Age-related alterations in the modulation of intracortical inhibition during stopping of actions

Behavioral assessment pre-TMS

The P(inhibit) was close to 50% and did not differ between young and older adults (F(1, 48) = 0.325, p = .571, η_p² = .007), allowing reliable estimation of the stop-signal reaction time (SSRT) with the integration method. There was a small (<1%), yet significant difference in P(inhibit) between the two sessions (F(1, 48) = 4.406, p = .041, η_p² = .084). The group by session interaction was not significant (F(1, 48) = 1.413, p = .240, η_p² = .029).

Older adults had significantly longer SSRTs compared to young adults (F(1, 48) = 8.854, p = .005, η_p² = .156). There was no main effect of session (F(1, 48) = 0.147, p = .703, η_p² = .003), and no interaction between age group and session (F(1, 48) = 0.755, p = .389, η_p² = .015). There was a trend for shorter go response times (GoRTs) in young compared to older adults (F(1, 48) = 3.607, p = .064, η_p² = .070). Furthermore, GoRTs significantly increased as a function of stop-signal probability (SSP, F(1.736, 83.338) = 93.037, p < .001, η_p² = .660) to a similar degree in young and older adults (F(1.736, 83.338) = 1.362, p = .260, η_p² = .028). Post hoc comparisons indicated that GoRTs significantly increased from 0% to 20% SSP (p < .001) and from 20% to 40% SSP (p < .001). All other main effects and interactions failed to reach significance (all F ≤ 1.865, all p ≥ .160).

Behavioral assessment during TMS

There was a trend for earlier participant-specific stop times during TMS for older compared to young adults (F(1, 48) = 3.795, p = .057, η_p² = .073). There was no session effect (F(1, 48) = 0.003, p = .955, η_p² = .000) and no interaction between age group and session (F(1, 48) = 0.515, p = .477, η_p² = .011). The P(inhibit) during TMS ranged on average between 65-70% and did not differ between age groups (F(1, 48) = 2.619, p = .112, η_p² = .052). There was no effect of session (F(1, 48) = 1.236, p = .272, η_p² = .025) and no interaction between age group and session (F(1, 48) = 0.653, p = .423, η_p² = .013).

TMS intensities and motor threshold

TMS intensities and motor threshold are summarized in Table 2. The 2-way ANOVA_RM revealed that the TS intensity was significantly higher in the SICI compared to LICI session (F(1, 48) = 6.289, p = .016, η_p² = .116). There was no significant difference in the TS intensity between age groups (F(1, 48) = 0.060, p = .807, η_p² = .001) and no interaction between age group and session (F(1, 48) = 0.003, p = .958, η_p² = .000). There were no group differences in CS intensity (t(48) = 0.006, p = .995) or tMT (t(43.687) = -0.344, p = .732).

Resting-state measures of CSE, SICI and LICI

CSE (represented by the unconditioned motor evoked potential (MEP)) in rest did not differ between age groups (Table 2) (F(1, 48) = 1.340, p = .253, η_p² = .027) or sessions (F(1, 48) = 2.204, p = .144, η_p² = .044). There was also no interaction between age group and session (F(1, 48) = 1.849, p = .180, η_p² = .037). There was an age effect on SICI in rest (t(13.533) = 2.101, p = .043) and LICI in rest (U = 212, p = .022), with lower inhibition in older compared to young adults.

Modulation of CSE and SICI during reactive inhibition

Changes in CSE across the three conditions in the SICI session are shown in Figure 1A. On average, a total of 30 ± 11 trials per condition were included to assess CSE in the analysis. The 2-way ANOVA_RM indicated that there was a main effect of age group (F(1, 48) = 4.307, p = .043, η_p² = .082) in which CSE was found to be lower in older versus young adults. Furthermore, a main effect of condition (F(1.714, 82.268) = 18.867, p < .001, η_p² = .282) and a significant interaction between age group and condition (F(1.714, 82.268) = 3.984, p = .028, η_p² = .077) was observed. Post-hoc comparisons indicated that MEP amplitudes were reduced in early compared to go (p = .006), and stop compared to go (p < .001) in young adults. The difference between early and stop was not significant (p = .141). In older adults, there were no differences in MEP amplitude between conditions (all p ≥ .199). CSE did not significantly differ between young and older adults in any of the trial types (all p ≥ .090).

Figure 1. Modulation of corticospinal excitability (CSE, panel A) and short-interval intracortical inhibition (SICI, panel B) across the three conditions in the SICI session. Statistically significant differences in the interaction term are described in the main text. CSE is measured as unconditioned motor evoked potential (MEP) amplitude. Inhibition was calculated as follows: [1 – (MEP_{CS + TS} / MEP_TS)] * 100. Error bars represent standard error of mean. ***p<0.001, **p<0.01, *p<0.05, ^p<0.1.

The modulation of SICI across the three conditions is shown in Figure 1B. On average, a total of 30 ± 11 unconditioned trials and a total of 30 ± 11 conditioned trials per condition were used to calculate SICI in the analysis. The 2-way ANOVA_RM indicated that there was a significant main effect of age group (F(1, 47) = 6.838, p = .012, η_p² = .127), with higher inhibition in young compared to older adults. Furthermore, a main effect of condition (F(2, 94) = 13.685, p < .001, η_p² = .226) and a significant interaction between age group and condition (F(2, 94) = 3.296, p = .041, η_p² = .066) was observed. Post-hoc comparisons showed that inhibition was higher in early compared to stop and go in both young and older adults (all p ≤ .003). SICI was significantly higher in young compared to older adults in the early (p = .026) and stop condition (p = .024), but not in the go condition (p = .694). Finally, planned comparisons indicated that there was more inhibition in stop compared to go in young (p = .036) but not in older adults (p = .231). These results suggest that the recruitment of SICI during reactive inhibition was affected in older adults. In spite of that, there was no significant correlation between the modulation of SICI (SICI go – SICI stop) and SSRT in young (r = .108, p = .615) or older adults (r = .151, p = .461) (Figure 2A).

Figure 2. (A) Relation between the stop-signal reaction time (SSRT) in the short-interval intracortical inhibition (SICI) session and the difference in SICI between go and stop conditions in young (black) and older adults (white). (B) Relation between the slope of the go response time (GoRT) and the slope of long-interval intracortical inhibition (LICI) in the early condition (300 ms after trial onset, untransformed data). The slope was calculated as a function of stop-signal probability.

Modulation of CSE and LICI as a function of stop-signal probability

The modulation of CSE as a function of SSP in the early condition in the LICI session (Figure 3A) was investigated with a 2-way ANOVA_RM. On average, a total of 18 ± 4 trials per condition were included to assess CSE in the analysis. There was no effect of age group (F(1, 50) = 0.132, p = .718, η_p² = .003), no effect of SSP (F(1.556, 77.805) = 1.113, p = .321, η_p² = .022) and no interaction between age group and SSP (F(1.556, 77.805) = 2.300, p = .119, η_p² = .044).

Figure 3. Modulation of corticospinal excitability (CSE, panel A) and long-interval intracortical inhibition (LICI, panel B) as a function of stop-signal probability (SSP) in the LICI session. CSE is measured as unconditioned motor evoked potential (MEP) amplitude. Inhibition was calculated as follows: [1 – (MEP_{CS + TS} / MEP_TS)] * 100. Untransformed LICI values are presented. Error bars represent standard error of mean. ***p<0.001, **p<0.01, *p<0.05, ^p<0.1.

On average, a total of 18 ± 4 unconditioned trials and a total of 18 ± 4 conditioned trials per condition were used to calculate LICI in the early condition. Analysis of the LICI data showed that inhibition was significantly higher in young compared to older adults (Figure 3B) (F(1, 49) = 20.153, p < .001, η_p² = .291). Furthermore, there was a main effect of SSP on LICI (F(1.787, 87.539) = 4.195, p = .022, η_p² = .079). Post-hoc tests revealed a trend for higher inhibition in the 40% SSP condition compared to the 0% SSP condition (p = .083). The interaction between group and SSP did not reach significance (F(1.787, 87.539) = 2.089, p = .135, η_p² = .041). Furthermore, there was no significant correlation between the slope of LICI and the slope of GoRTs in young (LICI: 2 ± 8, GoRT: 11 ± 8; r_S = .028, p = .895) or older adults (LICI: 10 ± 23, GoRT: 12 ± 11; r_S = .087, p = .665) (Figure 2B).

Behavioral assessment

An anticipated response version of the stop-signal task was used with different stop-signal probabilities to assess reactive and proactive inhibition (Figure 5A-B) [7,27,49]. Participants were seated on a chair and had to place their right hand on the table with their index finger on a switch. The switch was pressed when the finger was resting, requiring no active force production. On a computer screen (refresh rate = 60 Hz), a bar started to fill at a constant velocity and participants were instructed to stop the bar as close as possible to the horizontal target line (800 ms from trial onset) by releasing the switch, i.e. lifting their index finger (go trials). The color of the target line changed after each go trial providing feedback about task performance. The colors represent the absolute timing difference relative to the target: green: < 20 ms, yellow: < 40 ms, orange: < 60 ms and red: > 60 ms. In stop trials, the bar would automatically stop before reaching the target line and participants were instructed to cancel the movement of releasing the switch, i.e. not lift their finger and thus keep the switch pressed. A staircasing algorithm was used to ensure an equal number of successful and unsuccessful stop trials (i.e. P(inhibit) ≈ 50%). More specifically, the time point on which the bar automatically stopped filling on stop trials (i.e. stop time) was increased with 33 ms when stopping was successful and decreased with 33 ms when stopping was unsuccessful. To assess proactive inhibition, three SSPs were used (0%, 20% and 40%) (Figure 5B). The color of the bar indicated the probability of stops (light blue bar: 0%, blue bar: 20%, magenta bar: 40%).

Figure 5. Anticipated response version of the stop-signal paradigm with timing of TMS and conditions of interest. (A) In go trials, participants had to stop the bar as close as possible to the horizontal target at 800 ms from trial onset. The bar could be stopped by releasing the right index finger from the switch. In stop trials, the bar stopped automatically and participants had to cancel the movement of lifting their finger. (B) Proactive inhibition was assessed by manipulating the stop-signal probability. The color of the bar indicated the probability of stops (light blue: 0%, blue: 20%, magenta: 40%). (C) The solid vertical lines represent the start (0 ms) and finish time (1280 ms) of the trial. The target is indicated by the vertical dashed line at 800 ms from trial onset. The TMS test stimulus was delivered at 300 ms from trial onset (early) or 150 ms after the participant-specific stop time (late) in go and stop trials. The stop time was determined for each participant separately and was based on task performance without TMS. (D) Three different conditions were included in the statistical analyses: early, stop and go. GoRT: go response time, ST: stop time.

At the beginning of each session, all participants practiced the task (Familiarization, Figure 4). We instructed participants to aim for green or at least yellow lines after go trials to reinforce go task performance. The participants were told that it would not be possible to cancel the movement of lifting their finger on all stop trials. Finally, participants were instructed that the probability of stops was lower when the bar was dark blue compared to magenta and that only go trials would occur when the bar was light blue. After the practice blocks, participants were asked to describe the meaning of the three colors to make sure that the instructions were understood correctly.

Next, they completed one block with 0% SSP (25 go trials) and one block in which 20% and 40% SSP were randomly varied (150 trials). The number of stop trials was matched between the two SSPs resulting in 80 go and 20 stop trials for the 20% condition and 30 go and 20 stop trials for the 40% condition (Figure 4). Inter trial interval was set at 3.25 ms and the bar was reset to empty 1.28 s after trial onset.

TMS preparation and EMG

SICI and LICI were performed with two Magstim 200 units connected through a BiStim module (Magstim Company, Dyfed, UK). A figure-of-eight-coil (70 mm outer diameter) was placed over the motor hotspot of the right first dorsal interosseous (FDI) with the handle pointing posteriorly at an approximately 45° angle to the midsagittal plane. EMG activity was recorded within the right FDI with EMG surface electrodes (Ag/AgCl) using a belly-tendon montage (Bagnoli-16 EMG system, Delsys Inc., Boston, USA). EMG signals were amplified (gain = 1000), bandpass filtered (20-2000 Hz) and 50 Hz noise was eliminated (Humbug, Quest Scientific, North Vancouver, Canada). The signals were sampled at 5000 Hz and stored for offline analysis (CED Signal Version 4.03, Cambridge Electronic Design, UK).

An orthogonal 1x1 cm coordinate system was marked on a swimming cap with references to the left and right external auditory meatus, occiput and vertex. The motor hotspot was defined as the location from which five consecutive TMS pulses produced the highest mean MEP within relaxed right FDI. The motor hotspot was marked on the cap and captured with a neuronavigation system (Visor 2, ANT Neuro, the Netherlands) with the standard MNI template to ensure optimal positioning of the magnetic coil during the experiment.

For SICI, an ISI of 3 ms was used. The TS intensity was set to elicit an MEP of approximately 1 mV when the finger was on the switch. To determine the CS intensity, we first measured the task motor threshold (tMT). This threshold was defined as the minimum intensity required to elicit four out of eight MEPs when the finger was on the switch [24]. The CS intensity for SICI was then set to 90% tMT and systematically lowered with 2% until relative inhibition was on average 50% (average of five consecutive paired-pulses).

For LICI, an ISI of 100 ms was used. The TS and CS intensities were set to elicit an MEP of approximately 1 mV when the finger was on the switch.

TMS during task

Participants performed ten task blocks per session while TMS was applied (see Figure 4). Consistent with Cowie et al. [24], the go only trials (i.e. SSP = 0%) were presented in separate blocks. The 20 and 40% SSP conditions were randomly varied within the other eight task blocks. In the TMS task blocks, the TS was randomly applied at two different time points: early (i.e. 300 ms after trial onset / 500 ms before the target) or late (i.e. participant-specific stop time + 150 ms) (Figure 5C-D). The late stimulation point was chosen because active inhibition is expected 150 ms after presentation of the stop-signal [7]. The participant-specific stop time was previously determined from the no-TMS task block performed earlier with the following formula: 800 ms – [mean(GoRT) – mean(stop time)] – 50 ms. The average response time in go trials (GoRT) was defined as the time between trial onset and lifting the finger from the switch, and thus a GoRT of 800 ms reflects a response on the target. By using a participant-specific stop time, the degree of stopping difficulty is comparable across participants. By subtracting 50 ms, the bar stops earlier and the P(inhibit) during TMS increases, resulting in more successful stop trials and thus more trials that can be included in the analysis. Furthermore, it reduces the chance that the TMS pulse is applied after EMG onset in go trials. The participant-specific stop time during TMS was kept constant. To keep the participant alert, catch trials were introduced in which the bar stopped 125 ms after the participant-specific stop time and thus closer to the target. Consequently, stopping was more difficult in these trials.

In total, there were 714 trials (see Figure 4 for number of trials per block). For the early stimulation time point, 20 single-pulse and 20 paired-pulse trials were administered per SSP condition. For the late stimulation time point, 24 single-pulse and 24 paired-pulse trials were delivered in the 0% SSP condition, and 56 single-pulse and 56 paired-pulse in both the 20% and 40% SSP conditions. Out of these 56 trials, 28 were stop trials. More trials were included in the late compared to the early stimulation time point because there is a higher chance that former trials will be excluded due to EMG onset. There were 2, 232 and 54 non-stimulated go trials in the 0%, 20% and 40% SSP condition, respectively. Finally, there were 16 non-stimulated stop trials in the 20% SSP condition and 18 non-stimulated stop trials in the 40% SSP condition. These non-stimulated stop trials were the catch trials in which the bar stopped 125 ms after the participant-specific stop time. The trial duration was 3.25, 4 or 4.45 s depending on the time point of stimulation in the previous and current trial, such that there was at least 4 s between two consecutive TS pulses.

TMS during rest

Resting-measures of CSE and intracortical inhibition were assessed during the rest block, which was always performed halfway through the experiment (i.e. block 6, Figure 4). Participants were instructed to look at a white fixation cross presented in the center of a black screen while their finger was resting on the switch. A total of 15 single-pulse and 15 paired-pulse trials were conducted in a random order at the same intensities as described above.

Behavioral analysis

Behavioral analysis was performed on baseline performance prior to TMS. GoRTs shorter than 400 ms (i.e. early response) or longer than 1280 ms (i.e. no response) were considered errors and were removed from analysis. The average GoRT for each SSP was calculated. The amount of response slowing (GoRT) with increasing SSP served as a measure of proactive inhibition. The percentage of successful stop trials (P(inhibit)) was calculated across the 20 and 40% SSP. Next, the stop-signal reaction time (SSRT) was calculated across the 20% and 40% SSP conditions with the integration method [50]. More specifically, the number of failed stop trials was divided by the total number of stop trials to get P(respond). GoRTs were sorted in ascending order and the nth GoRT was obtained where n equals the number of go trials multiplied by P(respond) [50]. The SSRT was estimated by subtracting the average stop time from the nth GoRT. The SSRT reflects the speed of stopping and served as a measure of reactive inhibition.

Variables of interest for the task during TMS were the percentage of successful stop trials (i.e. P(inhibit)) and the participant-specific stop time. GoRTs and SSRT during TMS were not investigated since TMS may influence response times.

EMG data processing and analysis

All trials were visually inspected and excluded if there was activity in the period between the TS and the onset of the MEP. The root mean square (RMS) of the EMG signal 50 ms prior to the first TMS pulse was calculated for each trial (Matlab 2016b) and trials were discarded if the RMS exceeded 20 µV. Trials in which GoRTs were < 300 ms or > 1280 ms were removed. Peak-to-peak MEP amplitudes in the right FDI muscle were calculated for each trial and outliers (± 3 SD) were removed. The MEP amplitudes were averaged across all trials per trial type, at each stimulation time point. Conditions of interest were early, go and stop (Figure 5D). For the stop condition, we only included MEPs of successful stop trials. Next, percentage SICI/LICI was calculated for each condition using the following formula: Inhibition % = [1 – (MEP_{CS + TS} / MEP_TS)] * 100. Scores between 0% and 100% indicate inhibition while scores below 0% indicate facilitation. Normality of the data was visually inspected with Q-Q plots and did not hold for the LICI measures.

Behavioral data

Statistical analysis were performed with the Statistical Package for the Social Sciences (IBM SPSS, Version 24.0, Armonk, NY, USA). First, statistical analyses were performed on the behavioral data that were collected at the beginning of each session, prior to stimulation. The effect of age group (young, older) and session (SICI, LICI) on P(inhibit) was analyzed with a 2-way repeated measures analysis of variance (ANOVA_RM). To asses age-related differences in reactive inhibition, a 2 (age group) x 2 (session) ANOVA was performed with SSRT as dependent variable. Age-related differences in proactive inhibition were tested with a 2 (age group) x 2 (session) x 3 (SSP: 0, 20, 40) ANOVA_RM with GoRT as dependent variable. GoRTs were expected to increase as a function of SSP [27]. Second, we investigated the effects of age group and session on the participant-specific stop time and P(inhibit) during TMS using a two-way ANOVA_RM.

TMS intensities, motor threshold and resting measures of CSE, SICI and LICI

Age-related differences in TMS intensities, motor threshold and resting-measures of CSE and intracortical inhibition were investigated. The TS intensity and the size of the amplitude of the unconditioned MEPs in rest were tested with a 2 (age group) x 2 (session) ANOVA_RM. The CS intensity in the SICI session, the tMT in the SICI session and SICI in rest were analyzed with student t-tests with age group as independent variable. Due to deviation from the normality assumption, the effect of age on LICI in rest was tested with the non-parametric Mann-Whitney U-test.

Modulation of CSE and LICI as a function of SSP

The modulation of LICI as a function of SSP (indicated by the color of the bar) was tested with a 2-way ANOVA_RM with age group as between variable, SSP (0, 20, 40) as within variable, the unconditioned MEP amplitude in rest as covariate and LICI in the early condition as dependent variable. It was expected that the amount of LICI in the beginning of the trial would increase with increasing SSP. Because of deviation from the normality assumption, LICI was transformed using an exponential function. The same ANOVA_RM was performed on the unconditioned MEPs to assess the modulation of CSE.

Associations between proactive response slowing and the modulation of LICI were investigated with Spearman correlation coefficients in young and older adults separately. To get a single-value index of the modulation of LICI and GoRTs with increasing SSP, we calculated the slope with a simple linear regression (Matlab 2016b). The slopes were correlated to investigate the association between the modulation of LICI and response slowing.

The significance level was set at p < .05 for all tests. Greenhouse-Geisser correction was applied when the assumption of sphericity was violated. All post-hoc comparisons were conducted with Tukey HSD. All data are presented as mean ± standard deviation in the text.