Neurophysiological signatures of ageing: compensatory and compromised neural mechanisms

Abstract

Spatiotemporal patterns of neural oscillations change with ageing, even in the cognitively unimpaired individual. Whether these neurophysiological changes represent ageing-related vulnerabilities or mechanisms that support cognitive resilience remains largely unknown. In this study, we used magnetoencephalography imaging to examine age-related changes of resting-state whole-brain neurophysiology in a well-characterized cohort of cognitively unimpaired individuals (n = 70; age range 52-87 years). We quantified spatial patterns of age-related changes in band-limited spectral power within delta-theta (2-7 Hz), alpha (8-12 Hz) and beta (13-30 Hz) bands and the spectral aperiodic slope (15-50 Hz), and examined how spectral changes are associated with cognitive abilities in healthy ageing. In a subset of individuals (n = 40) who were evaluated with a uniform battery of cognitive tests, using a partial least square regression approach, we examined the associations between age-related spectral changes and cognitive performance. We found that, with advancing age, delta-theta and beta spectral power reduces, while alpha spectral power increases. A periodic slope also showed reductions with ageing. Better cognitive scores were positively correlated with delta-theta reductions and alpha power increases associated with ageing, suggesting that these may represent compensatory neural mechanisms. Beta power reductions and spectral aperiodic slope changes, in contrast, correlated negatively with higher cognitive scores, suggesting that these may represent compromised neural mechanisms of ageing. Our findings highlighted that the neurophysiological changes that occur during later decades of life were distinct from the previously known lifespan changes. This study demonstrates the trajectories of neurophysiological changes in cognitive ageing explicitly relating to conserved and impaired neural mechanisms with important implications for identifying specific spectral changes in neurodegenerative processes in the context of ageing.

Keywords: aperiodic slope; compromised and compensatory brain changes; healthy ageing; magnetoencephalography; neural oscillations.

Graphical Abstract

Figure 1

Spectral changes with ageing. Band-limited spectral power within delta–theta (2–7 Hz), alpha (8–12 Hz) and beta (13–30 Hz) frequency bands and spectral aperiodic slope within 15–50 Hz is examined in the full cohort (n = 70). The brain renderings display the average spectral power values for each frequency band. Delta–theta spectral power distribution showed a frontal predominant distribution (A), alpha showed an occipito-temporal predominant distribution (B) and beta showed a parieto-central predominant distribution (C). Exponent of the aperiodic spectral component showed reduced slopes in the frontal and temporal lobes (D). Spline fits for region-specific spectral data (Brainnetome atlas modules) showed the trajectory of change across age for each band-limited spectral power (E–G), and a linear model fit showed a low-grade increase of alpha power with ageing (J) but a low-grade decrease in delta–theta (I) and beta power with ageing (K). Spline fits showed the trajectory of aperiodic slope change across age (H), and a linear model fit showed decreasing aperiodic slope with ageing (L). Coloured lines in subplots E–L represent the spectral measures in each of the 48 modular-level parcellations defined in the Brainnetome atlas, and dotted lines represent the average across all regions. SP, spectral power.

Figure 2

Spatial patterns of band-limited spectral power and aperiodic slope changes with ageing. Band-limited spectral power is examined within delta–theta (2–7 Hz), alpha (8–12 Hz) and beta (13–30 Hz) frequency bands and within each tercile bin of the age distribution in the full cohort, which included the age bins of 50–64 years (n = 21), 65–74 years (n = 29) and 75–90 years (n = 20). The brain renderings depict the average spectral power as a percentage (%) in a regional analysis for the cortical regions defined in the Brainnetome atlas in each cohort. When considering the pattern of age across the three tercile bins, delta–theta showed a reduction of spectral power (A), alpha showed a slight increase in spectral power, especially from 50–64 range to 65–74 range (B), while beta showed a reduction of spectral power (C). The colour bars are scaled from minimal to maximal spectral power % values within each frequency band as shown. Spectral aperiodic slope showed a reduction from 50–65 to 75–90, especially over the frontal and inferolateral parietal regions (D). An LMM analysis examining the changes across the three tercile bins, within band-limited spectral power for delta–theta (2–7 Hz), alpha (8–12 Hz) and beta (13–30 Hz), showed region dependency of these changes, where delta–theta reductions are predominantly frontal (E), alpha increases are predominantly temporal (F) and beta reductions are predominantly frontal (G). Spectral aperiodic slope changes were widely distributed with inferolateral temporal and cingulate cortices showing the greatest reductions with age (H). The images in E–H show significant regions after thresholded at FDR 5%. SP, spectral power.

Figure 3

Associations between spectral changes and cognitive performance in healthy ageing: a median split analysis. A median split analysis identified above-median subgroup (n = 10) and below-median subgroup (n = 30) based on the performance on memory, executive and processing speed functions (A). This analysis included the subset of individuals (n = 40) who underwent uniform cognitive assessments memory, executive and processing speed functions. A ROI-based LMM analysis that examined the specific spectral measure between the two groups from median split. For band-limited frequency measures, above-median subgroup has lower delta–theta (t = −12,0, P < 0.0001), higher alpha (t = 3.91, P < 0.0001) and higher beta (t = 15.93, P < 0.0001) spectral power compared with the below-median subgroup (B). A similar LMM showed that above-median subgroup has higher aperiodic slopes (t = 24.70, P < 0.0001) compared with below-median agers (B). An LMM analysis examining the spatial patterns of changes between above- and below-median subgroups, within band-limited spectral power for delta–theta (2–7 Hz), alpha (8–12 Hz) and beta (13–30 Hz) and aperiodic slope, showed region dependency of these changes, where delta–theta reductions were frontally distributed (C), alpha increases were temporo-occipital (D) and beta reductions were widespread but most predominant in the frontal lobes (E). Aperiodic slope increases in above-median subgroup was widely distributed involving all regions of the cortex (F). The images show significant regions in the group contrast analysis between below-median (n = 30) and above-median (n = 10) subgroups and after thresholding at FDR 5%.

Figure 4

Associations between spectral changes and cognitive performance in healthy ageing: a PLSR analysis. A PLSR with two LDs combining the spectral changes and cognitive function in executive, memory and processing speed domains, as well as age, demonstrated the associations between neural and cognitive measures in the latent space depicted by LD1 versus LD2 for band-limited spectral power (A) and for aperiodic slope (B). In the three panels shown in subplot (A), each dot indicates a subject (n = 40) and the colour scale from green shades to red shades indicates low to high spectral power within each frequency band of delta–theta, alpha and beta, respectively, from left to right. Lower delta–theta spectral power and higher cognitive scores were associated with LD1 (left panel), higher alpha spectral power and higher cognitive scores were associated with LD2 (middle panel) and higher beta spectral power and higher cognitive scores were associated with LD1 (right panel), in each respective PLSR model. In subplot (B), each dot indicates a subject (n = 40) and the colour scale from light to dark indicates low to high aperiodic slope. Lower aperiodic slope values were associated with LD1, and higher cognitive scores were associated with LD2. Subplots (C)–(F) depict the regional patterns of factor loadings: first LD derived from PLSR of band-limited spectral power measures (C); first LD derived from PLSR of aperiodic slope (D); second LD derived from PLSR of band-limited spectral power measures (E); second LD derived from PLSR of aperiodic slope (F). SP, spectral power.