Developmental trajectory of transmission speed in the human brain

The structure of the human connectome develops from childhood throughout adolescence to middle age, but how these structural changes affect the speed of neuronal signaling is not well described. In 74 subjects, we measured the latency of cortico-cortical evoked responses across association and U-fibers and calculated their corresponding transmission speeds. Decreases in conduction delays until at least 30 years show that the speed of neuronal communication develops well into adulthood.


Supplementary Figure 1. Similar relation between age (years) and N1 latency (ms) at different stimulation currents.
To ensure that the relation between age and latency was not driven by the fact that some electrode pairs were stimulated with a current of 8 mA, while others had a current of 4, 6 or 7 mA, we calculated the average latency per subject across all connected electrodes. The relation between age and latency across all subjects shows a significant (two-sided Spearman's ρ, P FDR < .05) negative relation between age and latency for stimulation with both 4 mA (P FDR =.001) and 8 mA (P FDR <.001). 6 mA and 7 mA had too few samples for a robust fit.

Supplementary Figure 2. Relation between age and relative number of N1s in long-range connections.
Age (in years, x-axis) versus the relative number of N1s as a proportion of the total number of possible CCEPs within each connection (y-axis). Each dot represents one subject. The relative number of N1s per subject is determined by calculating -for each stimulus-pair with electrodes on the stimulus ROI -the ratio of N1s per the number of electrodes within the response ROI, and averaging over all these ratios. No significant relations between age and the relative number of CCEPs were observed (two-sided Spearman's ρ, P FDR < .05). .55,.47,.74,.88,.051,.051 and .051. Figure 3. Relation between age and variance in N1 peak latency in long-range connections. Age (in years, x-axis) versus the variance in N1 latency (in ms, y-axis). For each subject, we calculated the variance in latencies across connections in the same fiber pathway and tested whether increased variance relates to age. Each dot represents one subject. No significant relations between age and variance in latency were observed (two-sided Spearman's ρ, P FDR < .05). The FDR corrected P-values (from left-to-right and top-to-bottom) are: .19, .39, .23, .23, .05, .39, .39 and .54. Figure 4. Relation between mean N1 peak latency and variance in N1 peak latency in long-range connections. For each subject, we calculate the variance in N1 peak latencies across connections in the same fiber pathway and test whether increased latencies (in ms, x-axis) also have more variability (in ms, y-axis). Each dot represents the mean and variance in one subject. In 4 out of 8 long-range connections we observe a positive relation between the mean and variance in N1 latencies (two-sided Spearman's ρ, P FDR < .05, red asterisk indicates significance). Showing that increased latency often relates to more variability. The FDR corrected P-values (from left-to-right and top-to-bottom) are: .13, .10, <.001, <.01, <.01, .27, <.001 and .13

Supplementary
Supplementary Figure 5. Relation between mean N1 latency and N1 peak width. The relative temporal synchrony of the arriving signals in a measured electrode could be reflected in the width of the evoked potential. We therefore calculate the full width half max of the N1 peak, where the amplitude is 50% of the N1-peak amplitude (in ms, y-axis) and test whether this is related to the latency (in ms, x-axis). In all of the above connections, we observe a positive relation between the latency and the half N1 peak width (two-sided Spearman's ρ, P FDR < .05, red asterisk indicates significance). The FDR corrected P-values (from left-to-right and top-to-bottom) are: <.001, <.001, <.001, <.01, <.01, <.01, <.001 and <.001. Figure 6. The latency of N1 peaks measured on seizure onset zones (SOZ) or non-SOZs when stimulating elsewhere. For n=26 subjects this graph shows a box plot of the N1 latencies for electrodes measuring from the SOZ or non-SOZ areas. The central mark indicates the median latency, the bottom and top edges indicate 25th and 75th percentiles and the whiskers extend to extreme points excluding outliers (1.5 times more or less than the interquartile range). The asterisks display the subjects in whom a significant difference is found between latencies measured in SOZs and non-SOZs (two-sided Mann Whitney U test, P FDR < .05). In 9 subjects, the latency is significantly increased in the SOZ. In 2 subjects, the latency is significantly decreased in the SOZ. The FDR corrected P-values for the tests between SOZ and non-SOZ (from left-to-right) are: <.001, .08, .87, .11, .40, .01, .02, .11, .10, <.01, .31, .10, <.01, <.01, .34, .11, <.001, <.01, .04, <.001, .01, .21, .33, .11, .23, and .23.

Supplementary Figure 7. The latency of N1 peaks in other electrodes when stimulating electrodes on seizure onset zones (SOZ) or on other areas (non-SOZs).
For n=26 subjects this graph shows a box plot of N1 latencies after stimulating SOZ and non-SOZ. The central mark indicates the median, the bottom and top edges indicate 25th and 75th percentiles and the whiskers extend to extreme points excluding outliers (1.5 times more or less than the interquartile range). The asterisks display the subjects in whom a significant difference is found between latencies when stimulating SOZs and non-SOZs (two-sided Mann Whitney U test, P FDR < .05). In 8 subjects, we find a significant increase in latency in response electrodes when the SOZ is stimulated. In 3 subjects, we find a significant decrease in latency in response electrodes when the SOZ is stimulated. The FDR corrected P-values for the tests between SOZ and non-SOZ For each of the 74 subjects and stimulated electrode pair, we detected the N1 latencies and calculated the standard deviation across the N1 latencies when more than five N1 responses were detected (as with fewer responses, a standard deviation would not be robust). Each dot in a row represents a stimulated electrode pair in that subject and the x-axis indicates the standard deviation across the N1 latencies. If there would be volume conduction, N1 latency would be equal in all responses and the standard deviation would be zero, which is not the general case for these data.

Supplementary Figure 9. Electrode distributions across ages for each tract and end-point region of interests (ROIs).
The blue graphs show a histogram of the number of electrodes on top of the endpoint ROIs (y-axis) for each 10 years (x-axis). The yellow graph shows a histogram of the total number of participants for each 10 years (x-axis). The distribution of electrodes over age follows the distribution of subjects over age, implying that the electrodes are relatively uniformly distributed across the age ranges. Electrodes were assigned to several brain regions based on the label from the Destrieux atlas in Freesurfer.