key: cord-0430684-h5kn5kpo
authors: Morita, Tomoyo; Hirose, Satoshi; Kimura, Nodoka; Takemura, Hiromasa; Asada, Minoru; Naito, Eiichi
title: Hyper-adaptation in the Human Brain: Functional and structural changes in the foot section of the primary motor cortex in a top wheelchair racing Paralympian
date: 2022-03-19
journal: bioRxiv
DOI: 10.1101/2022.03.17.484824
sha: 605fad5982add33a05567559f6ebc1fba0edaa5e
doc_id: 430684
cord_uid: h5kn5kpo

The human brain has the capacity to drastically alter its somatotopic representations in response to congenital or acquired limb deficiencies and dysfunctions. The main purpose of the present study was to elucidate such extreme adaptability in the brain of an active top wheelchair racing Paralympian (participant P1) who has congenital paraplegia (dysfunction of bilateral lower limbs). Participant P1 has undergone long-term wheelchair racing training using bilateral upper limbs and has won a total of 19 medals in six consecutive summer Paralympic games as of 2021. We examined the functional and structural changes in the foot section of the primary motor cortex (M1) in participant P1 as compared to able-bodied control participants. We also examined the functional and structural changes in three other individuals (participants P2, P3, and P4) with acquired paraplegia, who also had long-term non-use period of the lower limbs and had undergone long-term training for wheelchair sports (but not top athletes at the level of participant P1). We measured brain activity in all the participants using functional magnetic resonance imaging (MRI) when bimanual wrist extension-flexion movement was performed, and the structural MRI images were collected. Compared to 37 control participants, participant P1 showed significantly greater activity in the M1 foot section during the bimanual task, and significant local GM expansion in this section. Significantly greater activity in the M1 foot section was also observed in participant P4, but not in P2 and P3, and the significant local GM expansion was observed in participant P2, but not in P3 and P4. Thus, functional or structural change was observed in an acquired paraplegic participant, but was not observed in all the paraplegic participants. The functional and structural changes typically observed in participant P1 may represent extreme adaptability of the human brain. We discuss the results in terms of a new idea of hyper-adaptation.

Introduction 54

Somatotopy is a fundamental functional structure for sensorimotor processing in the brain and is rich 55 in plasticity. Previous neuroimaging studies have shown that the human brain has the capacity to 56 drastically change its somatotopic representations in response to congenital or acquired limb 57 deficiencies and dysfunction (Flor et al., 1995 (Flor et al., , 2006 The main purpose of the present study was to elucidate such higher adaptability in the brain 61 of an active top, wheelchair racing Paralympian (participant P1), who had congenital paraplegia 62 (dysfunction of bilateral lower limbs). Participant P1 had received long-term wheelchair racing 63 training using bilateral upper limbs, since she was eight years old, and had won a total of 19 medals 64 in six consecutive summer Paralympic games as of 2021. We examined the functional and structural 65 changes in the foot section of the primary motor cortex (M1) in participant P1, as compared to able-66 bodied control participants. 67

We conducted functional and structural magnetic resonance imaging (MRI). In the functional 68 MRI experiment, prompted by the previous reports that somatotopic representation , 2020), we tested our hypothesis that 72 the M1 foot section of participant P1 was involved in sensory-motor processing of the hand, which is 73 rarely seen in able-bodied persons (Morita et al., 2021a) . In the present study, we were particularly 74 interested in the M1 because this is the executive locus of voluntary limb movement. In the structural 75 MRI experiment, we examined the change in volume of the gray matter (GM) in the foot section of 76 M1 of participant P1. It is known that long-term intensive hand/finger training, for manipulating 77 musical instruments, causes GM expansion in the M1 hand section (Gaser and Schlaug, 2003) . If the 78 M1 foot section of participant P1 had been used as the hand section through her long-term training 79 for wheelchair racing, using the upper limbs, we may expect the GM expansion (increase in GM 80 volume) in this section. In addition to these investigations, we also explored the change in white 81 matter (WM) in the brain of participant P1. If the GM of the M1 foot section of participant P1 is 82 expanding, we may also expect the development of the nerve fibers that connect between the M1 foot 83 section and other regions of the brain, resulting in expansion of WM which contains such developed 84 nerve fibers. 85

The secondary purpose of the present study was to test whether the functional and structural 86 changes expected in participant P1 are specific to this participant or common to other paraplegic 87 participants. We tried to recruit paraplegic participants who had a long-term non-use period of the 88 lower limbs and long-term wheelchair sports training and obtained three participants. One participant 89 had paraplegia at the age of one (P2), and the remaining two had paraplegia due to spinal cord injury 90 at the ages of 17 (P3) and of 21 (P4), respectively. They were acquired paraplegic persons having leg 91 non-use period of more than 30 years and long-term training for wheelchair sports (Table 1) . But 92 none of them were top athletes at the level of participant P1. 93

In both functional and structural MRI experiments, our region of interest (ROI) was the M1 94 foot section. It is known that the foot section of M1 is represented in the medial wall motor region, 95 which is closely located in the hand region of the cingulate motor area (CMA) in monkeys (He et al., 96 1995) and humans (Ehrsson et al., 2003; Naito et al., 2007; Amiez and Petrides, 2014) . In addition, 97 the trunk section of M1 is known to be closely located to the M1 foot section . 98 Therefore, we first defined the ROI in the foot section of M1 (M1 foot ROI) using the functional 99 MRI data obtained when 37 able-bodied control participants performed a right foot, right hand, and 100 trunk tasks. 101

In the functional MRI experiment, we scanned the brain activity when the four paraplegic 102 participants (P1, P2, P3, and P4), and the control participants performed bimanual wrist extension-103 flexion movements. We selected a bimanual task because moving both hands and arms 104 simultaneously is essential movement for pedaling a wheelchair. We first directly compared the brain 105 activity obtained from each paraplegic participant (P1, P2, P3, or P4) to that of the control 106 participants to explore regions, in which a paraplegic participant shows significantly greater activity 107 than the control participants, within the M1 foot ROI (contrast analysis; see below). Next, we tested 108

if the activity obtained from the ROI and the number of activated voxels identified in the ROI were 109 significantly greater in each paraplegic participant when compared with the control participants (ROI 110 analysis; see below). 111

In the structural MRI experiment, we collected the structural MRI images from all the 112 participants and performed voxel-based morphometry (VBM) analysis. We first explored regions, in 113 which a paraplegic participant shows significant change in GM volume as compared to the control 114 participants, within the M1 foot ROI (contrast analysis; see below limbs; nonetheless, there were somatic sensations (light touch and pin prick). These were evaluated 137 by a physiotherapist with more than 10 years of experience (NK, one of the authors). We confirmed 138 the handedness of the participants using the Edinburgh Handedness Inventory (Oldfield, 1971) , and 139 participants P1, P3, and P4 were right-handers, and participant P2 was ambidextrous (Table 1) . 140

Regarding the control participants, we recruited right-handed and -footed able-bodied adults (n = 37; 141

37.4 ± 10.9 [mean ± standard deviation] years old, range 25-59 years old, 25 female). They had 142 experience in various sports since their school days, but none of them were athletes participating in a 143 particular sport. We confirmed the handedness of the control participants using the Edinburgh 144

Handedness Inventory (95.9 ± 7.6, Oldfield, 1971 and at the end of entire experiment, we collected the structural MRI data from all participants. In the 158 functional MRI experiment, to define the foot section of M1, the 37 control participants performed 159 right foot, right hand, and trunk tasks (Figure 1(A) ), in addition to a main bimanual task ( Figure  160 2(A)). The paraplegic participants performed the bimanual task. After the bimanual task, both the 161 control and paraplegic participants also performed right-hand active and passive movement tasks, 162

which we intend to report in a future paper. 163

Before the fMRI experiment, we explained the tasks to be performed in the scanner to every 164 participant, and they experienced the tasks outside the MRI room to familiarize themselves with the 165 tasks. Thereafter, the participants entered the room and were placed in the MRI scanner. Their heads 166

were immobilized using sponge cushions and an adhesive tape, and their ears were plugged. The 167 participant's body parts (chest, pelvis, and shin) of the participants were fixed to the MRI bed using 168

Velcro to reduce their body movements during the task. When performing a task, the participants 169 were asked to close their eyes, relax their entire body, refrain from producing unnecessary 170 movements, and only think of the assigned task. 171

Each participant completed one experimental 160-s run for each task. The run comprised five task 172 epochs, each lasting 15 s (Figure 2(B) ). Considering each epoch, the participants continuously 173 exerted cyclic movements for each task in synchronization with cyclic audio tones. The details of 174 each task are described below. The task epochs were separated by 15-s baseline (rest) periods. Each 175 run also included a 25-s baseline period before the start of the first epoch. During the experimental 176 run, we provided the participants with auditory instructions that indicated the start of a task epoch 177 (three, two, one, start). We also provided a 'stop' instruction generated by a computer to notify the 178 participants of the end of each epoch. The participants heard the same cyclic audio tones; however, 179 they did not generate any movement during the rest periods. All the auditory stimuli were provided 180 through an MRI-compatible headphone. An experimenter who stood beside the scanner bed checked 181

if the participants were performing each task properly by visual inspection throughout the run. 182

Functional MRI images were acquired using T2*-weighted gradient echo-planar imaging (EPI) 184 sequences with a 3.0-Tesla MRI scanner (MAGNETOM Trio Tim; Siemens, Germany) and 32-185 channel array coil. We used a multiband imaging technique (multiband factor = 3), which was used 186 in our previous study (Amemiya et al., 2021) . Each volume consisted of 48 slices (slice thickness = 187

3.0 mm) acquired in an interleaved manner, covering the entire brain. The time interval between the 188 successive acquisitions from the same slice was 1000 ms. An echo time of 27 ms and a flip angle of 189 60 ° were used. The field of view was 192 × 192 mm, and the matrix size was 64 × 64 pixels. The 190 voxel dimensions were 3 × 3 × 3 mm along the x-, y-, and z-axes, respectively. We collected 160 191 volumes for each experimental run. 192

Regarding the structural MRI image, a T1-weighted magnetization-prepared rapid gradient echo 193

(MP-RAGE) image was acquired using the same scanner for each participant, which was used in the 194 following voxel-based morphometry (VBM) analysis. The imaging parameters were as follows: TR 195 =1900 ms, TE = 2.48 ms; FA = 9 °, field of view = 256 × 256 mm 2 , matrix size =256 × 256 pixels, 196 slice thickness = 1.0 mm; voxel size = 1 × 1 × 1 mm 3 , and 208 contiguous transverse slices. 197

We prepared the following three tasks (Figure 1 (A)) to functionally define the foot section of M1 in 199 the control participants. 200

The control participants performed alternating dorsi-and plantar-flexions of the right foot at a 202 frequency of 1 Hz (left panel in Figure 1 (A)). They were required to continuously perform these 203 movements in synchronization with 1-Hz cyclic tones while relaxing their left foot. A supporter was 204 placed under the right calf and the right leg was lifted off the bed to enable the participants to 205 generate these movements without their right heel touching the bed. The participants were instructed 206 to perform right foot movements within the range of their maximum dorsi-and plantar-flexion 207

angles. These were measured outside the scanner (average range of motion across participants was 208 approximately 68.6 ± 15.5 °). This task was used to identify the brain regions that were associated 209 with foot movement. 210

The control participants continuously exerted cyclic extension-flexion movements of their right wrist 212

in synchronization with 1-Hz cyclic tones. We prepared a device to control the range of wrist motion 213 (middle panel in Figure 1 (A)), which was used in our previous study (Morita et al., 2021a) . A 214 movable hand-rest was mounted on the device (middle panel in Figure 1 (A)), and the hand was fixed 215 on the hand-rest that indicated the wrist angle. Two stoppers were fixed onto the device to control the 216 range of the wrist motion across the task epochs and participants. They were positioned to prevent 217 the wrist from extending beyond the straight (0 °) position and flexing beyond 60 °. The participants 218 had to touch one of the stoppers (0 ° or 60 °) alternately with the hand-rest in synchronization with 219 the 1-Hz audio tones while making controlled and continuous wrist extension-flexion movements. 220

An example of kinematic data when performing this task is shown in (Morita et al., 2021a) . This task 221 was used to depict foot-specific M1 section distinctly from the CMA hand region. This is because the 222 foot section of M1 and the CMA hand region are closely represented in the medial wall (see 223 Introduction). 224

The control participants pushed up a 4-kg weight placed on their abdomen (right panel in Figure  226 1(A)). They were asked to repeatedly perform a set of push-ups and immediate relaxation in 227 synchronization with 0.8 Hz of tones, without moving their heads, and 0.8 Hz was chosen because in 228 our pilot experiment, some participants reported that 1-Hz was too fast. They indicated that 0.8 Hz 229

was comfortable to follow the movements. The participants were instructed to keep their breathing as 230 normal as possible during the scanning. The weight was placed on their abdomen at the start of each 231 task epoch started and removed when the epoch was completed. This task was used to identify the 232 foot-specific M1 section, distinctly from the M1 trunk section, because the foot movement could 233 potentially co-activated the M1 trunk section . 234

All the participants continuously exerted cyclic extension-flexion movements of their left and right 236 wrists in synchronization with the 1-Hz tones (Figure 2(A) ). The participants generated in-phase 237 extension-flexion movements of both hands. The range of the wrist motion was between 0 ° and 60 ° 238 as shown in the right-hand task (see above). To control the range of motion, the same device, used in 239 the right hand task, was used for each of the left and right hands. 240

To eliminate the effects of unsteady magnetization during the tasks, the first ten EPI images in each 242 fMRI run were discarded. The imaging data were analyzed using SPM 12 (Wellcome Centre for  243 Human Neuroimaging, London, UK) implemented in MATLAB (MathWorks, Sherborn, MA, USA). 244

The following preprocessing was done for each participant. SPM default parameters were used unless 245 otherwise specified. First, all of the EPI images were aligned to the first EPI image of the first session 246

with six degrees-of-freedom (translation and rotation about x-, y-, z-axes) rigid displacement. 247

Through this realignment procedure, we obtained the data related to the position of the head that 248 changed over time from the first frame and through the six parameters. All the participants had a 249 maximum displacement of less than 1.5 mm in the x-, y-, or z-plane and less than 0.1° of angular 250 rotation about each axis during each fMRI run. These values were comparable to those from our 251 previous study (see Morita et al., 2019) , and thus no data were excluded from the analysis. The T1-252 weighted structural image of each participant was co-registered to the mean image of all the realigned 253 EPI images by using affine transformation. Finally, the structural image and the realigned EPI images 254 were spatially normalized to the standard stereotactic Montreal Neurological Institute (MNI) space 255 (Evans et al., 1994) . Normalization parameters to align the structural image to MNI template brain 256

were calculated using the SPM12 normalization algorithm. The same parameters were used to 257 transform the realigned EPI images. The normalized EPI images were resliced to 2-mm isotropic 258 resolution, and the successful alignment was visually checked. Finally, the normalized images were 259 spatially smoothed using a Gaussian kernel with a full width at half maximum of 4 mm along the x-, 260 y-, and z-axes. 261

Following the preprocessing, we used a general linear model Worsley and 262 Friston, 1995) to analyze the fMRI data. We prepared a design matrix for each participant. 263

Considering this single-subject analysis, the design matrix contained a boxcar function for the task 264 epoch in the run, which was convolved with a canonical hemodynamic response function. To correct 265 the residual motion-related variance after the realignment, the six realignment parameters were also 266 included in the design matrix as regressors of no interest. In the analysis, we did not perform global 267 mean scaling to avoid inducing Type I errors in the evaluation of negative blood oxygenation level-268 dependent (BOLD) responses (deactivation) (Aguirre et al., 1998) . We generated an image showing 269 the task-related activity in each task for each participant, which was used in the subsequent analyses. 270

In this image, the effect of the cyclic tones was most likely eliminated because the participants heard 271 the sound consistently during the task epochs and rest periods. 272

The procedure is summarized in Figure 1 (B). We performed a second-level group analysis (Holmes 274 and Friston, 1998) using a one-sample t-test to identify the brain regions that were significantly 275 activated during the right foot task in the control participants as a whole. Here, we depicted the 276 regions exclusively activated during the right foot task. For this purpose, we identified the brain 277 regions that were activated during the right foot task, but not during the right hand and trunk tasks for 278 the group of the control participants. This was carried out using two exclusive masks of images 279

showing the right hand and trunk task-related activities. As for a mask image of the right hand task-280 related activity, we generated a mask image by performing one-sample t-tests to depict the brain 281 regions that were active during the right hand task using height threshold of p < 0.05 uncorrected. 282

The same analysis was done to generate a mask image of trunk task-related activity. We used these 283 two mask images (union of these two images) to exclude the regions in which activity increased 284 during the right hand and/or trunk task. We used the family wise error rate (FWE)-corrected extent 285 threshold (p < 0.05) in the entire brain for a voxel-cluster image generated at an uncorrected height 286 threshold of p < 0.005. We found a significant cluster of active voxels in the left (contralateral) 287 medial wall motor regions (foot section in the left hemisphere). 288

To define the M1 in the medial wall motor regions, we used the cytoarchitectonic maps for areas 4a 289 and 4p implemented in the SPM Anatomy toolbox (Eickhoff et al., 2005) . We defined the left M1 290 foot section (the foot section of the left M1) by depicting the overlapped region between the 291 functional cluster in the left medial wall and the cytoarchitectonic maps for areas 4a and 4p. We 292 checked the individual activity in the left M1 foot section during the right foot, hand, and trunk tasks 293 (see Supplementary Material and Figure S1 ). We also defined the right M1 foot section. We first 294 flipped the cluster horizontally from the left medial wall to the right hemisphere. Thereafter, we 295 defined the putative right M1 foot section by identifying the overlapped region between the flipped 296 cluster in the right medial wall and the cytoarchitectonic maps for areas 4a and 4p. Finally, we 297 defined the M1 foot ROI (845 voxels, voxel size = 2 × 2 × 2 mm) by combining the left and right M1 298 foot sections. 299 2.5.2 Functional Analysis of bimanual task 300

First, we explored regions, in which a paraplegic participant showed significantly greater activity 302 than the control participants during the bimanual task, within the M1 foot ROI. In this analysis, we 303 directly compared the brain activity obtained from each paraplegic participant to that of the control 304 participants (Figure 2(C) ). This was a one-to-many two-sample t-test that was identical to Crawford 305

and Howell t-test (Mühlau et al. 2009; Boucard et al. 2015) , which is a modification of the regular 306 independent two-sample t-test, allowing us to compare one sample with a group of multiple samples, 307

where the within-group variance is estimated from the latter group by assuming the within-group 308 variances in two groups are identical (Crawford and Howell, 1998) . In this analysis, we included the 309 age and sex of all participants as nuisance covariates (effect of no interest) because these factors 310

could have an influence on the evaluation of present between-group difference. This approach has 311 been widely used in a GLM analysis to evaluate the target factor by considering effects of the 312 nuisance covariates (e.g., Morita for a voxel-cluster image generated at an uncorrected height threshold of p < 0.005). 315

In addition to the above contrast analysis, we performed ROI analysis. We first tested if the activity 317 obtained from the M1 foot ROI during the bimanual task was significantly greater in a paraplegic 318 participant compared to the control participants (Figure 3(A) ). We calculated mean activity of the 319 voxels in the entire M1 foot ROI in each participant. Next, we tested if the number of activated 320 voxels identified in the ROI was significantly greater in a paraplegic participant compared to the 321 control participants. We counted the number of activated voxels (height threshold of p < 0.005 (T >  322 2.61)) in the M1 foot ROI for each participant (Figure 3 (B), (C)). In each ROI analysis, we 323 performed the Crawford and Howell t-test with Bonferroni correction (Crawford and Howell, 1998; 324 Crawford and Garthwaite, 2012) to compare the data obtained from each paraplegic participant with 325 that of the control participants (n = 37). Given that this test was repeated for four paraplegic 326 participants, we corrected the p-values based on the number of test repetitions (n = 4). We used the 327 threshold of p < 0.05 with Bonferroni correction. Finally, prompted by our previous report that the 328 M1 foot section in able-bodied persons are deactivated during hand movement (Morita et al., 2021a), 329

we also checked for any significant decrease in activity (less than 0) in the entire M1 foot ROI during 330 the bimanual task in the control participants as a whole by performing one-sample t-test ( Figure  331 3(A)). 332 333

A voxel-based morphometry (VBM) analysis was performed to examine the change in GM volume 336

in the M1 foot ROI in each paraplegic participant, compared to the control participants. First, we 337 visually inspected the anatomical images of all the participants and confirmed the absence of 338 observable structural abnormalities and motion artifacts. Subsequently, the data were processed using 339

Statistical Parametric Mapping (SPM12, Wellcome Centre for Human Neuroimaging). The following 340 steps were performed using the default settings in SPM12 as recommended by Ashburner (2010) . 341

First, the anatomical image obtained from each participant was segmented into GM, WM, 342

cerebrospinal fluid (CSF), and non-brain parts. Next, using Diffeomorphic Anatomical Registration 343

Through Exponentiated Lie Algebra (DARTEL), we generated GM and WM DARTEL templates 344 based on the anatomical images obtained from all the participants. Thereafter, we applied an affine 345 transformation to the GM and WM DARTEL templates to align them with their tissue probability 346 maps in the Montreal Neurological Institute (MNI) standard space. Subsequently, a segmented GM 347 image from each participant was warped non-linearly to the GM DARTEL template in the MNI 348 space (spatial normalization). The warped image was modulated by Jacobian determinants of the 349 deformation field to preserve the relative GM volume, even after spatial normalization. The 350 modulated image of each participant was smoothed with an 8-mm FWHM Gaussian kernel and 351 resampled to a resolution of 1.5 × 1.5 × 1.5 mm 3 voxel size. The methods described above were also 352 used in our previous study (Morita et al., 2021b) . 353

As in the contrast analysis for functional data, we performed one-to-many two-sample t-test by 354 directly comparing the GM volume obtained from each paraplegic participant to that of the control 355 participants (Figure 4(A) ). This analysis has been repeatedly used in the VBM analysis ( for a voxel-cluster image generated at an uncorrected height threshold of p < 0.005 (T > 2.73)). In control participants, using the same procedure described above. 370

In addition to the contrast analysis, we performed a ROI analysis. We tested if the GM volume of the 372 ROI in a paraplegic participant was significantly different (increase or decrease) from the control 373 participants (Figure 4(C) ). We calculated the GM volume of the M1 foot ROI in each participant. 374

According to Whitwell repetitions (n = 4). We used the threshold of p < 0.05 with Bonferroni correction. 381

The procedure for conducting the analysis prior to spatial normalization was identical to that in the 383 GM volume analysis. For spatial normalization, a segmented WM image from each participant was 384 warped non-linearly to the WM DARTEL template in the MNI space. The warped image was 385 modulated by Jacobian determinants of the deformation field to preserve the relative WM volume, even after spatial normalization. As in the GM analysis, the modulated image of each participant was 387 smoothed with an 8-mm FWHM Gaussian kernel and resampled to a resolution of 1.5 × 1.5 × 1.5 388 mm 3 voxel size. 389

As we used in the GM volume analysis, one-to-many two-sample t-test was performed by directly 390 comparing the WM volume obtained from each paraplegic participant to that of the control 391 participants ( Figure 5 ). We also included age, sex, and the intracranial volume as the nuisance 392 covariates in the analysis (see above). To restrict the search volume within the WM, we used a WM 393 mask image that was created based on our present data using the SPM Masking Toolbox (Ridgway et  394 al., 2009). Therefore, the voxels outside this mask were excluded from the analysis. We searched for 395 significant clusters of voxels showing WM expansion (increase of WM volume) in each paraplegic 396 participant when compared with the control participants. Because there was no specific anatomical 397 hypothesis available for this analysis, we explored the brain regions showing WM expansion in the 398 entire brain space. We used the FWE-corrected extent threshold of p < 0.05 in the entire brain for a 399 voxel-cluster image generated at the uncorrected height threshold of p < 0.005. We also checked for 400 the brain region showing WM atrophy (decrease of WM volume) in the entire brain, in each 401 paraplegic participant, as compared to the control participants using the procedure as described above. 402

To visualize the individual WM volume in the significant cluster identified in participant P1, 403

we calculated the WM volume of the cluster in each participant. As we did in the ROI analysis for 404 GM volume, we calculated proportion of WM volume of the significant cluster to the intracranial 405 volume in each participant (see above Results 410

When the brain activity in each paraplegic participant was directly compared to that of the control 412 participants, it was found that two of the paraplegic participants had a significant cluster of voxels 413

showing greater activity within the M1 foot ROI (Figure 2 (C)). One was participant P1, who had a 414 significant cluster (peak coordinates: x, y, and z = 8, -24, and 62; T = 10.15, 423 voxels) in the M1 415 foot ROI. The other was participant P4, who also showed a significant cluster (peak coordinates: x, y, 416 and z = 2, -32, and 64; T = 7.35; 136 voxels). Importantly, the significant clusters identified in these 417 participants were located in the precentral region (Figure 2(C) ), though the M1 foot ROI seemed to 418 extend to the right postcentral region. Participants P2 and P3 did not show any significant clusters. 419

Participant P2 merely had a total of 30 voxels having T-value greater than 2.73 (which corresponded 420 to height threshold p < 0.005) within the ROI (not shown in Figure 2 (C)). Participant P3 had no such 421 voxels in the ROI. 422

When we looked at the individual activity obtained from the M1 foot ROI, participants P1 and P4 423

showed an increase in activity beyond the range of distribution of the control data (Figure 3(A) ). 424

Indeed, when we performed the Crawford (Figure 3(A) ). One-sample t-431 test revealed significant decrease in activity in the control participants as a whole (t(36) = 3.14, p = 432 0.003). 433

Next, when the number of activated voxels identified in the M1 foot ROI (845 voxels) were counted, 434 724, 234, 32, and 582 voxels were found in the participant P1, P2, P3, and P4, respectively ( Figure  435 3(B), (C)). Participants P1 and P4 showed greater number of activated voxels beyond the range of 436 distribution of the control data (Figure 3(B) ). The Crawford and Howell t-test showed that the 437 numbers of the activated voxels in the participants P1 and P4 were significantly larger than those in 438 the control participants (P1; t(36) = 7.46, p = 1.7 x 10 -8 after Bonferroni correction, P4; t(36) = 5.86, 439 p = 2.1 x 10 -6 after Bonferroni correction). The number of the activated voxels in the participants P2 440 and P3 was not significantly different from those in the control participants (P2; t(36) = 1.94, p = 441 0.12 after Bonferroni correction, P3; t(36) = −0.33, p > 1 after Bonferroni correction). These results 442

(the significant differences only in P1 and P4) were replicated when we counted the number of 443 activated voxels using two different height thresholds of p < 0.01 (T > 2.35) and of p < 0.001 (T >  444 3.15) (not shown in Figure 3 (B)). 445

Viewed collectively, the series of analyses consistently showed that the participants P1 and P4 had 446 significantly greater activity in the M1 foot ROI during the bimanual task as compared to the control 447 participants. 448 449

When we explored the GM expansion (= increase in GM volume) in the M1 foot ROI in each 451 paraplegic participant as compared to the control participants, two of the paraplegic participants (P1 452

and P2) showed a significant cluster of voxels showing GM expansion in the ROI (Figure 4(A) ). As 453 in the functional clusters (Figure 2(C) ), the significant clusters identified in these participants were 454 located in the precentral region. In participant P1, the significant cluster of voxels showing GM 455 expansion was observed in the right side of the M1 foot ROI (peak coordinates = 6, -23, 54; T = 4.82; 456 240 voxels; Figure 4 (A)). Importantly, 75% of the expanded region (gray section in Figure 4 (B)) 457 overlapped with the region in which the participant showed a significant cluster of voxels showing 458 greater activity during the bimanual task than the control participants (yellow and gray sections in 459 Figure 4(B) ). When we calculated the dice coefficient (range from 0 to 1, with 1 meaning complete 460 overlap) to evaluate the spatial overlap, the value was 0.29. In participant P2, the significant cluster 461 of voxels showing GM expansion was observed in the left side of the M1 foot ROI (peak coordinates 462 = -15, -33, 69; T = 6.10; 395 voxels; Figure 4 (A)). Neither of the participants P3 and P4 showed 463 voxels having T-value greater than 2.73 (which corresponded to height threshold p < 0.005) in the 464 ROI. 465

We also examined the GM atrophy (= decrease in GM volume) in the M1 foot ROI. None of the 466 paraplegic participants showed any significant clusters of voxels showing a decrease of GM volume 467 within the ROI, compared to the control participants. A non-significant cluster of voxels (103 voxels  468 having T-value greater than 2.73) was found in participant P4 (not shown in Figure 4 (A)), and none 469 of the other participants (P1, P2 and P3) had such voxels in the ROI. 470

When we looked at the individual GM volume of the M1 foot ROI, none of paraplegic participants 471 showed a significant increase or decrease in GM volume when compared with the control participants, though participant P4 showed a relatively lower value of GM volume (Figure 4(C) ). 473

Thus, even though a significant cluster of voxels showing GM expansion was found in participants 474 P1 and P2 (Figure 4(A) ), the expansion was most likely localized in a limited section of the M1 foot 475 ROI, and the GM expansion was not observed throughout the M1 foot ROI. 476

Finally, when we explored the WM expansion (= increase of WM volume) in the entire brain, 478 participant P1 showed significant clusters of voxels indicating WM expansion in the bilateral medial 479 frontal regions (Figure 5(A) ) and in the bilateral structures along the optic radiation ( Figure 5(B) ). 480

Anatomical locations of these clusters are shown in Table 2 . None of the other participants (P2, P3, 481 and P4) exhibited significant WM expansion in any region of the brain. The right medial frontal 482 cluster seemed to be connected to the right M1 foot section in which the GM expansion was observed 483

(white section in Figure 5(A) ). When we visually inspected the individual WM volume in the cluster 484 identified in participant P1, the data obtained from participants P2, P3, and P4 were within the range 485 of distribution of the control data, whereas participant P1 showed the largest value beyond the range 486 of distribution of the control data in both frontal (right panel in Figure 5 (A)) and occipital (right 487 panel in Figure 5 (B)) clusters of the bilateral hemispheres of the brain. Finally, none of the 488 participants showed significant WM atrophy in any of the regions of the brain. 489 490 4

Discussion 491

Compared to the 37 control participants, participant P1 had a significant cluster of voxels showing 492 greater activity within the M1 foot ROI (Figure 2 (C)) during the bimanual task, and a significant 493 cluster of voxels showing GM expansion within the M1 foot ROI (Figure 4(A) ). Thus, functional and 494 structural changes were observed in participant P1 as expected. A significant cluster within the M1 495

foot ROI was also observed in participant P4, during the bimanual task, but not in P2 and P3 ( Figure  496 2(C)). On the other hand, a significant cluster showing GM expansion within the ROI was observed 497 in participant P2, but not in P3 and P4 (Figure 4(A) ). Hence, the significant functional or structural 498 change was not always observable in all the paraplegic participants, and only participant P1 exhibited 499 both functional and structural changes in the M1 foot ROI. 500

There are limitations to the current study. First, we could only recruit a limited number of paraplegic 501 participants owing to the restrictions imposed by COVID-19 although we could recruit a relatively 502 larger number of able-bodied control participants. The number of participants was too small to 503 conclude the difference between congenital and acquired paraplegia, the effect of the type of 504 wheelchair sport, or the effect of the duration of wheelchair sport training. In addition, there were 505 differences in the age and sex among the paraplegic participants. And their ages were somewhat 506 higher than those of the control participants, though neither participant's age was significantly 507 different from the control group (P1; p > 1, P2; p = 0.08, P3; p = 0.29, P4; p = 0.40 after Bonferroni 508 correction). As for the M1 foot ROI, we should have mapped the right M1 foot section by actually 509 measuring brain activity during a left foot task. In addition, it should be borne in mind that the M1 510 foot section of paraplegic individuals could be different from that of the control participants in terms 511 of its size and location. Finally, collecting more data per participant may have increased reliability of 512 the data. Despite these limitations, this study has several implications, as discussed below. 513 4.1 Using M1 foot section for sensory-motor processing of the hand 515

As reported in our previous study (Morita et al., 2021a) , we confirmed a significant decrease in 516 activity of the M1 foot ROI during the bimanual task in the control participants as a whole ( Figure  517 3(A)). Such activity decrease can be considered as cross-somatotopic inhibition (Zeharia et al., 2012; 518 Morita et al., 2021a; Naito et al., 2021) , in which the brain tries to suppress the occurrence of an 519 unintended foot movement during hand movement. In contrast, in participant P1, we found 520 significantly greater activity in the ROI (precentral region) during the task when compared to the 521 control participants (Figure 2(C) ). In the ROI analysis, we also confirmed significantly greater 522 activity (Figure 3(A) ) and significantly greater number of activated voxels (Figure 3(B 2014). If we consider this evidence, we may speculate that in participant P1 with congenital 541 paraplegia, the foot section had not developed as a foot section and developed as a section for other 542 body parts (e.g., hand) since the fetal stage, and that in participant P2, the foot representation was still 543 immature when he afflicted with paraplegia at the age of one. 544

In contrast, in the case of participant P4, the foot section would have developed as a normal foot 545 section until being afflicted with paraplegia at the age of 21. Thus, the use of the M1 foot section for 546 sensory-motor processing of the hand (Figures 2(C) and 3) should be considered as reorganization of 547 the already acquired foot representation after being affected by paraplegia, and this reorganization is 548 likely due to long-term training of upper limbs through wheelchair sports training. On the other hand, 549 participant P3, whose foot section would also have developed as normal foot section until being 550 afflicted with paraplegia at the age of 17, did not use the M1 foot section for sensory-motor 551 processing of the hand (Figures 2(C) and 3) . The difference between participants P3 and P4 implies 552 the possibility that the content of wheelchair sports training (purposes of using the hands) could be an 553 important factor to promote the use of the M1 foot section for sensory-motor processing of the hand. 554

From this perspective, we could point out that both participants P1 and P4 had long-term training for 555 wheelchair track racing and marathon ( Table 1 ) that are the sports in which the hands are used only 556

for wheelchair mobility. The M1 foot section in an able-bodied individual is mainly used for mobility 557 purposes (i.e., locomotion). Hence, it would be interesting to see if intensive training through the use 558 of the hands for this purpose promotes the conversion of the M1 foot section into the hand section. 559

However, these observations would be considered to be in the realm of speculation until they are 560 verified by future studies. 561

In the case of individuals born without one hand, it has been reported that the cerebellar hand section 562

for the missing hand, in addition to the hand section of the primary sensorimotor cortices, is involved 563 in the sensory-motor processing of the foot (Hahamy et al., 2017; Hahamy and Makin, 2019). 564

However, in the present study, none of our paraplegic participants showed a significant increase in 565 the activity in the cerebellar foot section during the bimanual task (see Supplementary Material and 566 Figure S2 ). This suggests the possibility of a difference in the latent potential for the conversion of 567 somatotopic representation between the hand and foot sections in the cerebellum. In addition, it is 568 known that long-term training of the hand (Krings et al., 2000) or foot (Naito and Hirose, 2014) 569 movement may facilitate efficient recruitment of the brain activity (show reduction of brain activity) 570

in its corresponding somatotopic sections. However, contrary to these previous reports, none of our 571 paraplegic participants efficiently recruited (showed less) activity in the hand sections of the multiple 572 motor areas (the M1, CMA, and cerebellum) during the bimanual task (see Supplementary Material  573 and Figure S3 ), although they had long-term training of the upper limbs during wheelchair sports 574 training (Table 1) . Further investigations are required to generalize these findings to a larger 575 population of individuals with long-term wheelchair sports training. 576 577

As expected, the contrast analysis revealed significant GM expansion within the M1 foot ROI in 579 participant P1 when compared with the control participants (Figure 4(A) ). Such significant GM 580 expansion was also observed in participant P2, but not in P3 and P4 (Figure 4(A) ). Hence, similar to 581 the above functional changes, the GM expansion within the M1 foot ROI may occur even in an 582 individual with acquired paraplegia, but does not always occur in all individuals with acquired 583 paraplegia even though they have long-term non-use period of the lower limbs and long-term 584 wheelchair sports training. 585

Importantly, such GM expansion was localized in a limited portion of the M1 foot ROI 586 (Figure 4(A) ), and the expansion was not observed throughout the M1 foot ROI as shown in the ROI 587 analysis (Figure 4(C) ). In addition, we could point out that the GM expansion was observed in the 588 right precentral region in participant P1, while it was observed in the left region in participant P2, 589 though we have no clear reasons for the difference. Although the exact physiological changes 590 underlying GM expansion are still unknown, axon sprouting, dendritic branching synaptogenesis, 591 neurogenesis, and changes in the glial number and morphology are suggested to be important 592 contributors to GM expansion (Zatorre et al., 2012) . Hence, our data suggest that these changes 593 would have occurred in the localized precentral regions of participants P1 and P2. 594

The finding that significant GM expansion was observed in participants P1 and P2 but not in 595 P3 and P4 (Figure 4(A) ) leads us to conjecture that GM expansion is likely to occur in persons who 596 became paraplegic at a very early stage of development. From the developmental perspective, it 597

should also be noted that neither of the participants P3 and P4, who became paraplegic as a result of 598 spinal cord injury during adulthood, showed significant GM atrophy in the M1 foot ROI ( Figure  599 4(C)). Despite GM atrophy is often reported in the primary sensory-motor cortices after spinal cord It is possible that GM atrophy would have been observed if the brains of participants P3 and P4 were 602 scanned immediately following their spinal cord injury. Therefore, there is a possibility that long-603 term wheelchair sports training might have contributed to improve the putative GM atrophy. In the 604 case of participant P4, since his M1 foot section was involved in sensory-motor processing of the 605 hand (Figures 2(C) and 3) , long-term training of the upper limbs through wheelchair sports training 606 could have restored his putative GM atrophy. In the case of participant P3 who did not use the M1 607 foot section for sensory-motor processing of the hand (Figures 2(C) and 3) , long-term training of 608 other body parts through wheelchair sports training could have restored the putative atrophy. 609

In participant P1, 75% of the GM-expanded region (gray section in Figure 4 (B)) overlapped 610 with the region in which the participant showed a significantly greater activity than the control 611 participants (yellow and gray sections in Figure 4 (B)) during the bimanual task. Such consistency 612 between functional and structural changes were only observed in participant P1. Hence, in participant 613 P1, the GM was expanded in the M1 foot section that was used for sensory-motor processing of the 614

hand. If we consider that GM expansion is deeply associated with use-dependent plasticity (Gaser 615 and Figure S4 ), suggesting that our 624 paraplegic participants use these hand sections to the same degree as the control participants use 625 them. Similarly, none of our paraplegic participants showed significant GM expansion in the 626 cerebellar foot section ( Figure S4 ). 627

Only participant P1 showed significant WM expansion in the bilateral medial frontal regions and in 629 the bilateral structures along the optic radiation ( Figure 5 ). Although the exact physiological changes 630 underlying WM expansion are not fully understood, white matter changes are thought to be related to 631 changes in the number of axons, axon diameter, the packing density of fibers, axon branching, axon 632 trajectories and myelination (Zatorre et al., 2012) . As with GM volume, use-dependent plasticity can 633 also be an important factor for WM expansion (Scholz et al., 2009 ). Thus, the long-term wheelchair 634 racing training must be associated with the WM expansion in this participant. 635 WM expansion in the bilateral structures along the optic radiation are most likely associated with 636 visual functions, which implies the possibility of the intensive use of visual system by participant P1. 637

On the other hand, the medial frontal regions are likely to contain nerve fibers that connect the 638 medial prefrontal cortex and the medial wall motor regions, including the M1 foot sections (probably 639 the superior branch of the superior longitudinal fasciculus; Parlatini et al., 2017) . Indeed, the right 640 medial frontal cluster seemed to connect to the right M1 foot section in which the participant showed 641 GM expansion (white section in Figure 5(A) ). Evidence has been accumulating to support the view 642 that the medial prefrontal cortex is important for generating highly-motivated behaviors ( 

In the brain of participant P1, the M1 foot section was used for sensory-motor processing of the hand 649 (Figures 2(C) and 3) , and this claim was further corroborated by the GM expansion in the foot 650 section that was activated during the hand sensory-motor processing (Figure 4(A), (B) ). The M1 foot 651 section is normally deactivated during the hand sensory-motor processing in able-bodied people 652 (Figure 3(A) ; . Hence, in this participant, the M1 foot section, which is normally 653 suppressed during the hand sensory-motor processing, has drastically changed to become used for 654 this sensory-motor processing other than its original function of foot motor control. We want to 655

propose that such phenomenon (a brain region that is not normally involved in a certain information 656

processing, but is rather suppressed during this information processing, in able-bodied persons, is 657 chronically disinhibited and is regularly used during this information processing), could be better 658 referred to hyper-adaptation. Because such adaptation may represent extreme adaptability of the 659 human brain, which is rarely seen in typically developed individuals and should be distinct from 660 normal adaptions as we often observe in force-field (e.g., Shadmehr and Mussa-Ivaldi et al., 1994) 661

and visuo-motor (e.g., Imamizu et al., 2000) adaptations and so on. 662

The use of the foot section for the hand sensory-motor processing was also observed in participant P4 663 (Figures 2(C) and 3), though no supportive evidence of significant GM expansion was observed in 664 this participant. Thus, it was likely that hyper-adaptation might also occur in this participant. In 665 participants P2 and P3, the activity in the M1 foot section during the bimanual task was comparable 666

to that observed in the control participants (Figures 2(C) and 3). However, this does not guarantee 667 that hyper-adaptation does not occur in their brains because their M1 foot sections might be used for 668

other sensory-motor processing rather than the hand, which was not directly assessed in the current 669 work. 670

Based on the above definition, tactile processing in the visual cortex of some blind individuals (e.g., 671 Sadato et al., 1996 Sadato et al., , 2002 ) could also be another example of hyper-adaptation because the human 672 visual cortex is normally suppressed during sensory-motor processing in sighted people (cross-modal 673 inhibition; Morita et al., 2019), but the visual cortex of some blind individuals has drastically 674 changed to become involved in tactile processing other than its original function of visual processing. 675

Hyper-adaptation may also occur when the brain manages to adapt to irreversible and severe injuries 676 and diseases through long-term sensory, motor and cognitive trainings. Further researches will reveal 677 what kind of adaptation should be termed hyper-adaptation, and neural mechanisms and psycho-678 behavioral factors that trigger and promote hyper-adaptation. 679 680 5

The authors declare that the research was conducted in the absence of any commercial or financial 682 relationships that could be construed as a potential conflict of interest. Sakamoto for giving us the opportunity to investigate the brain of the top wheelchair racing 698

Paralympian. We would like to thank Editage for the English language editing. 699

Data Availability Statement 700

The data that support the findings of this study are available on request from the corresponding 701

author. The data are not publicly available because they contain information that can compromise the 702 privacy of the participants in this study. 703

The study was conducted in accordance with the guidelines of the Declaration of Helsinki (1975) right hand, and trunk tasks to identify the brain regions that are activated during the right foot task, 885

but not during the right hand and trunk tasks. Please see the text for details. (C): Results from one-to-many two-sample t-test. in participants P1 and P4, but not in P2 and P3, were significantly larger than those in the control 907 participants. (A, B): Red diamonds, blue rectangles, orange circles, and pink triangles represent the 908 data obtained from paraplegic participants P1, P2, P3, and P4, respectively. Gray dots represent the 909 data obtained from the control participants. The plotted points for the control participants are 910 horizontally jittered to avoid over-plotting. Asterisks indicate significant differences from the contr 911 data (** p < 0.001, * p < 0.05 with Bonferroni correction). (C): Voxels having T-value greater than 912 2.61 (height threshold p < 0.005) within the M1 foot ROI in each participant. Please note that 913 participants are listed in order of greater number of activated voxels, from top to bottom. These 914 voxels are superimposed on the sagittal slices (x = −6, −4, −2, 0, 2, 4, and 6) of the MNI standard 915 brain. Abbreviations: M1, primary motor cortex; ROI, region-of-interest; a.u., arbitrary unit. 

Ultrasound study of fetal movements in singleton and twin pregnancies at 12-19 weeks

The inferential impact of global signal 715 covariates in functional neuroimaging analyses

Neurological 718 and behavioral features of locomotor imagery in the blind

Neuroimaging evidence of the anatomo-functional organization 721 of the human cingulate motor areas

VBM tutorial

How does 725 the human brain deal with a spinal cord injury?

Single-case research in neuropsychology: A 728 comparison of five forms of t-test for comparing a case to controls

Comparing an individual's test score against norms 731 derived from small samples

Absence of 733 localized grey matter volume changes in the motor cortex following spinal cord injury

Somatotopic mapping of the developing sensorimotor cortex in the preterm human brain

Organized toe maps in 739 extreme foot users

Changes in 741 grey matter induced by training

Imagery of voluntary movement of fingers, toes, and 743 tongue activates corresponding body-part-specific motor representations

A 746 new SPM toolbox for combining probabilistic cytoarchitectonic maps and functional imaging data

An MRI-based probabilistic 749 atlas of neuroanatomy

Phantom-753 limb pain as a perceptual correlate of cortical reorganization following arm amputation

Phantom limb pain: a case of maladaptive 756 CNS plasticity?

MRI 758 investigation of the sensorimotor cortex and the corticospinal tract after acute spinal cord injury: a 759 prospective longitudinal study

Analysis of fMRI time-series revisited

Brain structures differ between musicians and non-musicians

Representation of multiple body parts in the missing-hand territory of congenital one-766 handers

Remapping in cerebral and cerebellar cortices Is not restricted 768 by somatotopy

Topographic organization of corticospinal projections 770 from the frontal lobe: motor areas on the medial surface of the hemisphere

Generalisability, random effects and population inference

Brain sensorimotor system 775 atrophy during the early stage of spinal cord injury in humans

Voxel-based 778 morphometry studies of personality: issue of statistical model specification--effect of nuisance 779 covariates

Neuroscience of apathy and anhedonia: a transdiagnostic 781 approach

Human cerebellar activity reflecting an acquired internal model of a new tool

Circular analysis in 786 systems neuroscience: the dangers of double dipping

Cortical 789 activation patterns during complex motor tasks in piano players and control subjects. A functional 790 magnetic resonance imaging study

Phantom movements and pain 793 An fMRI study in upper limb amputees

Reorganization in the 795 primary motor cortex after spinal cord injury -A functional Magnetic Resonance (fMRI) Study

Reduced functional connectivity in early-stage drug-naive Parkinson's disease: a 799 resting-state fMRI study

Adaptation in the motor cortex following cervical spinal cord injury

Developmental changes in task induced brain 805 deactivation in humans revealed by a motor task

Examination of the development and aging of brain 808 deactivation using a unimanual motor task

Gray-matter expansion of social brain networks in 811 individuals high in public self-consciousness

Voxel-based morphometry in individual patients: a pilot study in early huntington disease

Efficient foot motor control by Neymar's brain

Existence of interhemispheric inhibition 818 between foot sections of human primary motor cortices: Evidence from negative blood oxygenation-819 level dependent signal

Human limb-821 specific and non-limb-specific brain representations during kinesthetic illusory movements of the 822 upper and lower extremities

Cortical 825 reorganization of lower-limb motor representations in an elite archery athlete with congenital 826 amputation of both arms

Depression-like 828 state induced by low-frequency repetitive transcranial magnetic stimulation to ventral medial frontal 829 cortex in monkeys

Disruption of functional organization within the primary motor cortex in children with autism

The assessment and analysis of handedness: The Edinburgh inventory

Functional segregation and integration within fronto-parietal networks

Issues 839 with threshold masking in voxel-based morphometry of atrophied brains

Critical period for cross-modal 842 plasticity in blind humans: a functional MRI study

Activation of the primary visual cortex by Braille reading in blind subjects

When the single matters more 847 than the group: very high false positive rates in single case Voxel Based Morphometry

Training induces changes in 850 white-matter architecture

Adaptive representation of dynamics during learning 852 of a motor task

Congenitally altered motor experience 854 alters somatotopic organization of human primary motor cortex

Investigating 857 neuroanatomical features in top athletes at the single subject level

How to determine leg dominance: The agreement between self-reported and 861 observed performance in healthy adults

The role of rat medial frontal cortex in 864 effort-based decision making

Normalization of cerebral volumes 867 by use of intracranial volume: implications for longitudinal quantitative MR imaging

Analysis of fMRI time-series revisited-again

An 872 embodied brain model of the human foetus

Plasticity in gray and white: 874 neuroimaging changes in brain structure during learning

Negative blood oxygenation level dependent 877 homunculus and somatotopic information in primary motor cortex and supplementary motor area