Similar neural pathways link psychological stress and brain-age in health and multiple sclerosis

Figures

• 

  Figure 1. fMRI stress paradigm The total paradigm comprised five stages. The stress stage was divided in stages 4a (“Evaluation”) and 4b (“Feedback”). In 4a, participants performed subtraction tasks across a series of trials. In each, they had to subtract operand Y from X as fast as possible by selecting the correct answer among a set of four options depicted on a screen with an MRI-compatible response box. X started with 43521. In all trials in 4a (and 4b), operand Y was randomly determined (range: 1–99). In the first project, 4a either lasted until a participant provided ten correct answers or 4 min if they failed to provide ten correct answers. In the second, 4a had a fixed duration of 4 min. To compensate, we only evaluated rCBF, heart rate, and mental arithmetic performance data acquired in the final 8 min of 4b and modeled the duration of 4a as covariate of no interest (CNI) “time-to-feedback” in respective regression analyses in this study. In 4a, the time provided to solve a task was 8 s. When a trial was correctly solved, X equaled the difference X minus Y from the preceding trial. X remained unchanged in case of a false or too slow answer. After 4a, 4b started immediately. 4b…

• 

  Figure 2. Within-participant stress responsivity and functional connectivity modelling (A) illustrates the computation of group masks for gray matter (GM), white matter (WM), cerebrospinal fluid (CSF), and lesions used in the fMRI analyses to constrain them to GM areas and to determine CNI. The masks were derived from voxel-wise tissue probability maps computed in the MNI standard space and manually determined lesion masks coregistered to the standard space. GM, WM and CSF tissue probability maps were first averaged across participants of a group and each voxel was classified to the tissue with the maximal averaged probability to give binary group masks for GM, WM, and CSF. For HPs, these masks were used in the fMRI analyses. In PwMS, a lesion group map was additionally computed denoting the sum of lesions per voxel across patients and coordinates with at least one lesion were classified as lesion voxel in a binary lesion mask. The binary GM, WM, and CSF group masks finally used in PwMS fMRI analyses were then determined by removing all lesion coordinates from the binary GM, WM and CSF masks determined in the step before. Importantly, however, to account for lesions in a highly sensitive…

• 

  Figure 3. Neural stress responses (A–C) shows neural stress responses. The bar graph in 3A shows the t-statistics for the difference in rCBF for stress (final 8 min of stage 4b) minus rest (stage 2) separately for both groups. Labels depicted above the baseline of 3A highlight the 17 regions yielding significant rCBF differences in HPs, labels depicted below the 14 in PwMS. Please note, that the coarse clustering of regions in 3A into frontal, temporal, etc. was conducted to increase the comprehensibility of the graph and is not necessarily optimal in anatomical terms. The render brains in 3B (HPs) and 3C (PwMS) summarize regions with significant stress responses. The scalebars in (B and C) provide the color coding for significant t-statistics obtained in the analysis of regional neural stress responsivity.

• 

  Figure 4. Functional connectivity and brain-PAD (A) Shows stress-responsive regions identified in main analysis 1 to ease interpretability of this figure. (B) The heatmaps in (B) depict the t-statistics for the association between FC of a given pair of stress-responsive regions on one hand and brain-PAD on the other. Specifically, the left half of (B) shows these t-statistics for HPs, the right half for PwMS. The top row shows associations for rest, the bottom row for those for stress. In the left half of (B), we illustrate (from left to right) the connection between (the center coordinates) of anterior insula and occipital regions predictive of brain-PAD and a scatter-diagram of the association between stress-related anterior insula – occipital FC and brain-PAD in HPs. In the right half of (B), we depict the same for PwMS. The scalebar in (B) provides the color coding for t-statistics obtained in the analysis of associations between FC and brain-PAD. (C) Highlights the two regions whose FC during stress shows the strongest connection with brain-PAD separately for each group as well as the FC-brain-PAD connection in terms of a scatter-graph. The p-values reported in (C) reflect the…

• 

  Figure 5. Regional brain aging (A) illustrates the brain-age prediction approach or how the machine learning algorithm predicts the brain-age for six selected participants respectively. Specifically, it determines a score for each voxel which individually biases the algorithm slightly to a younger or an older age and which can be understood as marker of regional brain aging. The (weighted) sum of voxel scores gives the brain-age of a person. Voxels highlighted in blue correspond to the 5% of all voxels that contribute most strongly to a younger brain-age prediction, orange ones the 5% contributing most strongly to an older prediction. The scalebar in (A) provides the color coding for the strength of these biases. (B) shows associations between voxel-wise brain-age scores and the whole-brain GM fraction for HPs and PwMS. In the upper row of 5B, superimposed on the axial and sagittal slices, we show significant voxel-wise t-statistics found for the link between voxel brain-age scores and GM fraction for HPs (orange palette) and for PwMS (blue palette) in independent analyses. The black arrowheads highlight the peak coordinate for which a significant negative overlapping association was found…

• 

  Figure 6. Functional connectivity and future brain-PAD in PwMS (A) Shows stress-responsive regions identified in main analysis 1 to ease interpretability of this figure. (B) The heatmaps in (B) depict the t-statistics for the association between FC at the baseline visit of pairs of stress-responsive regions and Δbrain-PAD on the left for rest and on the right for stress. The scalebar in (B) provides the color coding for t-statistics computed in the analysis of associations between FC and Δbrain-PAD. (C) Highlights (the connection between the center coordinates of) right middle frontal cortex right and cerebellar vermal lobule VI – VII, the two regions whose stress-stage FC showed the strongest connection with Δbrain-PAD in main analysis 4. (D) Presents a scatter-diagram of this association. The p-value reported in (D) reflects the type-I-error-rate for the strongest association between FC and Δbrain-PAD on an FWE-corrected level (pFWE) and its uncorrected equivalent on a single test-level (puncorr).

•