Connectivity dynamics and cognitive variability during aging

Aging is associated by cognitive changes, with strong variations across individuals. One way to characterize this individual variability is to use techniques such as magnetoencephalography (MEG) to measure the dynamics of neural synchronization between brain regions, and the variability of this connectivity over time. Indeed, few studies have focused on fluctuations in the dynamics of brain networks over time and their evolution with age. We therefore characterize aging effects on MEG phase synchrony in healthy young and older adults from the Cam-CAN database. Age-related changes were observed, with an increase in the variability of brain synchronization, as well as a reversal of the direction of information transfer in the default mode network (DMN), in the delta frequency band. These changes in functional connectivity were associated with cognitive decline. Results suggest that advancing age is accompanied by a functional disorganization of dynamic networks, with a loss of communication stability and a decrease in the information transmitted. This could be partly due to the loss of integrity of the network structure.

parietal networks). Our objectives were twofold: i) To study changes in dynamic connectivity 23 with age: Between young and old individuals, we hypothesized differences in functional 24 networks, as well as greater variability in the activity of these networks; ii) To investigate the 25 relationships between changes in dynamic connectivity and cognitive changes: We expected 26 that stability in synchronization and directionality of connectivity over time would be associated 27 with better cognitive performance with age, compared to high variability in these measures. 28 Preservation of this neural activity would help maintain cognitive abilities with age.

7
Data pre-processing 8 The Elekta Neuromag MaxFilter 2.2 has been applied to all MEG data (temporal signal space 9 separation (tSSS): 0.98 correlation, 10s window; bad channel correction: ON; motion 10 correction: OFF; 50Hz+harmonics (mains) notch). Afterwards, artifact rejection, filtering (0.3-11 100 Hz bandpass), re-referencing (i.e. using the algebraic average of the left and right mastoid 12 electrodes), temporal segmentation into epochs, averaging and source estimation were 13 performed using Brainstorm (Tadel et al., 2011). In addition, physiological artefacts (e.g. 14 blinks, saccades) were identified and removed using spatial space projection of the signal. In 15 order to improve the accuracy of the source reconstruction, the FreeSurfer (Fischl, 2012), 16 software was used to generate cortical surfaces and automatically segment them from the cortical structures from each participant's T1-weighted anatomical MRI. The advanced MEG   between two regions over time. If the phase difference varies little, the PLV is close to 1 (this 14 corresponds to high synchronisation between the regions), while the low association of phase 15 difference across regions is indicated by a PLV value close to zero. To ensure PLV results did 16 not reflect volume conduction artefacts, control analyses were conducted using phase lag index 17 (weighted PLI analyses). Because PLV is an undirected measure of functional connectivity, and 18 to investigate brain dynamics with complementary metrics, analyses of transfer entropy (TE) 19 have also been conducted. TE measures of how a signal a can predict subsequent changes in a 20 signal b (Ursino et al., 2020). It then provides a directed measure of a coupling's strength. If 21 there is no coupling between a and b, then TE is close to 0, while TE is close to 1 if there is a 22 strong coupling between a and b.

23
The range of each frequency band was based on the frequency of the individually observed   open access and scripts are available online. To assess differences between age groups in 1 demographic and functional connectivity variables, t-tests and ANOVAs were applied using 2 Jamovi software (https://www.jamovi.org/; version 1.6.23). Functional data (PLV, TE) were 3 analyzed using 2 (age group: young/old) x 4 (networks: DMN, SN, FPL, and FPR) x 6 4 (frequency bands: delta, theta, alpha, beta, gamma1, gamma2) repeated-measures ANOVAs to 5 determine which network and frequency band showed the greatest young/old changes. The

11
Age-related differences in cognitive performance 12 The main behavioral and demographic data from the Cam-CAN database are summarized in 13 Table 1.   advancing age (Figure 2). We observed a significant negative regression between such 2 variability and cognitive performance (VSTM, p = 0.009, r = -0.387). The rest of the analyses 3 was therefore focused on the DMN network, in the delta frequency band. 4 We then performed permutation t-tests on the DMN couplings between age groups. Different 5 couplings were found to be significantly more variable for the older group compared to the 6 younger group especially for interhemispheric and fronto-parietal couplings (Figure 3). In the 7 older group, the right frontoparietal coupling was found to be negatively correlated with 8 cognitive performance (MMSE test; r = -0.305, p = 0.039). In the older group, the 9 interhemispheric coupling (bilateral supramarginal regions) was found to be negatively 10 correlated with cognitive performance (VSTM test; r = -0.344, p = 0.021). These data suggest 11 an increase in variability in the overall DMN network in the delta frequency band, but also an Our main objective was to investigate changes in the stability and variability of brain 2 communication dynamics with age and the relationship of these changes with age-related 3 cognitive changes. Our connectome-based approach, based on MEG data in healthy young and 4 older participants from the Cam-CAN database, allowed us to investigate changes of 5 connectivity dynamics with aging. Two time-resolved connectivity aspects were studied: the 6 stability of synchronized communications over time, and directed connectivity. Brain activity 7 was studied at rest, as previous work suggested a link between the activity of specific networks 8 at rest and cognitive abilities (e.g. Nashiro et al., 2017). In this study, we first showed an 9 increased variability of phase synchrony over time with age, especially in the delta frequency 10 band. We also showed a reversal of the main direction of synchronized connectivity with age: 11 connectivity in the fronto-parietal direction was found to be increased in older participants, 12 whereas it was stronger in the parieto-frontal direction for younger participants. These analyses. 19 The study of oscillatory activity allowed us to specify age-related changes in the variability of 20 phase synchrony over time, and the specific frequency band associated with these differences.

21
Phase synchrony between brain regions is a critical parameter of neural communications (e.g.,

22
Fries, 2015). Indeed, with advancing age, changes in synchronized network communications are also consistent with previous M/EEG work reporting an overall slowing of brain activity 30 with advancing age (e.g., Celesia, 1986), with an increase of slow rhythms relative to faster 31 rhythms. Increased slow waves seem to be associated with the cognitive decline observed with 32 advancing age. Here, we show that this slowing of brain rhythms with age is associated with a 33 loss of stability in neuronal communications, and poorer performance.

34
In association with synchrony analyses, transfer entropy analyses allow the quantification of 35 directed connectivity (see Ursino et al., 2020). This quantifies the information flow between 36 brain regions more precisely than functional connectivity, thus allowing the detection of causal age. This reversal of information transfer between young and old participants was negatively 41 correlated with cognitive performance (especially for working memory and fluid intelligence).

42
The reversal of information transfer and decreased variability in phase synchrony observed here 43 may help furthering the age-related pattern described in the PASA model (Cabeza et al., 2018). Theoretically, the variability of brain communications has received little investigation, as it was 30 long considered as noise, but is now recognized as contributing to brain functions (Uddin et al.,31 2020). Here, we show that healthy aging is associated with an increased variability in 32 synchronized brain communications, and with changes of the main connectivity directions 33 between brain regions. Results highlight that even when brain networks are not engaged in a 34 particular cognitive activity, significant changes occur with age regarding connectivity 35 dynamics and information flow between regions of different functional brain networks.

36
Advancing age appears to be accompanied by a functional disorganization of dynamic 37 networks, with a loss of communication stability and a decrease in the information transmitted.

38
The study of dynamic connectivity contributes to a better understanding of the cognitive decline