GeroScience (2023) 45:477–489 https://doi.org/10.1007/s11357-022-00656-7 ORIGINAL ARTICLE Episodic memory dysfunction and hypersynchrony in brain functional networks in cognitively intact subjects and MCI: a study of 379 individuals Brenda Chino · Pablo Cuesta · Javier Pacios · Jaisalmer de Frutos‑Lucas · Lucía Torres‑Simón · Sandra Doval · Alberto Marcos · Ricardo Bruña · Fernando Maestú Received: 23 June 2022 / Accepted: 1 September 2022 / Published online: 15 September 2022 © The Author(s) 2022 Abstract Delayed recall (DR) impairment is one of test, we identified a cluster of differences in FC and the most significant predictive factors in defining the studied which connections drove such an effect in DR. progression to Alzheimer’s disease (AD). Changes in The moderation effect of APOE genotype between FC brain functional connectivity (FC) could accompany results and delayed recall was evaluated too. Higher this decline in the DR performance even in a resting FC in beta band in the right occipital region is associ- state condition from the preclinical stages to the diag- ated with lower DR scores in both groups. A signifi- nosis of AD itself, so the characterization of the rela- cant anteroposterior link emerged in the seed-based tionship between the two phenomena has attracted analysis with higher values in MCI. Moreover, APOE increasing interest. Another aspect to contemplate is the genotype appeared as a moderator between beta FC and potential moderator role of the APOE genotype in this DR performance only in the CI group. An increased association, considering the evidence about their impli- beta FC in the anteroposterior brain region appears to cation for the disease. 379 subjects (118 mild cognitive be associated with lower memory performance in MCI. impairment (MCI) and 261 cognitively intact (CI) indi- This finding could help discriminate the pattern of the viduals) underwent an extensive evaluation, including progression of healthy aging to MCI and the relation MEG recording. Applying cluster-based permutation between resting state and memory performance. B. Chino (*) J. Pacios · J. de Frutos‑Lucas · L. Torres‑Simón · S. Doval · Institute of Neuroscience, Autonomous University F. Maestú of Barcelona, Barcelona, Spain Department of Experimental Psychology, Complutense e-mail: brendanadia.chino@autonoma.cat University of Madrid, Madrid, Spain B. Chino · P. Cuesta · J. Pacios · J. de Frutos‑Lucas · J. de Frutos‑Lucas L. Torres‑Simón · S. Doval · R. Bruña · F. Maestú Centre for Precision Health, Edith Cowan University, Center for Cognitive and Computational Neuroscience, Joondalup, WA 6027, Australia Complutense University of Madrid, Madrid, Spain J. de Frutos‑Lucas P. Cuesta · R. Bruña Centro de Investigación Nebrija en Cognición (CINC), Department of Radiology, Rehabilitation, Universidad de Nebrija, Madrid, Spain and Physiotherapy, Complutense University of Madrid, Madrid, Spain A. Marcos Neurology Department, Hospital Clinico San Carlos, P. Cuesta · J. Pacios · A. Marcos · R. Bruña · F. Maestú Madrid, Spain Instituto de Investigación Sanitaria San Carlos (IdISSC), Madrid, Spain Vol.: (0123456789) 13 478 GeroScience (2023) 45:477–489 Keywords Delayed recall · MCI · Cognitively (e.g., del Río et al. [13]). Indeed, increased FC, normal · Functional connectivity · MEG measured during functional magnetic resonance imaging (fMRI), has been associated with reduced episodic memory in patients with MCI and AD [14]. Introduction However, fMRI-based FC estimates are an indirect measure of brain activity. Furthermore, they pro- Alzheimer’s disease (AD) is a progressive neurodegen- vide limited information on the frequency of neural erative disorder that can develop unnoticed for several oscillations, particularly those in the fast ranges, con- years prior to the manifestation of clinical symptoma- sidered one of the core neural mechanisms for cog- tology. Along with executive functioning, episodic nition, and more relevant to this study, for episodic memory impairments are the hallmark of the cognitive memory [15, 16]. Therefore, neurophysiological dysfunction associated with the disease [1, 2], which measures such as magnetoencephalography (MEG) accompany histopathological and morphological or electroencephalography (EEG), having a high changes. In particular, delayed recall (DR) of meaning- temporal resolution, are able to capture the brain ful information is notably affected in AD, with recall dynamics in the frequency domain providing an of stories showing the sharpest deterioration in longitu- enriched information for network analysis. dinal studies [3]. Moreover, episodic memory declines Considering all the evidence provided above, are well documented in people not yet diagnosed with in this study we explore the potential association AD [4] (see also Bäckman et al. [5] for meta-analytic between delayed recall performance and FC of rest- evidence), suggesting that individual performance ing-state brain activity in healthy volunteers and on tests evaluating this cognitive domain may help MCI patients. As a second step, the role of APOE to identify people either at risk of developing AD or genotype as a potential moderator of the relation even in prodromal stages as mild cognitive impairment between FC and delayed recall was evaluated, under (MCI) [3]. Moreover, the genotype for the apolipopro- the hypothesis that this genotype can increase brain tein E (APOE) is known to be the major genetic risk vulnerability to damage. We expect that subjects factor for AD, and carriers of ε4 allele show greater with MCI will show a differential functional connec- impairments on episodic memory performance in the tivity profile than cognitively intact (CI) participants form of poorer DR (e.g., Wolk et al. [6]). related to memory performance, indicating prelimi- On the neurophysiological level, the last decades nary signs of conversion in the AD continuum. have seen the emergence of functional connectivity (FC) and its identification as a potential neuromarker for the diagnosis of AD [7, 8]. Disrupted synchroni- Methods zation is considered as a sign of synaptic dysfunction in AD, consequence of abnormal neural excitation/ Participants inhibition balance [9]. Indeed, people with MCI tend to show hypersynchronization of anterior networks, The sample consisted of 379 individuals divided in combination with hypo-synchronization of pos- into 118 MCI (aged from 58 to 87) and 261 CI par- terior areas [10, 11]. Given that episodic memory ticipants (aged from 41 to 82). The participants were impairments could be accompanied by changes in recruited from the Hospital Universitario San Carlos brain FC from the preclinical stages of the disease [17] and from “Centro para Mayores del Distrito de to the diagnosis of AD itself, the characterization Chamartín”, both located in Madrid (Spain). Gen- of the functional relationship between the two phe- eral inclusion criteria were as follows: a modified nomena has attracted increasing interest. Rather than Hachinski score ≤ 4, a Geriatric Depression Scale focusing on task-related FC measures, which can be (short form) score ≤ 5, and T1, T2, and diffusion- affected by variability in experimental designs and weighted MRIs within 54 weeks before the MEG task difficulty [12], neuroimaging research in the last recordings (on average, the time period between the years has focused on the study of resting-state (RS) MEG and MRI recordings was 3 months) without an functional connectivity patterns and how they relate indication of infection, infarction, or focal lesions to individual differences in cognitive performance (rated by two independent experienced radiologists Vol:. (1234567890) 13 GeroScience (2023) 45:477–489 479 [18]). In addition, the criteria for the MCI diagnosis Neuropsychological assessment were established according to the NIA-AA clinical criteria [19]. For more information about the diag- All participants were screened using standardized nostic criteria for MCI, see López et al. [20]. For CI diagnostic instruments and received a thorough neu- participants, we exclude subjects with evidence of ropsychological assessment as formerly detailed in significant hippocampal atrophy in a T1-weighted López et al. [20]. The screening consisted of stand- MRI scan within 2 months before MEG acquisition, ardized tests that included the Spanish version of as hippocampal atrophy is considered a brain marker the Mini-Mental State Examination (MMSE; [22]), associated with neurodegeneration [21]. No one of the Geriatric Depression Scale-Short Form (GDS- the participants exhibited a history of psychiatric SF; [23]), and the Logical Memory (I and II) subtest or neurological disorders other than MCI. Further- (Wechsler Memory Scale III, WMS-III; [24]). more, we advised subjects to avoid medications that Due to its effectiveness as a measure of verbal epi- could affect MEG activity, such as benzodiazepines, sodic memory, logical memory (LM) is one of the for 48 h before recordings (A detailed list of the most frequently administered subtests in the Wechsler samplecharacteristics can be found in Table 1). Memory Scale-III (LM-WMS-III) [24]. In the LM test, the participants presented a text, and the memory abil- ity is divided into immediate recall, delayed recall, and Standard protocol approvals, registrations, and patient recognition. Our study only included the analysis of the consents delayed recall score, which consisted of free recall of the passages after a 20 to 30 min delay after the pres- All participants were native Spanish speakers and entation. The narrative nature of the task is sensitive to provided written informed consent. The Institutional discriminate between normal aging, MCI [25], and early Review Board Ethics Committee at Hospital Universi- dementia, due to its tight relationship with other high- tario San Carlos approved the study protocol, and the level cognitive functions such as episodic memory, con- procedure was performed following the Helsinki Dec- ceptual organization, and schema formation [26]. laration and National and European Union regulations. Table 1  Descriptive measures of the final sample Variable Whole sample CI MCI p values Sex (M/F) 135/244 91/170 44/74 Age 68.00 ± 8.59 65.96 ± 8.48 74.28 ± 5.26 < 0.001 Education 13.24 ± 5.54 14.30 ± 5.49 9.99 ± 4.33 < 0.001 Depression 2.23 ± 2.74 1.63 ± 2.33 4.05 ± 3.09 < 0.001 Delayed recall 40.46 ± 23.21 49.98 ± 17.18 11.14 ± 12.20 < 0.001 eTIV 1,384,457.31 ± 8190.08 1,395,275.80 ± 154,063.19 1,359,214.17 ± 148,921.36 0.224 Total GM 558,888.08 ± 545,633.32 567,118.08 ± 53,093.03 533,530.78 ± 51,447.00 < 0.001 Total WM 424,346.04 ± 64,360.58 432,579.60 ± 64,658.18 398,977.79 ± 56,705.69 < 0.001 Hippocampus 6829.66 ± 952.96 7132.30 ± 846.21 6123.50 ± 805.16 < 0.001 Left cingulum 0.41 ± 0.04 0.41 ± 0.03 0.39 ± 0.04 < 0.001 Right cingulum 0.41 ± 0.40 0.41 ± 0.03 0.39 ± 0.04 < 0.001 Forceps major 0.55 ± 0.03 0.55 ± 0.03 0.53 ± 0.04 < 0.001 Forceps minor 0.54 ± 0.04 0.54 ± 0.04 0.52 ± 0.03 < 0.001 We present values as mean ± standard deviation (SD) for the characteristics of the participants as well as variables used for correla- tion analyses. These include sex (male/female), age (in years), education (expressed in years of education), depression (measured by global depression scale), delayed recall (Wechsler Scale-Logical Memory II Index) and structural measures such as eTIV estimated total intracranial volume, GM total gray matter volume, WM cerebral white matter volume, hippocampus volume, left and right cin- gulum along the hippocampal cortex fractional anisotropy, forceps major fractional anisotropy, and forceps minor fractional anisot- ropy. Results are displayed for the whole sample and for each subsample of interest (CI and MCI) Vol.: (0123456789) 13 480 GeroScience (2023) 45:477–489 MRI acquisition and volumetric analyses Afterwards, we estimated the source level activ- ity for each individual. As source model, we used a We used a General Electric 1.5 T system with a high- 1 cm homogeneous grid of source positions defined resolution antenna and a homogenization PURE fil- in MNI space and labeled according to the automated ter (Fast Spoiled Gradient Echo sequence, TR/TE/ anatomical labeling (AAL) atlas. This source model TI = 11.2/4.2/450 ms; flip angle 12°; 1-mm slice consisted of 1202 positions in 78 cortical areas and thickness, 256 × 256 matrix, and FOV 25 cm) to was transformed to subject space using a linear trans- obtain T1-weighted images of our participants. The formation between the template and the T1-weighted resulting images were processed using the Freesurfer MRI of the participant. This image was also used to software (version 5.1.0) and its specialized tool for generate a single-shell head model defined by the automated cortical parcellation and subcortical seg- inner skull surface. Then, we combined the head mentation [27]. The measures that were included in model, the source model, and the sensor definition to further analyses were total gray matter, total cerebral create a lead field using a modified spherical solution. white matter, and hippocampus (in m ­ m3). The vol- As the last step, we used a linearly constrained mini- umes of bilateral structures were collapsed in order to mum variance beamformer as inverse method. obtain a single measure for each region. We estimated FC by means of the phase locking value (PLV), a phase synchronization metric that Diffusion tensor imaging evaluates the distribution of the phase difference between two-time series. Briefly, after source recon- The same scanner was also used to collect diffusion- struction, the dataset consisted of matrices of 1202 weighted images (DWI) (single-shot echo planar nodes by 4000 samples by epochs for each of the 4 sequence, TE/TR 96.1/12,000 ms; NEX 3 for increas- frequency bands studied here. Then, for each fre- ing the SNR; 2.4-mm slice thickness, 128 × 128 quency band and epoch, we calculated the PLV [30] matrix, and 30.7 cm FOV). We acquired 1 image with via the following procedure: firstly, we used the Hil- no diffusion sensitization (i.e., b0 images) and 25 bert transform to extract the instantaneous phase φj(t) DWI directions (b = 900 s/mm2). for each node j = 1…1202 and time t = 1…4000 ms: DWI images were processed using probabilistic ( ) fiber tractography which was run on the automated zj (t) = xj (t) + i ∙ Hilbert xj (t) = Aj (t) ∙ ei𝜑j (t) tool AutoPtx (https://​fsl.​fmrib.​ox.​ac.​uk/​fsl/​fslwi​ki/​ Secondly, we estimated the synchronization AutoP​tx) as in Verdejo-Román et al. [28]. Due to its between each pair of signals j and k by means of their relation to memory performance, we studied the rela- difference of phases 𝜑j (t) and 𝜑k (t) using the follow- tion between FC, DR, and fractional anisotropy (FA) ing expression: at the uncinate, forceps major, and forceps minor. |M | 1 ||∑ i(𝜑j (tm )−𝜑k (tm )) || Magnetoencephalography PLV = e M ||m=1 | | | | MEG data was recorded using a 306-channel whole- head MEG system (Vectorview, Elekta AG, Finland), where T = 4000 is the number of samples in the time placed in a magnetically shielded room located at the series (4 s per epoch at 1000 Hz sampling rate). Center for Biomedical Technology in Madrid, follow- Lastly, we averaged the PLV matrices across epochs ing the protocol described in de Frutos-Lucas et al. to obtain a more robust estimator of resting-state FC. [29]. First, we applied the Maxfilter software (tempo- This algorithm provided symmetrical whole-brain ral extension of the signal space separation method, matrices of 1202 × 1202 nodes per participant and correlation window of 10 s, and correlation limit of frequency band (theta, between 4 and 8 Hz; alpha, 0.9) to remove external noise. Then we used FielT- between 8 and 12 Hz; beta, between 12 and 30 Hz; trip software 28 to automatically scan the data for and gamma, between 30 and 45 Hz) [31]. Then, we artifacts, which were visually confirmed by an MEG calculated the nodal strength (also known as weighted expert. Artifact-free data were segmented in 4-s global connectivity), which is defined for each node epochs, plus 2 s of real data at each side as padding. as the sum of its FC with the rest of the nodes. To Vol:. (1234567890) 13 GeroScience (2023) 45:477–489 481 account for the number of links, the strength of each members of the cluster to obtain a representative FC node was then normalized by dividing the number of marker. Of note, this FC marker would be indicating links connected to it. This procedure resulted in one that the global FC of the possible significant clusters brain map of normalized node strengths per each par- appeared to be associated with memory performance. ticipant and frequency band. Seed‑based analyses (seed link FC) APOE genotype To examine whether the strength FC results were Genomic DNA was extracted from 10 ml blood caused by global or region-specific effects, we per- samples in ethylenediaminetetraacetic acid. Detec- formed complementary seed analyses, using the tion of APOE genotype was performed with previous clusters as seeds. For it, we calculated the TaqMan technology using an Applied Biosystems average FC of each source position with the sources 7900 HT Fast Real-Time PCR machine (Applied in the cluster. Then, we repeated the statistical CBPT Biosystems, Foster City, CA). See the genotyping analysis using these seed-based FC values instead of method previously described in Cuesta et al. [32] the strength FC values. for more information. All the sample was included independently of APOE genotype in the initial analysis. Then, to evaluate the potential moderation Correlations between FC and measures of white role of genotype, the participants were classified as matter integrity and brain volume APOE ɛ4 carriers and noncarriers (i.e., ɛ3ɛ3). Par- ticipants who presented less frequent allele com- We used the aforementioned cluster markers in sub- binations (i.e., ɛ2ɛ2, ɛ2ɛ3, ɛ2ɛ4, and ɛ4ɛ3ɛ4) were sequent correlation analyses with measures of AD- excluded from the sample. specific signatures. As to this, we used both the whole sample and a stratification of the cohort by diagnosis Statistical analyses (MCI and CI). To account for multiple comparisons, the resulting p values were corrected using a false dis- Functional connectivity strength (strength FC) covery rate (FDR). All statistical analyses were car- ried out using MATLAB R2020b (MathWorks Inc.). Cluster-based permutation test (CBPT) was car- ried out separately for each frequency band [33]. We defined a cluster as a set of spatially adjacent nodes Moderation analysis that presented a significant partial correlation (Spear- man correlation using age as a covariate, p < 0.001) Additionally, we analyzed the impact of FC on DR in the same direction between the strength FC values focusing on the possible influence exerted by APOE and each DR variable. In this framework, a cluster genotype. Each group was divided into ɛ4 carriers and can be considered as a functional unit. Only clusters ɛ4 noncarriers to evaluate multiple regression analy- including at least 1% of the grid (i.e., a minimum of sis (using age and years of education as covariates), 12 nodes) were considered. The Spearman rho values and we calculated the increase in variance explained were transformed into Fisher Z values, and the clus- after including the interaction into the model. Then, we ter-mass statistics were computed as the sum of the Z examined the effect of FC on delayed recall scores in the values of all nodes within the cluster. The p value for APOE genotype subgroups and used the Johnson-Ney- each cluster was calculated in a nonparametric fash- man technique to identify the threshold where the syn- ion, using a null distribution generated by the mass chronization shows a dysfunctional pattern in CI (CI33/ of the main cluster obtained over 5000 random per- CI34) and MCI (MCI33/MCI34). Statistical analyses mutations (shuffled versions) of the data [30]. Only were carried out using Process Macro, an extension for those clusters that resulted significant (p < 0.05) after SPSS that calculates X’s direct, indirect, and total effects this step were considered in further analyses. Then, on Y and unstandardized and standardized regression we used the average of the strength FC values of the Vol.: (0123456789) 13 482 GeroScience (2023) 45:477–489 Vol:. (1234567890) 13 GeroScience (2023) 45:477–489 483 ◂Fig. 1  A In dark blue, marked as primary β, is displayed the different between groups (ANCOVA with age brain region whose functional connectivity strength (strength and education as covariates, p value = 0.006, F FC) was found inversely correlated with delayed recall (DR). In light blue are depicted the region, marked as secondary β, value = 7.591), showing higher values for the MCI whose FC with the primary β cluster was found to inversely patients than for the CI individuals (Fig. 1C). correlate with DR. B Scatter plot shows the correlation between primary β cluster strength FC and DR computed with Lower occipital beta FC is differently associated with the whole sample (gray), MCI patients (red), and cognitively intact (CI) participants (green). Boxplot graphic shows the brain structure in MCI and CI patients average strength FC of the primary β cluster for each group. C Scatter plot shows the correlation between primary β < – > sec- To better understand our results, we explored the ondary β FC and DR computed with the whole sample (gray), relationship between the strength FC and structural MCI patients (red), and cognitively intact (CI) participants (green). Boxplot graphic shows the average seed link FC of measures (see Table 3). About MCI, white mat- the secondary β cluster for each group. In boxplot graphics, ter (r =  − 0.24; p = 0.01) and gray matter volume vertical blue line indicates 95% confidence interval, whereas (r =  − 0.22; p = 0.02) exhibited negative associa- salmon boxes depict avg ± sd tion with primary β electrophysiological activity. The same direction was observed with right cingu- coefficients, standard errors, t, p values, and R2 for the lum in the hippocampal area (r =  − 0.25; p = 0.01) models [34]. and forceps major fractional anisotropy (r =  − 0.24; p = 0.02). In contrast, only one inverse correlation was found for the FC of the link mβ-sβ related to Results the right gray matter volume (r =  − 0.19; p = 0.04). On the other hand, the CI group’s association Delayed recall is associated with decreased occipital in primary β with structural measures was only beta frequency band FC in MCI and CI participants expressed inversely with forceps minor fractional anisotropy (r =  − 0.14; p = 0.03). One significant (CBPT p value = 0.019) pri- mary cluster emerged in the beta band (primary β, Fig. 1A and B), located on the right occipi- Moderation effect of APOE‑ε4 genotype, beta FC tal region (detailed regions within the cluster are link in delayed recall scores shown in Table 2). The correlation between the strength FC of this cluster and DR performance After we had described how lower levels of DR was negative, indicating that the higher the strength related to a distinctive strength FC profile, we ana- FC, the lower the memory performance. In addi- lyzed the impact of APOE genotype as a modera- tion, the correlation remained significant when tor (W) between the pattern of FC in right occipital looking at the MCI (rho =  − 0.236; p = 0.011) and and left anterior areas in beta band (X) and the DR CI (rho =  − 0.207; p < 0.001) groups separately scores (Y) separately for each group. Our results (Fig. 1B). The strength FC value was found to be showed a significant regression coefficient differ- higher in the MCI patients than in the CI individ- ent from zero (b3 =  − 1428.07, t (209) =  − 3.10, uals, but the difference did not reach significance p = 0.002) for XW only for the CI group, meaning (ANCOVA with age and education as covariates, p that the effect of the network on episodic memory value = 0.150, F value = 2.079; see Fig. 1B). scores depends on the APOE genotype in cog- When exploring the seed-based FC in the brain, nitively intact participants. Once we defined the using the primary β cluster as seed, we found one moderator effect of the APOE genotype, we tried significant secondary cluster whose FC with the to identify what was the threshold in FC that dif- primary cluster showed to be negatively correlated ferentiated the episodic memory performance with DR. The correlation between DR and the seed in carriers vs noncarriers. The Johnson-Neyman link FC remained significant when looking at the CI technique allowed the exact calculation of the (rho =  − 0.165; p = 0.008) and MCI (rho =  − 0.312; conditions and the limit values in which a mod- p < 0.001) groups separately (Fig. 1A and C). How- erator obtains statistically significant slopes. Our ever, the average seed link FC was significantly results showed significant differences in FC values Vol.: (0123456789) 13 484 GeroScience (2023) 45:477–489 Table 2  Main cluster and seed analyses presented increased in FC at less DR performance Cluster mβDR % sβDR % ROIs Right middle occipital lobe 29.41 Left superior frontal gyrus. Medial 58.82 Right cuneus 15.38 Left middle frontal gyrus 44.12 Right superior occipital lobe 20.00 Left superior frontal gyrus 48.15 Right inferior occipital lobe 40.00 Left cingulate gyrus. Anterior part 42.11 Left superior frontal gyrus. Medial orbital 50.00 Left gyrus rectus 37.50 Left superior frontal gyrus. Orbital 66.67 Beta main cluster whose functional connectivity strength (FC) was significantly correlated with DR was used a seed in a seed-based analysis. List of regions of interests (ROIs) from the AAL atlas that were captured for each significant cluster. % depicts the percent- age of the ROI that fall within the cluster (only show the ROIs above 15%) Table 3  The relationship between the strength FC and structural measures MCI CI Structure r p value Structure r p value mβDR Total WM − 0.2431 0.0086 Forceps minor − 0.143 0.0253 Right cingulum − 0.2462 0.0122 Forceps mayor − 0.2365 0.0161 Total GM − 0.2152 0.0204 sβDR Right total GM − 0.1923 0.0386 Results for Spearman correlation analyses between the FC of the edge < main-β. seed2-β > and brain structural integrity scores. Total gray matter volume (GM) and total cerebral white matter volume (WM) in mm3; right cingulum along the hippocampal cortex frac- tional anisotropy; forceps major fractional anisotropy; forceps minor fractional anisotropy Fig. 2  Variation of the Spearman rho scores (age and education with age as covariate) obtained for four groups: CI/MCI with and as covariates) for the correlation between delayed recall (DR) without any allele APOE 4. As it can be seen, the influence of the and secondary β seed link FC with ε4 APOE carriage. In order allele 4 of the APOE is notorious in the CI ε4 group. The solid to show qualitatively the dependence of APOE in the relationship black line that connects all groups is just indicative of the Rho’s between DR and the FC of the link primary β <—> secondary β variance clusters, we have depicted the Spearman rho scores (computed Vol:. (1234567890) 13 GeroScience (2023) 45:477–489 485 higher than 0.116, which means that the effect Moreover, these results can also be interpreted as well of APOE genotype on episodic memory scores was as a hampered desynchronization. Beta desynchroni- expressed only when the value of FC was above zation is one of the main electrophysiological features 0.116 (See Fig. 2). of resting state brain activity, and a higher FC is com- patible with the idea of an interruption of the natural desynchronization process at rest, reflecting a dedif- Discussion ferentiation process [37]. Nevertheless, we will refer to our finding as hypersynchronization, as that is the Episodic memory dysfunction and altered functional direct interpretation of the actual results. connectivity patterns are among the most impor- Interestingly, when we tested the potential role of tant features associated with cognitive decline in the APOE ε4 allele as a moderator of the relationship AD and its prodromal stages [5]. Together with the between the identified pattern of FC (between right genetic risk posed by the presence of the ε4 allele of occipital and left anterior areas in beta band) and the APOE gene, these variables can provide valuable delayed recall memory, we found a significant effect information on the identification of individuals with a in the healthy ε4 carriers. In this subsample, the indi- higher risk of developing AD. With this aim, we eval- rect relationship between beta FC and memory per- uated these variables in a large sample of participants formance was exacerbated, whereas in MCI patients, either healthy or diagnosed with MCI. Firstly, we the role of APOE ε4 allele was nonsignificant, likely analyzed the relationship of FC and delayed recall, due to the more advanced stage of these patients in considering, in a second step, the potential modera- the AD continuum. These results would be indicating tor effect of the APOE genotype on this relationship. the existence, in the healthy ε4 carriers, of a selec- Results from our analysis identified a cluster of right tive vulnerability in the electrophysiological process occipital regions whose global beta band functional of beta hypersynchronization, producing a result that connectivity were negatively associated with delayed mirrored the one observed in pathological aging. The recall scores. Moreover, a post hoc seed-based analy- deleterious effect of having one or two copies of the sis showed that the left frontal cortex was the primary APOE ε4 allele has been largely described in the lit- region contributing to this effect. erature [38]. Hence, it does not come as a surprise Previous studies have linked hypersynchroniza- that in ε4 carriers’ functional alterations in the brain tion in the beta band with early stages of pathologi- are more strongly associated with cognitive outcomes cal aging [10]. In task, there is evidence showing that than in noncarriers, most likely due to a reduced MCI patients who later progress to AD present higher remodeling and repair capacity resulting from values of functional connectivity in the beta band than impaired lipid transportation processes in ε4 carriers. those who will remain stable [10]. Other studies have Unfortunately, genetic background is a non-modifia- confirmed that these values can be a good predictor ble risk factor for cognitive decline, and no therapy of progression from MCI to AD [35]. This hypersyn- can reduce the associated risk directly. However, chronization has been proposed to be derived from several modifiable lifestyle factors have been shown neuropathology. In particular, high accumulations of to exert an influence on FC [29, 39]. Given that this beta amyloid protein are especially toxic to inhibitory pattern of increased FC between right occipital and neurons, inducing an excess of excitability that can left anterior areas in beta band predicts lower cogni- cause spurious synchronization (Garcia-Marin et al., tive performance in ε4 carriers, future studies looking [36]; see also Maestú et al., [17] for a multicentric into the potential effect of several lifestyle factors to study). In this line, our results reveal that an increase decrease this pattern of FC are warranted. of functional connectivity between frontal and poste- Episodic memory performance has been tradi- rior regions is associated with reduced delayed recall tionally understood as the first clinical manifestation scores. Of importance, this functional coupling was of an underlying Alzheimer-type neuropathological greater in the MCI group than in the healthy partici- process [3–5]. However, the emergence of clinically pants, according to the idea of a progressive increase measurable memory damage occurs in a stage where of neuronal excitability along the continuum of the the brain damage is substantial. Here, we sought for disease, until the network breakdown at AD [35]. functional integrity markers associated with memory Vol.: (0123456789) 13 486 GeroScience (2023) 45:477–489 performance that can be used as proxies of the under- clearly significant in healthy ε4 allele carriers, it did lying neuropathological process. Since the FC mark- not reach significance in noncarriers. On the other ers can be computed in every participant, this approx- hand, the MCI patients showed an exacerbation of imation would allow the identification of individuals this malfunction both in ε4 carriers and noncarri- with a higher risk of developing AD [40, 41]. Here ers. This process could be understood as maladaptive we found that beta frontal-occipital hypersynchro- process of dedifferentiation [37]. The advance of the nization predicts a poor DR performance in healthy neuropathology would be altering the ability of the and pathological aging. In addition, this FC marker brain to focus its activity, reflected as a lack of beta seemed to be more pronounced in MCI patients. A desynchronization. This would be likely to the lack similar finding was reported by Canuet et al. [42] of inhibitory connection shown in the pathological in an independent sample with abnormal CSF p-tau process related to AD [42, 48, 49]. Moreover, higher levels. Furthermore, in a longitudinal study, beta FC synchronization was not associated to better cogni- was negatively associated with working memory and tive performance, discarding the classical interpreta- executive function, and, at the baseline, their levels tion as a compensatory mechanism. It is interesting were the highest in progressive MCI compared to sta- to note here that the occipital lobe has been usually ble MCI, showing a high accuracy (71%) to discrimi- less impaired until the last state of the AD, and the nate between the two groups [35]. A work similar to appearance of the beta cluster in our results could be ours showed that an increase in beta and gamma fre- explained by the effect of amyloid pathology in the quencies above 16 Hz correlated with lower cognitive hippocampus regions that results in a reduction in performance [43]. Finally, the augmentation of beta- neuronal input to occipital areas. This decrease that band FC with age in healthy aging has been described influences the brain organization at the functional in a large study [44]. In addition to these FC patterns, level is defined as a functional diaschisis [50]. many works have shown the inverse relationship of Finally, some limitations of our study should be beta power with working memory over parietal sites addressed in future research. The use of a cross-sec- [45, 46] and with immediate and delayed recall in tional design provides only a snapshot of brain activ- posterior areas [47], supporting the idea of a ham- ity, and the picture can be enhanced by the developing pered beta desynchronization at rest is associated of longitudinal studies aimed to track the individual with these domains. FC changes along the AD continuum. In addition, the When evaluating the relationship between the combination of these techniques with neuropatho- FC markers and brain structural integrity, the MCI logical markers of AD, as amyloid or tau pathology, patients showed a negative association between would improve the clarity of the results since we occipital FC with white and gray matter measures, would be able to distinguish whether these results are reinforcing the relationship between the FC markers specific of AD [51]. described in this study with the progression of the Nevertheless, our results strongly suggest that the dementia. This points out that the worsening of the lack of desynchronization (or hypersynchronization) neuropathology in these patients would be captured in beta band has the potential to be used for assess- by their even higher hypersynchronization when com- ment of pharmacological and non-pharmacological pared to controls. This endorses our view of the FC interventions. Moreover, the results can be transferred values here reported as possible markers that would from the costly MEG to the widely available EEG. help tracking the trajectory a specific patient along Altogether, they have the potential to become a widely the AD continuum. used noninvasive biomarker of the neuropathological Considering all the evidence shown so far, a ten- progression underlying the developing of dementia. tative explanation of the present findings is that the hypersynchronization in beta FC reflects a progres- sive deterioration of neuronal function that advances Conclusion at hand with the evolution of the neuropathological process associated with dementia. In healthy aging, The goal of the present study was to bridge the gap the key factor was the presence of the ε4 allele; between measures of cognitive performance and elec- while the inverse relation between FC and DR was trophysiological markers in healthy and pathological Vol:. (1234567890) 13 GeroScience (2023) 45:477–489 487 aging. The results suggest that beta hypersynchroni- 4. Chen P, Ratcliff G, Belle SH, Cauley JA, DeKosky ST, zation-associated markers could be useful to evalu- Ganguli M. Patterns of cognitive decline in presympto- matic Alzheimer disease: a prospective community study. ate brain health. This contrasts with the typical use of Arch Gen Psychiatry. 2001. https://​doi.​org/​10.​1001/​archp​ RS markers more focused in the assessment of alpha syc.​58.9.​853. band features. Here, our results indicate that the brain 5. Bäckman L, Jones S, Berger AK, Laukka EJ, Small BJ. seems to progressively loss the ability to desynchro- Cognitive impairment in preclinical Alzheimer’s disease: a meta-analysis. Neuropsychology. 2005. https://​doi.​org/​ nize beta FC and that this process would be starting 10.​1037/​0894-​4105.​19.4.​520. to become significant in preclinical stages, at least in 6. Wolk DA, Dickerson BC, Alzheimer’s Disease Neuroim- participants with AD risk factors such as the presence aging Initiative. Apolipoprotein E (APOE) genotype has of the APOE ε4 allele. dissociable effects on memory and attentional-executive network function in Alzheimer’s disease. Proc Natl Acad Sci U S A. 2010; https://​doi.​org/​10.​1073/​pnas.​10014​12107 Funding Open Access Funding provided by Universitat 7. Habeck C, Foster NL, Perneczky R, Kurz A, Alexopoulos Autonoma de Barcelona. P, Koeppe RA, et al. Multivariate and univariate neuro- imaging biomarkers of Alzheimer’s disease. Neuroimage. Declarations 2008. https://​doi.​org/​10.​1016/j.​neuro​image.​2008.​01.​056. 8. Han SD, Arfanakis K, Fleischman DA, Leurgans SE, Ethical Statement All participants were informed about the Tuminello ER, Edmonds EC, et al. Functional connectiv- aims of this study and gave written informed consent. The Insti- ity variations in mild cognitive impairment: associations tutional Review Board Ethics Committee at Hospital Universi- with cognitive function. J Int Neuropsychol Soc JINS. tario San Carlos approved the study protocol, and the procedure 2012. https://​doi.​org/​10.​1017/​S1355​61771​10012​99. was performed following the Helsinki Declaration and National 9. Maestú F, de Haan W, Busche MA, DeFelipe J. Neuronal and European Union regulations. excitation/inhibition imbalance: core element of a transla- tional perspective on Alzheimer pathophysiology. Ageing Conflict of interest The authors declare no competing interests. Res Rev. 2021. https://​doi.​org/​10.​1016/j.​arr.​2021.​101372. 10. Bajo R, Castellanos NP, Cuesta P, Aurtenetxe S, Garcia- Open Access This article is licensed under a Creative Com- Prieto J, Gil-Gregorio P, et al. Differential patterns of con- mons Attribution 4.0 International License, which permits nectivity in progressive mild cognitive impairment. Brain use, sharing, adaptation, distribution and reproduction in any Connect. 2012. https://​doi.​org/​10.​1089/​brain.​2011.​0069. medium or format, as long as you give appropriate credit to the 11. López-Sanz D, Bruña R, Garcés P, Martín-Buro MC, original author(s) and the source, provide a link to the Crea- Walter S, Delgado ML, et al. Functional connectivity dis- tive Commons licence, and indicate if changes were made. The ruption in subjective cognitive decline and mild cognitive images or other third party material in this article are included impairment: a common pattern of alterations. Front Aging in the article’s Creative Commons licence, unless indicated Neurosci. 2017. https://​doi.​org/​10.​3389/​fnagi.​2017.​00109. otherwise in a credit line to the material. If material is not 12. Rosazza C, Minati L. Resting-state brain networks: litera- included in the article’s Creative Commons licence and your ture review and clinical applications. Neurol Sci Off J Ital intended use is not permitted by statutory regulation or exceeds Neurol Soc Ital Soc Clin Neurophysiol. 2011. https://​doi.​ the permitted use, you will need to obtain permission directly org/​10.​1007/​s10072-​011-​0636-y. from the copyright holder. To view a copy of this licence, visit 13. del Río D, Cuesta P, Bajo R, García-Pacios J, López- http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. Higes R, del-Pozo F, et al. Efficiency at rest: magnetoen- cephalographic resting-state connectivity and individual differences in verbal working memory. Int J Psychophys- iol Off J Int Organ Psychophysiol. 2012; https://​doi.​org/​ 10.​1016/j.​ijpsy​cho.​2012.​08.​011 References 14. Pasquini L, Scherr M, Tahmasian M, Meng C, Myers NE, Ortner M, et al. Link between hippocampus’ raised local and eased global intrinsic connectivity in AD. Alzheimers 1. Corey-Bloom J. The ABC of Alzheimer’s disease: cogni- Dement. 2015. https://​doi.​org/​10.​1016/j.​jalz.​2014.​02.​007. tive changes and their management in Alzheimer’s disease 15. Buzsáki G, Draguhn A. Neuronal oscillations in cortical and related dementias. Int Psychogeriatr. 2002. https://​doi.​ networks. Science. 2004. https://​doi.​org/​10.​1126/​scien​ce.​ org/​10.​1017/​s1041​61020​30086​64. 10997​45. 2. Reed BR, Mungas DM, Kramer JH, Ellis W, Vinters HV, 16. Fell J, Axmacher N. The role of phase synchronization in Zarow C, et al. Profiles of neuropsychological impairment memory processes. Nat Rev Neurosci. 2011. https://​doi.​ in autopsy-defined Alzheimer’s disease and cerebrovascular org/​10.​1038/​nrn29​79. disease. Brain. 2007. https://​doi.​org/​10.​1093/​brain/​awl385. 17. Maestú F, Peña JM, Garcés P, González S, Bajo R, Bagic 3. Locascio JJ, Growdon JH, Corkin S. Cognitive test per- A, et al. A multicenter study of the early detection of syn- formance in detecting, staging, and tracking Alzheimer’s aptic dysfunction in Mild Cognitive Impairment using disease. Arch Neurol. 1995. https://​doi.​org/​10.​1001/​archn​ Magnetoencephalography-derived functional connectivity. eur.​1995.​00540​35008​1020. Vol.: (0123456789) 13 488 GeroScience (2023) 45:477–489 NeuroImage Clin. 2015. https://​doi.​org/​10.​1016/j.​nicl.​ 30. Maris E, Oostenveld R. Nonparametric statistical test- 2015.​07.​011. ing of EEG- and MEG-data. J Neurosci Methods. 2007. 18. Bai Y, Hu Y, Wu Y, Zhu Y, He Q, Jiang C, et al. A pro- https://​doi.​org/​10.​1016/j.​jneum​eth.​2007.​03.​024. spective, randomized, single-blinded trial on the effect 31. Lopes da Silva F. EEG and MEG: relevance to neurosci- of early rehabilitation on daily activities and motor func- ence. Neuron. 2013; https://​doi.​org/​10.​1016/j.​neuron.​ tion of patients with hemorrhagic stroke. J Clin Neurosci. 2013.​10.​017 2012. https://​doi.​org/​10.​1016/j.​jocn.​2011.​10.​021. 32. Cuesta P, Garcés P, Castellanos NP, López ME, Aurten- 19. Albert MS, DeKosky ST, Dickson D, Dubois B, Feld- etxe S, Bajo R, et al. Influence of the APOE ε4 allele and man HH, Fox NC, et al. The diagnosis of mild cognitive mild cognitive impairment diagnosis in the disruption of impairment due to Alzheimer’s disease: recommendations the MEG resting state functional connectivity in sources from the National Institute on Aging-Alzheimer’s Asso- space. J Alzheimers Dis JAD. 2015. https://​doi.​org/​10.​ ciation workgroups on diagnostic guidelines for Alzhei- 3233/​JAD-​141872. mer’s disease. Alzheimers Dement J Alzheimers Assoc. 33. Zalesky A, Fornito A, Bullmore ET. Network-based statis- 2011. https://​doi.​org/​10.​1016/j.​jalz.​2011.​03.​008. tic: identifying differences in brain networks. Neuroimage. 20. López ME, Turrero A, Cuesta P, Rodríguez-Rojo IC, 2010. https://​doi.​org/​10.​1016/j.​neuro​image.​2010.​06.​041. Barabash A, Marcos A, et al. A multivariate model of 34. Hayes AF. Introduction to Mediation, Moderation, and time to conversion from mild cognitive impairment to Conditional Process Analysis, Second Edition: A Regres- Alzheimer’s disease. GeroScience. 2020. https://​doi.​org/​ sion-Based Approach. Guilford Publications. 2017;713 10.​1007/​s11357-​020-​00260-7. 35. Pusil S, Eugenia Lopez M, Cuesta P, Bruna R, Pereda E, 21. Jack CR, Bennett DA, Blennow K, Carrillo MC, Feldman Maestu F. Hypersynchronization in mild cognitive impair- HH, Frisoni GB, et al. A/T/N: An unbiased descriptive ment: the “X” model. Brain. 2019. https://​doi.​org/​10.​ classification scheme for Alzheimer disease biomarkers. 1093/​brain/​awz320. Neurology. 2016. https://​doi.​org/​10.​1212/​WNL.​00000​ 36. Garcia-Marin V, Blazquez-Llorca L, Rodriguez JR, 00000​002923. Boluda S, Muntane G, Ferrer I, et al. Diminished periso- 22. Tombaugh TN, McDowell I, Kristjansson B, Hubley AM. matic GABAergic terminals on cortical neurons adjacent Mini-Mental State Examination (MMSE) and the Modi- to amyloid plaques. Front Neuroanat. 2009. https://​doi.​ fied MMSE (3MS): a psychometric comparison and nor- org/​10.​3389/​neuro.​05.​028.​2009. mative data. Psychol Assess. 1996. https://​doi.​org/​10.​ 37. Fornito A, Bullmore ET. Connectomics: A new paradigm 1037/​1040-​3590.8.​1.​48. for understanding brain disease. Eur Neuropsychopharma- 23. Greenberg SA. How to try this: the Geriatric Depression col. 2015. https://​doi.​org/​10.​1016/j.​euron​euro.​2014.​02.​011. Scale: Short Form. AJN Am J Nurs. 2007. https://​doi.​org/​ 38. Chew H, Solomon VA, Fonteh AN. Involvement of lipids 10.​1097/​01.​NAJ.​00002​92204.​52313.​f3. in Alzheimer’s disease pathology and potential therapies. 24. Tulsky DS, Chiaravalloti ND, Palmer BW, Chelune GJ. Front Physiol. 2020. https://​doi.​org/​10.​3389/​fphys.​2020.​ Chapter 3 - The Wechsler Memory Scale, Third Edition: 00598. A New Perspective. In: Tulsky DS, Saklofske DH, Heaton 39. Klados MA, Styliadis C, Frantzidis CA, Paraskevopou- RK, Bornstein R, Ledbetter MF, Chelune GJ, et al., edi- los E, Bamidis PD. Beta-band functional connectivity is tors. Clinical Interpretation of the WAIS-III and WMS- reorganized in mild cognitive impairment after combined III [Internet]. San Diego: Academic Press. 2003;93–139. computerized physical and cognitive training. Front Neu- https://​doi.​org/​10.​1016/​B978-​01270​3570-3/​50007-9 rosci. 2016. https://​doi.​org/​10.​3389/​fnins.​2016.​00055. 25. Chapman LL, White DA, Storandt M. Prose recall in demen- 40. Duke Han S, Buchman AS, Arfanakis K, Fleischman DA, tia. A comparison of delay intervals. Arch Neurol. 1997; Bennett DA. Functional connectivity networks associated https://​doi.​org/​10.​1001/​archn​eur.​1997.​00550​24005​3012 with chronic musculoskeletal pain in old age. Int J Geriatr 26. Dunn JC, Almeida OP, Barclay L, Waterreus A, Flicker L. Psychiatry. 2013. https://​doi.​org/​10.​1002/​gps.​3898. Latent semantic analysis: a new method to measure prose 41. Sperling RA, Aisen PS, Beckett LA, Bennett DA, Craft S, recall. J Clin Exp Neuropsychol. 2002. https://​doi.​org/​10.​ Fagan AM, et al. Toward defining the preclinical stages of 1076/​jcen.​24.1.​26.​965. Alzheimer’s disease: recommendations from the National 27. Fischl B, Salat DH, Busa E, Albert M, Dieterich M, Institute on Aging-Alzheimer’s Association workgroups Haselgrove C, et al. Whole brain segmentation: automated on diagnostic guidelines for Alzheimer’s disease. Alzhei- labeling of neuroanatomical structures in the human brain. mers Dement J Alzheimers Assoc. 2011. https://​doi.​org/​ Neuron. 2002. https://​doi.​org/​10.​1016/​s0896-​6273(02)​ 10.​1016/j.​jalz.​2011.​03.​003. 00569-x. 42. Canuet L, Pusil S, López ME, Bajo R, Pineda-Pardo JÁ, 28. Verdejo-Román J, Björnholm L, Muetzel RL, Torres- Cuesta P, et al. Network disruption and cerebrospinal fluid Espínola FJ, Lieslehto J, Jaddoe V, et al. Maternal amyloid-beta and phospho-tau levels in mild cognitive prepregnancy body mass index and offspring white mat- impairment. J Neurosci Off J Soc Neurosci. 2015. https://​ ter microstructure: results from three birth cohorts. Int J doi.​org/​10.​1523/​JNEUR​OSCI.​0704-​15.​2015. Obes. 2005. https://​doi.​org/​10.​1038/​s41366-​018-​0268-x. 43. Schlee W, Leirer V, Kolassa IT, Weisz N, Elbert T. Age- 29. de Frutos-Lucas J, Cuesta P, Lopez-Sanz D, Peral-Suarez related changes in neural functional connectivity and its A, Cuadrado-Soto E, Ramirez-Torano F, et al. The rela- behavioral relevance. BMC Neurosci. 2012. https://​doi.​ tionship between physical activity, apolipoprotein E epsi- org/​10.​1186/​1471-​2202-​13-​16. lon 4 carriage, and brain health. Alzheimers Res Ther. 44. Vysata O, Kukal J, Prochazka A, Pazdera L, Simko J, 2020. https://​doi.​org/​10.​1186/​s13195-​020-​00608-3. Valis M. Age-related changes in EEG coherence. Neurol Vol:. (1234567890) 13 GeroScience (2023) 45:477–489 489 Neurochir Pol. 2014. https://​doi.​org/​10.​1016/j.​pjnns.​2013.​ 49. López ME, Cuesta P, Garcés P, Castellanos PN, Aurten- 09.​001. etxe S, Bajo R, et al. MEG spectral analysis in subtypes of 45. Borhani S, Zhao X, Kelly MR, Gottschalk KE, Yuan F, mild cognitive impairment. Age Dordr Neth. 2014. https://​ Jicha GA, et al. Gauging working memory capacity from doi.​org/​10.​1007/​s11357-​014-​9624-5. differential resting brain oscillations in older individu- 50. Campo P, Garrido MI, Moran RJ, Maestú F, García- als with a wearable device. Front Aging Neurosci. 2021. Morales I, Gil-Nagel A, et al. Remote effects of hip- https://​doi.​org/​10.​3389/​fnagi.​2021.​625006. pocampal sclerosis on effective connectivity during work- 46. Finnigan S, Robertson IH. Resting EEG theta power cor- ing memory encoding: a case of connectional diaschisis? relates with cognitive performance in healthy older adults. Cereb Cortex. 2012. https://​doi.​org/​10.​1093/​cercor/​bhr201. Psychophysiology. 2011. https://​doi.​org/​10.​1111/j.​1469-​ 51. Ranasinghe KG, Verma P, Cai C, Xie X, Kudo K, Gao X, 8986.​2010.​01173.x. et al. Altered excitatory and inhibitory neuronal subpopu- 47. Fleck JI, Kuti J, Brown J, Mahon JR, Gayda-Chelder C. lation parameters are distinctly associated with tau and Frontal-posterior coherence and cognitive function in older amyloid in Alzheimer’s disease. eLife. 2022; https://​doi.​ adults. Int J Psychophysiol Off J Int Organ Psychophysiol. org/​10.​7554/​eLife.​77850. 2016. https://​doi.​org/​10.​1016/j.​ijpsy​cho.​2016.​07.​501. 48. de Haan W, Mott K, van Straaten ECW, Scheltens P, Stam Publisher’s note Springer Nature remains neutral with regard CJ. Activity dependent degeneration explains hub vulner- to jurisdictional claims in published maps and institutional ability in Alzheimer’s disease. PLoS Comput Biol. 2012. affiliations. https://​doi.​org/​10.​1371/​journ​al.​pcbi.​10025​82. Vol.: (0123456789) 13