Transformations in oscillatory activity and evoked responses in primary somatosensory cortex in middle age: a combined computational neural modeling and MEG study

Abstract

Oscillatory brain rhythms and evoked responses are widely believed to impact cognition, but relatively little is known about how these measures are affected by healthy aging. The present study used MEG to examine age-related changes in spontaneous oscillations and tactile evoked responses in primary somatosensory cortex (SI) in healthy young (YA) and middle-aged (MA) adults. To make specific predictions about neurophysiological changes that mediate age-related MEG changes, we applied a biophysically realistic model of SI that accurately reproduces SI MEG mu rhythms, containing alpha (7-14 Hz) and beta (15-30 Hz) components, and evoked responses. Analyses of MEG data revealed a significant increase in prestimulus mu power in SI, driven predominately by greater mu-beta dominance, and a larger and delayed M70 peak in the SI evoked response in MA. Previous analysis with our computational model showed that the SI mu rhythm could be reproduced with a stochastic sequence of rhythmic approximately 10 Hz feedforward (FF) input to the granular layers of SI (representative of lemniscal thalamic input) followed nearly simultaneously by approximately 10 Hz feedback (FB) input to the supragranular layers (representative of input from high order cortical or non-specific thalamic sources) (Jones et al., 2009). In the present study, the model further predicted that the rhythmic FF and FB inputs become stronger with age. Further, the FB input is predicted to arrive more synchronously to SI on each cycle of the 10 Hz input in MA. The simulated neurophysiological changes are sufficient to account for the age-related differences in both prestimulus mu rhythms and evoked responses. Thus, the model predicts that a single set of neurophysiological changes intimately links these age-related changes in neural dynamics.

Figure 1. Age-related changes in MEG SI mu rhythms

A, Left panels Regression plots of prestimulus power in the mu (7-29 Hz, top) and mu-beta (15-29 Hz, bottom) ranges calculated from TFR spectrograms as a function of age. Significant correlations (p < 0.05) are indicated by solid regression lines. Right panels. Bar plots of mean and s.e. for time-averaged TFR power (1 sec, n=200 trials) in each frequency range for the YA (red) and MA (blue). B. Average prestimulus TFRs for the YA (top) and MA (bottom). The unit of power is (Am)². C. Average prestimulus power spectral density plots for the YA (blue) and MA (red), s.e. bars. D. Mean (+ s.e.) TFR ratios of mu-alpha to mu-beta power for the YA (red) and MA (blue). The TFR ratio was > 1 for YA and < 1 for MA (p = 0.07).

Figure 2. Symmetry and amplitude of prestimulus MEG oscillations

A. Histogram of symmetry indices for the prestimulus (1000 ms) period over all trials for the YA (red) and MA (blue). B. Histogram of peak-to-trough amplitudes during the prestimulus period for the YA and MA. C. Mean prestimulus peak-to-trough amplitudes for each group (s.e. bars). Amplitudes were significantly greater for the MA (p = 0.017) D. Regression plot of mean amplitudes by age (p = 0.01). E. Example single-trial prestimulus waveforms (left) and corresponding TFR plots (right) for one YA and one MA showing the single trial variability and symmetry of the oscillation around zero in both groups with greater amplitude oscillations in the MA. The unit of power in the TFR is (Am)².

Figure 3. Age-related difference in the early MEG SI evoked response

Evoked SI ECD responses for the first 175 ms after stimulation for the YA (red) and MA (blue) (shaded region depicts s.e.). The MA group showed a significantly greater magnitude M70 peak (p = 0.01), longer latency of this peak (p = 0.03), and trends toward a greater ∼50ms response (p = 0.06) and slope from the M70 to 100 ms response (p < 0.09).

Figure 4. The effect of the delay between the rhythmic 10Hz FF and FB inputs on the simulated SI mu rhythm

A. A 5 ms FF-FB delay shows that both mu-alpha and mu-beta components emerge in the TFR. The unit of power is (Am)^2. B. A 50 ms FF-FB delay shows that only mu-alpha emerges significantly in the TFR. C. Corresponding waveform for 5 ms delay. D. Corresponding waveform for 50 ms delay. E. The average mu-alpha and mu-beta power calculated from the TFR are approximately equal with a 5 ms delay. F. The average mu-alpha power is greater than mu-beta with a 50 ms delay.

Figure 5. The effect of the delay between the rhythmic 10Hz FF and FB inputs on symmetry indices and TFR mu-alpha to mu-beta ratios

Increasing the FF-FB delay from 5 ms to 50 ms increased the TFR mu-alpha to mu-beta ratio from slightly <1 to a value >1. The increase to a 50ms delay, however, produced a nearly 5-fold increase in the TFR ratio, which was not consistent with changes from YA to MA (see Figure 1D), thus negating model prediction A.

Figure 6. The effect of the strength of the 10Hz FF and FB input on symmetry indices and TFR mu-alpha to mu-beta ratios

For a fixed 5 ms delay between the rhythmic FF and FB input we manipulated the strength of the inputs via changes in the (A) post-synaptic conductance, (B) number of spike in the input burst on each cycle of the input, and (C) variance of spikes in the input burst on each cycle of the input (see Supplementary Figure 3). A. Increasing the post-synaptic conductances of the rhythmic FB input (left), while holding FF post-synaptic conductances constant, resulted in a decrease in the symmetry indices (top) and a decrease in the TFR mu-alpha to mu-beta ratio (bottom). The opposite pattern was found when the post-synaptic conductances of the FF input was increased, while holding FB conductances constant (right). B. Increasing the number of inputs on each cycle of the rhythmic FB input (left), while holding FF number constant, led to a decrease in symmetry indices and a decrease in the TFR ratio; the opposite pattern was found when the number of FF inputs were increased (right), while holding FB number constant. C. Increasing the variance of the inputs on each cycle the rhythmic FB input (left), while holding the FF variance constant, had a minimal effect on the symmetry indices, but resulted in a relative decrease in the TFR ratios. Increasing the variance of the rhythmic FF input (right), while holding the FB variance constant, again resulted in a relatively constant symmetry index, but led to an increase the TFR ratio.

Figure 7. Simulating age-differences in the SI mu rhythm

Using the parameter regime outlined in Table 1, the computational model was able to reproduce each of the observed group comparison age-related characteristics of the MEG data: When averaging over multiple trials (n=25 each group), the model reproduced A) larger mean amplitudes of oscillations in the MA model data, and B) a TFR mu-alpha to mu-beta ratio >1 for YA and <1 for MA. C. Average power spectral density plots for the modeled YA (blue) and MA (red) data. D. Time-averaged TFR power estimates were significantly larger in the MA simulations in mu and mu-beta range (p<0.001). E. Individual trial waveforms and spectrograms showing single trial variability and symmetric oscillations, with increased amplitude oscillations and greater beta dominance in the simulated MA compared to YA model data (bottom). Panels A-D depict mean and s.e. across trials.

Figure 8. Simulating age-differences in the SI evoked response

Simulating an evoked response sequence with FF-FB-LFF input, as described in the results, during ongoing YA and MA mu rhythms reproduced the age-differences observed in the MEG evoked response, including a greater magnitude M70 peak (p = 0.01), a greater slope from the M70 to 100ms response (p < .0001) and a trend toward a decreased M50 response (p = 0.08) in the MA simulation. Mean and s.e. shown over n=30 trials per group.