Experience-dependent modulation of the visual evoked potential: testing effect sizes, retention over time, and associations with age in 415 healthy individuals

Experience-dependent modulation of the visual evoked potential (VEP) is a promising proxy measure of synaptic plasticity in the cerebral cortex. However, existing studies are limited by small to moderate sample sizes as well as by considerable variability in how VEP modulation is quantified. In the present study, we used a large sample (n = 415) of healthy volunteers to compare different quantifications of VEP modulation with regards to effect sizes and retention of the modulation effect over time. We observed significant modulation for VEP components C1 (Cohen’s d = 0.53), P1 (d = 0.66), N1 (d = −0.27), N1b (d = −0.66), but not P2 (p = 0.1), and in one time-frequency cluster (~30 Hz and ~70 ms post-stimulus; d = −0.48), 2-4 minutes after 2 Hz prolonged visual stimulation. For components N1 (d = −0.21) and N1b (d = −0.38), as well for the time-frequency cluster (d = −0.33), this effect was retained after 54-56 minutes. Moderate to high correlations (ρ = [0.39, 0.69]) between modulation at different postintervention blocks revealed a relatively high temporal stability in the modulation effect for each VEP component. However, different VEP components also showed markedly different temporal retention patterns. Finally, P1 modulation correlated positively with age (t = 5.26), and was larger for female participants (t = 3.91), with no effects of either age or sex on N1 and N1b potentiation. These results provide strong support for VEP modulation, and especially N1b modulation, as a robust measure of synaptic plasticity, but underscore the need to differentiate between components, and to control for demographic confounders.


Introduction 51
Due to the essential role of synaptic plasticity in learning and memory (Takeuchi,  In a standard VEP modulation paradigm, subjects are exposed first to reversing 69 checkerboard or grating stimuli which elicit VEPs, then to a prolonged (e.g. Normann 70 et al., 2007) or high-frequency version (e.g. Teyler et al., 2005) of the same stimulus, 71 and lastly, after some delay, to the initial stimulation again, which now typically 72 evokes a slightly modulated visual potential. Importantly, the mechanisms underlying 73 such experience-dependent VEP modulation seem to share many characteristics 74 with LTP, thus having earned the placeholder epithet LTP-like plasticity. In mice, both 75 of attention afforded the visual stimulus, especially during high frequency or 126 prolonged visual stimulation. Attention levels might be indexed by visual stimulation-127 driven steady state responses (Çavuş et al., 2012), or inversely by power in the alpha 128 range (8-13 Hz) ( to consider include changes from baseline to postintervention amplitudes in the C1, 140 P1, N1, N1b, and P2 components, as well as in the peak to peak difference P1-N1. 141 Furthermore, as the largest effects are not necessarily phase-locked, time-frequency 142 analyses of the post-stimulus EEG should be employed to complement time-domain 143 analyses. Since these components have not been directly compared in a large 144 sample of healthy individuals, it is currently unknown which of the many potential 145 indices of LTP-like synaptic plasticity is most sensitive and robust. Typical sample 146 sizes within the field might make some studies vulnerable to winner's curse and 147 random effects (Ioannidis, 2008 blocks, were presented on a 24 inch 144Hz AOC LCD screen with 1 ms grey-to-grey 171 response time (Fig. 1). All blocks, including the intervention block, consisted of a 172 reversing checkerboard pattern with a spatial frequency of 1 cycle/degree over a 173 ~28° visual angle. The reversal frequency was fixed at 2 reversals per second for the 174 intervention block, whereas the baseline and postintervention blocks had jittered 175 stimulus onset asynchronies of 500-1500 ms (mean = 1000 ms). All baseline and 176 postintervention blocks lasted ~40 seconds (i.e., 40 checkerboard reversals), while 177 the stimulation block lasted 10 minutes (i.e., 1200 reversals). Postintervention blocks 178 were presented at 2 min, 3 min 40 s, 6 min 20 s, 8 min, ~30 min, ~32 min, ~54 min, 179 and ~56 min after the intervention block. Through all blocks, the participants fixed 180 their gaze on a red dot in the centre of the screen, and pressed a key on a gaming 181 controller when its color changed from red to green. Between the seventh and eight, 182 and between the ninth and tenth blocks, participants underwent mismatch negativity 183 (Näätänen, Gaillard, & Mäntysalo, 1978) and prepulse inhibition (Graham & Murray,184 1977) tasks, respectively.  All channels were subjected to group-level time domain analysis, and the channel 237 with highest amplitudes and most pronounced VEP modulation (i.e., Oz) was 238 selected for all later analyses (Fig. 3). form. Alpha levels were adjusted to control for multiple comparisons according to the 283 effective number of independent comparisons, derived using eigenvalues of the 284 correlation matrix of the entire continuous data set (Li & Ji, 2005), yielding an 285 experiment-wide significance threshold at 0.0009. Regression models were fitted 286 using the general linear model, while controlling for baseline amplitudes, model fit is 287 indexed using Nagelkerke R 2 , and effect is expressed with t-values. 288

Results 289
The checkerboard reversal stimulation evoked the expected C1, P1, N1, and P2 290 components of the VEP (Fig. 2; see Table 2 for latencies and amplitudes). Initial 291 group level analyses demonstrated that, across VEP components, the highest 292 amplitudes and the largest modulation effects were exhibited at the occipital Oz 293 electrode ( Fig. 3A-B), which was accordingly selected for individual level analyses.

344
There were also differences between component amplitudes within assessments 345

357
The time-frequency analysis exploring the main effect of prolonged visual stimulation 358 yielded five significant clusters (Fig. 6). Results from analyses across assessments 359 using individual participants' values averaged within clusters are presented in Table  360 4. Notably, these revealed that only the first cluster exhibited modulation at all (t = -6.57, p = 1.7 x 10 -10 ), and P1 modulation (t = 6.43, p = 3.8 x 10 -10 ).

377
The regression model for P1 modulation (R 2 = 0.15), revealed effects of age (t = 378 5.26, p = 1.6 x 10 -7 ) and sex (t = 3.91, p = 9.7 x 10 -5 ), with greater modulation for 379 older participants and female participants, respectively. The regression model for P2 380 modulation (R 2 = 0.09) also showed an increased difference from baseline to 381 postintervention blocks for female participants (t = 5.08, p = 4.3 x 10 -7 ). The Regression models for N1 (R 2 = 0.04) and N1b (R 2 = 0.07) modulation did not 388 provide evidence for effects of age, sex, intervention block alpha power, or 389 intervention steady state power. Finally, for the attentional task, we only obtained hit 390 rate data for 45.8% of participants, due to error in the gaming controller. Thus, we 391 performed a set of control analyses to ensure that the participants for which 392 attentional data was not obtained did not differ from the participants for which 393 attentional data was obtained. These showed that there was no difference between 394 these groups in P1, N1, N1b, or P2 modulation, but only a nominal difference in C1 395 modulation (p = 0.04), and that clear VEPs were evoked for 96% of participants for 396 which attentional data was not obtained. Among participants for which attentional 397 data was obtained, the mean hit rate was 98.4%. Together, these results indicate 398 overall satisfying levels of attention. increases in the third and last postintervention assessments ( Fig. 4; Fig. 5 2017), constitutes a clear exception, and appears inconsistent with NMDAR-453 dependent LTP, which exhibits a gradual decay (Citri & Malenka, 2008). Along the 454 same lines, the P2 component appears to lack input specificity . 455 Thus, the effect of time on P2 amplitudes might seem to require some other 456 mechanism than LTP-like synaptic plasticity. On the other hand, the retention slope 457 of P1 is consistent with synaptic plasticity as underlying mechanism, although with a 458 complete decay between 6 and 30 minutes after prolonged visual stimulation, P1 459 modulation might reflect some short-term plasticity such as post-tetanic potentiation 460 (Citri & Malenka, 2008). previously observed in older participants (Spriggs et al., 2017), and with the more 469 general decline in neural plasticity associated with aging (Burke & Barnes, 2006). Further, regression models demonstrated larger P1 modulation, and larger increase 471 in P2 amplitudes, among female participants, a result that -like the effects of age -472 was independent of baseline amplitudes. Together, these results underscore the 473 need to differentiate between VEP components, and to control for demographic 474 variables like age and sex, especially in case-control studies of VEP modulation. power at ~70 ms post-stimulus. Moreover, we observed differential retention slopes, 516 effect sizes, and associations to age and sex for the modulation of VEP components, 517 strongly suggesting that VEP modulation is not a unitary phenomenon. Taken 518 together with results from a series of invasive studies in rodents, our current results 519 support the use of prolonged visual stimulation induced VEP modulation, and 520 especially N1b modulation, as a robust, non-invasive index of LTP-like cortical 521 plasticity in humans. 522 Tables  666  667  Table 1