Learning-related contraction of grey matter in rodent sensorimotor cortex is associated with adaptive myelination

From observations in rodents, it has been suggested that the cellular basis of learning-dependent changes, detected using structural magnetic resonance imaging (MRI), may be increased dendritic spine density, alterations in astrocyte volume, and adaptations within intracortical myelin. Myelin plasticity is crucial for neurological function and active myelination is required for learning and memory. However, the dynamics of myelin plasticity and how it relates to morphometric-based measurements of structural plasticity remains unknown. We used a motor skill learning paradigm to evaluate experience-dependent brain plasticity by voxel-based morphometry (VBM) in longitudinal MRI, combined with a cross-sectional immunohistochemical investigation. Whole brain VBM revealed non-linear decreases in grey matter (GM) juxtaposed to non-linear increases in white matter (WM) that were best modelled by an asymptotic time course. Using an atlas-based cortical mask, we found non-linear changes with learning in primary and secondary motor areas and in somatosensory cortex. Analysis of cross-sectional myelin immunoreactivity in forelimb somatosensory cortex confirmed an increase in myelin immunoreactivity followed by a return towards baseline levels. The absence of significant histological changes in cortical thickness further suggests that non-linear morphometric changes are likely due to changes in intracortical myelin for which morphometric WM volume (WMV) data significantly correlated with myelin immunoreactivity. Together, these observations indicate a non-linear increase of intracortical myelin during learning and support the hypothesis that myelin is a component of structural changes observed by VBM during learning.


44
Longitudinal structural magnetic resonance imaging (sMRI) of the human brain has revealed   Table S1). To discriminate the effects of motor learning from those of 147 time, the learning effects were evaluated using whole-brain VBM analysis of longitudinal sMRI data 148 on trained mice relative to non-trained controls (i.e., group by time interactions). Three different 6 regression models (linear, asymptotic and quadratic), representing three different time courses, were used and revealed statistically significant changes in both GMV and WMV (PFDR corr < 0.01). asymptotic model provided a much higher number of significant voxels than either the linear or 154 quadratic models and there were no significant linear changes in WM associated with learning 155 (Table 1). Areas well-known to be involved in motor learning were identified by VBM analysis, 156 following an asymptotic time course model in trained animals relative to non-trained controls (SI Table 1. Grey and white matter training by group interaction effects. Whole-brain between-group 161 analysis presenting the significant number of voxels (PFDR corr. < 0.01 and < 0.001) together with the 162 change in volume (mm 3

165
In addition, whole brain analysis indicated that, in some brain areas, one model fit the time courses

193
(MOp), secondary motor cortex (MOs) and primary somatosensory cortex (SSp) (Fig. 3A, B). Using 194 this cortical mask, the three different regression models were tested and compared (Table 2). This 195 analysis demonstrated that asymptotic modelling was clearly preferred and that statistically 196 significant decreases in GMV (PFDR corr < 0.001) overlapped with significant increases in WMV (PFDR 197 corr < 0.01) in trained animals relative to non-trained controls (Fig. 3A, B). These were observed in

213
To further constrain our analysis, structural data were extracted and analyzed using a non-biased 214 volume of interest (VOI) for sensorimotor cortex, based on fMRI maps of forepaw stimulation 9 addition, we created two additional VOIs based on known areas of reorganization of forelimb 219 representation using multielectrode recordings and skill reaching [27]. Structural data were 220 extracted and plotted for the caudal forelimb area (CFA) and the rostral forelimb area (RFA) 221 contralateral to the trained limb (Table 3; SI Appendix, Fig. S4) and similar non-linear changes were 222 observed. Changes in both GMV and WMV followed a quadratic/non-linear pattern rather than 223 linear (Δ AICc > 2) except for WM in RFA where it was not possible to discriminate which model fit 224 best (Δ AICc < 2).

263
In addition to myelin immunoreactivity, cortical thickness was also quantified in histological sections 264 in sensorimotor cortex where myelin was evaluated ( Fig. 4D-F). No significant difference was

274
Morphometric changes in WMV and myelin immunoreactivity in SSp-ul were observed to follow a 275 non-linear trajectory in which we observed significant increases followed by a total, or partial, return

303
based measurements of experience-dependent brain changes remain unknown. In this study, we 304 combined motor skill learning in mice with longitudinal sMRI and immunohistochemistry to study 305 the nature of structural changes that take place in the brain during learning.

306
Longitudinal in vivo sMRI acquired throughout learning a skilled, single-pellet forelimb reach task 307 revealed bilateral non-linear decreases in GM juxtaposed to non-linear increases in WM modelled by an asymptotic time course function. Specifically, using an atlas-based cortical mask, we found bilateral non-linear changes with learning in primary and secondary motor areas and in myelin immunoreactivity in the somatosensory cortex for the forelimb, followed by a return towards

339
In contrast with our findings of GMV decreases, the study of experience-dependent volumetric

353
The WM enlargement we observed by VBM and by myelin immunohistochemistry exhibits an initial

407
In this study, we observed the temporal dynamics of experience-dependent macrostructural brain 408 changes during motor skill learning, identified non-linear decreases in grey matter volume 409 juxtaposed to non-linear increases in white matter volume, and found that these changes are 410 associated with adaptive myelination in forelimb sensorimotor cortex. Our results empirically back 411 up the idea that myelination is a rapid initial and partly transient plastic change in learning and support the use of VBM on WM structural data to evaluate myelinated fibers in cortex.

Experimental design.
To study the structural changes that occur in the brain during the acquisition 417 of a novel motor skill, two independent sets of experiments were performed: i) Motor skilled training was combined with in vivo longitudinal sMRI: trained animals (SRT, n = ii) Motor skilled training was performed and animals at different time points during the learning 422 paradigm were sacrificed for a cross-sectional evaluation of myelin immunoreactivity (Fig. 4A).

428
Animal care. All procedures were in accordance with protocols approved by the Umeå Regional  Bias corrected, skull stripped brains were manually realigned in SPM8 to approximate the 494 orientation of the stereotaxic, population-averaged, tissue-segmented in vivo brain templates for 495 wild type C57Bl/6 mice; described and provided in Hikishima et al., 2017[52]. The origin was also 496 set to match the template. The longitudinally acquired scans for each subject were registered using 497 serial longitudinal registration SPM12 to create an average image for each subject. These averages 498 were then used to create a brain template encompassing all subjects using a serial longitudinal 499 registration of the average from each subject (SI Appendix, Fig. 7A, B). Our study-specific template 500 was subsequently coregistered and resampled (from 0.1 to 0.08 mm isotropic resolution) to C57Bl/6 501 template provided in Hikishima et al., 2017[52]. Next, the individual scans, from all subjects and 502 timepoints, were coregistered and resampled to the in vivo study-specific brain template at 0.08 503 mm isotropic resolution.

504
A two-stage process was used create our own study-and sequence-specific tissue probability

511
An initial study-specific in vivo brain template was created using the DARTEL toolbox of 512 SPMmouse, which improves registration with an inverse consistent, diffeomorphic transformation.

513
This process was repeated a second time, segmentation followed by DARTEL, but using the 514 preliminary TPMs generated from the first DARTEL step to create our study-and MR sequence-

521
The individual smoothed and modulated GM and WM tissue probability maps were thresholded at

533
To test for different patterns of change in GM and WM, we used three different regression models 534 (and their opposite function) representing three different time courses; 1) Linear, 2) Increase 535 followed by a stabilization (inverse-quadratic-asymptotic), and 3) Increase followed by a 536 renormalization (inverse-quadratic) as depicted in SI Appendix, Fig. S8. We tested all three 537 regression models for each subject in separate LME analysis to detect changes in GM and WM 538 volumes, both between and within groups.

539
To study changes in GM and WM with learning specific to cortical areas, we restricted our 540 analysis to M1, M2 and S1 regions using a mask based on the Turone Mouse Brain Template

542
Tissue processing and histology. Animals were anesthetized using 100 mg/kg pentobarbital       S2. Forelimb reach-and-grasp training dynamically modulates macrostructural brain plasticity. Training mice in the single-pellet forelimb reach task produces non-linear decreases in grey matter volume (GMV) (PFDR corr < 0.001) and non-linear increases in white matter volume (WMV) (PFDR corr < 0.01) (A), whereas a linear increase in GMV was observed in non-trained control animals with time (B), whole-brain statistical maps (PFDR corr < 0.01) are represented on a studyspecific in vivo template (AP -0.1 mm, DV -3.0 mm from Bregma; 0.08 mm isotropic, radiological display).

Fig. S3
. Whole brain structural analysis of non-linear decreases in grey matter volume juxtaposed to non-linear increases in white matter volume with learning. Changes in GMV and WMV modelled using the asymptotic model and overlayed on the in vivo MRI template created from this study. Whole-brain decreases in GMV (cold blue scale) and the increases in WMV (warm red scale), in coronal sections ranging from A/P Bregma -4.3 to -7.5 mm.   There are no differences in myelin immunoreactivity between trained animals and nontrained control animals (A) at the last training day (training day 12). In addition, no differences in myelin immunoreactivity were observed between non-trained control animals at experimental day 0 and at training day 12 (B). Fig. S7. Sagittal, coronal, and horizontal sections of an in vivo microscopic T1-weighted image from one individual scan during the longitudinal study (A) and the study-specific template created for mouse brain (B). Study-and MR sequence-specific brain tissue probability maps (C). Sagittal, coronal, and horizontal sections of tissue probability maps (TPMs) of grey matter (GM), white matter (WM), cerebrospinal fluid (CSF).