The interaction between endogenous GABA, functional connectivity, and behavioral flexibility is critically altered with advanced age

Figures

• 

  Fig. 1. Experimental procedures and parameters of interest. a Study outline with MRI/MRS (session 1), task familiarization (session 2) preceding the main experiment (session 3) including EEG during task performance. Edited MRS and T1-weighted images were used to extract tissue-corrected GABA levels and additional macromolecules (GABA+) from left and right sensorimotor and occipital voxels. The behavioral paradigm involved transitions between a stable (mirror-symmetric in-phase tapping of both index or middle fingers synchronously) and a less stable (anti-phase tapping, i.e., contralateral index and middle finger synchronously) motor state. Task familiarization included stimulus-response mapping and individual performance frequency adjustment. Performance in motor-state transitions was described with transition latency and error rate. The EEG signal was projected into source space based on the centroid coordinates of the GABA voxels. Phase angles were computed based on spectrally decomposed (Morlet wavelet transform) source time series. Phase angle differences between source signal pairs were used to compute connectivity (inter-site phase clustering, ISPC) between cortical sources. Phase…

• 

  Fig. 2. GABA+ spectroscopy results. a Sum of individual GABA voxels projected into MNI space overlaid on standard brain template. Color coding indicates overlay agreement in percentage of all available images within group. Neurological display (i.e., coronal and axial view with left side on the left and right side on the right of image). b Individual edited spectra for LEFT S/M1, RIGHT S/M1, and OCC voxel color-coded for older (blue) and young (yellow) participants. Darker lines present average spectra per group (orange—young, dark blue—older). c Boxplots (lower/upper whiskers represent smallest/largest observation greater than or equal to lower hinge ± 1.5 * inter-quartile range (IQR), lower/upper hinge reflects 25%/75% quantile, the lower edge of notch = median − 1.58 * IQR/ sqrt(n), middle of notch reflects group median) and distributions shown for the interaction effect of group and voxel on GABA+, which is driven by the differences between the occipital voxel and both sensorimotor voxels within the older in addition to the between age group differences for both sensorimotor voxels. Asterisks indicate significant effects of model-derived marginal mean contrasts corrected for multiple…

• 

  Fig. 3. Predictors for the behavioral outcome. a Error rate. Boxplots and distributions for overall error rate given separately for transition modes (in-phase, anti-phase) and age groups (blue: older, yellow: young). b Transition Latency. Color coding as in (a). c Effect of practice, i.e., number of trials (depicted as centered variable), are given for failed transitions, fully correct transitions, and cumulative error rate (from left to right), and transition latency (failed transitions excluded). Brown indicates transitions into in-phase mode, light pink depicting transitions into anti-phase. Frames around graphs indicate relevant modulation of the outcome over the number of trials, i.e., for failed transitions, cumulative error rate, and transition latency. Only in the case of failed transitions, older showed a significantly different modulation over time for transitions into anti-phase compared to the young with initially higher rate of trials with 100% error rate. Cumulative error rate showed a comparable increase across trials while transition latency decreased comparably in the two age groups and for both transition modes. d For both groups and transitions modes (into in-phase,…

• 

  Fig. 4. Statistical results of ISPC between left and right S/M1 source. a For group contrast time-locked to the individual mean transition time. Cluster-corrected z maps for the test of GROUP contrast [YOUNG–OLDER, t-test against 0, p < 0.05, 2-tailed]. Color coding in the time-frequency resolved zISPC plot indicates t-values. Dashed vertical lines at 0 ms on the time axis indicate the individual median latency, i.e., the time of transition. b Statistical results for transition mode contrast time-locked to the individual mean transition time. Inlays show bar plots, which represent group averages (±SEM) of zISPC for respective cluster ranges indicated by capital letters for (a) group and (b) transition mode contrasts. Scatter plots depict individual participants’ data for the group (yellow—young, blue—older) and transition mode (brown—in-phase, light rosé—anti-phase), respectively. Cluster-corrected z maps for the test of TRANSITION MODE contrasts [into in-phase–into anti-phase, t-test against 0, p < 0.05 2-tailed]. Bar plots show transition mode averages (±SEM) for respective cluster ranges. Phase-related connectivity (ISPC) data was analyzed in N = 22 older and N = 20 young participants.

• 

  Fig. 5. Association between cortico-cortical phase angle differences at the time of transition and subsequent performance error. a Single-trial data shown for low beta [15–22 Hz] range. b Single-trial data shown for high beta range [25–30 Hz]. Data points represent single-trial data for transitions into in-phase (brown) and into anti-phase (light pink) mode, solid line indicates average phase angle difference—behavior association during transitions into in-phase mode, dashed line indicates average phase angle difference—behavior association for transitions into anti-phase mode. c Mean phase angle differences in the low beta frequency band were significantly modulated by factors age group and relatively higher versus lower GABA+ concentration when binarized into above group median (dark purple, solid lines represents subsample mean) versus below group median (yellow shading, dashed lines for subsample mean). d Mean phase angle differences in the high beta frequency band show a comparable pattern as in the low beta band. Rose plots show the histogram of binned phase angle differences with mean direction (red line) and 95% CI (black circumference) for significant non-uniformity of…

• 

  Fig. 6. Results of Bayesian moderated mediation models. a Overview over posterior directions (pd) for indirect (mediation) effects on error rate and transition latency conditional on upper/lower quintiles of moderator age (depicted as YOUNG and OLDER) for the models estimated with the independent variable (IV) based on the three time × frequency clusters derived in the response locked ISPC analysis. Models were run separately for left and right S/M1 GABA+ as mediator. The pd can be interpreted as the maximum probability of the effect accounting for the evidence derived from the data. Color coding of pd represents likelihood (in %) and direction of effect, i.e., red shading for positive effects and blue shading for negative effects. A pd of 95, 97.5, 99.5, and 99.95% corresponds to the frequentist two-sided p-value at the thresholds 0.1˚, 0.05*, 0.01**, 0.001*** respectively. See respective Methods section for further information about effect descriptors and measures of uncertainty used in Bayesian statistics. b Probability density plots for effects of parameters in the outcome and the mediator models input to the mediation analysis. Depicted are the three models with significant…

• 

  Fig. 7. Results of Bayesian moderated mediation models. From left to right, total effect (pathτ) conditional on moderator age, association between connectivity and GABA+ (path α), association between GABA+ and transition latency (path β), simulation of the mediation effect on the connectivity—behavior association (direct effect, path τ’) for varying levels of GABA+ (low—yellow, medium—red, high—dark purple), and probability density ploy for the mediation effect conditional on age including the difference of young versus older [Y-O] mediation effects. In all three models, right hemispheric GABA+ concentration is modeled as mediator. a Model for peri-transition low beta connectivity—error rate association. As visible in the total effect, the young behaviorally benefit from stronger S/M1-S/M1 connectivity in the low beta range, while the older show higher errors with stronger connectivity. The association between connectivity and GABA+ (path α) shows opposing trends in the two age groups, a negative association in the young and a positive association in the older. The association between GABA+ and transition latency (path β) is positive in the young and negative in the older. Simulations…