Hippocampal-Subregion Mechanisms of Repetitive Transcranial Magnetic Stimulation Causally Associated with Amelioration of Episodic Memory in Amnestic Mild Cognitive Impairment

Abstract

Background:

Altered hippocampal subregions (HIPsub) and their network connectivity relate to episodic memory decline in amnestic mild cognitive impairment (aMCI), which is significantly limited by over-dependence on correlational associations.

Objective:

To identify whether restoration of HIPsub and its network connectivity using repetitive transcranial magnetic stimulation (rTMS) is causally linked to amelioration of episodic memory in aMCI.

Methods:

In the first cohort, analysis of HIPsub grey matter (GM) and its functional connectivity was performed to identify an episodic memory-related circuit in aMCI by using a pattern classification approach. In the second cohort, this circuit was experimentally modulated with rTMS. Structural equation modeling was employed to investigate rTMS regulatory mechanism in amelioration of episodic memory.

Results:

First, in the first cohort, this study identified HIPsub circuit pathology of episodic memory decline in aMCI patients. Second, in the second cohort, restoration of HIPc GM and its connectivity with left middle temporal gyrus (MTG.L) are causally associated with amelioration of episodic memory in aMCI after 4 weeks of rTMS. Especially important, the effects of HIPc GM changes on the improvement of episodic memory were significantly mediated by HIPc connectivity with MTG.L changes in aMCI.

Conclusion:

This study provides novel experimental evidence about a biological substrate for the treatment of the disabling episodic memory in aMCI patients. Correction of breakdown in HIPc structure and its connectivity with MTG can causally ameliorate episodic memory in aMCI.

Keywords

Amnestic mild cognitive impairment episodic memory functional connectivity grey matter hippocampalsubregion pattern classification repetitive transcranial magnetic stimulation

INTRODUCTION

Amnestic mild cognitive impairment (aMCI), characterized by a decline in episodic memory, is a high-risk factor for Alzheimer’s disease (AD) [1, 2]. Episodic memory deficit, a hallmark of AD [3 –5], involves not only local alterations of the hippocampus, but also to a dysfunctional hippocampus network [6, 7]. Our lack of understanding of the pathophysiology hampers the development of new interventions to ameliorate episodic memory and prevent the progression of aMCI to AD.

Neuroimaging has been thought to potentially reveal the neuroanatomical basis of episodic memory impairment in aMCI; however, so far, no consistent neural circuit or network has been identified for therapeutic targets to improve episodic memory. A potential interpretation for this inconformity involves functional heterogeneity in hippocampal subregions (HIPsub) [6 , 9]. Indeed, the hippocampus includes the memory processes of consolidation and emotional regulation [8, 10]. In a recent neuroimaging meta-analytic study, the neurofunctional topography of the hippocampus was characterized, and showed that the left HIPsub was composed of the anterior emotional region (HIPe), the middle cognitive region (HIPc), and the posterior perceptual region (HIPp) [8]. Therefore, converging evidence suggested that a posterior hippocampus that closely linked to memory performance may contribute to episodic memory decline in aMCI patients [11, 12]. However, most published neuroimaging studies have only investigated the correlational relationship between neuroimaging characteristics of the hippocampus and episodic memory impairment. They did not look at how neuroimaging characteristics causally affect episodic memory impairment in aMCI patients. To address these issues, a within-subjects design can be used to identify symptom-pathophysiology associations in which pathophysiology is experimentally manipulated to determine the effect on clinical symptoms [13, 14].

Recently, neuromodulation techniques, including repetitive transcranial magnetic stimulation (rTMS) have been broadly applied to evaluate the symptom-specific response to rTMS [13 –18]. Especially, rTMS targeting the parietal cortex within hippocampal connectivity networks could improve associative memory performance in healthy individuals [19]. rTMS was performed to promote the improvement of episodic memory by targeting the precuneus in aMCI patients [18], for which the precuneus is well-characterized as a component of hippocampal intrinsic connectivity networks [19, 20] and a critical vulnerability area for the episodic memory deficits observed in early AD [21, 22]. However, how rTMS modulates hippocampal network connectivity in episodic memory amelioration is still mostly unclear.

In this study, we propose a strategy to empirically identify an episodic memory-related circuit in aMCI patients using a pattern classification approach (SVM: support vector machine), and subsequently, in a targeted manner, this circuit was experimentally manipulated to assess causal links in a separate cohort. In several rTMS studies, the feasibility of manipulating network functional connectivity [13, 23] was previously indicated in a unique way that modulated the targeted network [13, 14] and simultaneously ameliorated behavior [14, 19]. It is reasonable to speculate that, if HIPsub circuit breakdown and episodic memory decline are causally related, the modulation of this circuit should be reflected in the amelioration of episodic memory.

The objective of this study was to identify HIPsub circuit pathology of episodic memory decline in a cohort of aMCI patients. After identifying the circuit pathology, we further investigated, in a separate cohort, the effect of rTMS at the HIPsub circuit that was most closely related to episodic memory decline. We hypothesized that if breakdown in the HIPsub circuit was causally linked to episodic memory decline in aMCI patients, rTMS restoration of the circuit abnormality should be reflected in episodic memory amelioration. Supplementary Figure 1 displays the data analysis pipeline conducted in this study.

MATERIALS AND METHODS

Participants

Data used in this study were obtained from our in-home database: Nanjing Brain Hospital-Alzheimer’s Disease Spectrum Neuroimaging Project (NBH-ADsnp) (Nanjing, China), which is constantly being updated. Relevant information of the NBH-ADsnp is summarized in the Supplementary Methods.

Network identification of altered HIPsub related to aMCI

This study was approved by the responsible Human Participants Ethics Committee of the Affiliated Brain Hospital of Nanjing Medical University (No. 2018-KY010-01 and No. 2020-KY010-02) (Nanjing, China). Written informed consent was obtained from all participants.

According to our criteria, a total of 129 elderly individuals participated in this study. Of the subjects, 2 healthy controls (CN) and 16 aMCI patients were excluded due to excessive head movement (see quality assurance section below), and incomplete or missing magnetic resonance imaging (MRI) data. The eligible subjects included in final analyses were 55 CN and 56 aMCI (Table 1). Detailed inclusion and exclusion criteria are provided in the Supplementary Methods.

Table 1

Demographics, clinical measures, and head rotation parameters of aMCI and CN subjects

Items		CN n = 55	aMCI n = 56
Age (y)		62.91 (5.94)	64.30 (7.70)
Gender (male/female)		23/32	15/41
Education level (y)		12.51 (2.51)	11.15 (2.87) ^*
MMSE		28.58 (1.43)	27.29 (1.59) ^*
MoCA		25.05 (2.42)	22.73 (2.88) ^*
MDRS		141.46 (2.33)	136.45 (6.67) ^*
HAMD		1.82 (2.26)	3.80 (3.64) ^*
ITV		1130.24 (114.65)	1077.72 (118.39) ^*
Episodic memory tests
AVLT-IR	raw score	19.15 (4.36)	15.13 (4.29) ^*
	Z score	0.35 (0.94)	–0.51 (0.92) ^*
AVLT-5 min-DR	raw score	6.35 (2.20)	4.23 (2.29) ^*
	Z score	0.34 (0.93)	–0.55 (0.96) ^*
AVLT-20 min-DR	raw score	6.30 (1.94)	4.23 (2.29) ^*
	Z score	0.40 (0.73)	–0.68 (0.99) ^*
AVLT-total	raw score	31.79 (7.61)	22.79 (8.08) ^*
Composite Z scores of episodic memory		0.37 (0.77)	–0.58 (0.85) ^*
Head rotation parameters
FD_VanDijk		0.05 (0.03)	0.04 (0.03)
FD_Power		0.18 (0.08)	0.16 (0.09)
FD_Jenkinson		0.09 (0.04)	0.08 (0.05)

Data are presented as the mean (standard deviation, SD). CN, healthy controls; aMCI, amnestic mild cognitive impairment; MMSE, Mini-Mental State Examination; MoCA, Montreal Cognitive Assessment; MDRS, Mattis Dementia Rating Scale; HAMD, Hamilton Depression Scale; ITV, Intracranial volume; AVLT-IR, Auditory Verbal Learning Test–immediate recall; AVLT-5 min-DR, Auditory Verbal Learning Test–5-minute delayed recall; AVLT-20 min-DR, Auditory Verbal Learning Test–20-minute delayed recall; FD, framewise displacement. ^*Significant differences were found between CN and aMCI patients. MMSE, MoCA, and MDRS are displayed as raw scores. In this study, the composite Z scores were used to indicate the level of each cognitive domain.

Validation of causal relationship between altered HIPsub and episodic memory decline

A total of 24 patients with aMCI who were recruited for a clinical trial (No. ChiCTR1900022287) were included in this study from the NBH-ADsnp database.

rTMS was used to stimulate the precuneus of aMCI patients for 4 weeks in a sham-controlled design. rTMS (or sham) was applied daily, at the same time each day, in a 25-session course every Monday to Friday. Clinical measures, episodic memory tests, and MRI data were collected at baseline (pre-rTMS or sham intervention), after 2 weeks of rTMS or sham, and at the end of 4 weeks of rTMS or sham. A total of 24 aMCI patients were enrolled in the study (Notes: After baseline MRI scans and cognitive assessments, 4 aMCI patients temporarily dropped out of the study due to personal reasons), of which 20 aMCI patients were randomly divided into real rTMS (10 aMCI) or sham (10 aMCI) and 12 aMCI patients completed the trial of 4 weeks of rTMS (Notes: 4 aMCI patients completed 2-week rTMS treatment and did not want to continue the experiment due to job change. After completing 1-week rTMS treatment, 4 subjects did not want to continue to participate due to the long journey). Demographics, clinical measures, and episodic memory are summarized in Supplementary Table 1.

Neuropsychological assessment

Neuropsychological assessments were as described in our previously published studies [6 , 24–27]. All subjects underwent a standardized clinical interview and comprehensive neuropsychological assessments that were performed by neuropsychologists (Drs. Xue, Qi, and Liu), including Mini-Mental State Examination (MMSE), Montreal Cognitive Assessment (MoCA), Mattis Dementia Rating Scale (MDRS), Auditory Verbal Learning Test-immediate recall (AVLT-IR), Auditory Verbal Learning Test–5-min delayed recall (AVLT-5 min-DR), and Auditory Verbal Learning Test–20-min delayed recall (AVLT-20 min-DR). These tests were used to evaluate general cognitive function and episodic memory, respectively.

MRI data acquisition, structural MRI, and fMRI data preprocessing

Detailed MRI data acquisition parameters in NBH-ADsnp and the image processing procedure are summarized in the Supplementary Methods.

Quality assurance

To ensure reproducibility, test–retest reliability, and replicability on the fMRI metrics, we performed three strict quality controls: controlling for brain atrophy effect, head motion effect, and a strict multiple comparison correction.

Brain atrophy effect

Given that significant grey matter (GM) atrophies in aMCI patients [7, 24] have been reported, the anatomical differences between groups may affect these differences on the functional connectivity (FCs) of HIPsub. To clarify this issue, we computed global intracranial volumes (ITV) based on native GM, white matter, and cerebrospinal fluid (CSF) in CN and aMCI patients by using in-home MATLAB codes. We investigated the between-group differences on the network connectivity of HIPsub using ITV as an additional covariate in the general linear model (GLM) analysis.

Head motion effect

In this study, we used three approaches to control the head motion effect both at the individual and at the group levels. Firstly, we excluded aMCI patients with excessive head motion (cumulative translation or rotation > 3.0 mm or 3.0°). Then, we used a Friston 24-parameter model to regress out head motion effects from the realigned data [28]. Secondly, we also performed a ‘scrubbing’ procedure to scrub frames (volumes) with an excessively high whole-brain root mean square signal change over time in the preprocessed fMRI data for each individual [29 –31]. Furthermore, all volumes with a framewise displacement (FD) greater than 0.2 mm as nuisance covariates were regressed out, and any scan with 50% of volumes removed were discarded as described in a previous study [14]. Overall, 1 CN and 7 aMCI were excluded because of excessive head movement. No significant differences were observed in the head motion parameters between qualified CN and aMCI patients (Table 1).

Strict multiple comparison correction strategy

To ensure the reproducibility, test–retest reliability, and replicability on the fMRI metrics, we performed a strict multiple comparison correction [32], that is, statistical maps were thresholded using the permutation test with Threshold-Free Cluster Enhancement (TFCE) [33] and the false discovery rate (FDR) (p < 0.05), as implemented in DPABI [34]. For cluster-extent permutation tests, voxel thresholds of two-tailed p < 0.02 (Z > 2.3) were set. Finally, a two-tailed p < 0.05 threshold was set (1,000 permutations in FDR evaluation).

Definition of hippocampal subregions

Our definition of HIPsub referred to recent studies from Robinson et al. [8] and Bai et al. [10]. These studies used coactivation-based parcellation to reveal a subspecialization in the hippocampus. Their findings showed that the right hippocampal segmentation is ambiguous, therefore, we selected only the left hippocampal subregions as regions of interest (ROI) (Fig. 1a) based on the recent study published by Bai and colleagues [10]. The left hippocampus was defined as three subregions, which were appropriately located at the anterior involving in emotional processes (HIPe), middle involving in cognitive processes (HIPc), and posterior involving in perceptual function (HIPp). It is worth noting that the overlapping areas between adjacent clusters were removed (Fig. 1b) as presented by Bai and colleagues [10], which may affect the subsequent analysis. Non-overlapping clusters were selected as ROIs for further analysis (Fig. 1c). Furthermore, several studies have confirmed that HIPe is related to emotional processing [8, 10]. In this study, to more accurately show the causal relationship between restoration in the HIPsub circuit (HIPc and HIPp) and episodic memory amelioration, we mainly discussed the changes in memory function of aMCI after rTMS treatment and its possible mechanisms. Therefore, we put only results of two subregions (HIPc and HIPp) and excluded results of this subregion (HIPe) in the main body. To show the analysis results more comprehensively, the results of HIPe are also included in the Supplementary Material.

Fig. 1

Schematic diagram of hippocampal subregions (sagittal views) in the left hemisphere. A) Hippocampal subregions were referred to recent studies published by Robinson et al. [8] and Bai et al. [10], who used coactivation-based parcellation to reveal a subspecialization in the hippocampus by a data-driven method. Hippocampal subregions included HIPe (blue), HIPc (red), and HIPp (green). B) The magenta area and yellow area indicate overlapping areas between HIPe and HIPc, and between HIPc and HIPp, respectively. C) HIPe, HIPc, and HIPp indicated for further analysis. Noteworthy, the overlapping areas between adjacent clusters were removed as referred to Bai et al. [10], which may affect the subsequent analysis. HIPe, hippocampal emotional region; HIPc, hippocampal cognitive region; HIPp, hippocampal perceptual region.

Functional connectivity analyses

For each individual, the average time courses for all voxels within each HIPsub were extracted as the reference time course. Then, we performed voxelwise cross-correlation analysis between the averaged time courses of all voxels within the seed HIPsub region and each voxel in the remainder of the whole brain within the group-specific GM mask. Finally, to increase the normality of the correlation coefficients, Fisher’s z-transform analysis was performed.

rTMS protocol

Previous studies have confirmed that rTMS can promote the improvement of episodic memory by targeting the precuneus in aMCI patients [18], for which the precuneus is well characterized as a component of hippocampal intrinsic connectivity networks [19, 20] and a critical vulnerability area for the episodic memory deficits observed in early AD [21, 22]. Therefore, this study performed rTMS by targeting the precuneus in this study. rTMS was delivered using a Magstim Rapid2 magnetic stimulator with a 70-mm figure-8-shaped coil. rTMS was used to stimulate the precuneus of all aMCI patients. The Pz site of the 10–20 electroencephalogram system was used to locate the precuneus, and the tip of the intersection of the two coil loops was placed at the Pz site to stimulate the precuneus [18].

rTMS was applied, using trains of 1000 stimuli at a frequency of 10 Hz and at an intensity of 100% of the motor threshold (MT). The MT was defined as the lowest intensity producing motor evoked potentials of greater than 50μV in at least 5 out of 10 trials in the relaxed first dorsal interosseous muscle of the contralateral (right) hand [35]. Participants received 25 sessions of either rTMS or sham stimulation over the precuneus. Each day, stimulation sessions consisted of a stimulation of 4 s with an interval of 56 s. The entire session lasted approximately 25 min each daily. The sham rTMS blocks were performed with the coil held close to the precuneus but angled away.

TMS protocol adverse events

None of the participants reported any adverse events during the rTMS trial.

Statistical analysis

All data in this study were tested for normality. The Shapiro–Wilk test was used to assess data normal distribution. To note, all data (age, education level, neuropsychological characteristics, ITV, head rotation parameters, altered HIPsub GM values, and altered HIPsub connectivity) in this study exhibited a normal/Gaussian distribution, except gender. Therefore, all statistical analysis used parametric testing.

Demographics and neuropsychological data

Two-sample T-test and chi-square tests were used to compare differences in demographic data, clinical, episodic memory, ITV, and head rotation parameters between aMCI patients and CN subjects (p < 0.05).

Network identification of altered HIPsub related to episodic memory impairment

We used a two-sample t-test to investigate differences in GM volumes of HIPsub between aMCI patients and CN subjects before rTMS treatment after controlling for age, sex, and education.

Using GLM analysis, we assessed the differences in the FCs of HIPsub based on structural GM alteration between aMCI patients and CN subjects before rTMS treatment after controlling for age, sex, education, ITV, and mean FD (TFCE-FDR-corrected p < 0.05 and cluster size > 405 mm³). Subsequently, masks were made based on brain regions of between-group differences on the FCs of HIPsub in aMCI patients. These masks were used for the analysis of pre- versus post-rTMS (pre-sham- versus post-sham-rTMS) fMRI data from study #2 (i.e., validation of causal relationship between altered HIPsub and episodic memory decline).

Pattern classification based on the altered HIPsub GM and FC

We applied a SVM approach to test how well GM and network connectivity of altered HIPsub could distinguish aMCI patients from CN subjects, A linear SVM classifier was performed using LIBSVM software (Software available at https://www.csie.ntu.edu.tw/∼cjlin/libsvm). A leave-one-out cross-validation (LOOCV) strategy was used to assess the generalization of this SVM classifier and to assess its accuracy, sensitivity, and specificity in this study. Briefly, if there are N samples in total, in each LOOCV experiment, the N-1 samples are viewed as the training set, and the omitted one is used as a test subject to computing the classification error. LOOCV accuracy was yielded by averaging all accuracies obtained at each tested subject.

Validation of causal relationship between altered HIPsub and episodic memory decline with rTMS related to aMCI

To empirically validate the causal relationship between the changes in GM and network connectivity of HIPsub and the changes in episodic memory in aMCI patients after rTMS treatment, paired t-tests were employed to calculate the changes in GM and FC of HIPsub pre- versus post-rTMS (or pre-sham- versus post-sham-rTMS) after which the changes were correlated with the changes in episodic memory of that individual between baseline assessment and at 4 weeks post-rTMS (or post-sham-rTMS) in aMCI patients after controlling for age, sex, GM (only for FC comparison), and education.

rTMS regulatory mechanism in amelioration of episodic memory

To validate the mechanism of rTMS improving episodic memory, we performed one mediation analysis whether network connectivity changes of HIPsub mediates the effects of GM change on episodic memory during rTMS treatment. Therefore, we performed structural equation modeling (SEM) for GM change, network connectivity change of HIPsub, and episodic memory change after 4 weeks of rTMS treatment. Detailed mediation analysis method is provided in the Supplementary Methods.

Sham versus real rTMS comparison

Of the 12 aMCI patients with full clinical assessments, usable sMRI, and fMRI scan data at baseline and 4 weeks post-rTMS (or sham), 8 patients had been randomized to active rTMS and 4 patients received the sham rTMS. We performed a two-sample t-test to investigate the differences in the changes in GM and FC of HIPsub between pre-post real rTMS and pre-post sham rTMS. Pre-real-rTMS (or sham-rTMS) maps were subtracted from post-real-rTMS (or sham-rTMS) maps to generate maps of FC or GM changes for each subject.

Improvement of statistical power with a low sample size, it is potentially possible that the statistical significance of the observed effects may not remain robust.

RESULTS

Network identification of altered HIPsub related to aMCI patients

Structural and functional alterations of HIPsub in aMCI patients

As shown in Fig. 2A, compared with CN, aMCI patients showed significantly altered GM volumes of HIPc and HIPp (p < 0.001).

Fig. 2

Comparison of GM volumes and functional connectivity in HIPsub (HIPc and HIPp) between aMCI patients and CN before rTMS treatment. A) A structural diagram of the HIPsub in the left hemisphere. Bar chart showing comparison results of GM volumes in the HIPsub between aMCI patients and CN before rTMS treatment after controlling for age, sex, and education. Supplementary Figure 7 presents the results from HIPe GM in aMCI compared to CN. B) HIPc-subregion and brain different regions of the HIPc functional connectivity between CN and aMCI patients. Bar chart shows the quantitative comparison of functional connectivity in these regions. C) HIPp-subregion and brain different regions of the HIPp functional connectivity between CN and aMCI patients. Bar chart shows the quantitative comparison of functional connectivity in these regions. ^*p_TFCEhboxFDR < 0.05. Supplementary Figure 8 presents the results from HIPe functional connectivity in aMCI patients compared to CN. Results were shown after controlling for age, sex, education, ITV, and FD at a threshold of p < 0.05 using TFCE-FDR correction with cluster size > 405 mm³. CN, healthy controls; aMCI, amnestic mild cognitive impairment; HIPc, hippocampal cognitive region; HIPp, hippocampal perceptual region; GM, grey matter; TFCE, threshold-free cluster enhancement; FDR, false discovery rate; ITV, intracranial volume; FD, framewise displacement; CEREpos.L, left cerebellum posterior lobe; CEREpos.R, right cerebellum posterior lobe; MTG.L, left middle temporal gyrus; MTG.R, right middle temporal gyrus; PHG.L, left parahippocampa gyrus; IFGorb.R, right inferior frontal gyrus, orbital part; LING.L, left, lingual gyrus; LING.R, right lingual gyrus; MFG.L, left medial frontal gyrus; FusG.L, left fusiform gyrus. ^***p < 0.001.

Figure 2B and Supplementary Table 2 show that, in the HIPc network, compared with CN, aMCI patients showed decreased FC in bilateral CERpos and IFGorb.R, increased FC in bilateral MTG and PHG.L (p_TFCEhboxFDR < 0.05, cluster size > 405 mm³). Figure 2C and Supplementary Table 2 show that, in the HIPp network, compared with CN, aMCI patients showed decreased FC in MFG.L and PHG.L, increased FC in bilateral LING, and bilateral MTG (p_TFCEhboxFDR < 0.05, cluster size > 405 mm³). All results were controlled for age, sex, education, ITV, and FD.

Supplementary Figures 7–9 present the results from HIPe in aMCI patients compared to CN.

Classification of aMCI patients based on the altered HIPsub GM volumes and functional connectivities in aMCI patients

The SVM classifier’s classification accuracy was 88.3%. As shown in Fig. 3, the SVM classifier’s receiver operating characteristic curve shows a high power to discriminate aMCI patients from CN on an individual subject basis, with an AUC of 94.7%, 87.3% sensitivity, and 92.7% specificity.

Fig. 3

Classification of individuals as aMCI versus CN by MRI-based “classifier”. ROC curve shows the classification power in MRI-based “classifier” of aMCI from CN. Note: The values of ACC, AUC, sensitivity, and specificity in lower right of the figure present the optimum values under the optimum combined index score (red point). aMCI, amnestic mild cognitive impairment; CN, healthy controls; AUC, area under the ROC curve; ACC, accuracy; Opt, optimum; ROC, receiver operating characteristic; MRI, magnetic resonance imaging.

Validation of causal relationship between altered HIPsub and episodic memory decline with rTMS

Changes of HIPsub GM pre- versus post-rTMS (or sham rTMS)

As shown in Fig. 4A, compared with pre-rTMS, aMCI patients showed significantly increased GM volumes of HIPc and HIPp at 4 weeks post-rTMS (p < 0.001), while no differences were observed in HIPc and HIPp GM volumes pre- versus post-sham rTMS (p > 0.05).

Fig. 4

The effects on HIPsub volumes and its network functional connectivity of aMCI patients after 4 weeks of real rTMS or sham rTMS treatment controlling for age, sex, and education. A) Bar chart showing the quantitative effects on HIPsub volumes of aMCI patients after 4 weeks of real rTMS or sham rTMS treatment after controlling for age, sex, and education. B) HIPc seed, brain regions of HIPc functional connectivity changes (TFCE-FDR correction), and quantitative changes on HIPc functional connectivity of aMCI patients after 4 weeks of real rTMS or sham rTMS treatment. C) HIPp seed, brain regions of HIPp functional connectivity changes (TFCE-FDR correction), and quantitative changes on HIPp functional connectivity of aMCI patients after 4 weeks of real rTMS or sham rTMS treatment. ^*p < 0.05, ^***p < 0.001. Supplementary Figure 9 presents the results from HIPe functional connectivity in aMCI patients after 4 weeks of rTMS treatment. bef-real-aMCI, amnestic mild cognitive impairment before real rTMS treatment; aft-real-aMCI, amnestic mild cognitive impairment after real rTMS treatment; bef-sham-aMCI, amnestic mild cognitive impairment before sham rTMS treatment; aft-sham-aMCI, amnestic mild cognitive impairment after sham rTMS treatment; HIPc, hippocampal cognitive region; HIPp, hippocampal perceptual region; TFCE, threshold-free cluster enhancement; FDR, false discovery rate; PHG.L, left parahippocampa gyrus; MTG.L, left middle temporal gyrus; FusG.L, fusiform gyrus; STG/MTG.R, right superior/middle temporal gyrus.

Changes of HIPsub FC pre- versus post-rTMS (or sham rTMS)

As shown in Fig. 4B, in the HIPc network, compared with pre-rTMS, aMCI patients showed significantly increased FC in MTG.L 4-week post-rTMS (p < 0.05). Moreover, as shown in Fig. 4C, in the HIPp network, compared with pre-rTMS, aMCI patients significantly increased FC in MTG.R and FusG.L, and decreased FC in STG.R/MTG.R at 4 weeks post-rTMS (p < 0.05). All results were controlled for age, sex, education, ITV, and FD.

No differences were observed in the HIPc and HIPp connectivity pre- versus post-sham rTMS.

Sham rTMS versus real rTMS comparison

Supplementary Table 3 shows the results when comparing changes in HIPsub at post-pre real rTMS and post-pre sham rTMS.

Changes of episodic memory pre- versus post-rTMS (or sham rTMS)

As shown in Fig. 5A, B, episodic memory (AVLT) in aMCI patients returned to normal after 4 weeks of real rTMS treatment (p < 0.001) while no differences were observed after 4 weeks of sham TMS treatment (p > 0.05).

Fig. 5

Relationships between the changes of HIPc GM volume and its functional connectivity, and the changes of episodic memory in aMCI patients after 4 weeks of rTMS treatment. A, B) Line chart showing the changes of episodic memory in aMCI patients before rTMS, after 2 weeks and 4 weeks of rTMS treatment compared to CN after controlling for age, sex, and education. C) Relationships between the changes of HIPc GM volume and the changes of episodic memory in aMCI patients after 4 weeks of rTMS treatment. D) Relationships between changed HIPsub functional connectivity and episodic memory in aMCI subjects after 4 weeks of rTMS treatment. Results are presented after controlling for age, sex, and education. FDR control was used for multiple comparisons correction. ^***p < 0.001. CN, healthy controls; aMCI, amnestic mild cognitive impairment; HIPc, hippocampal cognitive region; MTG.L, left middle temporal gyrus; FC, functional connectivity; FDR, false discovery rate; aMCI-B, aMCI at baseline; aMCI-2W-TMS, aMCI after 2 weeks of rTMS; aMCI-4W-TMS, aMCI after 4 weeks of rTMS. AVLT, Auditory Verbal Learning Test. AVLT-IR, Auditory Verbal Learning Test-immediate recall; AVLT-5 min-DR, Auditory Verbal Learning Test–5-min delayed recall, AVLT-20 min-DR, Auditory Verbal Learning Test–20-min delayed recall.

Correlations between changes of GM, FC of HIPsub, and episodic memory pre- versus post-rTMS

As shown in Fig. 5C, the change in HIPc GM volume in aMCI patients positively correlated with the change in the total AVLT score [r = 0.783, p = 0.022, confidence interval (CI): 0.406–0.991] before and after 4 weeks of rTMS treatment (p_FDR < 0.05).

As shown in Fig. 5D, the change in FC between HIPc and MTG.L in aMCI patients positively correlated with the change in the total AVLT score (r = 0.803, p = 0.016, CI: 0.508–0.975) before and after 4 weeks of rTMS treatment (p_FDR < 0.05). All results were controlled for age, sex, and education.

rTMS regulatory mechanism in amelioration of episodic memory

As shown in Fig. 6A, mediator analysis showed changes in FC between HIPc and MTG.L significantly mediated the relationship between GM change and episodic memory change (AVLT score) before and after 4 weeks of rTMS treatment (χ² = 0.391, p = 0.532; RMR < 0.001; GFI = 1; NFI = 1; CFI = 1; IFI = 1). Significantly positive correlations were observed between GM change volume and episodic memory change (β= 0.280, p < 0.001) and between changes in FC of HIPc-MTG.L and episodic memory (β= 0.652, p < 0.001). Furthermore, a significant mediating effect was observed in changes in FC of HIPc-MTG.L on the association between GM change and episodic memory change (β= 0.457, p < 0.001).

Fig. 6

Structural equation modeling results and rTMS regulating hypothetical model among changed GM and FC of HIPc, and episodic memory. A) Structural equation modeling (SEM) shows changed HIPc-MTG.L functional connectivity partially mediates the association between changed HIPc GM volume and episodic memory (AVLT scores) in aMCI patients after 4 weeks of rTMS treatment (β= 0.4571, p < 0.001). All observed variables are adjusted for age, gender, and education. Rectangles represent observed variables. Mediation results are presented as standardized regression coefficients. Standardized parameter estimates are shown. Regression weights are indicated by arrows. ^***p < 0.001. B) indicating a rTMS regulating hypothetical model. aMCI, amnestic mild cognitive impairment; HIPe, hippocampal emotional region; HIPc, hippocampal cognitive region; HIPp, hippocampal perceptual region; PreCUN, precuneus; MTG.L, left middle temporal gyrus; FC, functional connectivity; GM, grey matter; AVLT, Auditory Verbal Learning Test; rTMS, repetitive transcranial magnetic stimulation.

DISCUSSION

To our knowledge, the present study was the first to answer questions about the causal relationship between altered neuroimaging characteristic of HIPc and clinical episodic memory deficits in aMCI patients by combining two mature technologies (MRI and rTMS), which established a symptom-pathophysiology relationship. In particular, this study identified a biological target or circuit that mediates amelioration of episodic memory in aMCI. Taken together, the findings of this study promote our in-depth understanding of why some aMCI patients in clinical trials show a symptomatic response to rTMS respond whereas others do not.

Our findings showed that aMCI patients displayed abnormal HIPc and HIPp structure and distinctly altered patterns of HIPc and HIPp network connectivity, which shows a high power to discriminate aMCI from CN on an individual subject basis, with an ACC of 88.3%, AUC of 94.7%, 87.3% sensitivity, and 92.7% specificity. Our results are consistent with the findings presented in our previous study [6]. Indeed, in several studies, it was suggested that each subregion of the hippocampus is specialized to process a unique aspect of the event and has distinctly anatomical and functional connectivity [6 , 37]. Our findings suggest that each subregion of the hippocampus has a selective topography of pathological involvement during disease progression of aMCI patients.

Our findings showed that episodic memory performance in aMCI patients returned to normal after 4 weeks of rTMS treatment. These findings are in agreement with the data presented in a previous study [18], which showed that rTMS targeting the precuneus can improve episodic memory performance in aMCI patients. Indeed, a great number of studies have well shown that the precuneus is a component of hippocampal intrinsic connectivity networks [6 , 20], a key node area for episodic memory deficits observed in early AD [21, 22], and a vulnerable region for the progress of aMCI to AD [18]. When combining the above-mentioned evidence and our findings, it is reasonable to speculate that the precuneus is an ideal target for tailored rTMS intervention to ameliorate episodic memory decline in aMCI patients. In particular, the changes in HIPc GM volume and its connectivity with MTG in aMCI patients by 4 weeks of rTMS treatment targeting the precuneus within the HIPc network positively correlated with the episodic memory change while the changes in the HIPp network did not. These findings suggested that these changes were highly selective to the targeted brain regions, that is, the HIPc network circuit involved in cognitive processes that were linked directly to episodic memory, whereas the HIPp network did not. Our results support the hypothesis that the efficacy of neuromodulation depends on symptom-specific treatment targets with different underlying brain circuits [13]. Furthermore, a recent neuromodulation study verified that rTMS targeting the parietal cortex within hippocampal networks could improve associative memory performance in healthy individuals [19]. The subregion of the hippocampus network that is closely linked to memory performance may contribute to episodic memory impairment in aMCI patients [11 , 19].Therefore, converging evidence suggests that the precuneus-HIPc-MTG may be a biological substrate for the treatment of the disabling episodic memory in aMCI patients.

Our most fascinating finding was that HIPc connectivity with MTG changes mediated the effects of changes in HIPc GM on episodic memory during rTMS intervention, which is highly supported by the theoretical hypothesis of previous studies [12]. These findings suggest a neuroimaging-based regulatory mechanism of rTMS (see Fig. 6B). That is, rTMS first modulates changes in neural activity of the precuneus [19, 20], and subsequently induces the localized long-term neuronal plasticity change of the HIPc structure itself (i.e., HIPc GM change) by communicating relevant information within the HIPc network [19, 38]. This further promotes the change of its connectivity with MTG [6], then, HIPc connectivity with MTG changes mediates the effects of changed HIPc GM on episodic memory, and finally ameliorates episodic memory in aMCI.

In this study, several strengths should be highlighted. Methodological aspect: First, the hippocampus is divided into HIPe, HIPc, and HIPp, which is helpful in revealing a subspecialization in the hippocampus. Second, strict quality control is used in MRI data preprocessing. Third, we performed a strict multiple comparison correction strategy to ensure reproducibility, test–retest reliability, and replicability on the fMRI metrics. Fourth, a pattern classification approach was used to identify GM and network connectivity of altered HIPsub as closely related to aMCI, which shows a high power to discriminate aMCI from CN on an individual subject basis, with high AUC, sensitivity, and specificity. Clinical significance aspect: First, a causal relationship was established between HIPc and its connectivity with MTG and episodic memory decline in aMCI patients, which takes away the field from purely correlational studies. Second, a quantifiable and engageable target of the precuneus within the HIPc network was established to modulate episodic memory decline in aMCI patients. Third, in this study, we established a precision-medicine model for the treatment of the AD disease spectrum. A clinical treatment trial design can target disease-specific pathophysiology to improve clinical manifestations (episodic memory decline) in the AD disease spectrum. Fourth, in this study, we established a model of response and non-response to rMTS based on individual level explanation, which links neuromodulation to a biological outcome (neuroimaging characteristics: functional connectivity and GM volume) rather than to symptomatic response alone. Fifth, in this study, we established a rTMS regulatory mechanism, indicating that functional connectivity changes mediate the effects of GM change on episodic memory in response to rTMS.

Study limitations

Our study has also some limitations. First, in this study, a relatively small sample size was used to validate the causal relationship between altered HIPc and its connectivity with MTG and episodic memory decline (rTMS experiment). Conventional rTMS clinical trials often use clinical response as the sole readout of rTMS efficacy. In this case, our sample size is insufficient to draw conclusions about the role of HIPc and its connectivity with MTG in episodic memory amelioration. In this study, we combined the clinical response with GM volume and functional connectivity. It is reasonable to speculate that our findings were adequately powered to discover the close causal associations between HIPc and its connectivity with MTG and amelioration of episodic memory. Second, our findings suggest a target biological substrate for amelioration of episodic memory in aMCI patients, but in this study, we do not demonstrate that HIPc network-targeted rTMS is the sole intervention pathway.

Conclusions

In conclusion, this study provides novel experimental evidence about a biological substrate for the treatment of the disabling episodic memory in aMCI patients. We found that by directly modulating the precuneus-HIPc-MTG circuit, episodic memory deficits can be ameliorated.

Footnotes

ACKNOWLEDGMENTS

This study was supported by the National Key Research and Development Program of China (No. 2018YFC1314300), the National Natural Science Foundation of China (No. 81701675; 81971255); the Key Project supported by Medical Science and technology development Foundation, Nanjing Department of Health (No. JQX18005); the Cooperative Research Project of Southeast University-Nanjing Medical University (No. 2018DN0031); the Key Research and Development Plan (Social Development) Project of Jiangsu Province (No. BE2018608; BE2019610); Jiangsu Provincial Medical Talent project (ZDRCA2016075); the Innovation and Entrepreneurship Training Program for College Students in Jiangsu Province (No.201810312061X; 201910312035Z); and Key Scientific Research Projects of Colleges and Universities in Henan Province (No:18A190003).

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Petersen

, Negash

(2008) Mild cognitive impairment: An overview. CNS Spectr 13, 45–53.

Albert

, DeKosky

, Dickson

, Dubois

, Feldman

, Fox

, Gamst

, Holtzman

, Jagust

, Petersen

, Snyder

, Carrillo

, Thies

, Phelps

(2011) The diagnosis of mild cognitive impairment due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 270–279.

Sperling

, Aisen

, Beckett

, Bennett

, Craft

, Fagan

, Iwatsubo

, Jack

Jr. , Kaye

, Montine

, Park

, Reiman

, Rowe

, Siemers

, Stern

, Yaffe

, Carrillo

, Thies

, Morrison-Bogorad

, Wagster

, Phelps

(2011) Toward defining the preclinical stages of Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 280–292.

Simon

, Yokomizo

, Bottino

(2012) Cognitive intervention in amnestic mild cognitive impairment: A systematic review. Neurosci Biobehav Rev 36, 1163–1178.

Bourrelier

, Kubicki

, Rouaud

, Crognier

, Mourey

(2015) Mental rotation as an indicator of motor representation in patients with mild cognitive impairment. Front Aging Neurosci 7, 238.

Chen

, Duan

, Shu

, Wang

, Long

, Liu

, Liao

, Shi

, Chen

, Zhang

(2016) Differential contributions of subregions of medial temporal lobe to memory system in amnestic mild cognitive impairment: Insights from fMRI study. Sci Rep 6, 26148.

Chen

, Zhang

, Li

(2015) Can multi-modal neuroimaging evidence from hippocampus provide biomarkers for the progression of amnestic mild cognitive impairment? Neurosci Bull 31, 128–140.

Robinson

, Barron

, Kirby

, Bottenhorn

, Hill

, Murphy

, Katz

, Salibi

, Eickhoff

, Fox

(2015) Neurofunctional topography of the human hippocampus. Hum Brain Mapp 36, 5018–5037.

Davachi

(2006) Item, context and relational episodic encoding in humans. Curr Opin Neurobiol 16, 693–700.

10.

Bai

, Wei

, Xie

, Wang

, Ji

, Wang

, Tian

(2019) Hippocampal-subregion functional alterations associated with antidepressant effects and cognitive impairments of electroconvulsive therapy. Psychol Med 49, 1357–1364.

11.

Ludowig

, Trautner

, Kurthen

, Schaller

, Bien

, Elger

, Rosburg

(2008) Intracranially recorded memory-related potentials reveal higher posterior than anterior hippocampal involvement in verbal encoding and retrieval. J Cogn Neurosci 20, 841–851.

12.

Poppenk

, Moscovitch

(2011) A hippocampal marker of recollection memory ability among healthy young adults: Contributions of posterior and anterior segments. Neuron 72, 931–937.

13.

Siddiqi

, Taylor

, Cooke

, Pascual-Leone

, George

, Fox

(2020) Distinct symptom-specific treatment targets for circuit-based neuromodulation. Am J Psychiatry 177, 435–446.

14.

Brady

Jr. , Gonsalvez

, Lee

, Ongur

, Seidman

, Schmahmann

, Eack

, Keshavan

, Pascual-Leone

, Halko

(2019) Cerebellar-prefrontal network connectivity and negative symptoms in schizophrenia. Am J Psychiatry 176, 512–520.

15.

Rossi

, Hallett

, Rossini

, Pascual-Leone

, Safety of TMS Consensus Group (2009) Safety, ethical considerations, and application guidelines for the use of transcranial magnetic stimulation in clinical practice and research. Clin Neurophysiol 120, 2008–2039.

16.

Turriziani

, Smirni

, Zappala

, Mangano

, Oliveri

, Cipolotti

(2012) Enhancing memory performance with rTMS in healthy subjects and individuals with Mild Cognitive Impairment: The role of the right dorsolateral prefrontal cortex. Front Hum Neurosci 6, 62.

17.

Rabey

, Dobronevsky

(2016) Repetitive transcranial magnetic stimulation (rTMS) combined with cognitive training is a safe and effective modality for the treatment of Alzheimer’s disease: Clinical experience. J Neural Transm (Vienna) 123, 1449–1455.

18.

Koch

, Bonni

, Pellicciari

, Casula

, Mancini

, Esposito

, Ponzo

, Picazio

, Di Lorenzo

, Serra

, Motta

, Maiella

, Marra

, Cercignani

, Martorana

, Caltagirone

, Bozzali

(2018) Transcranial magnetic stimulation of the precuneus enhances memory and neural activity in prodromal Alzheimer’s disease. Neuroimage 169, 302–311.

19.

Wang

, Rogers

, Gross

, Ryals

, Dokucu

, Brandstatt

, Hermiller

, Voss

(2014) Targeted enhancement of cortical-hippocampal brain networks and associative memory. Science 345, 1054–1057.

20.

Kahn

, Andrews-Hanna

, Vincent

, Snyder

, Buckner

(2008) Distinct cortical anatomy linked to subregions of the medial temporal lobe revealed by intrinsic functional connectivity. J Neurophysiol 100, 129–139.

21.

Buckner

, Andrews-Hanna

, Schacter

(2008) The brain’s default network: Anatomy, function, and relevance to disease. Ann N Y Acad Sci 1124, 1–38.

22.

Raichle

, MacLeod

, Snyder

, Powers

, Gusnard

, Shulman

(2001) A default mode of brain function. Proc Natl Acad Sci U S A 98, 676–682.

23.

Eldaief

, Halko

, Buckner

, Pascual-Leone

(2011) Transcranial magnetic stimulation modulates the brain’s intrinsic activity in a frequency-dependent manner. Proc Natl Acad Sci U S A 108, 21229–21234.

24.

Chen

, Shu

, Wang

, Liu

, Shi

, Zhang

(2015) The interaction of APOE genotype by age in amnestic mild cognitive impairment: A voxel-based morphometric study. J Alzheimers Dis 43, 657–668.

25.

Xue

, Yuan

, Yue

, Xu

, Wang

, Wu

, Ji

, Zhou

, Zhao

, Rao

, Yang

, Xiao

, Chen

(2019) Distinct disruptive patterns of default mode subnetwork connectivity across the spectrum of preclinical Alzheimer’s disease. Front Aging Neurosci 11, 307.

26.

Chen

, Chen

, Shu

, Chen

, Ward

, Wang

, Liu

, Antuono

, Li

, Zhang

, Alzheimer’s Disease Neuroimaging Initiative (2019) Predicting progression from mild cognitive impairment to Alzheimer’s disease on an individual subject basis by applying the CARE index across different independent cohorts. Aging (Albany NY) 11, 2185–2201.

27.

Chen

, Shu

, Wang

, Zhan

, Liu

, Zhang

(2020) Intrinsic connectivity identifies the sensory-motor network as a main cross-network between remitted late-life depression- and amnestic mild cognitive impairment-targeted networks. Brain Imaging Behav 14, 1130–1142.

28.

Friston

, Williams

, Howard

, Frackowiak

, Turner

(1996) Movement-related effects in fMRI time-series. Magn Reson Med 35, 346–355.

29.

Sheline

, Raichle

, Snyder

, Morris

, Head

, Wang

, Mintun

(2010) Amyloid plaques disrupt resting state default mode network connectivity in cognitively normal elderly. Biol Psychiatry 67, 584–587.

30.

Power

, Barnes

, Snyder

, Schlaggar

, Petersen

(2012) Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion. Neuroimage 59, 2142–2154.

31.

Van Dijk

, Sabuncu

, Buckner

(2012) The influence of head motion on intrinsic functional connectivity MRI. Neuroimage 59, 431–438.

32.

Chen

, Lu

, Yan

(2018) Reproducibility of R-fMRI metrics on the impact of different strategies for multiple comparison correction and sample sizes. Hum Brain Mapp 39, 300–318.

33.

Smith

, Nichols

(2009) Threshold-free cluster enhancement: Addressing problems of smoothing, threshold dependence and localisation in cluster inference. Neuroimage 44, 83–98.

34.

Yan

, Wang

, Zuo

, Zang

(2016) DPABI: Data processing & analysis for (resting-state) brain imaging. Neuroinformatics 14, 339–351.

35.

Rossini

, Burke

, Chen

, Cohen

, Daskalakis

, Di Iorio

, Di Lazzaro

, Ferreri

, Fitzgerald

, George

, Hallett

, Lefaucheur

, Langguth

, Matsumoto

, Miniussi

, Nitsche

, Pascual-Leone

, Paulus

, Rossi

, Rothwell

, Siebner

, Ugawa

, Walsh

, Ziemann

(2015) Non-invasive electrical and magnetic stimulation of the brain, spinal cord, roots and peripheral nerves: Basic principles and procedures for routine clinical and research application. An updated report from an I.F.C.N. Committee. Clin Neurophysiol 126, 1071–1107.

36.

Libby

, Ekstrom

, Ragland

, Ranganath

(2012) Differential connectivity of perirhinal and parahippocampal cortices within human hippocampal subregions revealed by high-resolution functional imaging. J Neurosci 32, 6550–6560.

37.

Zeineh

, Holdsworth

, Skare

, Atlas

, Bammer

(2012) Ultra-high resolution diffusion tensor imaging of the microscopic pathways of the medial temporal lobe. Neuroimage 62, 2065–2082.

38.

Feurra

, Galli

, Pavone

, Rossi

(2016) Frequency-specific insight into short-term memory capacity. J Neurophysiol 116, 153–158.