Twenty-two healthy adults (8 female) volunteered for the study. Seven participants were not considered due to possible activation of ventral roots (see Lumbar-evoked potentials). Therefore, the data presented here are representative of the 15 (4 female) volunteers fulfilling all study requirements (males: 11 subjects, 31 ± 6 years, height 178 ± 6 cm, weight 82 ± 8 kg; females: 4 subjects, 28 ± 1 years, height 166 ± 8 cm, weight 64 ± 7 kg). All included participants were free from neurological illness and musculoskeletal injury in the lower-limbs for the last 6 months, were not taking any medications known to affect the nervous system and had no contraindications to transcranial magnetic stimulation (TMS), which was assessed via a health questionnaire (modified from Rossi et al. (2009). Before testing, all participants were fully informed of the procedures and possible risks, and each participant provided written inform consent. The study was approved by the Ethical committee of the University of Jyväskylä (10.01.2020) and was conducted with accordance with the Declaration of Helsinki (2013).

An a priori sample size estimation was conducted using G*Power software (version 3.1, University of Dusseldorf, Germany), based on data presented by Yacyshyn et al. (2016) for α = 0.05 and power = 0.80. The estimated sample size needed was 18 participants to assess torque × time delay interaction between unconditioned and conditioned LEPs.

Detailed description of Torque, M-max, TMS, Lumbar stimulation and EMG can be found in the subsections below.

Participants visited the laboratory on one occasion. To assess responses in the RF muscle, participants were sat in a custom-built chair with a calibrated load cell (Faculty of Sport and Health Sciences, University of Jyväskylä, Finland) with hip and knee at 90° flexion and the shin strapped with a non-elastic restraint ~ 2 cm superior to the ankle malleoli. The voltage signal originating from the load cell was calibrated and converted into torque (N·m). All measures were performed on the right (i.e., dominant) leg, assessed by self-report of which foot they primarily kick a ball (van Melick et al. 2017).

Once the participant was secured to the dynamometer, the maximum compound action potential (M-max) was assessed in a relaxed condition. Two maximal voluntary contraction (MVC) trials were performed 60 s apart. Prior to the MVC, two contractions at ~ 50% and ~ 80% of estimated MVC were performed as a warm-up. Verbal encouragement and visual feedback were provided to motivate participants to produce maximal effort. Thereafter, target contraction intensities (25%, 50% and 75% of MVC) were displayed on the screen as visual feedback for the participant.

Placement of the lumbar stimulation electrodes was assessed to avoid activating spinal nerve roots (see Lumbar-evoked potentials). Thereafter, stimulator intensity was adjusted to produce a LEP of 50% of the M-max at rest, and this stimulation intensity was used throughout the experiment. TMS coil placement was defined as the location producing the largest MEP in the RF, and stimulator output intensity was standardized to evoke ~ 200 ms SP from the stimulator artefact to the resumption of the voluntary EMG signal, during brief voluntary contractions at each torque.

During the session, unconditioned and conditioned LEPs were delivered during the same voluntary contraction. Unconditioned LEP consisted of a single stimulation delivered at the lumbar level. Conditioned LEPs consisted of a paired stimulation of TMS followed by lumbar stimulation separated by predetermined and randomly ordered time delays (60, 90, 120 and 150 ms). Participants were instructed to contract to, and briefly hold, one of the three different contraction intensities (25, 50 and 75% of MVC) in a randomized order. Once the participant reached the required level, an unconditioned LEP was delivered followed by a conditioned LEP at one of the different time delays (Fig. 1). The contractions were held for 5–8 s and stimuli were delivered 2–3 s apart. Sets of five unconditioned, followed by conditioned LEPs, were given per time delay and per torque level as a single block, giving a total of 60 unconditioned and conditioned stimuli. To avoid fatigue (see Results), 30, 45 and 60 s rest was given between contractions at 25%, 50% and 75% of MVC, respectively, and 60, 120 and 180 s rest was given between the sets of 5 contractions. At the end of the protocol, M-max and MVC were reassessed.

One participant’s mean (solid) and individual (dashed) trials that represent the experimental design of one set of unconditioned and conditioned lumbar stimulation at different time delays taken from 25% MVC trials. TMS transcranial magnetic stimulation, LS lumbar stimulation

Percutaneous electrical stimulation of the femoral nerve (3.2 cm cathode/anode arrangement; Polar Neurostimulation Electrodes, Espoo, Finland) was performed to elicit M-max in RF (1 ms square pulse duration; Digitimer DS7AH, Hertfordshire, UK). Electrodes were placed 2 cm apart and placed at each side of the femoral nerve, located by palpation and identification of the femoral artery (Walker et al. 2016). M-max was elicited by gradually increasing stimulator output intensity until the EMG response plateaued. To ensure supramaximality, this intensity was further increased by 50% (mean ± standard deviation intensity: 257 ± 151 mA).

Single TMS pulses were delivered using a Magstim 2002 magnetic stimulator (Magstim Co., Ltd., Whitland, UK) connected to a concave double-cone coil, positioned over the left cortical hemisphere for RF with a posterior-to-anterior current orientation. The hotspot was defined, at rest, as the position eliciting the largest MEP recorded in the EMG using the same intensity (i.e., 50–70% stimulator output) producing a visible MEP. The coil position was marked on the scalp, once the hotspot was found, to maintain the same position throughout the protocol. Stimulus intensities were set to evoke a silent period of ~ 200 ms for all contraction intensities (Table 1).

LEPs were elicited with a constant-current stimulator (1 ms square pulse duration; Digitimer DS7AH, Hertfordshire, UK) via self-adhesive electrodes (Polar Neurostimulation Electrodes, Espoo, Finland). The cathode (5 × 10 cm) was centered over the first lumbar vertebra (L1) and the anode (circular shape; 3.2 cm diameter) was placed on the midline of the vertebral column ~ 5 cm above the top edge of the cathode as described by Škarabot et al. (2019a).

The intensity of stimulation (309 ± 108 mA) was standardized to 50% of the M-max evoked in the resting position. Potential activation of ventral roots was assessed by examining the onset latency of the LEP with an increase in stimulator intensity (Petersen et al. 2002) and tracking LEP amplitude during increased voluntary contraction (Taylor et al. 2002). Should the ventral roots be activated by the stimulation procedures, onset latency would have shortened with an increase in stimulator intensity and LEP amplitude would have been the same during increased voluntary contraction (Petersen et al. 2002; Taylor et al. 2002, 2006; Škarabot et al. 2019a).

Dorsal root activation was assessed via paired LS with 50 ms time delay (Fig. 2), where the amplitude of the second LEP was compared to the first. Evidence of dorsal root activation would be a decrease in the second LEP due to post-activation depression at the motor-neuron pool (Hofstoetter et al. 2018). All remaining participants showed no sign of the responses described and reported that they found LS to be tolerable.

Data extracted from one participant showing that spinal root activation did not occur. A When increasing the intensity of stimulator output there was no reduction in latency. B A lumbar stimulated doublet with 50 ms interval, showing similar amplitudes between the stimulations

Muscle activity was recorded using adhesive Ag/AgCl electrodes (3 × 2 cm, BlueSensor N, Ambu, Penang, Malaysia) from m.Bicep Femoris (BF) and RF according to SENIAM Guidelines (Hermens et al. 2000). Skin was shaved, abraded with sandpaper, and wiped with alcohol before setting the electrodes in bipolar arrangement with 2 cm center-to-center distance. Impedance was set < 2kΩ, and the reference electrode was positioned above the patella. EMG data were amplified (1000 ×), bandpass filtered (16–1000 Hz; Neurolog System, Digitimer Ltd, UK)) and sampled online at 3000 Hz using CED Power1401-3 (Cambridge Electronic Design Ltd, Cambridge, UK).

Torque was sampled at 1000 Hz, amplified by a custom-built amplifier (ForAmps 1 v1.2, University of Jyväskylä, Finland) and converted by a 16-bit A/D board (CED Power1401-3, Cambridge Electronics Design, Cambridge, UK) in combination with Spike2 software (version 6.10, Cambridge Electronic Design, Cambridge, UK).

Offline analyses were performed with Spike software (version 6.10, Cambridge Electronic Design, Cambridge, UK) to manually obtain M-max amplitude, MVC, MEP Silent Period and unconditioned LEP onset latencies. The other outcome measures were analyzed by a customized MATLAB script (version R2020b, The MathWorks, Inc., Natick, USA). Peak-to-peak amplitude of LEPs and MEPs was analyzed automatically between latencies-of-interest following peripheral nerve stimulation, lumbar stimulation or TMS (Taylor et al. 1999), respectively. Torque was averaged over the 100 ms before the stimulator artefact. SP duration was determined, through visual inspection, as the time from the stimulator artefact to the return of voluntary EMG (Damron et al. 2008).

SPSS software (version 26.0, SPSS Inc., Chicago, USA) was used for all statistical methods. Means and standard deviation (SD) were calculated and reported throughout. Normality of the data was tested with the Shapiro–Wilk test and confirmed by z-score with an acceptance of + 2 to -2 (e.g. skewness score/skewness scoreSE and kurtosis score/kurtosis scoreSE). Data that did not fulfil those requirements were Log10 transformed, which then fulfilled the requirements for Normality. Paired t-tests were used to assess possible effects of fatigue between M-maxpre and M-maxpost, MVCpre and MVCpost, and to evaluate unconditioned LEP amplitude at different torque levels in the control measurements (shown in Fig. 3). One-way analysis of variance (ANOVA) was used to assess potential differences between the three contraction intensities in control measures: Unconditioned LEP latencies, MEP amplitude and MEP Silent Period (shown in Table 1). To determine whether Normalized [Conditioned/Unconditioned LEP*100] LEPs responded differently at the tested time delays between the three different torque levels, two-way repeated measures ANOVA was employed. When sphericity assumptions were violated, Greenhouse–Geisser corrections were used. Post-hoc Bonferroni adjustments were used when significant main effects were found. When comparing Unconditioned and Conditioned LEP at each time delay, the Benjamin–Hochberg test corrected for multiple paired t test comparisons with a 10% false discovery rate. Effect sizes are represented as partial eta-squared values (ηp2 = small: 0.01, medium: 0.06, large: 0.14) for the factors of the ANOVA and as Hedge’s g for between-group effect sizes for these relative changes (g = small: < 0.3, medium: 0.3–0.8, large: > 0.8). Αlpha was set at 0.05.

Mean (± SD) and individual values of unconditioned LEP response normalized to M-max at different contraction intensities. Increases in LEP amplitude with increases in torque shows that the stimulation was evoked trans-synaptically