Breakdown of TMS evoked EEG signal propagation within the default mode network in Alzheimer’s disease

(cid:1) AD patients show a remarkable breakdown of signal propagation within the DMN tested by TMS-EEG recordings. (cid:1) These disruptions are not detectable stimulating other areas (left dorsolateral prefrontal cortex) or for different networks. (cid:1) Individual connectivity impairment is s associated with the level of cognitive impairment, as measured by the CDR-SB.


Introduction
Alzheimer's disease (AD) is well-known as the principal cause of dementia worldwide (Mayeux and Stern, 2012).Although several pathophysiological models were suggested, AD aetiology remains unclear and brain-related dysfunctions still need to be fully understood.Investigating brain connectivity in patients affected by AD is one of the most promising approaches adopted to explore the disease's neurophysiological and cognitive dysfunctions.In recent years, several shreds of evidence highlighted the notion that loss of synaptic function could be an early event antecedent to neuronal degeneration.Such impairment of synaptic mechanisms could play a key role in the onset of cognitive dysfunction in AD by disrupting large scale connectivity (E.P. Casula et al., 2022;Koch et al., 2018Koch et al., , 2022)).The first signs of cognitive damage appear only when a substantial synaptic loss has occurred in vulnerable brain regions (Reddy et al., 2010).AD causes a dysfunction of connectivity, that can involve specific resting-state brain networks (RSNs).Metabolic 18F-fluorodeoxyglucose positron emission tomography (FDG-PET) and functional Magnetic Resonance Imaging (fMRI) data showed a prominent dysfunction of the so-called Default Mode Network (DMN) (Buckner et al., 2008;Dennis and Thompson, 2014;Raichle et al., 2001;Wu et al., 2011).Core AD symptoms such as memory deficit have been considered as a direct consequence of such DMN brain connectivity disruption (Koch et al., 2018;Mevel et al., 2011;Morris, 1996;Veldsman et al., 2017).Quantifying the link among synaptic functioning (or dysfunction) within specific brain networks such as the DMN directly in vivo in AD patients has not been possible so far.Different techniques such as FDG-PET (Ding et al., 2019), fMRI (Sperling, 2011) and electroencephalography (EEG) (Horvath et al., 2018) have been used to investigate the relationship among synaptic dysfunction and network connectivity in AD patients.However, these techniques provide only an indirect estimate of network dysfunction, being limited by the low temporal resolution.
Here, we used a novel multimodal approach to track neural signal propagation within the DMN in AD patients, to shed light on this crucial aspect of the disease and to identify new insights about its pathophysiology.We took advantage of a combination of transcranial magnetic stimulation with electroencephalography (TMS-EEG) as an emerging method to directly probe local and widespread cortical dynamics, through the recording of post-synaptic potentials following depolarisation generated by TMS.By applying a series of TMS pulses over the precuneus (PC), the central key of the DMN, we sought to determine how neural signal was propagating from the site of stimulation within the network in each patient.We then used functional and structural MRI to integrate these neurophysiological signals within the individual anatomical and functional network.

Participants and assessment procedure
Twenty AD patients (10 female, 56 to 85 years, 69.3 mean, 7.04 standard deviation) were recruited for this study between January 2021 and September 2022.Seventeen age-matched healthy volunteers (HV) (10 female, 50 to 82 years, mean ± sd 69.9, 8.71 years) were recruited.AD patients were recruited at the Memory Clinic of the Santa Lucia Foundation (Rome, Italy).AD patients' characteristics are shown in the Table 1.To be considered eligible for the study, patients underwent a complete clinical investigation, an MRI scan and a complete neuropsychological evaluation (Magni et al., 1996).Lumbar puncture was performed to investigate the cerebrospinal fluid (CSF) biomarker evidence of AD amyloid and tau pathology.
Patients were eligible if they had an established diagnosis of probable mild-to-moderate AD according to the International Working Group recommendations (Dubois et al., 2016).The study was approved by the review board and ethics committee of the Santa Lucia Foundation (CE/PROG.716;20/09/2019) and was conducted following the principles of the Declaration of Helsinki and the International Conference on Harmonisation Good Clinical Practice guidelines.All patients or their relatives or legal representatives provided written informed consent.Patients could withdraw at any point without prejudice.
AD patients' assessment consisted of three different examinations: (1) a neurophysiological assessment, performed with a TMS-EEG recording; (2) a structural and functional assessment performed with an MRI scanning; (3) a neuropsychological assessment.HV underwent a TMS-EEG recording as in AD patients.To evaluate the presence of side effects or discomfort during neurophysiological and neuroimaging assessments all the participants had to fill in a questionnaire at the end of each evaluation.

TMS-EEG recordings and analysis
A Magstim Rapid 2 magnetic stimulator (Magstim Company, Whitland, Wales, UK) was used to deliver biphasic waveform TMS pulses, with a pulse width of ∼0.1 ms.The stimulator was connected to a figure-of-eight-coil with a diameter of 70mm that was directly positioned over the subject's scalp.During TMS-EEG neurophysiological assessment 80 single pulses were applied at a random interstimulus interval (ISI) of 2 seconds with a variation of 20%.To avoid auditory and somatosensory responses caused by TMS, we followed an accurate procedure previously tested in our studies (Cristofari et al., 2023;Mancuso et al., 2024Mancuso et al., , 2023;;Rocchi et al., 2021) and detailed in the supplementary materials.The intensity of TMS single pulses was initially set at 100% of the adjusted Resting Motor Treshold (RMT) and then eventually adjusted in order to achieve a TMS-evoked potential (TEP) with a peak-to-peak amplitude of at least 4 μV (see Koch et al., 2022).This approach was adopted since some studies provide evidences about the necessity to use individualized intensity to stimulate cortical areas during TMS-EEG in HV and AD patients (Casarotto et al., 2022;E.P. Casula et al., 2022;Koch et al., 2022).TMS pulses were delivered over the PC and the left dorsolateral prefrontal cortex (L-DLPFC) on the participants' scalp via neuronavigation of the individual highresolution T1w image collected for each patient (see below for details).The T1w images were imported in the neuronavigation system (SofTaxic; E.M. S., Bologna s.r.l.) coupled with a Polaris Vicra infrared camera, and were coregistered to digitalized landmarks for online monitoring of the head and coil position.The TMS targets (PC, L-DLPFC) were individualized for each patient using a non-linear transformation of MNI coordinates (PC: x, y, z: 0, -62, 64; L-DLPFC: x, y, z: -47, 25, 34) to subject space through a home-based script using FSL (refer to the 'Structural and functional MRI assessment' paragraph for details).The EEG was acquired using TMS-compatible equipment (BrainAmp 32MRplus, BrainProducts GmbH, Munich, Germany) and was continuously recorded from 64 sites positioned according to the 10-20 International System, using TMS-compatible Ag/AgCl pellet electrodes mounted on an elastic cap.TMS-EEG data were preprocessed offline with Brain Vision Analyzer (Version 2.2.0, Brain Products GmbH, Gilching, Germany) following an optimized pipeline previous used by our group in many studies (Elias P. Casula et al., 2022a;Casula et al., 2024;L et al., 2022;Maiella et al., 2022;Tăuƫan et al., 2023).Pre-processed EEG data were imported in Brainstorm (Tadel et al., 2011) (V. 08-Dec-2022), software running in a MATLAB environment (MathWorks Inc., Natick, MA, Version R2020B) to perform different analysis.Indeed, we performed an analysis focusing on TEPs; successively we performed the source analysis to reconstruct the topographical origins of the TMS-EEG signals; finally, we calculate the connectivity between areas within the DMN through a time-frequency coherence analysis.

Statistical analysis
Statistical analysis was performed with SPSS (Version 22.0.) and with Brainstorm toolbox (Tadel et al., 2011) working in a MATLAB environment (MathWorks Inc., Natick, MA, Version R2020B).Before undergoing parametric or nonparametric statistical procedures, normality distribution of data residuals was assessed with Shapiro-Wilks' test; homoscedasticity was assessed with Levene's test.Analysis of cortical excitability in the spatial domain was performed with a nonparametric, clusterbased permutation statistics comparing TEPs recorded in each electrode for the two groups from 50 ms before the TMS pulse to 200 ms after it using Monte Carlo estimates by randomly permuting the original conditions data for 3.000 times (Casula et al., 2018;Maris and Oostenveld, 2007).Analysis of cortical activity in the temporal domain was conducted in the following way.We first extracted the source activation time series from -50 to 200 ms after TMS for each site and network condition.Afterwards, we performed multiple independent Student's t-test comparing two surrogate distributions of the two groups constructed by randomly permuting the original conditions data 3.000 times.To further reduce the occurrence of type I errors, p-values were corrected with false discovery rated method (Casula et al., 2016).Then we used the same procedure to analyse the source activation time series within six ROIs of the DMN (see cortical source analysis paragraph) comparing the two groups after PC stimulation.
To investigate connectivity, we first performed a coherence analysis, a complex-valued metric that measures the linear relationship of two signals in the frequency domain.We compared Precuneus ROI with the other 5 ROIs after the PC stimulation, considering time-series in six frequency bands (delta, theta, alfa, beta, gamma l, gamma h) from 50 ms before the TMS pulse to 200 ms after it.This analysis was conducted in the same way of time-domain source activity analysis.To assess linear relationships between clinical and neurophysiological data, we chose to perform a bivariate correlation between our clinical measure, i.e.CDR-SB, and our neurophysiological primary outcome, i.e. cortical activation over DMN.Moreover, the relationship between the structural and neurophysiological outcomes was assessed by investigating the linear relation between the cortical activation over the DMN and the Fractional Anisotropy values extracted from the Cingulum and the Corpus Callosum.Finally, the same procedure was used to evaluate the relationship between the DMN cortical activation and the functional connectivity within the DMN and the FPN.Correlations were computed with Pearson's or Spearman's coefficient (two-tailed) depending on the distribution of data.

Neurophysiological results
TMS-EEG experimental procedures were well tolerated by AD patients and age-matched HV. Figure 1 depicts the TEPS recorded after PC stimulation.Figure 1 (a) shows how for both groups TEPs lasted around 200 ms after the TMS pulse and were characterized by a series of positive and negative peaks with an amplitude ranging from -2.5 to 2.5 μV.This activation dynamic, in terms of waveform and amplitude, was similar for the two groups.Specifically, figure 1 (b) shows significant differences between the two groups for the following electrodes: CP5 (t(35) =2.3, p =0.02) from 90 to 105 ms; CP3 (t(35) =2.5, p =0.01) from 90 to 110 ms; CP2 (t(35) =2.3, p =0.02) from 100 to 130 ms; P5 (t(35) =2.4,p =0.01) from 92 to 107 ms; P3 (t(35) =2.3, p =0.02) from 93 to 110 ms; Pz (t(35) =2.2, p =0.03) from 95 to 116 ms. Figure 2 depicts the TEPS recorded after L-DLPFC stimulation.For both groups TEPs lasted around 200 ms after the TMS pulse and were characterized by a series of positive and negative peaks with an amplitude ranging from -2.5 to 2.5 μV.This activation dynamic, in terms of waveform and amplitude, was similar for the two groups.Figures 3,4 and 5 depict the results of the comparison between AD and HV for the cortical sources networks and ROIs considered for the analysis.Figure 3 represents the cortical sources activation after PC stimulation.Overall, cortical sources activation tracks a poster-anterior medial spreading dynamics of cortical activity in the HV group, while cortical activations remain constrained within the stimulated area (PC) in the AD group.The comparison shows significant differences in DMN source activity between the two groups in three time windows (25-45 ms, t(35)=2.22,p=0.03; 65-82 ms, t(35) = 2.12, p = 0.03; 109-136 ms, t(35) = 2.11, p = 0.04); no differences were observable for the FPN source activity.Figure 4 represents the cortical sources activation after L-DLPFC stimulation.Overall, cortical sources activation strongly involves areas around the stimulation site, extending posteriorly toward the posterior parietal cortex after 50 ms.The comparison shows no significant differences for the FPN source activity between the two groups.Figure 5 represents the cortical sources activation after the PC stimulation for the six ROIs in which we subdivided the DMN.The a significant difference for two time windows in the Frontal ROI (15-31 ms, t(35) = 1.4,p = 0.04; 50-86 ms, t(35) = 2.4, p = 0.01); five time windows in the Parietal left ROI (11-34 ms, t(35) = 2.9, p = 0.01; 51-60 ms, t(35) = 1.3, p = 0.04; 77-86 ms, t(35) = 1.9, p = 0.04; 100-110 ms, t(35) = 1.16, p = 0.04; 125-150 ms, t(35) = 2.17, p = 0.02); two time windows in the Parietal right ROI (11-85 ms, t(35) = 2.35, p = 0.02; 101-132 ms, t(35) = 2.3, p = 0.03); two time windows in the Temporal left ROI (20-27 ms, t(35) = 1.5, p = 0.04; 120-144 ms, t(35) = 1.6, p = 0.04).Statistics analysis did not determine any significant difference between groups in the right Temporal and Precuneus ROIs.We then performed a coherence analysis in the time-frequency domain.The results of this analysis are presented in figure 6.The Precuneus ROI was compared to the other ROIs, investigating the synchronization between different areas from -50 to 200 ms respect to the TMS pulse.The precuneus ROI showed a greater synchronized activity with the medial frontal ROI in HV than AD in the following frequency bands and time windows: delta from 176 to 200 ms (t( 35

Structural and functional MRI results
The relationships between structural, functional, neurophysiological and clinical outcomes are shown in Figure 6.In particular, we calculated the functional connectivity within the main hubs of DMN (Precuneus, Medial Prefrontal Cortex, right and left Angular Gyrus) and we correlated it to the global DMN source activation.Panel (a) depicts the results of this analysis.The results showed a negative correlation (r = -0.495,p = 0.013) between the mentioned measures.Moreover, to investigate the specificity of this result, we analysed the correlation between the functional connectivity calculated within the main hubs of FPN (right and left DLPFC and PPC) and the global DMN source activation.The results highlighted no significant correlation between the neurophysiological measure and the functional connectivity in the FPN nodes (r = -0.274,p = 0.121).The same procedure was used to evaluate the relationship between the neurophysiological and structural outcomes.Specifically, the white matter integrity (calculated by Fractional Anisotropy) was extracted for each subject from the Cingulum (main tract) and the Corpus Callosum (control tract) using a deterministic fibre tracking algorithm and then correlated to the global DMN source activation.Panel (b) depicts the results of this analysis.The results showed that the DMN source activation and the white matter integrity of the Cingulum were negatively correlated (r = -0.563,p = 0.01), whereas no significant correlation was found with the white matter integrity of the Corpus Callosum (r = -0.104,p = 0.663, Figure 5), demonstrating the specificity of our results.Furthermore, we explored whether the structure of the Cingulum was related to the functional connectivity within the DMN, thus focusing the analysis on the relationship between structural and functional outcomes.Panel (c) depicts the results of this analysis.A significant positive correlation emerged between the white matter integrity of the Cingulum and the DMN functional connectivity (r = 0.452, p = 0.045).Finally, the link between neurophysiological and clinical outcomes was studied by correlating the DMN and FPN source activation and the CDR-SB scores.The results revealed a negative correlation between DMN source activation and CDR-SB scores (r = -0.453,p = 0.003), while no significant correlation was found between the FPN source activation and the CDR-SB scores (r = 0.213, p = 0.198).

Discussion
Here, we used a novel multimodal approach to track neural signal propagation within the DMN in AD patients.Indeed, we pointed to the PC as a key hub area of the DMN being connected with important areas forming the DMN such as the anterior cingulate cortex and the inferotemporal cortex.We used TMS pulses to probe with a millisecond time resolution the propagation of evoked EEG signal across the different brain networks reconstructed using fMRI and DTI tractography.In AD patients a probe TMS pulse was able to enhance local evoked activity over the PC thereby unmasking underlying hyperexcitability (E.P. Casula et al., 2022).In contrast, although the evoked signal was higher in AD patients, it was confined locally and not able to propagate efficiently within the key nodes of the DMN.This is also consistent with previous findings observed in young healthy subjects showing that more posterior sites have more reactive TEPs and less propagation (Rosanova et al., 2009).Concomitant fMRI and tractography recordings helped to define how this impaired signal propagation was related to the same connectivity matrices derived from BOLD signal and transferred by specific white matter bundles forming the cingulum.Moreover, we found that the above described breakdown of signal propagation was specific for the DMN.TMS-EEG evoked signal propagation within the FPN did not vary between AD patients and age matched HVs.
Cortical reactivity assessment measured by applying TMS-EEG on the PC revealed increased excitability of the EEG evoked signal below the stimulation site.Such hyperexcitability has been reported in several studies across different areas of the brain (Pasquini et al., 2019;Zott et al., 2019) and recently described also over the PC (Casula et al., 2022).However, here we sought to investigate how EEG signal propagated within the entire DMN while stimulating the PC.We found a striking difference in terms of source-level DMN global activation between AD patients and HV.While the stimulation of the PC in HV produces a robust global activation of the whole DMN, TMS-EEG evoked signal propagation in AD patients remains restricted to the area of stimulation.Indeed, we expanded our analysis by segmenting the DMN into different regions, to investigate if signal propagation varied across the different brain areas connected with the PC.This analysis revealed that signal propagation was particularly reduced in AD patients as compared to HV between the PC and the frontal and temporal nodes of the DMN.Moreover, we also analysed the coherence frequency bands synchronization within the DMN.We found that AD patients showed decreased coherence the precuneus and the medial frontal cortex and between the precuneus and the temporal poles in the theta and gamma bands.This is consistent with previous studies reporting deceased activity of cortical oscillations in AD patients (Elias P. Casula et al., 2022b;Fide et al., 2023;Kumar and Ray, 2023).This is relevant since gamma band oscillatory activity is involved in the formation of synaptic plasticity, and it is decreased in AD even before the occurrence of cognitive decline (Iaccarino et al., 2016).Moreover, different studies showed the relationship between the delta, theta, beta oscillations synchronization disorders (Koenig et al., 2005;Stam et al., 2003) and the cognitive test scores in AD (Zhang et al., 2021).The altered communication between the parietal and frontal areas could represent a substrate of the AD patient's impairments in cognitive and behavioural domains.Indeed, the information's integration between parietal and frontal regions is crucial for a large number of cognitive domains, i.e. memory (Ramanan et al., 2019) and executive functions (Cristofori et al., 2019).Not surprisingly, these results resemble the breakdown of connectivity initially identified during sleep using a similar TMS-EEG approach by Massimini et al. (2005) and later confirmed in disorders of consciousness (Massimini et al., 2012(Massimini et al., , 2010)).Future studies could test whether there is a parallelism between sleep related alterations in connectivity in patients with AD.
We also found that DMN connectivity correlated with the CDR-SB which is used as the primary outcome measure to assess cognition and functional independence in several RCT, indicating that a more altered connectivity was associated with worse cognitive functions.DMN aberrancies involvement in AD progression are widely reported in plenty of studies that adopt different models (Ji et al., 2018) and several techniques (Koch et al., 2018;Pascoal et al., 2019;Qian et al., 2019;Yildirim and SONCU BÜYÜKİŞCAN, 2019).Aβ accumulation in AD is well-known as one of the main biomarkers able to identify the presumed presence of the disease; DMN is considered as a remarkable network for the Aβ anomaly accumulation, indicating it as a system particularly damaged by the neuropathology and probably related to the cognitive problems observable in patients (Ossenkoppele et al., 2019).We found consistency with these previous results about the DMN-strong divergencies that occur between AD patients and HV.

From local hyperexcitability to impaired long-range connectivity
The PC cortical hyperexcitability detected with our TMS-EEG analysis could involve directly the degeneration of inhibitory GABAergic cells (Cheng et al., 2020;Mattson, 2020;Toniolo et al., 2020), including fast-spiking GABAergic interneurons (Kann et al., 2014).In animal models of AD, hyperactivity of cortical neurons has been associated with decreased GABAergic inhibition (Busche et al., 2008).Additionally, reduced GABAergic terminals have been reported in AD patients as well as in APP-PS1 transgenic mice, indicating that the loss of GABAergic terminals may lead to the hyperactivity of the neurons in contact with Aβ plaques (Garcia-Marin et al., 2009).On the other hand, GABA agonists (e.g., benzodiazepines) may re-establish local inhibition, reducing hyperexcitability and improving cognition in AD animal models (Busche et al., 2008;Sun et al., 2009;Verret et al., 2012).Hence, the hyper-excitability reported here may involve the degeneration of GABAergic interneurons (Cheng et al., 2020;Mattson, 2020;Toniolo et al., 2020;Verret et al., 2012).Moreover, it has been shown that these mechanisms of hyperexcitability are also attenuated by microglia cells (Badimon, 2020;Merlini et al., 2021).Microglia-driven neuro suppression might play a complementary role in restricting excessive neuronal activation that cannot be sufficiently suppressed by inhibitory neurons alone (Badimon, 2020).Interestingly, in our study, we did not find any correlation between the above-mentioned hyperexcitability and the breakdown of DMN connectivity.Hence, it is possible that impaired network communication involves different neurotransmitters apart from GABA.
Previous works showed that long-term cholinergic treatment with AChEIs shows enhanced connectivity of the DMN and the interrelated hippocampus (Blautzik et al., 2016;Goveas et al., 2011).Moreover, serotoninergic drugs have been associated with increased DMN connectivity with the PC and posterior cingulate cortex in AD patients (Klaassens et al., 2019).Moreover, several experimental studies performed both in primates and in humans revealed that dopaminergic fibres arising from the meso-cortical tract innervate the parietal cortex and the PC (van Kempen et al., 2022).Hence it is plausible that the breakdown of DMN connectivity could be the consequence of neurotransmitters alterations that have been consistently described in different studies.We recently proposed (Koch and Martorana, 2023) that the reduction of functional connectivity could eventually derive from the degeneration of mesocorticolimbic dopaminergic pathway activity or in parallel from the degeneration of locus coeruleus, a condition that occurs early in Alzheimer's disease.
Disconnection between posterior and anterior regions in AD could also be ascribed to the main pathological characteristics of the disease, including tau protein deposition, which has been strongly related to the mechanisms of disconnection within and between networks (Adams et al., 2019;Zhou et al., 2012).However, it remains to be better elucidated how the different neurotransmitters and the neuropathological deposition of tau and amyloid contribute to the specific DMN alterations reported here.

TMS-EEG as a tool to assess brain connectivity
EEG has been used for many years to acquire a direct, non-invasive view of human brain activity in conditions of physiological and pathological aging including AD.Although findings related to EEG indexes of connectivity become valuable over the years, their use is often disorganized and there is still no agreement as to which connectivity analysis could be elected as more affordable to detect the longitudinal neurodegenerative changes in AD (Babiloni et al., 2018(Babiloni et al., , 2009;;Brunovsky et al., 2003).In this background, our findings show that the combination of TMS-EEG is capable to dig more directly into connectivity dysfunction and represents a valuable novel approach to measuring dynamics of brain networks activity in AD patients.Moreover, TMS pulses may be applied over specific nodes of the network accordingly to the underlying individual resting-state fMRI network activity, thereby providing the chance to identify the most prominent alterations in each patient.At this regard it has to be noted that a limitation of the current study is that we set the reference intensity of TEPs to a threshold of 4 microvolts which may be too low to always guarantee a strong evoked signal.Similarly, the latency of the neurophysiological response, which varies depending on the area to be stimulated, should also be considered.Whenever possible, an early assessment of the area's response (i.e.10-50 ms after the TMS pulse) would be optimal to accurately study cortical reactivity.
We also found that the breakdown of DMN connectivity detected with TMS-EEG was consistent with structural and functional MRI DMN analyses.Previous fMRI and DTI studies highlighted that the DMN and the cingulum are strongly implicated in the onset and progression of AD (Bendlin et al., 2010;Bozzali et al., 2012;Bressler and Menon, 2010;Catheline et al., 2010).The dysfunction and degeneration of these neural networks contribute to the cognitive deficits observed in individuals with the disease, particularly in areas related to memory, self-referential thinking, and internal mentation.(Brettschneider et al., 2015;Zhou et al., 2012).
Even though recent studies have applied multimodal neuroimaging approaches to detect sensitive biomarkers for AD (Teipel et al., 2012;Zimny et al., 2011), there are no studies integrating neuroimaging and neurophysiological approaches to identify multimodal biomarkers useful in clinical diagnosis to predict disease progression and evaluating treatment effects.In the current study, we used a multimodal approach to detect whether the intrinsic functional connectivity within the nodes of DMN, the white matter integrity of the cingulum, and the TMS-evoked signal propagation are potentially correlated as effective AD biomarkers.By examining this relationship, we found a significant correlation between the neuroimaging and neurophysiological biomarkers, showing that higher TMS-evoked signal propagation over DMN corresponds to a lower functional and structural integrity over DMN and cingulum.On the other hand, our findings suggest that the DMN functional connectivity is positively linked to the integrity of the cingulum.The interconnection between restingstate networks and specific white matter tracts has been previously demonstrated in the healthy human brain (van den Heuvel et al., 2009;Honey et al., 2009;Park and Friston, 2013;Skudlarski et al., 2008).Specifically, the presence of direct structural white matter pathways defined as cingulum tract, connecting the regions of DMN, has already been proved (Greicius et al., 2009;Heuvel et al., 2008;Teipel et al., 2010).As demonstrated by our results, this correlation between the integrity of the cingulum and DMN connectivity remains stable also in AD patients, despite the disrupted connectivity within the DMN and the white matter abnormalities typical of the disease.Moreover, our results confirm previous findings indicating the association between the decreased white matter integrity in the cingulum and the altered connectivity within the DMN (Agosta et al., 2011;Zhang et al., 2014) but also highlight the specificity of the white matter and the network involved in the disease.

Figure 2
Figure 2 Cortical excitability results after left dorsolateral prefrontal cortex (l-dlpfc) single pulse transcranial magnetic stimulation (TMS).Panel (a) shows the butterfly plot representing the TMSevoked potentials (TEPs).Panel (b) top figure represents the statistical plot that shows no significative channels over time.Bottom figure shows the statistical maps depicting the results of the permutation comparing AD and HV.No significative effects were detected.

Figure 3
Figure 3 Cortical source activations after precuneus (PC) stimulation for both healthy volunteers (HV) and Alzheimer Disease patients (AD).Panel (a) shows the global activation for different time points for both HV and AD.Cortical sources activation tracks a poster-anterior spreading only for HV, while AD activations remain localized to the stimulated area.Panel (b) shows the time-series of the DMN cortical sources activations for the two groups and (AD in blue, HV in yellow).Multiple independent Student's t-test significant differences between groups are evident only in Default Mode Network in three time windows (25-45 ms, t(35)=2.22,p=0.03; 65-82 ms, t(35) = 2.12, p = 0.03; 109-136 ms, t(35) = 2.11, p = 0.04).On the left, it is possible to examine a topographical representation of the network analysed.

Figure 4
Figure 4 Cortical source activations after left dorsolateral prefrontal cortex (L-DLPFC) stimulation for both healthy volunteers (HV) and Alzheimer Disease patients (AD).Panel (a) shows the global activation for different time points for both HV and AD.Cortical sources activation strongly involves areas around the stimulation site, extending posteriorly after 50 ms.Panel (b) shows the time-series of the FPN cortical sources activations for the two groups and (AD in blue, HV in yellow).No multiple independent Student's t-test significant differences were found for both networks and groups.On the left, it is possible to examine a topographical representation of the network analysed.

Figure 7
Figure 7 Relationship between structural assessment, functional evaluations, and neuropsychological outcomes; Top section of panel (a) shows the negative correlation (r = -0.495,p = 0.013) between DMN cortical source time series and functional connectivity (FC) within the nodes of DMN (i.e.PC and medial PFC), whereas the bottom section of panel (a) illustrates the lack of correlation between the same DMN cortical source time series and the FC of other cortical nodes that are not part of the DMN (i.e.PC and middle PFC, which are hubs of Fronto-parietal network), (r = -0.274,p = 0.121).Top section of panel (b) shows the negative correlation (r = -0.563,p = 0.01) between the same DMN cortical source time series and the values of fractional anisotropy (FA) extracted from Cingulum; on the other hand, bottom section of panel (b) illustrates the lack of correlation (r = -0.104,p = 0.663) between DMN cortical source time series and FA extracted from corpus callosum, thus demonstrating the specificity of these results.Finally, panel (c) shows the negative correlation (r = -0.453,p = 0.0029) between DMN cortical source time series and clinical outcome as measured by the Clinical Dementia Rating Scale -Sum of Boxes (CDR-SB), whereas bottom section of panel (c) shows the lack of correlation (r = 0.213, p = 0.198) between FPN cortical time series and CDR-SB.

•
Alzheimer's disease (AD) patients show a remarkable breakdown of signal propagation within the Default Mode Network (DMN) tested by TMS of the precuneus with simultaneous EEG recordings.
Precuneus and parietal left ROIs showed a greater oscillatory synchronized activity in HV than AD in the following frequency bands and time windows: beta from 166 to 171 ms (t(35) =2, p =0.04); gamma l from 123 to 128 ms ((t(35) =2.2, p =0.03); gamma h from 123 to 129 ms (t(35) =3.1, p =0.01).Precuneus and temporal left ROIs showed an higher oscillatory synchronized activity in HV than AD in the following frequency bands and time windows: beta from 21 to 41 ms (t(35) =2.3, p =0.03); gamma l from 10 to 21 ms (t(35) =3.2, p =0.01) and from 30 to 42 ms (t(35) =2, p =0.01).Finally, precuneus and temporal right ROIs showed a stronger oscillatory synchronized activity in HV than AD in the following frequency band and time window: beta from 158 to 174 ms (t(35) =2.6, p =0.01).No significative differences were found for what concern the comparative statistical analysis between Precuneus and Parietal right ROIs.