1 INTRODUCTION

The acquisition of novel motor skills is associated with structural and functional reorganization of the primary motor cortex (M1) (Adkins, Boychuk, Remple, & Kleim, 2006; Ljubisavljevic, 2006; Ruffino, Papaxanthis, & Lebon, 2017). In humans, transcranial magnetic stimulation (TMS) has been used to demonstrate corticomotor reorganization following a variety of motor skill training tasks. For example, individuals trained to perform piano exercises show greater excitability of M1 digit representations for the trained hand compared to untrained controls (Pascual-Leone et al., 1995). This is consistent with experimental studies that demonstrate M1 reorganization in response to upper limb motor training (Classen, Liepert, Wise, Hallett, & Cohen, 1998; Lotze, Braun, Birbaumer, Anders, & Cohen, 2003; Muellbacher, Ziemann, Boroojerdi, Cohen, & Hallett, 2001; Pascual-Leone, Grafman, & Hallett, 1994; Pascual-Leone et al., 1995). Cross-sectional studies reveal similar findings, with braille readers and tennis players possessing larger hand representations in M1 than people not proficient in these skills (Pascual, Wassermann, Sadato, & Hallett, 1995; Pascual-Leone et al., 1993; Pearce, Thickbroom, Byrnes, & Mastaglia, 2000). These observations support animal work demonstrating greater M1 synaptogenesis following reaching and manipulation training than nonskilled repetitive movement (Allred & Jones, 2004; Bury & Jones, 2002; Greenough, Larson, & Withers, 1985; Kleim et al., 2002, 2004).

The degree of M1 reorganization associated with changes in motor performance may depend upon the region being trained. Training-induced M1 changes have been shown to be graded from distal to proximal, with hand representations showing greater reorganization than upper arm representations (Krutky & Perreault, 2007). The potential for region-specific differences in M1 reorganization is supported by findings that interventions, such as theta burst stimulation (Martin, Gandevia, & Taylor, 2006) and ipsilateral hand movement (Sohn, Jung, Kaelin-Lang, & Hallett, 2003), induce more effective corticomotor inhibition in distal than proximal muscle representations. Structural and functional differences across M1 representations may contribute to this region-dependent reorganization. For example, distal muscles, involved in fine motor control, possess larger and more excitable M1 representations than those for proximal muscles that contribute to limb positioning (Palmer & Ashby, 1992; Wassermann, McShane, Hallett, & Cohen, 1992). These findings suggest that neural contributions to improved motor performance may differ between muscle groups (Krutky & Perreault, 2007). However, as studies of motor training have focussed primarily on the upper limb (Adkins et al., 2006; Carson, Ruddy, & McNickle, 2016; Ljubisavljevic, 2006), it remains unclear whether the potential for corticomotor reorganization following training is limited in other proximal and axial regions, such as the lower back.

Lower back muscles are functionally distinct from those of the limbs, providing a greater contribution to postural functions than fine motor control (Kaigle, Holm, & Hansson, 1995; Tsao, Druitt, Schollum, & Hodges, 2010; Wilke, Wolf, Claes, Arand, & Wiesend, 1995). Trunk muscles have a greater proportion of ipsilateral projections from M1 than upper limb muscles (Strutton et al., 2004), and motor control of these regions involves different pathways in the basal ganglia (Visser et al., 2008). There are also fewer corticospinal projections to the lower back than to the hand, with these projections arising from a smaller cortical area (Cheyne, Kristeva, & Deecke, 1991; Penfield & Boldrey, 1937). Animal studies suggest that lower back muscles receive greater input from subcortical sites when compared to upper limb muscles, which demonstrate cortical dominance (Deliagina, Beloozerova, Zelenin, & Orlovsky, 2008; Galea, Hammar, Nilsson, & Jankowska, 2010; Lemon, Kirkwood, Maier, Nakajima, & Nathan, 2004). These features of back muscle control might underlie differences in the potential role and influence of corticomotor reorganization during motor skill acquisition. Although this holds implications for rehabilitation and motor retraining, no studies have examined corticomotor adaptations following short-term motor training, or the relationship between these factors, in the lower back.

This study aimed to (a) determine whether short-term motor training (i.e., a lumbopelvic tilt visuomotor task) induced reorganization of the corticomotor representations of lower back muscles, (b) quantify the degree of variation of corticomotor responses to visuomotor training in the lower back, and (c) determine whether improvements in task performance were correlated with corticomotor reorganization. A rapid TMS mapping technique was employed to probe corticomotor reorganization (Cavaleri, Schabrun, & Chipchase, 2018; van de Ruit, Perenboom, & Grey, 2015). The short duration of this technique allowed investigation of transient changes in the size and distribution of corticomotor representations that might be induced by short-term training, but not feasibly assessed using conventional methods. Given previous observations regarding the corticomotor reorganization of proximal upper limb muscles following motor training, we hypothesized that lower back representations would exhibit only small changes in excitability or organization following motor training.

2 MATERIALS AND METHODS

2.2 Experimental protocol

Participants attended a single experimental session and were requested to refrain from caffeine, alcohol, or exhaustive exercise from 12 hr prior to testing. Before testing, a random number generator was used to randomly allocate participants to either the experimental task (lumbopelvic tilt motor training task) or a control task (repeated finger abduction) involving no lumbopelvic movement. The experimental session was divided into three phases: Baseline, Testing, and Recovery (Figure 1). During the Baseline phase, the corticomotor representation of the lumbar erector spinae (LES) muscle was mapped twice using TMS. This was done to examine the within-session reliability of map features in the absence of the motor training task. During the Testing phase, participants performed one of the two training tasks for a total of 15 min, which was divided into three 5-min training blocks. Maps of the LES representation were acquired after each training block to examine potential changes in map features with motor training. During the Recovery phase, two maps were acquired 15 and 30 min after the final training block in order to investigate retention or resolution of any corticomotor adaptations (Figure 1).

The protocol in this study reflected short-term motor training, in which improvement (“fast learning”) can be induced by a limited number of trials over a time scale of less than one hour (Karni et al., 1998; Luft & Buitrago, 2005). In contrast, long-term motor training (not assessed) involves slowly evolving incremental performance gains (“slow learning”) triggered by consistent practice over multiple sessions constituting a total timeframe of hours or days (Karni et al., 1998; Luft & Buitrago, 2005).

2.3 Lumbopelvic tilt motor training task

Participants sat with their thighs supported, feet flat on the floor, and knees and hips flexed to 90°. Participants allocated to perform the lumbopelvic tilt motor training task had a triaxial accelerometer (CXL10LP3, Crossbow technologies; Weight: 46 g; Dimensions: 1.9 cm × 4.8cm × 2.5cm) affixed to the highest point of the right iliac crest as determined by a trained physiotherapist using palpation. Accelerometer data were sampled at 2 kHz using a Power 1401 data acquisition system and Spike2 software (Cambridge Electronic Design).

Real-time output of accelerometer motion in the sagittal plane was displayed as a blue line using Spike2 on a computer monitor (53 cm resolution of 1,920 × 1,080) at eye level, two meters in front of the participant. Anterior and posterior tilting of the pelvis was indicated by downward and upward movement of the blue line, respectively. Participants were first instructed to move the blue line to draw a smooth sinusoidal pattern by alternately tilting their pelvis anteriorly and posteriorly as far as possible for 30 s. A red "target" sinusoidal wave was then generated and scaled to 85% of the participant's maximum range in each direction.

Participants allocated to perform the finger abduction (control) task had the accelerometer affixed along the posterior aspect of the right index finger, with the center of the accelerometer positioned over the proximal interphalangeal joint. The forearm was supported in midpronation/supination with the elbow flexed at 90°. Downward and upward movement of the blue trace reflected finger adduction and abduction, respectively. The red target wave was generated as described above. This setup controlled for attention without training lumbopelvic movement or lower back muscle control.

During the recorded trial blocks, the red target wave was presented for 5 min, with the amplitude of the wave altered unpredictably every cycle (25%, 50%, 75%, or 100% of the target range). Participants were instructed to tilt their pelvis (pelvic tilt group) or move their finger (control group) to match the red target wave as accurately as possible. TMS mapping was performed after each 5 min block. A sample of the output during lumbopelvic tilt training for a representative participant is presented in Figure 2.

3 RESULTS

3.3 Testing and recovery phases

3.3.1 Both groups demonstrated improved motor performance

Both the pelvic tilt and control group demonstrated improved performance in the form of reduced motor error over time (Time: F_{1.3, 35.6} = 18.5, p < .001; Figure 3). Motor error was significantly reduced from Training Block 1 to Training Block 2 (p = .002) and from Training Block 2 to Training Block 3 (p = .03). Although the control group had a higher degree of motor error at each time point (Group: F_{1, 28} = 9.3, p = .01, all post hoc p = .01), there was no difference between groups in terms of the change in motor error over time (Group × time: F_{1.3, 35.6} = 1.7, p = .20).

3.3.2 Neither group demonstrated corticomotor reorganization

As shown in Figure 4, motor training did not elicit changes in map area (Time: F_{5, 130} = 0.7, p = .62; Group: F_{1, 26} = 0.2, p = .65; Group × time: F_{5, 130} = 0.5, p = .76) or map volume (Time: F_{5, 130} = 0.15, p = .71; Group: F_{1, 26} = 0.05, p = .83; Group × time: F_{5, 130} = 0.5, p = .75) in either group. Similarly, CoG did not change in either group following repeated exposure to the motor training tasks (Time: F_{5, 130} = 1.2, p = .30; Group: F_{1, 26} = 0.6, p = .46; Group × time: F_{5, 130} = 1.0, p = .38; Figure 4).

3.3.3 Corticomotor responses to motor training were variable, and performance did not differ between "responders" and "nonresponders"

The SD of the Δarea during the Baseline phase was 2.0 cm², and the SD of the Δvolume was 84.2 mV. Participants were therefore categorized as responders in terms of map area if their individual Δarea > 2.0 cm² (positive responder) or <−2.0 cm² (negative responder), and responders in terms of map volume if Δvolume > 84.2 mV (positive responder) or <−84.2 mV (negative responder). All other participants were classified as nonresponders. Although 80% of participants in the pelvic tilt group demonstrated reduced motor error between Training Blocks 1 and 2, only 53% (33% positive, 20% negative) were classified as responders in terms of map area and 53% (40% positive, 13% negative) in terms of map volume over this period. Similarly, 93% of participants in the pelvic tilt group reduced motor error between Blocks 1 and 3, but only 47% (27% positive, 20% negative) were categorized as responders in terms of map area and 53% (33% positive, 20% negative) in terms of map volume. In the control group, 47% (27% positive, 20% negative) of participants were classified as responders for both map area and map volume following Training Block 3. Changes in CoG from baseline were less than 1 cm for 87% of participants in the pelvic tilt group and 93% of participants in the control group.

There was no difference in task performance between participants in the pelvic tilt group classified as responders, and those classified as nonresponders in terms of map area (Time: F_{2, 26} = 13.2, p < .001; Response: F_{1, 13} = 0.5, p = .48; Response × Time: F_{2, 26} = 0.2, p = .84) or map volume (Time: F_{2, 26} = 13.7, p < .001; Response: F_{1, 13} = 0.9, p = .37; Response × Time: F_{2, 26} = 0.3, p = .76).

3.3.4 There was no relationship between changes in pelvic tilt task performance and corticomotor reorganization

There was no relationship between changes in pelvic tilt motor error and changes in map area (Block 1–2: r = −.08, p = .78; Block 2–3: r = −.32, p = .25; Block 1–3: r = −.23, p = .41; Figure 5a), map volume (Block 1–2: r = −.15, p = .59; Block 2–3: r = −.10, p = .72; Block 1–3: r = −.16, p = .58; Figure 5b), or CoG displacement (Block 1–2: r = .03, p = .92; Block 2–3: r = −.14, p = .61; Block 1–3: CoG: r = −.04, p = .89; Figure 5c) at any time point.

4 DISCUSSION

These data are the first to test for an association between motor performance and reorganization of the corticomotor representations of lower back muscles. Although performance improved during a lumbopelvic tilt visuomotor task, no concomitant changes in map area, volume, or center of gravity were identified by TMS mapping. After three training blocks, only 47% (27% positive, 20% negative) of participants in the pelvic tilt group were classified as responders in terms of map area and 53% of participants (33% positive, 20% negative) in terms of map volume. This suggests that corticomotor adaptations to motor training are variable and may not underpin short-term motor performance improvements in the lower back.

The protocol used in this study enabled exploration of the effects of short-term motor training on the corticomotor representations of lower back muscles. A visuomotor tracing task was selected as this type of training has been shown to induce improved motor performance in the hand, upper arm, and ankle joints (Classen et al., 1998; Lotze et al., 2003; Muellbacher et al., 2001; Pascual-Leone et al., 1994, 1995; Perez, Lungholt, Nyborg, & Nielsen, 2004). Following task familiarization, three training blocks constituting a total of 15 min of motor training were employed and sufficient to reduce motor error. Previous research has demonstrated that as little as 6 min of skill training induces increases in MEP amplitude, with sessions beyond 20 min potentially attenuating corticomotor reorganization due to overlearning or loss of focus (Floyer-Lea & Matthews, 2004; Jensen, Marstrand, & Nielsen, 2005; van de Ruit & Grey, 2019; Willerslev-Olsen, Lundbye-Jensen, Petersen, & Nielsen, 2011). Overlearning describes a process whereby initial increases in MEP amplitude with task performance return to baseline once task mastery has been achieved (Muellbacher et al., 2001). Sequential improvement in performance across three training blocks, but lack of any relationship with TMS map features at any time point, suggests that the findings of this study are not the result of overlearning. Inclusion of the finger abduction group controlled for the attention aspects of the task without training lumbopelvic movement or promoting improved lower back muscle control.

The findings of this study contrast previous reports of corticomotor reorganization following upper limb motor training (Classen et al., 1998; Lotze et al., 2003; Muellbacher et al., 2001; Pascual-Leone et al., 1994, 1995). A potential explanation for this discrepancy is differences in the neural control underlying hand and lower back movement. Hand muscles are involved in dextrous movements requiring precise control, whereas lower back muscles perform primarily postural functions (Deliagina et al., 2008). Although corticospinal projections to lower back muscles are present in humans (Ferbert, Caramia, Priori, Bertolasi, & Rothwell, 1992), TMS-evoked responses are small and generally require facilitation via muscle contraction even at high stimulation intensities (Chang et al. 2019; Ferbert et al., 1992; Nathan, Smith, & Deacon, 1996; Tsao, Galea, & Hodges, 2008). Animal studies show greater involvement of the brainstem, vestibulospinal and reticulospinal tracts in control of muscles of the lower back than the hand, which are primarily under cortical control (Deliagina et al., 2008; Galea et al., 2010; Lawrence & Kuypers, 1968; Lemon et al., 2004). The findings of the present study suggest that improved motor performance in the lower back may involve adaptations in these subcortical networks rather than the motor cortex. Treatment techniques promoting corticomotor reorganization may therefore have less impact on motor performance in the lower back than the upper limb. This is supported by recent evidence that peripheral electrical stimulation modifies corticomotor excitability when applied to the upper limb (Barsi, Popovic, Tarkka, Sinkjær, & Grey, 2008; Chipchase, Schabrun, & Hodges, 2011; Golaszewski et al., 2012) but not when applied to the lower back (Elgueta-Cancino, Massé-Alarie, Schabrun, & Hodges, 2019).

Another explanation for these findings relates to the nature of the training task itself. The present study used a visuomotor tracing task that required continuous participant attention. In contrast, many studies of upper limb training have used ballistic motor training tasks requiring repeated nonskilled muscle activation (e.g., rapid finger tapping) (Hammond & Vallence, 2006; Ljubisavljevic, 2006; Muellbacher et al., 2001). There is evidence to suggest that different neurophysiological mechanisms may underpin these forms of motor training. For example, inhibitory repetitive transcranial magnetic stimulation over M1 disrupts retention of skill improvements during a ballistic thumb and finger opposition task (Baraduc, Lang, Rothwell, & Wolpert, 2004; Muellbacher et al., 2002), but has no effect on a more complex skill requiring finger movement toward a target while compensating for varying levels of externally applied force (Baraduc et al., 2004). This suggests that M1 may be more engaged during ballistic tasks, potentially having less influence over visuomotor training. During visuomotor training, neural adaptations may occur in regions beyond M1. This is supported by functional neuroimaging studies in both humans and animals demonstrating increases in supplementary motor area, premotor cortex, parietal cortex, and cerebellum activity during complex visuomotor tasks (Catalan, Honda, Weeks, Cohen, & Hallett, 1998; Floyer-Lea & Matthews, 2004, 2005; Hardwick, Rottschy, Miall, & Eickhoff, 2013).

There is also evidence to suggest that the difficulty of a visuomotor task may influence the degree of corticomotor reorganization that occurs. Previous research suggests that simple tasks can be performed using subcortical circuits alone, eliciting no changes in cortical reorganization (Dhawale, Wolff, Ko, & Ölveczky, 2019; Kawai et al., 2015). Similarly, motor control may be transferred from cortical to subcortical sites once task mastery has been achieved (Hwang et al., 2019). This suggests that the results of the present study could be attributed to a lack of difficulty or complexity in the selected task. However, previous research has shown consistent corticomotor reorganization during similar tracing tasks involving the neck, upper arm, hand, and ankle joints (Lotze et al., 2003; Pascual-Leone et al., 1994, 1995; Perez et al., 2004; Rittig-Rasmussen, Kasch, Fuglsang-Frederiksen, Jensen, & Svensson, 2013). Further, as task mastery was not achieved (93% of participant demonstrated improvements in task performance over time), the results of the present study are unlikely due to a lack of task complexity. Nevertheless, further studies of short-term motor training in the lower back, involving ballistic motor tasks or visuomotor tasks of varying difficulties, are required to confirm whether the nature of the training task influences the location or extent of cortical reorganization.

Task-specific corticomotor reorganization may also explain the difference in the results of this study and others involving trunk musculature. For example, repeated voluntary activation of the multifidus or transverse abdominis muscles induces corticomotor reorganization over a two-to-three week period (Massé-Alarie, Beaulieu, Preuss, & Schneider, 2016; Tsao, Galea, & Hodges, 2010). The tracing task used in the present study required coordinated activity of multiple muscles and had greater visuomotor demands than repeated focussed muscle activation, potentially eliciting reorganization in different premotor, subcortical, and spinal centers. Alternatively, given the relatively small influence of corticospinal projections to the lower back, repeated practice over an extended period of time may be required to drive corticomotor adaptations. Indeed, emerging research suggests that task duration may influence the degree and location of cortical reorganization that occurs (Dhawale et al., 2019; Hwang et al., 2019) Further research is required to determine whether reorganization of lower back muscle representations in M1 follows ongoing exposure to complex visuomotor tasks (over periods of days to weeks), or if different schedules of training elicit distinct patterns of corticomotor reorganization. This would provide insight into any potential differences in the neurophysiological mechanisms underpinning short-term and long-term motor training. It is also important to note that previous research has been conducted on participants with lower back pain (Massé-Alarie et al., 2016; Tsao, Galea, et al., 2010), whose corticomotor responses to motor training may differ from those of the healthy participants tested in this study (Schabrun, Burns, Thapa, & Hodges, 2017).

Emerging evidence suggests that the relationship between corticomotor excitability and motor performance may be more variable than once thought. Vallence, Kurylowicz, and Ridding (2013) reported that 28% of participants demonstrated no change in MEP amplitude following a ballistic thumb abduction training task, and more recently, van de Ruit et al. (2019) found that only 18%–36% of participants were classified as responders to a finger visuomotor training task. Consistent with the findings of the present study, these investigations showed no relationship between improved motor performance and corticomotor reorganization. This response variability is also observed following noninvasive brain stimulation paradigms, where "inhibitory" or "excitatory" protocols elicit the predicted corticomotor response in 30%–50% of participants (Fratello et al., 2006; Martin et al., 2006; Müller-Dahlhaus, Orekhov, Liu, & Ziemann, 2008; Wiethoff, Hamada, & Rothwell, 2014). This variability has been attributed to factors such as differences in anatomy, genetics, and history of synaptic activity (Fratello et al., 2006; Martin et al., 2006; Müller-Dahlhaus et al., 2008; Ridding & Ziemann, 2010; Wiethoff et al., 2014). The present study extends previous findings by demonstrating no differences in motor performance between individuals who exhibit corticomotor reorganization during training and those who do not. This suggests that, while the neural mechanisms underpinning improved motor performance may vary between participants, this variability does not contribute to observable differences in motor outcomes.

This study employed a rapid TMS mapping technique, rather than averaging a set of MEP amplitudes over a single cranial site, in order to evaluate corticomotor responses to motor training (Cavaleri et al., 2018; van de Ruit et al., 2015). The rapid mapping technique has been previously validated and has shown good-to-excellent within-session reliability when used to map upper limb muscle representations (Cavaleri et al., 2018; van de Ruit et al., 2015; van de Ruit, 2015). The reliability and baseline variance observed in this study are consistent with findings in upper limb representations (van de Ruit & Grey, 2019), and mapping has been shown to produce results consistent with those obtained after averaging MEP responses at the motor cortical hotspot (Summers et al., 2019). Differences between the findings of this study and previous research are therefore unlikely attributable to the mapping technique employed. Despite a rigorous approach toward data collection and analysis, this study is not without limitations. It was not possible to blind the researchers performing TMS mapping to the motor training task as the pelvic tilt and control tasks required different accelerometer placements. This study was also restricted to a small sample of healthy participants. Further research with larger sample sizes, participants with back pain, and a variety of training tasks is required to determine whether the findings reported here have clinical implications.

5 CONCLUSIONS

No relationship between corticomotor reorganization and improvements in performance during a lumbopelvic tilt visuomotor task was identified. Substantial variability was observed in terms of corticomotor responses to motor training, with approximately 50% of participants showing no changes in map volume or area despite significant improvements in task performance. These findings suggest that short-term improvements in lower back motor performance during a visuomotor task may be driven by changes in remote subcortical and/or spinal networks rather than adaptations in corticomotor pathways. However, further research using tasks of varying complexities, schedules, and durations is required to confirm this hypothesis.

CONFLICT OF INTEREST

None.

AUTHOR CONTRIBUTIONS

There are no conflicts of interest, additional acknowledgements, or affiliations to report. Rocco Cavaleri, Lucy Chipchase, Hugo Massé-Alarie, and Paul Hodges were involved in the conception of the study. Rocco Cavaleri, Hugo Massé-Alarie, Muath Shraim, and Paul Hodges were involved in data collection. All authors were involved in the design, writing, and editing of the study and manuscript. The final manuscript was approved by all authors.