Skip to main content
Download PDF

Effects of Alzheimer’s disease plasma marker levels on multilayer centrality in healthy individuals

Alzheimer's Research & Therapy volume 17, Article number: 8 (2025) Cite this article

Abstract

Background

Changes in amyloid beta (Aβ) and phosphorylated tau brain levels are known to affect brain network organization but very little is known about how plasma markers can relate to these measures. We aimed to address the relationship between centrality network changes and two plasma pathology markers: phosphorylated tau at threonine 231 (p-tau231), a proxy for early Aβ change, and neurofilament light chain (Nfl), a marker of axonal degeneration.

Methods

One hundred and four cognitively unimpaired individuals were divided into a high pathology load (33 individuals; HP) group and a low pathology (71 individuals; LP) one. All participants underwent a magnetoencephalography (MEG) recording, a neuropsychological evaluation and plasma sampling. With the MEG recordings, a compound centrality score for each brain source was calculated that considered both intra- and inter-band links. For each group, the relationship between this centrality score and the two plasma markers was studied by means of correlation analyses. Furthermore, the relationship between the centrality score and the plasma markers among the HP and LP groups was compared. Lastly, we investigated whether hubs were more intensely affected by these changes.

Results

Increasing concentrations of p-tau231, which is a proxy of Aβ pathology, were associated with greater theta centrality score of posterior areas that increased their connectedness in the theta range with the remaining areas, regardless of the latter’s frequency range. The opposite relationship was found for left areas that decreased their centrality score in the gamma frequency range. These results only emerged for HP individuals, who showed a significantly different relationship between centrality and p-tau231 compared to LP individuals. Hubs’ centrality score in the theta band was significantly more affected by p-tau231 levels compared to less central regions.

Conclusions

Early brain network reorganizations in cognitively unimpaired individuals are associated with elevated plasma p-tau231, a proxy for very early Aβ changes, only among individuals who show signs of a higher pathology load. Posterior centrality score increases in the theta band are congruent with previous literature and theoretical models, while gamma centrality score losses could be associated with inhibitory neuron dysfunction. Hubs were more intensely affected by p-tau231, and changed to a higher degree, thus corroborating hubs’ vulnerability.

Background

The discovery of Alzheimer’s disease (AD) preclinical phase, marked by neuropathological changes without associated clinical symptoms up to 20 years before dementia onset, highlighted the need for early and non-invasive biomarkers [1].

With the emergence of the first pathological hallmarks, neuronal dysfunction manifests in the form of functional changes [2, 3]. Electrophysiological recordings of individuals at the different stages of the continuum reveal a progressive disruption of functional connectivity (FC), marked by initial increases followed by a state of hypoconnectivity, first observed in posterior regions [4,5,6,7]. More specifically, Kudo et al., [8] found frequency-specific changes along the continuum, marked by increases in low-frequency bands and decreases in higher frequencies, prior to neurodegeneration and cognitive decline. This inverted u-shaped pattern in FC alterations is, at least partly, related to amyloid beta (Aβ) accumulation [9,10,11], generating a positive feedback loop between Aβ and hyperexcitation and hyperconnectivity [11,12,13]. This early relationship between neuronal activity and Aβ might underlie the early dysfunction within the continuum of areas with higher neuronal activity [9].

In the context of network neuroscience, centrality measures represent the nodes’ (in this case, brain regions’) importance for the functioning of the network [14, 15]. Regarding the prodromal and dementia phases, a progressive loss of centrality in posterior areas and an increase in anterior ones, affecting predominantly association areas, in a disease severity dependent fashion has been described [16,17,18,19,20]. In the preclinical phase, however, divergent results emerge, potentially due to the use of different centrality metrics [21, 22]. Studies performed using degree centrality report a positive association with Aβ-positron emission tomography (PET) levels along with an increase in temporo-occipital hubness [11, 23], whereas eigenvector centrality studies report reductions associated with Aβ pathology, affecting mostly posterior regions [14, 24]. Duan et al. [25] and Taguas et al., [26] report centrality alterations, of different signs depending on the measure, in different frequency bands for mild cognitive impairment and AD patients. Seeking a more integrative approach, Yu et al. [20] and Taguas et al., [26] show that combining centrality information from different bands uncovers changes in hubs (i.e. highly central nodes) that don’t appear when addressing single frequency band networks.

PET and cerebrospinal fluid markers (CSF) are poor candidates for early detection and to track disease progression (i.e. high costs, invasiveness and low accessibility). Consequently, emphasis has been placed on plasma biomarkers that have proven reliable as indicators of AD pathology [27]. Indeed, the Alzheimer’s Association Workgroup states that, in the near future, plasma biomarkers will receive regulatory approval to diagnose, on their own AD pathology [28]. High levels of neurofilament light (NfL) chains are indicative of axonal injury and neurodegeneration. Thus, in the context of AD, elevated NfL levels are considered a biomarker of non-specific processes involved in AD pathophysiology [28]. Although this is not a specific biomarker of AD some authors report significant increases early on in the continuum [27]. The recently discovered species of phosphorylated tau at threonine 231 (p-tau231) presents abnormal levels, both in plasma and CSF, that can be detected as early on in the continuum as the preclinical phase. Its levels increase in parallel to and tracking Aβ increases even before PET positivity for this marker has been achieved [29,30,31,32,33]. As a consequence, the Alzheimer’s Association Workgroup categorized it as a Core 1 biomarker, for its reactivity to Aβ species [28]. As such, in this paper p-tau231 has been considered a marker of Aβ.

To our knowledge, this study is the first to address the relationship between NfL and p-tau231 plasma levels, with a multilayer compound centrality metric in cognitively unimpaired individuals. The compound centrality score is based on functional connectivity measures from magnetoencephalography (MEG) recordings and the rationale behind it is twofold. Firstly, a multilayer approach will provide an integrative perspective of the network properties, while the development of a compound score will contribute to bridge the above-mentioned discrepancies found in the literature stemming from the use of different centrality measures. We have three main hypotheses to test with our analysis: First of all, we hypothesize that plasma markers will affect centrality of the brain by showing significant increases in centrality for slower frequency bands and decreases for high frequency bands, as reported by different studies for later stages. Second, we hypothesize that hubs will be more severely impaired; that is, regions with higher levels of activity and relevance will become more affected by early neuropathological changes, and thus, will show larger associations to plasma markers levels. Lastly, considering that plasma markers are very early indicators of neuropathology, the lack of established thresholds and current inability to diagnose based on plasma biomarkers, and that our sample consists of unimpaired individuals, we hypothesize that significant alterations in network centrality will be observable only above a certain threshold of plasma markers, as lower values will be most likely meaningless from a clinical perspective.

Methods

Participants

This study was carried out in a sample of 104 participants with available plasma markers and a valid MEG recording. Volunteers for this study are part of a larger project (“Study of the anatomo-functional connectome of AD-relatives: an early intervention on cognition and lifestyles”) longitudinally following participants’ evolution over time in subsequent waves. Participants for this initiative are unimpaired adults recruited from local hospitals, via advertisements in the Fulbright alumni association, in the “Asociación Española de Ingenieros de Telecomunicación Delegación de Madrid,” as well as in public media. This project aims to study the earliest electrophysiological signs of Alzheimer’s disease pathology by combining different neuroimaging, neuropsychological and plasma measurements for each participant at different stages. The data included in this work belongs to the first follow-up evaluation for which plasma markers (NfL and p-tau231) were acquired for the first time. We divided the sample into individuals with high levels of pathology (HP) and individuals with a low pathological (LP) load. Given the lack of an established cut-off for either plasma measure, we established the cut-off at the 50th percentile of both distributions. Individuals considered within the HP group were those who exhibited a value over the median of both NfL and p-tau231 distributions. All participants underwent a thorough neuropsychological assessment to ensure a healthy cognitive status in the different domains, which included as a screening test the Montreal Cognitive Assessment [34], the trail making test [35], the word list test and digit backward span subscales from the Wechsler Adult Intelligence Scale III [36], and the Rey-Osterrieth complex figure [37]. In order to consider participants as “cognitively unimpaired” their corrected performance in all tests was required to be within 1.5 standard deviations from the average of their age and formal education groups. Table 1 summarizes relevant demographic and clinical information of the sample.

Table 1 Sample demographic information. Values are presented as Mean ± Standard Deviation. p-tau231 values are presented in pg/mL and NfL values in ng/mL. MoCA: Montreal cognitive assessment; TMT: trail making test, versions A and B; Digits-b: digits backward; WLT: word list test, (t) total immediate recall and (d) delayed recall; Rey: Rey-Osterrieth complex figure, (i) immediate and (d) delayed recall
Full size table

Each participant had an available magnetic resonance imaging (MRI) scan acquired during the first assessment. During the second assessment, all participants underwent an MEG scan and a blood sample extraction for plasma markers determination. The average time span between the baseline MRI and follow-up MEG scan was 2 years and 8 months (std = 10 months). Moreover, the time span for the HP (mean = 2 years and 3 months; std = 11 months) and LP (mean = 2 years and 6 months; std = 8 months) group was not significantly different (t = -1.619; p = 0.108). Additionally, both the neuropsychological evaluation and the blood extraction were all performed within the two weeks after the follow-up MEG. Moreover, for the blood extraction, participants were instructed to attend between 8 am and 12 pm and fasting since their previous dinner.

Exclusion criteria for the current study included: (1) history of psychiatric or neurological disorders or drug consumption in the last week that could affect MEG activity; (2) family history of dementia other than AD; (3) evidence of infection, infarction, or focal lesions in a T2-weighted MRI scan; (4) alcoholism or chronic use of anxiolytics, neuroleptics, narcotics, anticonvulsants, or sedative-hypnotics; (5) cognitive impairment and (6) unusable MEG recording or T1-weighted image.

The “Hospital Clínico San Carlos” Ethics Committee approved this study, and the procedure was performed following internationally accepted guidelines and regulations.

Plasma sample determinations

For plasma separation, samples were centrifuged for 10 min at 2000 g at + 4 °C. The resulting plasma aliquots were stored in a -81ºC freezer, until samples from all participants were collected. Plasma storage was done within the first three hours after extraction. Plasma p-tau231 and NfL concentrations were quantified by competitive enzyme-linked immunosorbent assay, using commercial kits (MyBiosource, Inc., USA, MBS724296, and human NF-L Elabscience, respectively), according to the manufacturers’ instructions. Tests were performed in duplicate, and an automated microplate reader (Biochrom ASYS UVM 340, Cambridge, UK) measured the optical density at 450 nm with Mikrowin 2000 software (Berthold Technologies, Germany).

MEG recordings and preprocessing

Four minutes of ongoing brain activity during resting-state under eyes closed condition were acquired from each participant at the Center for Biomedical Technology (Madrid, Spain) using a 306-sensor Vectorview MEG system (Elekta AB, Stockholm, Sweden). Continuous head position information during the recording was registered through four coils placed over the forehead bilaterally and both mastoids of each participant. To this aim, coil position and head shape information were digitized using a Fastrack Polhemus system (Polhemus, Colchester, VT, USA). Electrooculographic and electrocardiographic activity were registered by placing two sets of bipolar electrodes. The recordings were performed inside a magnetically shielded room, and participants were instructed to stay as still as possible and relax. Data was acquired at a sampling rate of 1000 Hz and filtered online between 0.1 and 330 Hz. Before preprocessing of the signal, the contribution of head movements and distant sources outside the brain were removed from the signal by using the spatiotemporal expansion of the signal space separation method [38], implemented by Neuromag Software (MaxFilter version 2.2, correlation 0.90, time window 10 s).

After the spatiotemporal expansion of the signal space separation, MEG signal was preprocessed using Fieldtrip to detect muscular, ocular and jump artifacts [39]. Ocular and cardiac components were removed from the signal by using an independent component analysis based algorithm. After data segmentation, only subjects with at least 20 valid segments of 4 consecutive seconds of artifact-free activity were kept for further analyses. Given the elevated redundancy of the data after applying the signal space separation method [40], only data from magnetometers were considered for subsequent analyses.

Source reconstruction (individual MRI)

Individual T1 structural brain images were used to reconstruct source space MEG signals. To do so, MRI was acquired in a General Electric 1.5 T system with a high-resolution antenna together with a homogenization phased array uniformity enhancement filter (fast spoiled gradient echo sequence, TR/TE/TI = 11.2/4.2/450 ms; ageflip angle 12º; slice thickness 1 mm, 256 × 256 matrix and FOV 25 cm). An expert radiologist carefully reviewed and inspected each image obtained to ensure image quality and verify the absence of any relevant structural alterations.

The source model was formed by placing a regular grid of 4560 sources spaced 1 cm apart forming a cube. Only sources corresponding to cortical regions according to the Anatomical Atlas Labelling [41] were reconstructed and employed for further analyses, resulting in a total of 1210 brain sources. Subsequently, this source model was linearly transformed into each participant’s individual T1-weighted image. Single-shell head models were created using SPM12 brain segmentation, which were visually inspected to ensure the quality of the segmentation [42]. By combining head model and source model information, the lead field matrix was calculated using a modified spherical solution.

Sensor space signals were filtered between 2–45 Hz using a 450th-order finite input response filter designed with a Hann window. To avoid phase distortion, a two-pass filtering approach was used. In addition, to mitigate edge effects, 2000 samples of real data padding were added to each side of the signal. Time series of each source location were obtained using Linearly Constrained Minimum Variance beamformer as an inverse model using Fieldtrip functions [39] running in Matlab (Matlab 2019b). After source reconstruction, broadband signals were band-pass filtered into the five classical frequency bands, namely: delta (2–4 Hz), theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz) and gamma (30–45 Hz) for all subsequent analyses.

Functional connectivity and graph construction

Amplitude envelope correlation with leakage correction was employed to estimate functional connectivity between each pair of the 1210 brain sources, using the absolute value of the Pearson correlation between time-series envelopes as previously described [43]. Typically, FC approaches in the literature have only considered connections between nodes in isolated frequency bands. Consequently, functional coupling is typically only estimated between nodes’ activities in the same frequency band, overlooking the relevance of cross-frequency couplings for various brain dynamics [44]. In our approach amplitude envelope correlation was estimated for a given source so that its activity in each frequency band was correlated with that of every other source in all the frequency bands. This approach allows us to take into consideration not only the classical intra-band links (i.e. connection between two nodes in the same frequency band, e.g. source 1 delta – source 2 delta coupling), but also inter-band links (i.e. connection between two nodes in different frequency rhythms, e.g. source 1 delta – source 2 alpha coupling). By doing this, we obtained a set of 1210 (sources) by 5 (bands) connections for each node in each band, resulting in a large and comprehensive connectivity matrix for each participant with 6050 (1210 nodes by 5 bands) by 6050 (1210 nodes by 5 bands). These FC matrices were employed as the input for the graph analyses conducted in the present work. For an in-depth description of the method, see Taguas et al., [26].

A graph (G) is typically expressed as a set of nodes, denoted by the matrix (V); and the connections between those nodes, also called edges (E); G = (V,E). The connections between nodes are commonly expressed in a weight matrix (W). In this matrix, the weight (w) of the connection between the node i and the node j is captured by the element wij of the graph and represents the distance between two nodes in the graph.

In the context of a brain network study, the different nodes represent the regions of the brain and edges are usually expressed by a coupling or connection metric that can either be functional or structural. In our study, each of the 1210 sources was considered a node of the network. Furthermore, each frequency band was used as a different layer in our multilayer network, thus resulting in a total set of 6050 nodes (1210 nodes by 5 layers). A common problem in brain studies using network theory is the computational complexity of weighted networks which has typically led to the use of binarized networks. In this approach, an arbitrary threshold is employed to discard every connection below that value, maintaining only those above it. However, this methodology is not devoid of limitations, since the arbitrary nature of the threshold heavily influences the results obtained from the graph [45], and a lack of clear consensus for how to establish an appropriate threshold hinders this approach. In our work we opted to use weighted (not thresholded) undirected graphs in which the FC matrices previously described were employed as the weight of each edge in our graph.

Centrality calculation

Centrality has been typically studied using different metrics and indicators, each of them capturing slightly different aspects of weight distribution in brain communication. In our work we decided to take an integrative approach by combining three different network centrality metrics in a single measure that was employed for all the analyses. All the metrics were calculated for each network node of each subject using the NetworkX library (version 2.6.3) and SciPy library (version 1.7.1) in Python (version 3.10.4).

In particular, we used three graph measures as centrality indicators: node strength, eigenvector centrality and betweenness centrality. The strength of a node is determined by the number of connections a node has with other nodes of the network and the weight of its edges. It is calculated as the sum of the weights of the links incident to node \(i\), where \({w}_{ij}\) is the weight of the link between nodes \(i\) and \(j\).

$${s}_{i}={\sum }_{j\ne i}{w}_{ij}$$
(1)

The eigenvector centrality considers not only the number and weight of connections of a node, but also the relevance of the nodes they are connecting to. In essence, a node is considered more central if it is connected to nodes that are also highly central themselves. It is calculated following Eq. 2, in which \(M\) is the adjacency matrix, \(\lambda\) the eigenvalues and \(v\) the eigenvectors. The element \(i\) of the eigenvector \(v\) that is associated with the largest eigenvalue \({\lambda }_{1}\) holds the value for the eigenvector centrality of node \(i\).:

$$Mv=\lambda v$$
(2)

Finally, the betweenness centrality of a node estimates its ability to act as a bridge or intermediary between other nodes, by determining the proportion of shortest paths in the brain that runs through a given node. Equation 3 describes betweenness centrality, where \(V\) is the set of nodes, \(\sigma (j,k)\) is the number of shortest paths, and \(\sigma (j,k|i)\) is the number of those paths passing through some node \(i\) other than \(j,k\). As betweenness centrality takes link weights as a distance measure, we used the inverse of the weights (which, in our case, indicate functional proximity).

$${B}_{i}={\sum }_{j, k\in V}\frac{\sigma (j,k|i)}{\sigma (j,k)}$$
(3)

To reduce computation time, betweenness centrality was estimated using the algorithm developed by Brandes & Pich [46] with the value of k set to 250.

This resulted in three vectors per subject, each of them with a length of 6050 (composed by 1210 values, one for each source, in each of the five frequency bands). To obtain a comprehensive measure of centrality, we combined the three metrics into one, hereafter referred to as centrality score. To do so, we first z-scored each of the metrics independently, considering their highly different numeric ranges. Afterwards, these vectors were averaged for each subject, obtaining a vector of the same length for each subject containing the centrality score employed in all the subsequent analyses.

Statistical analysis

Statistical analyses were performed using Matlab 2019b and carried out in a stepwise manner for the HP and LP groups independently. For each group, we first evaluated whether changes in brain network centrality were associated with the p-tau231 and NfL levels. We conducted a Spearman correlation analysis, including age as a covariate, for each individual source in the brain, studying the statistical association between each plasma marker level (NfL and p-tau231) with the centrality score of each individual brain source. This analysis was conducted for each frequency band. To account for multiple comparisons, we conducted two-sided cluster-based permutation tests, using a Montecarlo approach [47]. Significant correlations were grouped based on their spatial contiguity (sources touching each other are considered as contiguous), obtaining a cluster size defined as the sum of the individual statistics of each source in it. We conducted 10.000 permutations on the original data to create a surrogate distribution of random cluster sizes to compare our original cluster size to the surrogate size distribution. An alpha level of 0.05 was used to deem the original cluster as significant. For the emerging significant clusters, an additional Spearman correlation analysis was performed, using sex as a covariate, to ensure that the results remained significant.

The next step of our analysis was to evaluate whether HP and LP individuals’ relationship between pathology markers and centrality values were different. For this, we used a Chow test [48], that aims to determine whether the coefficients of two regression models are equal; that is, whether one regression model or two separate ones fit the dataset best. Equation 4 shows the formula to calculate the F statistic for this test, where RSSp refers to the residual sum of squares from the regression that considers both subgroups pooled in a single model. RSS1 and RSS2 represent the residual sum of squares for the model performed on each subsample; k is the number of parameters in the model and N1 and N2 the number of observations in each subsample.

$$F =\frac{(RSSp-(RSS1+RSS2))/k}{(RSS1+RSS2)/(N1+N2-2k)}$$
(4)

Lastly, we aimed to study whether the most important regions of the network (i.e. the hubs) were more intensely affected by the plasma pathology. Since we hypothesized that plasma marker levels would affect centrality of certain regions, our intention was to test our hypothesis that more central regions would be more intensely affected by biomarker levels. To do so, firstly we calculated the grand-average centrality of each source, as a measure of its hubness. This was done by simply averaging the centrality scores of each source considering all frequency bands for each subject. This value was then averaged again across subjects to obtain a single centrality score value for each source for the whole population, thus resulting in a 1210 by 1 vector. These values reflect the specific relevance of each node for maintaining brain communication in the population. Thus, a higher value would indicate that the specific source has a high centrality in the whole sample and thus can be considered a hub, while a lower value would indicate the opposite. Finally, we studied the statistical association between the rho value obtained for each source in the first analysis step, (i.e. how intensely associated are its centrality and the plasma marker through our whole sample) with the subsequently calculated grand-average of that source (i.e. a measurement of its total hubness). This was done by calculating the correlation between the two mentioned vectors; namely, grand-average centrality (1210 values) and rho-value obtained in step one of the statistical analyses (1210 values). Positive associations in this second analysis would confirm our hypothesis, since they would indicate that hubs (regions with larger grand-average centrality) would be more intensely affected by plasma marker alterations (i.e., would show larger correlations, -rho values- between centrality and plasma level). This analysis was only carried out if we found a significant cluster in step one of our analyses. Thus, only if the specific plasma marker (NfL or p-tau231) is associated with centrality in a specific frequency band, we would proceed to step two of our analysis.

Results

Association between plasma markers and centrality score

Firstly, NfL levels’ association with source centrality was studied by means of correlation analyses. As a result of said analyses no clusters emerged for any group in any of the bands tested. Henceforth, we did not find any significant association between centrality score and NfL levels.

Regarding p-tau231 levels and its association with centrality score, five clusters emerged for the HP group, of which only two were significant and retained significance after multiple comparisons correction. In the theta band, we observed a significant cluster (rhosum = 81.7; pcluster = 0.013) in which p-tau231 plasma levels and the centrality score were positively correlated (Fig. 1a). This cluster included mostly bilateral regions covering posterior aspects of the brain, which presented a positive correlation (rhoaverage = 0.587; p = 4.196*10–4) between their theta centrality score and p-tau231 (Fig. 1c). This correlation remained unchanged after controlling for the effect of sex. These included mainly areas over parietal, occipital and posterior temporal regions such as the angular gyrus, inferior parietal lobe, inferior and middle occipital cortices, fusiform gyri and left precuneus among others. Centrality score in the gamma band was also significantly associated with p-tau231 levels (rhosum = -66.08; pcluster = 0.018) although in a negative direction (Fig. 1b). Negative correlations between these two metrics were mainly observed over left hemispheric regions including a great proportion of the left temporal lobe (both medial and lateral structures) and extending into ventral and posterior aspects of the frontal and parietal lobe (such as the inferior frontal and posterior frontal gyri). These regions presented a negative correlation (rhoaverage = -0.621, p = 1.445*10–4) between their gamma centrality score and p-tau231 (Fig. 1d), which remained unchanged after controlling for sex. Additionally, three non-significant clusters emerged for the HP group, a positive cluster in delta (rhosum = 5.063, pcluster = 0.532) and two negative ones, in alpha (rhosum = -22.812; pcluster = 0.101) and beta (rhosum = -6.302; pcluster = 0.504). For the LP group only a non-significant positive cluster in beta emerged (rhosum = 4.693; pcluster = 0.450).

Fig. 1
figure 1

Regions showing significant correlation between the centrality score and p-tau231, and the relationship between the clusters’ average centrality score (CS) and p-tau231 in the high pathology (HP) group. In the upper panel: (a) regions that show a significant correlation with the theta band centrality in the HP group and (b) regions that show a significant correlation with the gamma band centrality in the HP group. Red colors denote positive correlation coefficients while blue colors indicate negative correlation coefficients. The centrality score was calculated from MEG resting state recordings, combining classical centrality metrics. In the lower panel: scatterplots showing the relationship in HP individuals (purple) and low pathology (LP) individuals (yellow) between p-tau231 and the (c) average cluster centrality score in the theta band, and (d) the average cluster centrality score in the gamma frequency band. The upper panel images were obtained using MRIcroGL (version 1.2.20220720) with data obtained from Matlab2019b. The scatterplots were also created with Matlab2019b

Full size image

After corroborating that p-tau231 levels were only significantly associated in the HP group and not in the LP group we wanted to verify whether there was a structural change in the relationship between p-tau231 and centrality below and above a certain threshold of the marker. To this aim, two chow tests were conducted (one for each significant cluster) to assess if the model fits significantly better taking into consideration the marker threshold. Both tests revealed a significant structural change in the data relationship below and above the threshold: (Theta cluster: F2,100 = 3.46, p = 0.035; Gamma cluster: F2,100 = 3.23, p = 0.044). These significant results indicate that model fit between p-tau231 level and cluster mean centrality is significantly better when adjusting the two subgroups separately compared to the whole sample, meaning that the relationship between the two variables is significantly different in both groups.

Correlation analyses were performed between neuropsychological measures and the clusters’ centrality score, as well as plasma pathology markers. No significant association was found (data not presented).

The association between plasma markers and centrality score is affected by the hubness of the region

We conducted the analysis to study how the mean relevance of each brain region in the network (i.e. its grand-average centrality score) influenced the relationship between p-tau231 levels and theta/gamma centrality score for each source. To this aim we conducted two correlations, as previously explained, one for each band showing a significant association between the centrality score and the plasma marker.

Results for the theta band showed a highly significant association between the mean relevance of each brain region, and how their centrality changes in association with p-tau231 levels (rho = 0.499; p = 9.2 × 10–77). This positive association indicates that the more relevant a node is in the network (thus, the more it acts as a hub), the larger is the association between its centrality in the theta band and p-tau231 level. This relationship can be further observed in Fig. 2, where it can be appreciated how the association (rho values) between p-tau231 levels and theta band centrality is larger in those sources with a higher grand average centrality in the sample. Thus, increasing linearly with the general relevance of the node in the network.

Fig. 2
figure 2

Influence of the total relevance of each node within the network (i.e. grand-average centrality; GA) on the correlation between centrality score and p-tau231 levels. a Probability distribution for the correlation coefficients obtained from the correlation analysis between p-tau231 and the centrality score (i.e. measure combining classical centrality measures) of increasingly relevant nodes (i.e. increasingly greater grand-average) in the network, divided in increasing 10 percentile steps. Gray vertical lines indicate mean correlation coefficient (ρ-val) within each grand average percentile step. b Scatter plot of the correlation coefficient (ρ-val) of all nodes as a function of their grand-average centrality. Both the ridgeplot and the scatterplot were created using Matlab2019b

Full size image

In the gamma band, we did not observe a significant association between these two variables (rho = 0.006; p = 0.81).

Discussion

This pioneer study is, to the best of our knowledge, the first effort to address the relationship between network centrality alterations and the plasma pathology markers p-tau231, a marker of Aβ pathology, and NfL indicative of neurodegeneration, assessing how the combined elevated or low levels of both within our sample differently relate to centrality. It is also the first one to assess multilayer centrality recorded with MEG in an unimpaired population. Our results reveal a twofold relationship between p-tau231 concentrations and the functional network properties among individuals with higher levels of plasma AD biomarkers. With increasing concentrations of p-tau231, posterior areas hold greater relevance within the network through their theta oscillations. In contrast, left hemispheric regions show the opposite pattern, exhibiting a lower level of importance within the network through their activity in the gamma frequency band. Furthermore, the relationship between these regions’ centrality and p-tau231 is statistically different from the relationships exhibited by individuals with a low pathology load. Lastly, a subsequent analysis revealed that the most relevant areas for the network are also the ones that present the greatest association between their integrative centrality score and p-tau231.

In recent years, plasma p-tau231 has proven highly relevant for the study of the preclinical stage of AD given its early increases that parallel those of Aβ species. Indeed, it shows elevated levels even before Aβ-PET positivity is achieved, is a good predictor of elevated Aβ and is discriminant of the different stages of the continuum [29,30,31,32,33]. As such, it has been proposed as a Core 1 biomarker of Aβ pathology in the 2024 criteria revision proposed by the Alzheimer’s Association Workgroup [28]. Consequently, in this study, p-tau231 has been considered a marker of Aβ pathology. Our results show heightened and reduced centrality in the theta and gamma frequency bands in the HP group, respectively, in relation to plasma p-tau231 levels, as initially hypothesized based on the reports on connectivity trajectories across frequencies in the earliest stages of the AD continuum [8]. Regarding the theta band, posterior areas that include the bilateral inferior parietal lobe, angular gyri, middle temporal gyri and inferior and middle occipital regions, become more relevant within the network by increasing their connectedness within the 4-8 Hz range with the remaining areas, regardless of the latter’s frequency range, as p-tau231 levels increase. This is congruent with previous literature, where alterations in the theta frequency band have been extensively described. Across the continuum, progressive increases in relative theta power [49, 50], along with enhanced synchrony in this band have been consistently observed [5, 8, 51,52,53]. Furthermore, Hatz et al., [52] found the theta connectivity between central and parieto-occipital regions to be the most discriminant characteristic among healthy controls, mild cognitive impairment, and AD patients. Finally, these posterior regions have been extensively studied in the context of AD, as they closely overlap with regions Aβ pathology accumulation, and display centrality alterations across the continuum [14, 16, 20, 25, 54]. Furthermore, these posterior regions with increased theta centrality mirror the multilayer results reported by Taguas et al. [26] in higher-frequency bands among MCI patients, who also show increased centrality in low frequency bands in anterior regions. Lastly, the average centrality score in the theta frequency band in these regions from HP and LP individuals was better explained by two regression models, indicating that the positive relationship between theta centrality and p-tau231 only appears after a certain amount of AD pathology has been acquired. Consequently, our results expand previous findings, demonstrating that not only local and long-range synchronization are affected, but network structure is reconfigured this early in the continuum in association with novel AD plasma markers in a disease specific fashion, affecting the expected regions and frequency bands. This network reconfiguration is highly relevant and should be further studied as it could underlie the functional and cognitive changes observed in the AD continuum, providing a deeper understanding of the pathology.

On the other hand, network centrality from regions in the left hemisphere, with a predominance for the frontal and temporal cortex, exhibit a negative correlation with p-tau231 levels among HP individuals. The relevance of these areas within the network diminishes as gamma activity participation decreases in the overall network communication structure in relation to p-tau231 concentrations. Again, the negative relationship between gamma’s centrality score in these regions and p-tau231 among HP individuals differs from that of LP individuals, suggesting once more that this relationship only emerges after a given level of AD pathology is achieved. Gamma activity has been extensively linked to parvalbumin inhibitory neurons, whose alterations suppress gamma activity [55,56,57]. On a related note, Verret et al., [57] demonstrated increases in hyperconnectivity associated with gamma activity reductions in an experimental model of AD, Cuesta et al., [58] reported decreased gamma connectivity in association to epileptogenic activity, and various works yield promising findings regarding the use of invasive and non-invasive techniques to ameliorate of AD pathogenesis and symptomatology by promoting neural activity in this frequency band [59, 60]. Consequently, the diminished relevance of these areas within the network in the gamma frequency band could be a reflection of ongoing Aβ pathology and neuronal hyperactivity, commonly observed at the beginning of the disease [10, 11].

The lack of association between NfL, as opposed to p-tau231, with the centrality score might be due to different factors. Firstly, it might be that these individuals are still in a very early stage of the continuum for NfL levels to be discriminant. In their respective studies, Mattsson et al., [61] and Nyberg et al., [62] found significant differences in NfL levels in individuals further along the continuum, but failed to find them in preclinical individuals, and argue that this could be due to the minimal axonal injury found during this stage, the disease-inespecificity of this axonal-damage marker and/or the increases in this marker associated with age. Although some studies report significant differences in NfL levels in the preclinical stage, their sample size is very extensive [27]. Hence, the reduced number of subjects in this study might represent a limiting factor.

Subsequently, to study whether hubs are more vulnerable, we calculated a grand-average measure for the HP individuals. It considers the relevance of each node in all frequency bands as the average across subjects to determine the importance of every node in the functional network. Our results are in agreement with previous studies, suggesting that those areas showing a more hub-like behavior are also more intensely affected by p-tau231 early on in the continuum. These regions showed the highest association between the increased centrality score and p-tau231 plasma levels and include areas such as the bilateral inferior parietal gyrus, temporal regions, and occipital areas, which have been defined as hubs, as well as vulnerable to AD pathology, both in network as well as in classic FC studies [4, 11, 20, 63].

On the other hand, while we describe that the most relevant areas are the ones that exhibit a larger increase in centrality associated to the p-tau231 levels, previous literature in later stages of the continuum have described decreases in centrality in these same regions [4, 17, 20]. This pattern of an initial increase that gives way to a decrease in centrality can be understood through the theoretical model proposed by Stam [18], which explains why hubs exhibit an enhanced vulnerability in diseases like AD following this inverted U-shape centrality pattern. Accordingly, in the early stages, subordinate nodes in the hierarchy start sending information directly to hubs. This reorganization increases their centrality but, if maintained, produces hub overload, which makes them lose centrality in favor of other areas. This pattern of posterior hub disruption accompanied by an anteriorization of the most central areas has been described in later stages of the continuum [4, 17, 20]. Moreover, it is congruent with computational models that demonstrate that hub vulnerability can be activity-dependent; excessive neural firing, which has been demonstrated in the early stages of AD, is associated with Aβ increases and begins in some of these regions [11, 19, 64]. This neural hyperexcitability induces synaptic dysfunction that would potentially translate into increased connectivity and centrality, that eventually give way to a reduction in connectivity, more random networks, and selective damage to hubs [18, 64].

Given the aforementioned treatment of p-tau231 as an Aβ marker, the relationship between centrality and p-tau231 among HP individuals, that is maximum in hubs in the theta frequency band, might indicate that these regions exhibit incipient pathological changes on a molecular level that underlie the ones observed on a functional level, thus suggesting that these individuals are entering the AD continuum.

The importance of performing multilayer analyses to yield a more comprehensive understanding of brain function has already been discussed in the literature [20]. Likewise, the use of an integrative centrality score allows the study of complementary aspects of the individual and classically studied measures, which, on their own, have been known to yield conflicting results [21, 22].

This is an unprecedented study aimed at providing an integrative understanding of brain network patterns in cognitively unimpaired individuals. However, this study is not without shortcomings. As previously mentioned, the sample size could represent a limiting factor, preventing the finding of a relevant relationship between NfL levels and brain dynamics. Remarkably, this study is part of an initiative aimed at longitudinally following these individuals, with the next time-point assessment already approved and plans for cohort enrichment. Additionally, it should be noted that the current categorization of our sample into the HP and LP groups based on the concentrations of both, p-tau231 and NfL, are not reflective and should not be interpreted as a PET-based or CSF-based distinction. Instead, these categories only refer to the distribution of subjects’ concentrations within our sample. In this line, future assessment rounds should include, in addition to plasma biomarkers, other well-established techniques to quantify AD pathology to allow for a better classification of individuals. Nevertheless, plasma biomarkers might represent a more ecological, less invasive, more cost-efficient, and widely accessible alternative. In this line, the following assessment round has been approved and financed and will begin in the following months. It will include plasma determinations, including p-tau231 and other Core 1 markers to compare their performance in relation with electrophysiological parameters. In this follow-up, an attempt should be made to use an even shorter measurement timespan than the already brief two-week period used in this study, to minimize potential confounding variables and preserve the integrity of the relationship between MEG, plasma and neuropsychological measures. Lastly, longer recordings (a minimum of 10 min and even longer) would be advisable for future projects to maximise intra and inter-subject reproducibility, as recommended by Liuzzi et al., [65].

Conclusions

Overall, our study takes a pioneering approach to study functional networks in unimpaired individuals with varying concentrations of pathology markers, being the first to combine the use of MEG data, plasma biomarkers and cognitively unimpaired individuals, with novel and integrative multilayer network analyses. This allowed us to overcome the oversimplification of single-frequency band studies and combine the properties of various centrality measures to obtain a more comprehensive understanding of brain connectivity patterns. Our results are consistent and expand previous literature, showing early alterations in network communication associated with elevated plasma biomarkers in cognitively unimpaired individuals, that emerge after sufficient AD pathology has been acquired. These alterations are observed particularly in hubs, which are vulnerable areas to AD pathology.

Data availability

The data and the algorithms that support the findings of this study are available from the corresponding author, upon reasonable request.

Abbreviations

Aβ:

Amyloid beta

AD:

Alzheimer’s disease

CS:

Centrality score

CSF:

Cerebrospinal fluid

FC:

Functional connectivity

GA:

Grand-average

HP:

High pathology

LP:

Low pathology

MEG:

Magnetoencephalography

MRI:

Magnetic resonance imaging

NfL:

Neurofilament light chain

PET:

Positron emission tomography

p-tau231:

Hyperphosphorylated tau 231

References

  1. 2023 Alzheimer's disease facts and figures. Alzheimers Dement. 2023;19(4):1598–695. https://doiorg.publicaciones.saludcastillayleon.es/10.1002/alz.13016.

  2. Busche MA, Chen X, Henning HA, Reichwald J, Staufenbiel M, Sakmann B, et al. Critical role of soluble amyloid-β for early hippocampal hyperactivity in a mouse model of Alzheimer’s disease. Proc Natl Acad Sci. 2012;109(22):8740–5.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  3. Dubois B, Hampel H, Feldman HH, Scheltens P, Aisen P, Andrieu S, et al. Preclinical Alzheimer’s disease: definition, natural history, and diagnostic criteria. Alzheimers Dement. 2016;12(3):292–323.

    Article  PubMed  Google Scholar 

  4. López-Sanz D, Bruña R, Garcés P, Martín-Buro MC, Walter S, Delgado ML, et al. Functional connectivity disruption in subjective cognitive decline and mild cognitive impairment: a common pattern of alterations. Front Aging Neurosci. 2017;9. Available from: https://www.frontiersin.org/articles/10.3389/fnagi.2017.00109. Cited 2023 Jan 25.

  5. Nakamura A, Cuesta P, Kato T, Arahata Y, Iwata K, Yamagishi M, et al. Early functional network alterations in asymptomatic elders at risk for Alzheimer’s disease. Sci Rep. 2017;7(1):6517.

    Article  PubMed  PubMed Central  Google Scholar 

  6. Pusil S, López ME, Cuesta P, Bruña R, Pereda E, Maestú F. Hypersynchronization in mild cognitive impairment: the ‘X’ model. Brain. 2019;142(12):3936–50.

    Article  PubMed  Google Scholar 

  7. Ramírez-Toraño F, Bruña R, de Frutos-Lucas J, Rodríguez-Rojo IC, de Marcos Pedro S, Delgado-Losada ML, et al. Functional connectivity hypersynchronization in relatives of alzheimer’s disease patients: an early E/I balance Dysfunction? Cereb Cortex. 2021;31(2):1201–10.

    Article  PubMed  Google Scholar 

  8. Kudo K, Ranasinghe KG, Morise H, Syed F, Sekihara K, Rankin KP, et al. Neurophysiological trajectories in Alzheimer’s disease progression. 2024. Available from: https://elifesciences.org/reviewed-preprints/91044v2. Cited 2024 Mar 27.

  9. Palmqvist S, Schöll M, Strandberg O, Mattsson N, Stomrud E, Zetterberg H, et al. Earliest accumulation of β-amyloid occurs within the default-mode network and concurrently affects brain connectivity. Nat Commun. 2017;8(1):1214.

    Article  PubMed  PubMed Central  Google Scholar 

  10. Quevenco FC, van Bergen JM, Treyer V, Studer ST, Kagerer SM, Meyer R, et al. Functional Brain Network Connectivity Patterns Associated With Normal Cognition at Old-Age, Local β-amyloid, Tau, and APOE4. Front Aging Neurosci. 2020;12. Available from: https://www.frontiersin.org/articles/10.3389/fnagi.2020.00046. Cited 2023 Jan 25.

  11. Sepulcre J, Sabuncu MR, Li Q, El Fakhri G, Sperling R, Johnson KA. Tau and amyloid β proteins distinctively associate to functional network changes in the aging brain. Alzheimers Dement. 2017;13(11):1261–9.

    Article  PubMed  Google Scholar 

  12. Canter RG, Penney J, Tsai LH. The road to restoring neural circuits for the treatment of Alzheimer’s disease. Nature. 2016;539(7628):187–96.

    Article  PubMed  Google Scholar 

  13. Schultz AP, Chhatwal JP, Hedden T, Mormino EC, Hanseeuw BJ, Sepulcre J, et al. Phases of hyperconnectivity and hypoconnectivity in the default mode and salience networks track with amyloid and tau in clinically normal individuals. J Neurosci. 2017;37(16):4323–31.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  14. Lorenzini L, Ingala S, Collij LE, Wottschel V, Haller S, Blennow K, et al. Eigenvector centrality dynamics are related to Alzheimer’s disease pathological changes in non-demented individuals. Brain Commun. 2023;5(3):fcad088.

    Article  PubMed  PubMed Central  Google Scholar 

  15. Rubinov M, Sporns O. Complex network measures of brain connectivity: uses and interpretations. Neuroimage. 2010;52(3):1059–69.

    Article  PubMed  Google Scholar 

  16. Binnewijzend MAA, Adriaanse SM, Van Der Flier WM, Teunissen CE, De Munck JC, Stam CJ, et al. Brain network alterations in Alzheimer’s disease measured by Eigenvector centrality in fMRI are related to cognition and CSF biomarkers. Hum Brain Mapp. 2014;35(5):2383–93.

    Article  PubMed  Google Scholar 

  17. Engels MM, Stam CJ, Van Der Flier WM, Scheltens P, De Waal H, Van Straaten EC. Declining functional connectivity and changing hub locations in Alzheimer’s disease: an EEG study. BMC Neurol. 2015;15(1):145.

    Article  PubMed  PubMed Central  Google Scholar 

  18. Stam CJ. Modern network science of neurological disorders. Nat Rev Neurosci. 2014;15(10):683–95.

    Article  PubMed  CAS  Google Scholar 

  19. Yu M, Gouw AA, Hillebrand A, Tijms BM, Stam CJ, Van Straaten ECW, et al. Different functional connectivity and network topology in behavioral variant of frontotemporal dementia and Alzheimer’s disease: an EEG study. Neurobiol Aging. 2016;42:150–62.

    Article  PubMed  Google Scholar 

  20. Yu M, Engels MMA, Hillebrand A, Van Straaten ECW, Gouw AA, Teunissen C, et al. Selective impairment of hippocampus and posterior hub areas in Alzheimer’s disease: an MEG-based multiplex network study. Brain. 2017;140(5):1466–85.

    Article  PubMed  Google Scholar 

  21. Wang Z, Qiao K, Chen G, Sui D, Dong HM, Wang YS, et al. Functional connectivity changes across the spectrum of subjective cognitive decline, amnestic mild cognitive impairment and alzheimer’s disease. Front Neuroinformatics. 2019;24(13):26.

    Article  Google Scholar 

  22. Zuo XN, Ehmke R, Mennes M, Imperati D, Castellanos FX, Sporns O, et al. Network Centrality in the Human Functional Connectome. Cereb Cortex. 2012;22(8):1862–75.

    Article  PubMed  Google Scholar 

  23. Xie Y, Liu T, Ai J, Chen D, Zhuo Y, Zhao G, et al. Changes in centrality frequency of the default mode network in individuals with subjective cognitive decline. Front Aging Neurosci. 2019;20(11):118.

    Article  Google Scholar 

  24. Ingala S, Tomassen J, Collij LE, Prent N, Van ‘T Ent D, Ten Kate M, et al. Amyloid-driven disruption of default mode network connectivity in cognitively healthy individuals. Brain Commun. 2021;3(4):fcab201.

    Article  PubMed  PubMed Central  Google Scholar 

  25. Duan F, Huang Z, Sun Z, Zhang Y, Zhao Q, Cichocki A, et al. Topological network analysis of early alzheimer’s disease based on resting-state EEG. IEEE Trans Neural Syst Rehabil Eng. 2020;28(10):2164–72.

    Article  PubMed  Google Scholar 

  26. Taguas I, Doval S, Maestú F, López-Sanz D. Toward a more comprehensive understanding of network centrality disruption in amnestic mild cognitive impairment: a MEG multilayer approach. Alzheimers Res Ther. 2024;16(1):216. https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s13195-024-01576-8. PMID:39385281;PMCID:PMC11462918.

    Article  PubMed  PubMed Central  Google Scholar 

  27. Alzheimer’s Disease Neuroimaging Initiative, Hu H, Chen KL, Ou YN, Cao XP, Chen SD, et al. Neurofilament light chain plasma concentration predicts neurodegeneration and clinical progression in nondemented elderly adults. Aging. 2019;11(17):6904–14.

    Article  PubMed Central  Google Scholar 

  28. Jack Jr CR, Andrews JS, Beach TG, Buracchio T, Dunn B, Graf A, Carrillo MC. Revised criteria for diagnosis and staging of Alzheimer's disease: Alzheimer's Association Workgroup. Alzheimers Dement. 2024;20(8):5143–69.

  29. Ashton NJ, Pascoal TA, Karikari TK, Benedet AL, Lantero-Rodriguez J, Brinkmalm G, et al. Plasma p-tau231: a new biomarker for incipient Alzheimer’s disease pathology. Acta Neuropathol (Berl). 2021;141(5):709–24.

    Article  PubMed  CAS  Google Scholar 

  30. Ashton NJ, Benedet AL, Pascoal TA, Karikari TK, Lantero-Rodriguez J, Brum WS, et al. Cerebrospinal fluid p-tau231 as an early indicator of emerging pathology in Alzheimer’s disease. eBioMedicine. 2022;76:103836.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  31. Milà-Alomà M, Ashton NJ, Shekari M, Salvadó G, Ortiz-Romero P, Montoliu-Gaya L, et al. Plasma p-tau231 and p-tau217 as state markers of amyloid-β pathology in preclinical Alzheimer’s disease. Nat Med. 2022. Available from: https://www.nature.com/articles/s41591-022-01925-w. Cited 2024 Mar 26.

  32. Suárez-Calvet M, Karikari TK, Ashton NJ, Lantero Rodríguez J, Milà-Alomà M, Gispert JD, et al. Novel tau biomarkers phosphorylated at T181, T217 or T231 rise in the initial stages of the preclinical Alzheimer’s continuum when only subtle changes in Aβ pathology are detected. EMBO Mol Med. 2020;12(12):e12921.

    Article  PubMed  PubMed Central  Google Scholar 

  33. Tissot C, Therriault J, Kunach P, Benedet AL, Pascoal TA, Ashton NJ, et al. Comparing tau status determined via plasma pTau181, pTau231 and [18F]MK6240 tau-PET. eBioMedicine. 2022;76:103837.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  34. Nasreddine ZS, Phillips NA, Bédirian V, Charbonneau S, Whitehead V, Collin I, ... Chertkow H. The Montreal Cognitive Assessment, MoCA: a brief screening tool for mild cognitive impairment. J Am Geriatr Soc. 2005;53(4):695–699.

  35. Reitan R. Validity of the Trail Making test as an indicator of organic brain damage. Percept Mot Skills. 1958;8:271–6.

    Article  Google Scholar 

  36. Wechsler D. WMS-III. Wechsler Memory Scale. 3rd ed. San Antonio, TX: Psychological Corporation; 1997.

    Google Scholar 

  37. Osterreith P, Rey A. Le test de copie d’une figure complexe. Arch Psychol. 1944;30:205–20.

  38. Taulu S, Simola J. Spatiotemporal signal space separation method for rejecting nearby interference in MEG measurements. Phys Med Biol. 2006;51(7):1759–68.

    Article  PubMed  CAS  Google Scholar 

  39. Oostenveld R, Fries P, Maris E, Schoffelen JM. FieldTrip: open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data. Comput Intell Neurosci. 2011;2011(1):156869.

    PubMed  Google Scholar 

  40. Garcés P, López-Sanz D, Maestú F, Pereda E. Choice of Magnetometers and Gradiometers after Signal Space Separation. Sensors. 2017;17(12):2926.

    Article  PubMed  PubMed Central  Google Scholar 

  41. Rolls ET, Huang CC, Lin CP, Feng J, Joliot M. Automated anatomical labelling atlas 3. Neuroimage. 2020;206:116189.

    Article  PubMed  Google Scholar 

  42. Friston KJ. Statistical parametric mapping: the analysis of functional brain images. 1st ed. Amsterdam Boston: Elsevier / Academic Press; 2007.

    Google Scholar 

  43. Brookes MJ, Woolrich M, Luckhoo H, Price D, Hale JR, Stephenson MC, et al. Investigating the electrophysiological basis of resting state networks using magnetoencephalography. Proc Natl Acad Sci. 2011;108(40):16783–8.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  44. Jensen O, Mazaheri A. Shaping Functional Architecture by Oscillatory Alpha Activity: Gating by Inhibition. Front Hum Neurosci. 2010;4. Available from: http://journal.frontiersin.org/article/10.3389/fnhum.2010.00186/abstract. Cited 2023 Nov 14.

  45. Van Wijk BCM, Stam CJ, Daffertshofer A. Comparing brain networks of different size and connectivity density using graph theory. Sporns O, editor. PLoS One. 2010;5(10):e13701.

    Article  PubMed  PubMed Central  Google Scholar 

  46. Brandes U, Pich C. Centrality estimation in large networks. Int J Bifurc Chaos. 2007;17(07):2303–18.

    Article  Google Scholar 

  47. Maris E, Oostenveld R. Nonparametric statistical testing of EEG- and MEG-data. J Neurosci Methods. 2007;164(1):177–90.

    Article  PubMed  Google Scholar 

  48. Chow GC. Tests of equality between sets of coefficients in two linear regressions. Econometrica. 1960;28(3):591.

    Article  Google Scholar 

  49. Jelic V, Johansson SE, Almkvist O, Shigeta M, Julin P, Nordberg A, et al. Quantitative electroencephalography in mild cognitive impairment: longitudinal changes and possible prediction of Alzheimer’s disease. Neurobiol Aging. 2000;21(4):533–40.

    Article  PubMed  CAS  Google Scholar 

  50. Jeong J. EEG dynamics in patients with Alzheimer’s disease. Clin Neurophysiol. 2004;115(7):1490–505.

    Article  PubMed  Google Scholar 

  51. Briels CT, Schoonhoven DN, Stam CJ, De Waal H, Scheltens P, Gouw AA. Reproducibility of EEG functional connectivity in Alzheimer’s disease. Alzheimers Res Ther. 2020;12(1):68.

    Article  PubMed  PubMed Central  Google Scholar 

  52. Hatz F, Hardmeier M, Benz N, Ehrensperger M, Gschwandtner U, Rüegg S, et al. Microstate connectivity alterations in patients with early Alzheimer’s disease. Alzheimers Res Ther. 2015;7(1):78.

    Article  PubMed  PubMed Central  Google Scholar 

  53. Ruiz-Gómez SJ, Hornero R, Poza J, Maturana-Candelas A, Pinto N, Gómez C. Computational modeling of the effects of EEG volume conduction on functional connectivity metrics Application to Alzheimer’s disease continuum. J Neural Eng. 2019;16(6):066019.

    Article  PubMed  Google Scholar 

  54. Skouras S, Falcon C, Tucholka A, Rami L, Sanchez-Valle R, Lladó A, et al. Mechanisms of functional compensation, delineated by eigenvector centrality mapping, across the pathophysiological continuum of Alzheimer’s disease. NeuroImage Clin. 2019;22:101777.

    Article  PubMed  PubMed Central  Google Scholar 

  55. Güntekin B, Erdal F, Bölükbaş B, Hanoğlu L, Yener G, Duygun R. Alterations of resting-state Gamma frequency characteristics in aging and Alzheimer’s disease. Cogn Neurodyn. 2023;17(4):829–44.

    Article  PubMed  Google Scholar 

  56. Perez C, Ziburkus J, Ullah G. Analyzing and modeling the dysfunction of inhibitory neurons in alzheimer’s disease. Cymbalyuk G, editor. PLoS One. 2016;11(12):e0168800.

    Article  PubMed  PubMed Central  Google Scholar 

  57. Verret L, Mann EO, Hang GB, Barth AMI, Cobos I, Ho K, et al. Inhibitory interneuron deficit links altered network activity and cognitive dysfunction in Alzheimer model. Cell. 2012;149(3):708–21.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  58. Cuesta P, Ochoa-Urrea M, Funke M, Hasan O, Zhu P, Marcos A, et al. Gamma band functional connectivity reduction in patients with amnestic mild cognitive impairment and epileptiform activity. Brain Commun. 2022;4(2):fcac012.

    Article  PubMed  PubMed Central  Google Scholar 

  59. Bobola MS, Chen L, Ezeokeke CK, Olmstead TA, Nguyen C, Sahota A, et al. Transcranial focused ultrasound, pulsed at 40 Hz, activates microglia acutely and reduces Aβ load chronically, as demonstrated in vivo. Brain Stimulat. 2020;13(4):1014–23.

    Article  CAS  Google Scholar 

  60. Martorell AJ, Paulson AL, Suk HJ, Abdurrob F, Drummond GT, Guan W, et al. Multi-sensory gamma stimulation ameliorates alzheimer’s-associated pathology and improves cognition. Cell. 2019;177(2):256–271.e22.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  61. Mattsson N, Andreasson U, Zetterberg H, Blennow K, for the Alzheimer’s Disease Neuroimaging Initiative. Association of plasma neurofilament light with neurodegeneration in patients with Alzheimer disease. JAMA Neurol. 2017;74(5):557.

    Article  PubMed  PubMed Central  Google Scholar 

  62. Nyberg L, Lundquist A, Nordin Adolfsson A, Andersson M, Zetterberg H, Blennow K, et al. Elevated plasma neurofilament light in aging reflects brain white‐matter alterations but does not predict cognitive decline or Alzheimer’s disease. Alzheimers Dement Diagn Assess Dis Monit. 2020;12(1). Available from: https://onlinelibrary.wiley.com/doi/10.1002/dad2.12050. Cited 2023 Nov 9.

  63. Buckner RL, Sepulcre J, Talukdar T, Krienen FM, Liu H, Hedden T, et al. Cortical hubs revealed by intrinsic functional connectivity: mapping, assessment of stability, and relation to Alzheimer’s disease. J Neurosci. 2009;29(6):1860–73.

    Article  PubMed  PubMed Central  CAS  Google Scholar 

  64. De Haan W, Mott K, Van Straaten ECW, Scheltens P, Stam CJ. Activity dependent degeneration explains hub vulnerability in Alzheimer’s disease. Sporns O, editor. PLoS Comput Biol. 2012;8(8):e1002582.

    Article  PubMed  PubMed Central  Google Scholar 

  65. Liuzzi L, Gascoyne LE, Tewarie PK, Barratt EL, Boto E, Brookes MJ. Optimising experimental design for MEG resting state functional connectivity measurement. Neuroimage. 2017;155:565–76.

    Article  PubMed  Google Scholar 

Download references

Acknowledgements

We would like to thank all the participants that have selflessly given us their time and made this study possible.

Funding

This study was funded by the Spanish Ministry of Science and Innovation [RTI2018-098762-B-C31 and PID2021-122979OB-C21]; the GAIN, Axencia Galega de Innovación, [IN607B2021/12] co-funded by the European Union, and Programa INVESTIGO, [TR349V-2022–10000052-00] co-funded by the European Union. Complimentary, it was supported by predoctoral grants by the Spanish Ministry of Universities [PRE2019-087612 and FPU18/00517] to AGC and MCG, respectively, and an eBrain-Health Grant [101058516] to IT.

Author information

Author notes

  1. Alejandra García-Colomo and David López-Sanz contributed equally to this work.

Authors and Affiliations

  1. Center for Cognitive and Computational Neuroscience, Complutense University of Madrid, Pozuelo de Alarcón, 28223, Spain

    Alejandra García-Colomo, Ignacio Taguas, Martín Carrasco-Gómez & Fernando Maestú

  2. Department of Experimental Psychology, Cognitive Psychology and Speech & Language Therapy, Complutense University of Madrid, Pozuelo de Alarcón, 28223, Spain

    Alejandra García-Colomo, David López-Sanz & Fernando Maestú

  3. Department of Legal Medicine, Psychiatry and Pathology, Complutense University of Madrid, Madrid, 28040, Spain

    Ignacio Taguas

  4. Department of Electronic Engineering, Universidad Politécnica de Madrid, Madrid, 28040, Spain

    Martín Carrasco-Gómez

  5. Translational Neuroscience Research Group, Galicia Sur Health Research Institute (IIS-Galicia Sur), SERGAS-UVIGO, CIBERSAM, Vigo, 36312, Spain

    Carlos Spuch & María Comis-Tuche

  6. Health Research Institute of the Hospital Clínico San Carlos (IdISSC), Madrid, 28240, Spain

    Fernando Maestú

Authors
  1. Alejandra García-Colomo
    View author publications

    You can also search for this author inPubMed Google Scholar

  2. David López-Sanz
    View author publications

    You can also search for this author inPubMed Google Scholar

  3. Ignacio Taguas
    View author publications

    You can also search for this author inPubMed Google Scholar

  4. Martín Carrasco-Gómez
    View author publications

    You can also search for this author inPubMed Google Scholar

  5. Carlos Spuch
    View author publications

    You can also search for this author inPubMed Google Scholar

  6. María Comis-Tuche
    View author publications

    You can also search for this author inPubMed Google Scholar

  7. Fernando Maestú
    View author publications

    You can also search for this author inPubMed Google Scholar

Contributions

Conception and design of the study: AGC, DLS, FM. Acquisition and data analysis: DLS, AGC, IT, MCG , MCT. Drafting a significant portion of the manuscript: AGC, DLS, CS, FM.

Corresponding authors

Correspondence to Alejandra García-Colomo or Fernando Maestú.

Ethics declarations

Ethics approval and consent to participate

The “Hospital Clínico San Carlos” Ethics Committee approved this study, and the procedure was performed following internationally accepted guidelines and regulations, which included informed consent from all participants.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

García-Colomo, A., López-Sanz, D., Taguas, I. et al. Effects of Alzheimer’s disease plasma marker levels on multilayer centrality in healthy individuals. Alz Res Therapy 17, 8 (2025). https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s13195-024-01654-x

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doiorg.publicaciones.saludcastillayleon.es/10.1186/s13195-024-01654-x

Keywords

Alzheimer's Research & Therapy

ISSN: 1758-9193