several limitations. First, the sample size was relatively small and the rTMS stimulation and sham were applied to separate cohorts, both may contribute variations to the observed effects. Second, the diagnosis of AD in our study was based on clinical assessment and did not incorporate pathologic markers of AD (e.g. amyloid or tau imaging). However, a clinical diagnosis of AD based on the NINCDS-ADRDA criteria has been shown to be highly reliable. ³⁰

Conclusions

In summary, left DLPFC/TL rTMS induced increased low fluctuation neural oscillations and functional connectivity in brain regions involved in frontal–limbic and cerebello–thalamo–cortical pathways, suggesting a possible neuronal mechanism of the well-observed beneficial effects of high-frequency rTMS.

ORCID iD

Yuanyuan Qin https://orcid.org/0000-0002-0673-3200