Professional Documents
Culture Documents
Cui Et Al., 2019
Cui Et Al., 2019
DOI 10.3233/JAD-181296
IOS Press
f
3
roo
4 Cognitive Improvements in Amnestic Mild
5 Cognitive Impairment: A Randomized
rP
6 Controlled Study
tho
7 Hailun Cuia,1 , Rujing Rena,1 , Guozhen Linb , Yang Zoua , Lijuan Jiangc , Zhengde Weid , Chunbo
8 Lic,e,f,2,∗ and Gang Wanga,2,∗
9
a Department of Neurology & Institute of Neurology, Ruijin Hospital, Shanghai Jiao Tong University School of
10 Medicine, Shanghai, China
b Department of Psychiatry, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China
Au
11
12
c Shanghai Key Laboratory of Psychotic Disorders, Shanghai Mental Health Center, Shanghai Jiao Tong
16
17
f CAS Center for Excellence in Brain Science and Intelligence Technology (CEBSIT), Chinese Academy of Science,
Shanghai, China
cte
18
20 Abstract.
21 Background: Repetitive transcranial magnetic stimulation (rTMS) is thought to be effective in alleviating cognitive symptoms
co
22
24
in patients with amnestic mild cognitive impairment (aMCI), but the mechanisms related to network modification are poorly
23
25
understood.
Objective: Here we tested rTMS efficacy and explored the effect of rTMS-induced changes in the default mode network
(DMN) and their predictive value for treatment response.
Un
ISSN 1387-2877/19/$35.00 © 2019 – IOS Press and the authors. All rights reserved
2 H. Cui et al. / Hypoconnected DMN Improves Cognition in aMCI
26 Methods: Twenty-one subjects clinically diagnosed with aMCI were recruited to complete a 10-session randomized and sham-
27 controlled rTMS treatment targeting the right dorsolateral prefrontal cortex. Resting-state functional magnetic resonance
28 imaging in tandem with neuropsychological assessments were administered before and after the intervention. Changes in
29 functional connectivity of the DMN and relevant brain regions, as well as the correlations between baseline functional
30 connectivity and clinical rating scales were calculated in order to elucidate the mechanism of treatment response to rTMS
31 therapy.
32 Results: Compared to the sham group, the rTMS group achieved improvement of neuropsychological performance and
33 significant functional connectivity changes within the DMN. Group × Time interactions were found between posterior cin-
gulate gyrus and right fusiform gyrus (F(1,19) = 17.154, p = 0.001), and also left anterior cingulate gyrus (F(1,19) = 3.908,
f
34
35 p = 0.063), showing an rTMS-induced deactivation of functional connectivity within the DMN. Baseline functional con-
roo
36 nectivity analysis of seeds within the DMN in the rTMS group revealed negative correlation with AVLT-Recognition score
37 changes.
38 Conclusion: rTMS-induced hypoconnectivity within DMN is associated with clinical cognitive improvements in patients
39 with aMCI. Further, pre-rTMS baseline activity of the DMN at rest may be a predictor for favorable rTMS treatment
40 response.
rP
Keywords: Alzheimer’s disease, amnestic mild cognitive impairment, default mode network, repetitive transcranial magnetic
41 stimulation, resting-state functional MRI
42
Trial Registration: Chinese Clinical Trial Registry (ChiCTR); ChiCTR-IPR-16009063. Trial Name: Repetitive transcra-
tho
43
44
nial magnetic stimulation on cognition in mild cognitive impairment and early stage of Alzheimer’s disease. URL:
45
http://www.chictr.org.cn/showproj.aspx?proj = 15273.
27 Mild cognitive impairment (MCI) has been sug- and has been shown to indeed influence intrinsic 56
28 gested to represent a distinct status intermediate brain connectivity through functional magnetic res- 57
29 between prodromal Alzheimer’s disease (AD) [1–3], onance imaging (fMRI) [11–13]. It is hypothesized 58
30 and more severe, debilitating, cognitive impair- that the memory improvement observed in patients 59
d
31 ment. Among variable presentations, amnestic MCI with MCI and AD undergoing rTMS treatment could 60
(aMCI), characterized by impaired episodic memory be attributed to the reversal of neural connectivity
cte
32 61
33 retention and retrieval, is of special interest both for changes which coincide with dysfunction; however, 62
34 its high risk of progression to AD and its potential to little is known concerning exact pattern(s) of func- 63
35 be a treatment window in which symptomatic relief tional rewiring of the brain connectome underlying 64
36 can be achieved [4, 5]. In the face of unfavorable off- any mechanism of improvement [7, 14, 15]. To 65
rre
37 target effects of the currently available medications tackle this question, it is very important to iden- 66
38 for AD symptoms, with little alteration of disease tify the commonly disturbed network(s) influencing 67
39 course or effect on higher-order brain dysfunction, brain dysfunction, if not also upstream of neuronal 68
40 clinicians are turning to non-invasive physiotherapies deterioration and death in AD patients, before any 69
which hold promise to defer AD morbidity through therapeutic intervention. In aMCI and AD, the con-
co
41 70
42 functional neural regulation. The most promising and sensus of literature in the field is that the most 71
43 well-tolerated means of AD intervention in recent prominently affected resting-state network with dis- 72
44 literature is the technique of repetitive transcranial ease alteration of measurable baseline activity is the 73
Un
45 magnetic stimulation (rTMS) [6, 7]. default mode network (DMN) [16, 17]. 74
46 The magnetic field produced by rTMS penetrates Hyperactivated at rest while deactivated through 75
47 the scalp and generates a sufficient electromotive task-related events, the DMN is presumed to be 76
48 force likely to depolarize superficial axons and involved in passively controlled states during which 77
49 influence neural activation patterns [8–10]. The the mind is internally oriented toward free thinking, 78
50 potential effects of rTMS to modify neuronal plastic- past reminiscence, and envisioning the future rather 79
51 ity across entire functional brain circuits and thereby than goal-directed tasks [18]. The earliest evidence 80
52 affect cortical excitability has promoted the appli- of its role in AD pathogenesis came from the obser- 81
53 cation of rTMS in a variety of neuropsychological vation of hypometabolism affecting regions that are 82
H. Cui et al. / Hypoconnected DMN Improves Cognition in aMCI 3
83 central components of the DMN, such as the pos- (rsfMRI) examination. 135
90 within the DMN able to predict cognitive decline, via memory clinic of Ruijin Hospital, Shanghai Jiao Tong 139
f
91 observation of metabolism and regional blood flow, University School of Medicine. Inclusion criteria are 140
roo
92 either of which can be used to evaluate the probability as follows: 1) age between 50–85 years old, no gen- 141
93 for MCI to develop into AD [23, 24]. In addition to der limitation, right-handed; 2) 6 years of education 142
94 the PCG, other major nodes within the DMN include at least; 3) met the previously published NIA-AA 143
95 the medial prefrontal cortex, lateral inferior parietal criteria (2011) for “MCI due to AD”, with memory 144
96 lobes, and medial temporal structures, which are all deficits as the only complaint of cognitive impair- 145
rP
97 interconnected and may be affected pathologically ment, reported by the patient, caregiver, or physician; 146
98 [25, 26]. Clinical trials applying rTMS to AD or MCI 4) Clinical Dementia Rating scale (CDR) memory 147
99 have found positive effects on memory, behavior, and score of 0.5; 5) Hachinski Ischemia Score (HIS) less 148
100 executive functions; however, there have been lim- than 4; 6) absence of dementia according to the DSM- 149
tho
101 ited studies revealing any possible mechanisms and IV for AD; 7) stable medication for at least 3 months. 150
102 therapeutic effects that rTMS exerts on brain func- Significant neurological or psychiatric diseases that 151
103 tional connectivity of the DMN to date [7, 27–29]. might result in cognitive dysfunction, unstable sys- 152
104 Measurable changes in regions distant from target temic condition, and MRI/rTMS contraindications 153
105 sites of rTMS have been demonstrated previously to (such as metal implants or claustrophobia) were 154
Au
106 represent modulation of neural activity. Thus, it is regarded as disqualified subjects and excluded. 155
107 reasonable to hypothesize that functional connectiv- The complete procedures of both rTMS and 156
108 ity within the DMN could be affected by targeting fMRI acquisition were approved by the Institutional 157
109 elsewhere using rTMS and that it may be possible to Review Board of Ruijin Hospital, Shanghai Jiao Tong 158
110 reveal mechanisms of cognitive improvement follow- University School of Medicine and conducted as the 159
112 modulation. consents from all the subjects were obtained before 161
cte
118 eral brain networks concerning cognitive, affective controlled pilot study carried out from October 165
119 and sensory processing [30–34]. Direct targeting 2016 to June 2017. All subjects involved were 166
120 of DLPFC has previously been proven to affect randomly assigned to two groups (1:1) following 167
121 episodic memory ability, likely through underlying simple randomization procedures (computerized ran- 168
co
122 functional pathways that connect to the subcorti- dom numbers), namely the rTMS group and the 169
123 cal memory related system, and therefore DLPFC sham group. Each subject received a 10-session 170
124 is one reasonable site to elicit therapeutic effects daily treatment for approximately 2 weeks (two 5- 171
125 in diseases showing memory impairments includ- consecutive-sessions with a 2-day break in between) 172
Un
126 ing MCI and AD [7, 27, 35]. We hypothesized that and a two-month follow-up neuropsychological 173
127 DMN functional connectivity changes are involved reassessment. A senior technician blinded for study 174
128 in the therapeutic mechanisms of rTMS and could design and neuropsychological assessments was des- 175
129 help explain individual responses through base- ignated to be the constant performer of the TMS 176
130 line functional connectivity assessment. To test this treatment (both for rTMS and sham group) using a 177
131 hypothesis, we tested the effectiveness of rTMS to MagPro R30 (Magventure, Denmark) with a figure- 178
132 improve functional memory in aMCI patients and of-eight coil model MCF-B70 (97 mm in diameter). 179
133 validated resulting changes in the DMN through Each daily stimulation session consisted of thirty 180
134 resting-state functional magnetic resonance imaging trains of 5 s stimuli delivered at 10 Hz spaced-out by 181
4 H. Cui et al. / Hypoconnected DMN Improves Cognition in aMCI
182 25 s of no stimulation (total number of stimuli: 1500). as the within-subjects factor. All analyses were 232
183 For subjects of the rTMS group, rTMS was delivered performed using SPSS software. Two-sided proba- 233
184 with an intensity of 90% of the resting motor thresh- bility values of p < 0.05 were considered statistically 234
185 old (RMT) rightly over the right DLPFC. RMT was significant. 235
f
190 and was able to induce a minimum of 50 mv of Each subject obtained MRI data in two sessions 237
roo
191 motor evoked potential (MEP) in five out of ten trials that occurred at baseline (pre-rTMS, less than a 238
192 recorded by the electrode pad located at the contralat- week before the treatment started) and afterwards 239
193 eral target muscle of the left first dorsal interosseous (less than 2 hours after the treatment finished). Dur- 240
194 muscle (FDI). The right DLPFC is located 5.5 ing the scanning, subjects were instructed to lay 241
195 cm anteriorly to the hotspot of each subject [36]. on their backs with their heads fastened in foam 242
rP
196 Subjects of the sham group received the same record- padding and wearing earplugs for noise reduction. 243
197 ing sound with the coil 90◦ perpendicular to the MRI data were acquired using a 3.0-T Siemens Verio 244
198 skull, with no practical magnetic field penetrated the MRI scanner (Erlangen, Germany) with a 12-channel 245
tho
ical images were scanned along the sagittal plane 247
200 Neuropsychological assessment and statistical with parameters as follows: repetition time/echo 248
201 analysis time (TR/TE) = (1900 ms/3.43 ms), field of view 249
202 General cognitive status and the treatment response size = 0.9 × 0.9 × 10.0 mm3 , slice thickness = 1 mm, 251
Au
203 were assessed by a researcher concealed from the and slice number = 160 slices. rsfMRI images were 252
204 allocation sequence using a multiple of neuropsycho- obtained using a gradient-recalled echo-planar imag- 253
205 logical scales at baseline (pre-intervention), 2 weeks ing (EPI) sequence with the following specifications: 254
206 (immediately after completing the 10-session course (TR/TE) = (2000 ms/25 ms), FOV = 240 × 240 mm2 , 255
207 of treatment, post-intervention), and at a two-month FOV phase = 100%, voxel size = 3.8 × 3.8 × 50.0 256
follow-up (8 weeks after the therapeutic session fin- mm3 , slice thickness = 5 mm, and 32 sagittal slices.
d
208 257
209 ished). Specifically, each patient completed global The functional run lasted for 316 s with 155 volumes 258
cte
210 cognitive function evaluation through Mini-Mental acquired. During the scanning, subjects passively 259
211 State Exam (MMSE) and Addenbrooke’s Cognitive viewed a fixation cross, not focused on anything in 260
212 Examination (ACE-III). Tests used to assess individ- particular and were kept awake during the resting- 261
213 ual domains of cognition included the logic memory state data acquisition. 262
214 test (short story test immediate recall), Auditory Ver- Data prepossessing in this study was per- 263
rre
215 bal Learning Test (AVLT) – immediate free recall/ formed using Data Processing Assistant for 264
216 (5 min/20 min) delayed free recall/recognition con- Resting State fMRI (DPARSF) V4.3, which is 265
217 taining 12 items, Trail Making Test A/B (TMTA/B), based on Statistical Parametric Mapping (SPM12; 266
218 digital span (forward/backward), as well as Geriatric http://www.Fil.ion.ucl.ac.uk/spm) and the toolbox 267
co
219 Depression Scale (GDS). Different summary scores of Data Processing & Analysis of Brain Imaging 268
220 derived from raw AVLT results were used in this (DPABI, http://rfmri.org/DPABI) [39, 40] including 269
221 study as the primary endpoint to evaluate the base- slice timing, head motion correction (nuisance 270
222 line and intervention effect of rTMS treatment both covariates including white matter signal, cere- 271
Un
223 in rTMS and sham groups [37, 38]. Demographics brospinal fluid signal, linear trend, and Friston 272
224 and baseline neuropsychological tests were analyzed 24-parameter model of head-motion parameter 273
225 using Fisher’s exact test for categorical variables and model were regressed out), normalization, spatial 274
226 two-sample two-tail Student’s t-test/non-parametric smoothing (FWHM = 6 mm) and band-pass filtering 275
227 (Mann-Whitney) test for continuous variables. The (0.01 Hz<f<0.08 Hz). These processed images were 276
228 intervention effect was investigated by conduct- used for seed-based connectivity analysis and the 277
229 ing mixed-design analysis of variance (ANOVA) subsequent correlation study. We did not use global 278
230 with Group (rTMS/sham) as the between-subjects signal regression because of possible distortion of 279
231 factor and time (pre/post-intervention, follow-up) group differences in functional connectivity [41, 42]. 280
H. Cui et al. / Hypoconnected DMN Improves Cognition in aMCI 5
281 An additional quality control measure for the wise connectivity maps generated through PCG. The 331
282 motion artifacts was performed using mean frame- overall correlations for each group before and after 332
283 wise displacement (FD) according to Power et al. the rTMS intervention were then forwarded to a 333
284 [43] in order to reflect the temporal derivative of mixed-design analysis of variance (ANOVA), with 334
285 movement parameters [44]. No participant was group (rTMS group/sham group) as the between- 335
286 classified as an outlier nor excluded from the main subjects factor, and time (pre/post-rTMS) as the 336
287 analysis. Images from the two groups did not differ in within-subjects factor. Mean time courses of each 337
288 mean FD (rTMS group/pre-intervention M = 0.104, region of significance were extracted and redefined as 338
f
289 SD = 0.055, range = 0.161, minimum = 0.051, max- specified regions of interest (ROI) for further analy- 339
roo
290 imum = 0.213; rTMS/post-intervention M = 0.101, sis. The functional connectivity between PCG and the 340
291 SD = 0.061, range = 0.173, minimum = 0.036, max- newly defined ROIs were reexamined using mixed- 341
292 imum = 0.209; sham/pre-intervention M = 0.094, design ANOVA through SPSS software for stricter 342
293 SD = 0.050, range = 0.131, minimum = 0.027, max- comparison (p < 0.05, Bonferroni corrected). Subse- 343
294 imum = 0.158; sham/post-intervention M = 0.095, quent contrasts were performed in order to examine 344
rP
295 SD = 0.307, range = 0.098, minimum = 0.042, maxi- the connectivity differences between each group 345
296 mum = 0.139). No significant group difference was (post-intervention scan: rTMS/sham) as well as dif- 346
297 found in motion parameter (F (3, 18.342) = 0.085, ferences between each scanning time point (pre/post 347
tho
and AVLT score changes (AVLT-D5, AVLT-D20, 349
299 Functional connectivity analysis AVLT-I, and AVLT-R) were also calculated. In all 350
300 The function of rTMS on regional connectivity son corrections for the group maps were performed 352
301 could be a large-scale modulation affecting a mul- by estimating the smoothness of the data after pre- 353
Au
302 tiplicity of regions within different networks, thus possessing procedure (3dFWHMx) ahead of 5000 354
303 we first performed a seed-based functional connec- Monte Carlo simulations using AlphaSim correction 355
304 tivity within the DMN by placing a 6-mm seed in the to determine statistical thresholds for voxel cluster 356
305 PCG at the given coordinates at the Montreal Neu- size needed for the achievement of a family-wise 357
306 rological Institute template coordinates of (0, –51, ␣ < 0.05 at a voxel-wise p < 0.005. 358
29) in order to calculate areas across the whole brain To investigate the possible causes for poten-
d
307 359
308 closely associated with it [45]. PCG is commonly tial discrepancies of cognitive improvement among 360
cte
309 identified as a critical node in the traditional DMN the patients receiving therapeutic rTMS, correlation 361
310 [17, 46, 47]. fMRI and PET studies have proven analysis was carried out to investigate the relationship 362
311 the essential role of PCG within DMN, both for its between the improved neuropsychological exams and 363
312 wide connection among other regions, such as the the initial status of brain represented by the base- 364
313 medial prefrontal cortex, anterior cingulate cortex, line functional connectivity within DMN. Regions of 365
rre
314 inferior parietal lobule and medial temporal regions DMN proved to have changed through the course of 366
315 (e.g., hippocampal/parahippocampal gyri), as well rTMS therapy and adjacent regions overlapping with 367
316 as its vulnerability of being impaired during neu- one another observed through a close look into the 368
317 ropathological process like AD [16, 48]. The average clusters with significant differences were extracted 369
co
318 blood oxygen level–dependent (BOLD) signal time as ROIs and the correlation within the mean time 370
319 course within the ROI of PCG was then correlated to course of each seed ROI was calculated. The bilat- 371
320 every voxel in the brain for each subject using Pear- eral PCG, the adjacent precuneus (PCUN), inferior 372
321 son correlation coefficients so as to identify specific parietal lobule (IPL), and ventral medial prefrontal 373
Un
322 regions within DMN that might have been changed. cortex (vMPFC) were defined as ROIs, and the base- 374
323 The correlation coefficients were then converted to line functional connectivity strength of each pair of 375
324 z scores using the Fisher r-to-z transformation in ROIs were then forwarded into correlation analysis 376
325 order to generate the resting-state functional connec- with the alterations in AVLT-I and AVLT-R (pre- 377
326 tivity z-map of DMN. We performed a one-sample post/pre). Statistics for the value were performed 378
327 t test showing the general activation of the baseline through GraphPad Prism software and validated by 379
328 FC within DMN in both groups (rTMS/sham). An SPSS. The FDR correction was implemented for mul- 380
329 independent-sample t test was used to exclude the tiple comparison using the Benjamini and Hochberg 381
384 Subjects characterization (r = 0.321, p = 0.336), age (r = 0.126, p = 0.712), edu- 433
385 A total of 25 aMCI patients were initially recruited p = 0.602) was not found, but baseline AVLT-R corre- 435
386 for receiving rTMS/sham stimulation, and 21 of them lated significantly with AVLT-R change (r = –0.698, 436
387 finished the whole session and rsfMRI scanning p = 0.017). 437
f
389 incomplete course of rTMS treatment, another two Functional imaging results 438
roo
390 excluded for abnormal structural findings unqualified
391 for inclusion criteria (Supplementary Figure 1). No Changes in functional connectivity induced by 439
392 significant differences were found between the two rTMS 440
393 groups regarding demographic and neuropsycholog- At baseline, the DMN map generated through 441
394 ical characteristics (Supplementary Table 1). the seed of PCG showed no significant difference 442
rP
between the two groups (rTMS/sham) (Supplemen- 443
395 rTMS improves short-term memory tary Figure 2). The functional connectivity pattern 444
396 Analysis of the effects of rTMS on neu- ing the ventral/dorsal medial prefrontal cortex, 446
tho
397 ropsychological performance improvements revealed PCG/retrosplenial cortex (Rsp) and adjacent pre- 447
398 significant Group × Time interactions (p < 0.05) on cuneus with the addition of lateral parietal cortex [21, 448
399 the memory domain. At the end of intervention 37, 38]. As for the alterations in functional connec- 449
400 (post-rTMS), significant improvement on AVLT- tivity between the PCG node and whole brain area, 450
401 Immediate free recall (AVLT-I) was observed in the clusters revealed significant Group × Time interac- 451
Au
402 rTMS group (p = 0.002) rather than the sham group tion (p < 0.05, FWE corrected), modeled as the effect 452
403 (p = 0.215) comparing to baseline level (Fig. 1C); of interest. Information regarding the identified clus- 453
404 Scores on AVLT-5 min delayed free recall (AVLT- ters are visualized in Supplementary Table 2. Clusters 454
405 D5) (Fig. 1A) and AVLT-20 min delayed free recall of size with volume greater than 540 mm3 were saved 455
406 (AVLT-D20) (Fig. 1B) were significantly increased as a single map (ROI) and were literally defined on 456
in the rTMS group (p < 0.001 and p = 0.004, respec- the voxels that showed peak activation compared with
d
407 457
408 tively), but unchanged in the sham group (p = 0.443 other voxels in the same cluster, namely under this 458
cte
409 and 0.504, respectively). A better performance trend condition, and they were left anterior cingulate gyrus 459
410 was observed in AVLT-Recognition (AVLT-R) com- (BA32, ACG.L), left superior frontal gyrus (BA8, 460
411 pared to the sham group (p = 0.086) (Fig. 1D) after SFG.L), right fusiform gyrus (BA37, FFG.R), and 461
412 the intervention. As for the follow-up reassessment left supplementary motor area (BA4, SMA.L). We 462
413 (8 weeks after the therapeutic regimen finished), extracted the mean time series of each newly defined 463
rre
414 significant improvements still existed in the rTMS ROI and calculated its correlation with PCG. Results 464
415 group concerning AVLT-I (p = 0.004), AVLT-D5 were analyzed by mixed-design analysis of variance 465
416 (p = 0.002), AVLT-D20 (p = 0.006), and AVLT-R (ANOVA) again for further validation and correction 466
417 (p = 0.011). Improvements with time were found in (Bonferroni corrected for four comparisons, thresh- 467
co
418 ACE-III, TMTA without group differences. No other old at 0.05/4 = 0.0125). With the exception of ACG.L 468
419 significant results appeared in the rest of test scores. (F(1,19) = 3.908, p = 0.063), FFG.R (F(1,19) = 17.154, 469
420 Moreover, no significant correlation in the rTMS p = 0.001), SFG.L (F(1,19) = 11.230, p = 0.003), and 470
421 group was found between percent change in AVLT- SMA.L (F(1,19) = 8.634, p = 0.008) showed signifi- 471
Un
422 I (where symptomatic improvement was defined as cant Group × Time interaction. Simple main effects 472
423 a positive change in test score: (post-pre)/pre) and analysis was carried out within these connectiv- 473
424 gender (r = 0.067, p = 0.844), education (r = 0.521, ity pairs (PCG-FFG.R/SFG.L/SMA.L) to determine 474
425 p = 0.101), nor baseline AVLT-I score (r = 0.011, the difference between each group (rTMS versus 475
426 p = 0.975), although age (r = –0.797, p = 0.003) and sham) at each time point (pre and post). As shown 476
427 resting motor threshold (RMT) (r = 0.603, p = 0.050) in Fig. 2A, the connectivity pair PCG-ACG.L was 477
428 did correlate significantly with AVLT-I, suggesting approaching significance (p = 0.063). A paired t-test 478
429 that younger patients with less activated corti- showed a tendency toward decrease after interven- 479
430 cal excitability (higher RMT) would gain more tion in the rTMS group (p = 0.094) and the two 480
H. Cui et al. / Hypoconnected DMN Improves Cognition in aMCI 7
f
roo
rP
tho
d Au
cte
Fig. 1. Differences in AVLT scores at each time point (post-rTMS, two-month follow-up) from baseline measurements. A) AVLT-5 min
delayed free recall (AVLT-D5), (B) AVLT-20 min delayed free recall (AVLT-D20), (C) AVLT-immediate free recall (AVLT-I) and (D) AVLT-
rre
Recognition (AVLT-R) changes; bars colored black for rTMS group while red for sham group. *p < 0.05, **p < 0.01. Error bars indicate
SEM.
481 groups were widely divergent in functional connec- functional connectivity (from z(r) = –0.121 to 0.166, 495
co
482 tivity after the therapy (z(r) = 0.179 (rTMS group) p = 0.024) without a baseline difference (p = 0.108) 496
483 to 0.540 (sham group), p = 0.028) without baseline (Fig. 3). The analysis of intervention effect on PCG- 497
484 difference (p = 0.897). Moreover, there was a slight SFG.L portrayed no significant changes in the rTMS 498
485 non-significance (r = –0.600, p = 0.051) for the corre- group while the same analysis of the sham group 499
Un
486 lation of changes in PCG-ACG.L connectivity with revealed a dramatic increase in functional correlation 500
487 improvements in AVLT-D5 rather than AVLT-D20 (z(r) = –0.142 to 0.353, p = 0.002) (Supplementary 501
488 (r = –0.452, p = 0.163) and other tests (Fig. 2B). The Figure 3). PCG has also been found to be corre- 502
489 connectivity between PCG and FFG.R varied through lated with SMA.L outside the DMN district with a 503
490 the course of intervention, where the rTMS group significant interaction effect. After the intervention, 504
491 showed a dramatic decrease in functional connectiv- the rTMS group had a significantly higher functional 505
492 ity after rTMS therapy (from z(r) = 0.151 to –0.234, connectivity (z(r) = 0.361 to –0.0198, p = 0.038) 506
493 p = 0.003), while in the sham group, an opposite trend compared to the sham group without baseline dif- 507
494 could be observed with a significant increment in ference (p = 0.452) (Supplementary Figure 4). No 508
8 H. Cui et al. / Hypoconnected DMN Improves Cognition in aMCI
f
roo
rP
tho
Fig. 2. Changes of the functional connectivity generated through PCG demonstrated significant Group × Time interactions for ACG.L with
deactivation effect (p < 0.05, Bonferroni corrected). Clusters of significance were demonstrated in the sagittal, coronal and axial views
(p < 0.05, family-wise error corrected) with mean correlation z(r) of each connectivity pair. A) The functional connectivity between posterior
Au
cingulate gyrus (PCG) and left anterior cingulate gyrus (ACG.L) showed a post-regimen difference of significance (p < 0.05) with a trend for
deactivation in the rTMS group (p = 0.094). Color bars are presented with t values. Bars show means and error bars represent SEM *p < 0.05,
**p < 0.01. B) Marginally significant correlation between changes in PCG-ACG.L connectivity and improvements in AVLT-D5, but not for
AVLT-D20, were seen. Dashed line indicates 95% confidence interval.
509 significant correlation of changes in connectiv- and other previous studies (Fig. 4A). Significant 532
ity between PCG and FFG.R/SFG.L/SMA.L with correlation between baseline functional connectivity
d
510 533
511 rTMS-induced alterations in AVLT scores was found. and neuropsychological test improvements in rTMS 534
In conclusion, all ROIs within the DMN area showed groups after FDR correction was found. Specifi-
cte
512 535
513 intervention-related deactivation (i.e., they presented cally, the baseline functional connectivity between 536
514 with reduction in functional connectivity) while the PCG.R and PCUN.R (r = –0.848, p < 0.05), IPL.L 537
515 only outlier ROI (SMA.L) exhibited a hyperactiva- (r = –0.817, p < 0.05), and PCUN.L (r = –0.850, 538
516 tion tendency after rTMS intervention. p < 0.05), was significantly negatively correlated 539
rre
517 Less activated baseline functional connectivity not the sham group (Fig. 4B-D). However, the 541
518 within DMN indicates a better response toward functional connectivity between PCG.R-PCG.L 542
519 rTMS therapy was not significantly correlated with AVLT-R 543
co
520 In order to investigate the potential mechanisms functional connectivity between IPL.L and PCG.L, 545
521 for a difference in response regarding cognitive PCUN.L, and vMPFC.L were also negatively 546
522 improvement among the patients receiving thera- correlated with AVLT-R with an expected trend 547
Un
523 peutic rTMS, we further analyzed the relationship (respectively for PCG.L, r = –0.753, p = 0.053; for 548
524 between the improving neuropsychological assess- PCUN.L, r = –0.617, p = 0.081; and for vMPFC.L, 549
525 ments (AVLT-I/R) and the initial status of brain r = –0.624, p = 0.189) (Fig. 4E, F, H). No signif- 550
526 represented by the baseline connectivity within icant correlations were found between AVLT-I 551
527 DMN. We mainly focused on four pairs of ROIs improvement and ROIs, although a negative corre- 552
528 including bilateral precuneus (PCUN.L/R), inferior lation with vMPFC.R-IPL.L (r = –0.765, p = 0.107, 553
529 parietal lobe (IPL.L/R), ventral medial prefrontal FDR corrected) approached significance (Supple- 554
530 cortex (vMPFC.L/R) as well as posterior cingulate mentary Figure 5A). Among the ROIs showing 555
531 gyrus (PCG.L/R) based on the present rTMS results significant Group × Time interaction changing 556
H. Cui et al. / Hypoconnected DMN Improves Cognition in aMCI 9
f
roo
rP
tho
Fig. 3. Changes of the functional connectivity generated through PCG demonstrated significant Group × Time interactions for FFG.R with
Au
deactivation effect (p < 0.05, Bonferroni corrected). Clusters of significance were demonstrated in the sagittal, coronal and axial views
(p < 0.05, FWE corrected) with mean correlation z(r). The functional connectivity between PCG and right fusiform gyrus (FFG.R) showed
significant within-group (p < 0.01) as well as between-group (p < 0.01) differences of functional connectivity reduction. Color bars are
presented with t values. Bars show means and error bars represent SEM. *p < 0.05, **p < 0.01.
after rTMS therapy, only PCG-SMA.L showed a rTMS effects on resting-state DMN
d
557 578
559 579
560 Figure 5B). the DMN have decreased FC with PCG, and the 580
561 DISCUSSION prefrontal cortex that constitutes the anterior DMN 583
rre
562 As is well-appreciated, to date, there has been lim- FC within DMN during resting state is responsible 585
563 ited research focusing on the longitudinal changes for the cognitive function improvement. This find- 586
564 of DMN network activity in aMCI and the resting- ing might have far-reaching significance and could 587
566 non-invasive methods of therapeutic intervention. nomena which had indicated the susceptibility to 589
567 Here, in the present study, our results not only AD-related pathology due to excessive metabolic 590
568 provide evidence for the effectiveness of rTMS ther- demand by neural activity and leads to further deduc- 591
Un
569 apy to ameliorate memory deficits of aMCI patients tions regarding the abnormalities of the DMN in 592
570 through improved neuropsychological performance, aMCI patients [16, 49, 50]. During the process of 593
571 but also reveal a potential mechanism associated with aging, as well as the pathological development of cog- 594
572 connectivity-based DMN alterations via rsfMRI. nitive impairment, the DMN is deeply involved, but 595
573 Moreover, our results show for the first time that changes over time have been seen to reflect altered 596
574 less activated baseline functional connectivity (FC) (increased/decreased) internal connections revealed 597
575 within the DMN indicates a better likelihood of by FC, varying according to different stages [21, 598
576 response to rTMS therapy, which itself also generally 51–57]. While a decreased FC in DMN is generally 599
f
roo
rP
tho
d Au
Fig. 4. Correlations between symptoms improvements and baseline functional connectivity between pairs of ROIs within DMN (p < 0.05,
FDR corrected). Symptoms improvement is percent change in AVLT-R (post-pre/pre). The black dots are individual data from subjects of
rTMS group, while the red triangle mark represented subjects of sham group. Black and red lines are the lines of best fit for the two group
cte
along with r and p values, respectively. A) Anatomical location of each ROI selected within DMN: posterior cingulate gyrus (PCG.L/R),
inferior parietal lobe (IPL.L/R), precuneus (PCUN.L/R) and ventral medial prefrontal cortex (vMPFC.L/R). B-D) The baseline functional
connectivity between PCG.R and PCUN.R (r = –0.848, p < 0.05), IPL.L (r = –0.817, P < 0.05), PCUN.L (r = –0.850, p < 0.05) revealed negative
correlation of significance with AVLT-R improvement in rTMS group, but not in sham group. G) The functional pair between PCG.L-R
and AVLT-R change (r = –0.624, p = 0.161). E, F, H) IPL.L-PCG.L/PCUN.L/vMPFC.L negatively correlated with AVLT-R with an expected
trend (r = –0.753, p = 0.053; r = –0.617, p = 0.081; and r = –0.624, p = 0.189, respectively).
rre
601 during the stage of AD, explanations vary for the probably already exists and persists from the pro- 616
602 bidirectional changes of connectivity seen in different dromal stage of AD, and remains as the disease 617
co
603 studies with aMCI patients, especially an enhanced progresses to aMCI. Thus, the so-called compen- 618
604 interaction with PCG. Some hypothesize that the satory increase in FC within certain parts of the 619
605 increased FC of certain components within DMN DMN is unlikely to make up for degraded cog- 620
606 could be explained as a “compensatory process” of nitive performance presenting with memory loss 621
Un
607 the brain aiming to make up for the less activated and executive dysfunction. A recent study focus- 622
608 brain regions engaged in the episodic memory pro- ing on asymptomatic individuals with AD family 623
609 cess [51, 58, 59]. Still, other studies questioned the history illustrated a similar scenario [61]. When 624
610 permanence of the response as well as the abil- divided into two separate groups based on subjec- 625
611 ity of enhanced DMN FC to facilitate memory tive cognitive decline (SCD) complaints, SCD [–] 626
612 function [60]. Our finding suggests that this compen- subjects showed lower posterior DMN-medial tem- 627
613 sation (hyperconnectivity within DMN) may not be poral memory system (MTMS) connectivity, and 628
614 effective as expected, and is possibly counterproduc- posterior DMN-MTMS connectivity was negatively 629
615 tive. As aforementioned, the functional abnormality correlated with immediate memory over time [61]. 630
H. Cui et al. / Hypoconnected DMN Improves Cognition in aMCI 11
631 Another longitudinal study also demonstrated the Baseline functional connectivity predicts 659
632 close relationship between the increasingly overac- treatment response 660
635 ther indicate that deactivation of the overactivated possible response to rTMS through a resting-state 662
636 DMN in aMCI patients can efficiently reverse mem- fMRI scan because rsfMRI is the most applica- 663
637 ory deficits and is potentially beneficial for the ble, easily-controlled means of imaging that can be 664
638 maintenance and normalization of a normally acti- widely implemented regardless of varying clinical 665
f
639 vated DMN. Meanwhile, there is a growing body of situations. In the present study, the subjects that pre- 666
roo
640 evidence underscoring the unwanted consequences sented with better therapeutic effects were those with 667
641 associated with chronic hyperconnectivity. In the relatively low resting-state functional connectivity at 668
642 context of the growing awareness of the dispro- baseline (no significant correlation was seen in the 669
643 portionately higher A deposition within highly sham group). All baseline connectivity pairs of ROIs 670
644 connected networks (e.g., posterior DMN) due to chosen, including among bilateral PCG, PCUN, IPL, 671
rP
645 elevated metabolism [63, 64], it is possible that the and vMPFC demonstrated obvious negative correla- 672
646 increased brain metabolism caused by local hyper- tion with the improvement in cognitive performance 673
647 connectivity would lead to sites of neurodegeneration seen for the rTMS group, indicating the importance 674
648 resulting in a cascade of network failure [21, 65], of a relative deactivation of DMN preceding a bet- 675
tho
649 although the relationship between elevated activity 676
650 and amyloid pathology could be reciprocal [66, 67]. patients with less activated DMN during the resting 677
651 Therefore, the risk of progressive cognitive decline state are more likely to be at an earlier stage of disease 678
652 in AD patients with A/tau-associated neuronal with relatively favorable prognosis, while those with 679
653 toxicity in hub regions of DMN could be accen- overactivated DMN are more likely to be approach- 680
Au
654 tuated through chronic hyperconnectivity [21]. The ing a later stage of MCI and consequent advanced 681
655 hypothetical developmental sequence of dementia deterioration. Based on the results obtained through 682
656 concerning cognitive decline and the corresponding rsfMRI in the present study, we surmise that the deac- 683
657 changes in activation of certain components of the tivated pattern within certain regions of the DMN 684
658 DMN is illustrated in Fig. 5. during resting state might help ease the difficulty in 685
686
688
ment. 697
ical stage of AD, undergo specific symptoms such as memory activities requiring initiative, such as memorizing and 701
loss and simultaneously present with hyperactivated DMN within calculating. Although we discovered a hyperactivated 702
certain circuit components. Based on work in this study, the mech- connection between PCG and SMA.L (a key node of 703
anism of rTMS ameliorating memory deficits is to deactivate the
abnormally overactivated DMN toward its normal activation sta- the sensory-motor network, SMN) (Supplementary 704
tus. Note that this diagram represents a hypothetical model for the Figure 4) and a positive correlation between base- 705
neuropsychological-neurofunctional continuum of AD but does line FC and neuropsychological score improvement 706
not affirm the exact time point of transition from overactivation (Supplementary Figure 5B), this potentially inspiring 707
to deactivation, as well as its relative extent of change. In clinical
practice, situations of different individuals can vary considerably result, indicating the possibility of a rTMS-induced 708
due to variable clinical backgrounds and other confounding factors. useful compensatory circuit beyond DMN (SMN) 709
12 H. Cui et al. / Hypoconnected DMN Improves Cognition in aMCI
713 includes variable degrees of collaboration and com- ral Science Foundation of China (No. 81671043), 761
714 prehension by patients that lead to uncertain impact the “Shuguang Program (16SG15)” supported by 762
715 on the coherence of the assessment. Cognitive fluc- Shanghai Education Development Foundation, the 763
716 tuation is common in aMCI and a reliable outcome Shanghai Municipal Commission of Health and 764
717 obtained through carefully devised task-related fMRI Family Planning (20174Y0021), and the Shanghai 765
f
718 cannot be guaranteed. One of the most important Municipal Education Commission - Gaofeng Clinical 766
roo
719 advantages of rsfMRI imaging, on the other hand, Medicine Grant (20172001). 767
720 is the ability to scan patients who are impaired in The authors would like to thank all the patients 768
721 cognitive function and unable to actively respond to for their participation in this study and Dr. Eric B. 769
722 any task-based scanning paradigm [56]. Meanwhile, Dammer, School of Medicine, Emory University, 770
723 although the changes in DMN we saw have been Atlanta, USA, for critical reading and comments 771
rP
724 consistently reported and replicated across diverse on this manuscript. We thank the three anonymous 772
725 clinical cohorts via different analytical methods, an reviewers for their constructive comments and sug- 773
727 have ignored co-morbid disease status or aging pro- Authors’ disclosures available online (https:// 775
tho
728 cesses that beset patients with cognitive decline, www.j-alz.com/manuscript-disclosures/18-1296r2). 776
738 [1] Petersen RC, Smith GE, Waring SC, Ivnik RJ, Kokmen E, 782
739 apart from mild memory loss. Tangelos EG (1997) Aging, memory, and mild cognitive 783
impairment. Int Psychogeriatr 9(Suppl 1), 65-69. 784
cte
[2] Morris JC, Storandt M, Miller JP, McKeel DW, Price JL, 785
Rubin EH, Berg L (2001) Mild cognitive impairment rep- 786
740 Conclusion resents early-stage Alzheimer disease. Arch Neurol 58, 787
397-405. 788
741 In the present study, our results provide evidence [3] Vos SJ, Verhey F, Frolich L, Kornhuber J, Wiltfang J, 789
Maier W, Peters O, Ruther E, Nobili F, Morbelli S, Frisoni
rre
790
742 for the effectiveness of applying rTMS towards the GB, Drzezga A, Didic M, van Berckel BN, Simmons A, 791
743 treatment of aMCI patient symptoms and also explore Soininen H, Kloszewska I, Mecocci P, Tsolaki M, Vel- 792
744 connectivity changes induced by rTMS using longi- las B, Lovestone S, Muscio C, Herukka SK, Salmon E, 793
745 tudinally measured rsfMRI. Our results extend the Bastin C, Wallin A, Nordlund A, de Mendonca A, Silva 794
D, Santana I, Lemos R, Engelborghs S, Van der Mussele S, 795
understanding of one potential pathogenic mecha-
co
746
Alzheimer’s Disease Neuroimaging Initiative, Freund-Levi 796
747 nism in development of cognitive AD symptoms, Y, Wallin AK, Hampel H, van der Flier W, Scheltens P, 797
748 consistent with the hypothesis of the long-term Visser PJ (2015) Prevalence and prognosis of Alzheimer’s 798
749 unfavorable consequences for the engagement of disease at the mild cognitive impairment stage. Brain 138, 799
1327-1338.
Un
800
750 additional neural resources represented by hypercon- [4] Dubois B, Albert ML (2004) Amnestic MCI or prodromal 801
751 nectivity [16, 71]. A negative correlation between Alzheimer’s disease? Lancet Neurol 3, 246-248. 802
752 the improvement in cognitive function and connectiv- [5] Petersen RC, Morris JC (2005) Mild cognitive impairment 803
as a clinical entity and treatment target. Arch Neurol 62, 804
753 ity of certain regions within the DMN demonstrates
1160-1163; discussion 1167. 805
754 that a less abnormally hyperactivated DMN is more [6] Sparing R, Mottaghy FM (2008) Noninvasive brain stimula- 806
755 likely to reach clinical remission and indicates the tion with transcranial magnetic or direct current stimulation 807
756 possibility of defining classes of aMCI patients (TMS/tDCS)-From insights into human memory to therapy 808
of its dysfunction. Methods 44, 329-337. 809
757 translationally through DMN functional connectivity [7] Nardone R, Tezzon F, Holler Y, Golaszewski S, Trinka 810
758 under the resting state. E, Brigo F (2014) Transcranial magnetic stimulation 811
H. Cui et al. / Hypoconnected DMN Improves Cognition in aMCI 13
812 (TMS)/repetitive TMS in mild cognitive impairment and [22] Luo X, Li K, Jia YL, Zeng Q, Jiaerken Y, Qiu T, Huang P, 877
813 Alzheimer’s disease. Acta Neurol Scand 129, 351-366. Xu X, Shen Z, Guan X, Zhou J, Wang C, Xu JJ, Zhang M, 878
814 [8] Brys M, Fox MD, Agarwal S, Biagioni M, Dacpano G, Alzheimer’s Disease Neuroimaging Initative (2019) Altered 879
815 Kumar P, Pirraglia E, Chen R, Wu A, Fernandez H, Wagle effective connectivity anchored in the posterior cingulate 880
816 Shukla A, Lou JS, Gray Z, Simon DK, Di Rocco A, Pascual- cortex and the medial prefrontal cortex in cognitively intact 881
817 Leone A (2016) Multifocal repetitive TMS for motor and elderly APOE epsilon4 carriers: A preliminary study. Brain 882
818 mood symptoms of Parkinson disease: A randomized trial. Imaging Behav 13, 270-282. 883
819 Neurology 87, 1907-1915. [23] Chetelat G, Desgranges B, de la Sayette V, Viader F, 884
820 [9] Perera T, George MS, Grammer G, Janicak PG, Pascual- Eustache F, Baron JC (2003) Mild cognitive impair- 885
821 Leone A, Wirecki TS (2016) The Clinical TMS Society ment: Can FDG-PET predict who is to rapidly convert to 886
f
822 consensus review and treatment recommendations for TMS Alzheimer’s disease? Neurology 60, 1374-1377. 887
roo
823 therapy for major depressive disorder. Brain Stimul 9, [24] Hirao K, Ohnishi T, Hirata Y, Yamashita F, Mori T, 888
824 336-346. Moriguchi Y, Matsuda H, Nemoto K, Imabayashi E, Yamada 889
825 [10] Nardone R, Brigo F, Holler Y, Sebastianelli L, Versace V, M, Iwamoto T, Arima K, Asada T (2005) The prediction 890
826 Saltuari L, Lochner P, Trinka E (2018) Transcranial mag- of rapid conversion to Alzheimer’s disease in mild cogni- 891
827 netic stimulation studies in complex regional pain syndrome tive impairment using regional cerebral blood flow SPECT. 892
828 type I: A review. Acta Neurol Scand 137, 158-164. Neuroimage 28, 1014-1021. 893
rP
829 [11] Sack AT, Linden DE (2003) Combining transcranial mag- [25] Wang K, Liang M, Wang L, Tian L, Zhang X, Li K, Jiang T 894
830 netic stimulation and functional imaging in cognitive brain (2007) Altered functional connectivity in early Alzheimer’s 895
831 research: Possibilities and limitations. Brain Res Brain Res disease: A resting-state fMRI study. Hum Brain Mapp 28, 896
832 Rev 43, 41-56. 967-978. 897
833 [12] Shafi MM, Westover MB, Fox MD, Pascual-Leone A (2012) [26] Sorg C, Riedl V, Perneczky R, Kurz A, Wohlschlager AM 898
834 Exploration and modulation of brain network interactions (2009) Impact of Alzheimer’s disease on the functional con- 899
tho
835 with noninvasive brain stimulation in combination with neu- nectivity of spontaneous brain activity. Curr Alzheimer Res 900
836 roimaging. Eur J Neurosci 35, 805-825. 6, 541-553. 901
837 [13] Dayan E, Censor N, Buch ER, Sandrini M, Cohen LG (2013) [27] Turriziani P, Smirni D, Zappala G, Mangano GR, Oliveri M, 902
838 Noninvasive brain stimulation: From physiology to network Cipolotti L (2012) Enhancing memory performance with 903
839 dynamics and back. Nat Neurosci 16, 838-844. rTMS in healthy subjects and individuals with Mild Cogni- 904
840 [14] Hsu WY, Ku Y, Zanto TP, Gazzaley A (2015) Effects of non- tive Impairment: The role of the right dorsolateral prefrontal 905
Au
841 invasive brain stimulation on cognitive function in healthy cortex. Front Hum Neurosci 6, 62. 906
842 aging and Alzheimer’s disease: A systematic review and [28] Anderkova L, Rektorova I (2014) Cognitive effects of 907
843 meta-analysis. Neurobiol Aging 36, 2348-2359. repetitive transcranial magnetic stimulation in patients 908
844 [15] Koch G, Bonni S, Pellicciari MC, Casula EP, Mancini M, with neurodegenerative diseases - clinician’s perspective. 909
845 Esposito R, Ponzo V, Picazio S, Di Lorenzo F, Serra L, J Neurol Sci 339, 15-25. 910
846 Motta C, Maiella M, Marra C, Cercignani M, Martorana [29] Lee J, Choi BH, Oh E, Sohn EH, Lee AY (2016) Treatment 911
847 A, Caltagirone C, Bozzali M (2018) Transcranial magnetic of Alzheimer’s disease with repetitive transcranial magnetic 912
d
848 stimulation of the precuneus enhances memory and neural stimulation combined with cognitive training: A prospec- 913
849 activity in prodromal Alzheimer’s disease. Neuroimage 169, tive, randomized, double-blind, placebo-controlled study. 914
cte
861 nard DA, Shulman GL (2001) A default mode of brain perceptual decision making. Proc Natl Acad Sci U S A 113, 926
862 function. Proc Natl Acad Sci U S A 98, 676-682. 6059-6064. 927
863 [19] Scherr M, Pasquini L, Benson G, Nuttall R, Gruber M, [34] George MS, Post RM (2011) Daily left prefrontal repeti- 928
864 Neitzel J, Brandl F, Sorg C, Alzheimer’s Disease Neu- tive transcranial magnetic stimulation for acute treatment 929
865 roimaging I (2018) Decoupling of local metabolic activity of medication-resistant depression. Am J Psychiatry 168, 930
Un
942 [37] Tierney MC, Szalai JP, Snow WG, Fisher RH, Nores A, components: Correlation, anticorrelation, and causality. 1007
943 Nadon G, Dunn E, St George-Hyslop PH (1996) Predic- Hum Brain Mapp 30, 625-637. 1008
944 tion of probable Alzheimer’s disease in memory-impaired [53] Sheline YI, Morris JC, Snyder AZ, Price JL, Yan Z, 1009
945 patients: A prospective longitudinal study. Neurology 46, D’Angelo G, Liu C, Dixit S, Benzinger T, Fagan A, Goate 1010
946 661-665. A, Mintun MA (2010) APOE4 allele disrupts resting state 1011
947 [38] Estevez-Gonzalez A, Kulisevsky J, Boltes A, Otermin P, fMRI connectivity in the absence of amyloid plaques or 1012
948 Garcia-Sanchez C (2003) Rey verbal learning test is a use- decreased CSF Abeta42. J Neurosci 30, 17035-17040. 1013
949 ful tool for differential diagnosis in the preclinical phase [54] Hafkemeijer A, van der Grond J, Rombouts SA (2012) 1014
950 of Alzheimer’s disease: Comparison with mild cognitive Imaging the default mode network in aging and dementia. 1015
951 impairment and normal aging. Int J Geriatr Psychiatry 18, Biochim Biophys Acta 1822, 431-441. 1016
f
952 1021-1028. [55] Hafkemeijer A, Altmann-Schneider I, Oleksik AM, van de 1017
roo
953 [39] Chao-Gan Y, Yu-Feng Z (2010) DPARSF: A MATLAB tool- Wiel L, Middelkoop HA, van Buchem MA, van der Grond J, 1018
954 box for “pipeline” data analysis of resting-state fMRI. Front Rombouts SA (2013) Increased functional connectivity and 1019
955 Syst Neurosci 4, 13. brain atrophy in elderly with subjective memory complaints. 1020
956 [40] Yan CG, Wang XD, Zuo XN, Zang YF (2016) DPABI: Brain Connect 3, 353-362. 1021
957 Data processing & analysis for (resting-state) brain imaging. [56] Sheline YI, Raichle ME (2013) Resting state functional con- 1022
958 Neuroinformatics 14, 339-351. nectivity in preclinical Alzheimer’s disease. Biol Psychiatry 1023
rP
959 [41] Saad ZS, Gotts SJ, Murphy K, Chen G, Jo HJ, Martin A, 74, 340-347. 1024
960 Cox RW (2012) Trouble at rest: How correlation patterns [57] Quiroz YT, Schultz AP, Chen K, Protas HD, Brickhouse M, 1025
961 and group differences become distorted after global signal Fleisher AS, Langbaum JB, Thiyyagura P, Fagan AM, Shah 1026
962 regression. Brain Connect 2, 25-32. AR, Muniz M, Arboleda-Velasquez JF, Munoz C, Garcia G, 1027
963 [42] Gotts SJ, Saad ZS, Jo HJ, Wallace GL, Cox RW, Martin Acosta-Baena N, Giraldo M, Tirado V, Ramirez DL, Tariot 1028
964 A (2013) The perils of global signal regression for group PN, Dickerson BC, Sperling RA, Lopera F, Reiman EM 1029
tho
965 comparisons: A case study of Autism Spectrum Disorders. (2015) Brain imaging and blood biomarker abnormalities 1030
966 Front Hum Neurosci 7, 356. in children with autosomal dominant Alzheimer disease: A 1031
967 [43] Power JD, Barnes KA, Snyder AZ, Schlaggar BL, Petersen cross-sectional study. JAMA Neurol 72, 912-919. 1032
968 SE (2012) Spurious but systematic correlations in func- [58] Zhang HY, Wang SJ, Xing J, Liu B, Ma ZL, Yang M, Zhang 1033
969 tional connectivity MRI networks arise from subject motion. ZJ, Teng GJ (2009) Detection of PCC functional connectiv- 1034
970 Neuroimage 59, 2142-2154. ity characteristics in resting-state fMRI in mild Alzheimer’s 1035
Au
971 [44] Yan CG, Cheung B, Kelly C, Colcombe S, Craddock RC, disease. Behav Brain Res 197, 103-108. 1036
972 Di Martino A, Li Q, Zuo XN, Castellanos FX, Milham [59] Qi Z, Wu X, Wang Z, Zhang N, Dong H, Yao L, Li K 1037
973 MP (2013) A comprehensive assessment of regional varia- (2010) Impairment and compensation coexist in amnestic 1038
974 tion in the impact of head micromovements on functional MCI default mode network. Neuroimage 50, 48-55. 1039
975 connectomics. Neuroimage 76, 183-201. [60] Hillary FG, Roman CA, Venkatesan U, Rajtmajer SM, Bajo 1040
976 [45] Brier MR, Thomas JB, Snyder AZ, Benzinger TL, Zhang D, R, Castellanos ND (2015) Hyperconnectivity is a fundamen- 1041
977 Raichle ME, Holtzman DM, Morris JC, Ances BM (2012) tal response to neurological disruption. Neuropsychology 1042
d
978 Loss of intranetwork and internetwork resting state func- 29, 59-75. 1043
979 tional connections with Alzheimer’s disease progression. [61] Verfaillie SCJ, Pichet Binette A, Vachon-Presseau E, Tabrizi 1044
cte
986 Graph theoretical analysis of structural and functional sys- [62] Wang Z, Liang P, Jia X, Jin G, Song H, Han Y, Lu J, Li 1051
987 tems. Nat Rev Neurosci 10, 186-198. K (2012) The baseline and longitudinal changes of PCC 1052
988 [48] Leech R, Sharp DJ (2014) The role of the posterior cingulate connectivity in mild cognitive impairment: A combined 1053
989 cortex in cognition and disease. Brain 137, 12-32. structure and resting-state fMRI study. PLoS One 7, e36838. 1054
990 [49] Cirrito JR, Yamada KA, Finn MB, Sloviter RS, Bales [63] Sperling RA, Laviolette PS, O’Keefe K, O’Brien J, Rentz 1055
co
991 KR, May PC, Schoepp DD, Paul SM, Mennerick S, DM, Pihlajamaki M, Marshall G, Hyman BT, Selkoe DJ, 1056
992 Holtzman DM (2005) Synaptic activity regulates inter- Hedden T, Buckner RL, Becker JA, Johnson KA (2009) 1057
993 stitial fluid amyloid-beta levels in vivo. Neuron 48, Amyloid deposition is associated with impaired default net- 1058
994 913-922. work function in older persons without dementia. Neuron 1059
995 [50] Vlassenko AG, Vaishnavi SN, Couture L, Sacco D, Shan- 63, 178-188. 1060
Un
996 non BJ, Mach RH, Morris JC, Raichle ME, Mintun MA [64] Sheline YI, Raichle ME, Snyder AZ, Morris JC, Head D, 1061
997 (2010) Spatial correlation between brain aerobic glycolysis Wang S, Mintun MA (2010) Amyloid plaques disrupt rest- 1062
998 and amyloid-beta (Abeta) deposition. Proc Natl Acad Sci U ing state default mode network connectivity in cognitively 1063
999 S A 107, 17763-17767. normal elderly. Biol Psychiatry 67, 584-587. 1064
1000 [51] Bai F, Zhang Z, Yu H, Shi Y, Yuan Y, Zhu W, Zhang X, [65] Dumont M, Wille E, Stack C, Calingasan NY, Beal MF, 1065
1001 Qian Y (2008) Default-mode network activity distinguishes Lin MT (2009) Reduction of oxidative stress, amyloid 1066
1002 amnestic type mild cognitive impairment from healthy deposition, and memory deficit by manganese superoxide 1067
1003 aging: A combined structural and resting-state functional dismutase overexpression in a transgenic mouse model of 1068
1004 MRI study. Neurosci Lett 438, 111-115. Alzheimer’s disease. FASEB J 23, 2459-2466. 1069
1005 [52] Uddin LQ, Kelly AM, Biswal BB, Castellanos FX, Milham [66] Brody DL, Magnoni S, Schwetye KE, Spinner ML, 1070
1006 MP (2009) Functional connectivity of default mode network Esparza TJ, Stocchetti N, Zipfel GJ, Holtzman DM (2008) 1071
H. Cui et al. / Hypoconnected DMN Improves Cognition in aMCI 15
1072 Amyloid-beta dynamics correlate with neurological status [70] Chhatwal JP, Schultz AP, Johnson KA, Hedden T, Jaimes S, 1090
1073 in the injured human brain. Science 321, 1221-1224. Benzinger TLS, Jack C, Jr., Ances BM, Ringman JM, Mar- 1091
1074 [67] Myers N, Pasquini L, Gottler J, Grimmer T, Koch K, Ort- cus DS, Ghetti B, Farlow MR, Danek A, Levin J, Yakushev 1092
1075 ner M, Neitzel J, Muhlau M, Forster S, Kurz A, Forstl H, I, Laske C, Koeppe RA, Galasko DR, Xiong C, Masters 1093
1076 Zimmer C, Wohlschlager AM, Riedl V, Drzezga A, Sorg C CL, Schofield PR, Kinnunen KM, Salloway S, Martins RN, 1094
1077 (2014) Within-patient correspondence of amyloid-beta and McDade E, Cairns NJ, Buckles VD, Morris JC, Bateman 1095
1078 intrinsic network connectivity in Alzheimer’s disease. Brain R, Sperling RA, Dominantly Inherited Alzheimer Network 1096
1079 137, 2052-2064. (2018) Preferential degradation of cognitive networks dif- 1097
1080 [68] Sheline YI, Barch DM, Price JL, Rundle MM, Vaishnavi ferentiates Alzheimer’s disease from ageing. Brain 141, 1098
1081 SN, Snyder AZ, Mintun MA, Wang S, Coalson RS, Raichle 1486-1500. 1099
f
1082 ME (2009) The default mode network and self-referential [71] Hillary FG, Grafman JH (2017) Injured brains and adap- 1100
roo
1083 processes in depression. Proc Natl Acad Sci U S A 106, tive networks: The benefits and costs of hyperconnectivity. 1101
1084 1942-1947. Trends Cogn Sci 21, 385-401. 1102
1085 [69] Jones DT, Machulda MM, Vemuri P, McDade EM, Zeng G,
1086 Senjem ML, Gunter JL, Przybelski SA, Avula RT, Knopman
1087 DS, Boeve BF, Petersen RC, Jack CR, Jr. (2011) Age-related
1088 changes in the default mode network are more advanced in
rP
1089 Alzheimer disease. Neurology 77, 1524-1531.
tho
d Au
cte
rre
co
Un