Handedness, motor skills and maturation of the corticospinal tract in the adolescent brain

Abstract

With anatomical magnetic resonance imaging, the signal intensity of the corticospinal tract (CST) at the level of the internal capsule is often paradoxically similar to that of grey matter. As shown previously in histological studies, this is likely due to the presence of very large axons. We measured the apparent grey‐matter density (aGMd) of the putative CST (pCST) in a large cohort of adolescents (n = 409, aged 12–18 years). We tested the following hypotheses: (1) The aGMd in the pCST shows a hemispheric asymmetry that is, in turn, related to hand preference; (2) the maturation of the CST during adolescence differs between both sexes, due to the influence of testosterone; (3) variations in aGMd in the pCST reflect inter‐individual differences in manual skills. We confirmed the first two predictions. Thus, we found a strong left > right hemispheric asymmetry in aGMd that was, on average, less marked in the 40 left‐handed subjects. Apparent GMd in the pCST increased with age in adolescent males but not females, and this was particularly related to rising plasma levels of testosterone in male adolescents. This finding is compatible with the idea that testosterone influences axonal calibre rather than myelination. The third prediction, namely that of a relationship between age‐related changes in manual skills and maturation of the pCST, was not confirmed. We conclude that the leftward asymmetry of the pCST may reflect an early established asymmetry in the number of large corticomotoneuronal fibres in the pCST. Hum Brain Mapp, 2009. © 2009 Wiley‐Liss, Inc.

INTRODUCTION

During the last century, numerous histological studies have shown that, in humans, the corticospinal tract (CST) contains between 500,000 and 1,100,000 fibres of different sizes, myelinated or not, descending from motor, premotor or sensory cortices. Importantly, the CST includes a small population of heavily myelinated, large‐diameter fibres [from 7.4 μm to 22 μm, Nyberg‐Hansen and Rinvik, 1962; Terao et al., 1994]. These fibres are thought to originate from the giant pyramidal “Betz cells” of Brodman's area 4 (BA4), and likely correspond to the fast, direct corticomotoneuronal functional component of the CST involved in motor control [Lemon, 1993; Porter, 1985].

It has been established that, depending on the MR imaging sequence, the CST can be more or less apparent on T1‐weighted images. A previous neuroradiological and histological work demonstrated that focal differences in T2, PD or T1‐weighted (T1W) intensities occur in the posterior third of the internal capsule, most likely due to the distinctive microstructure of the CST [Yagishita et al., 1994]. On silver‐stained brain slices, the presence of the CST in the posterior third of the internal capsule [Kretschmann, 1988] was associated with a pale area containing widely spaced, large axons with thick myelin‐sheaths. The location of this pale area matched that of a T1W hypo‐intense patch previously reported by others [Mirowitz et al., 1989]. Kitajima extended this finding by showing that the CST and the fibers of the external sagittal lamina [see Catani et al., 2003], shared similar histological and magnetic‐resonance (MR) properties [Kitajima et al., 1996]. Accordingly, both tracts tend to appear darker than the surrounding white matter on T1W images.

An interesting consequence of this T1W hypo‐intensity of the putative CST (pCST), probably due to the large axons, is that the pCST tends to be iso‐intense with grey matter and may be misclassified as such by automated tissue‐classification algorithms [Levy‐Cooperman et al., 2008]. In other words, depending on the strength of this effect, some hypointense white‐matter structures can paradoxically appear as grey matter, and this can have implications for analyses based on local tissue‐densities in a stereotaxic space [i.e. voxel based morphometry (VBM); Wright et al., 1995]: we noticed that this phenomenon caused the pCST to become evident on the average stereotaxic grey‐matter map of a large cohort of adolescent siblings aged from 12 to 18 years from the Saguenay Youth Study (SYS, T1W dataset, Fig. 1). We were able to segment the average left and right pCST and then to measure the individual “apparent grey‐matter densities” (aGMd) in these two volumes of interest in a sample of 409 adolescents (see Fig. 2).

Figure 1.

Left: The corticospinal tract (outlined in blue) as it appears on the grey‐matter average of 409 non‐linearly normalised brains (MNI coordinates: x = 23 mm, y = −15 mm, z = 7 mm). Right: superimposition with the myeloarchitectonic maps from Rademacher et al. [ 2001].

Figure 2.

A: Slice of the average grey matter map (z = 10 mm) showing the left putative corticospinal tract at the level of the internal capsule, with the outline of the automatically determined region of interest for that slice in red. B: Watershed transform of the same slice, with isocontours of the probabilistic map of the corticospinal tract from Rademacher et al. [ 2001]. A warmer colour reflects a lower rank of the corresponding watershed basin (smaller seed or peak value). This transform of the image allowed the segmentation of the pCST by adding the 30 highest pixels to the seed value within the pCST patch (in red). C: Final volumes of interest, computed over 26 slices, and rendered over the average grey matter density map.

We took advantage of this phenomenon and tested three hypotheses:

Firstly, we predicted that the aGMd in the pCST shows a hemispheric asymmetry that is, in turn, related to hand preference. Several reports of a leftward asymmetry of the CST exist [Dubois et al., 2008; Galaburda et al., 1978; Nathan et al., 1990; Rademacher et al., 2001; Westerhausen et al., 2007; but see White et al., 1997], and it may be that the very adaptable structure and function of the CST [Lemon and Griffiths, 2005] would reflect the bias towards right‐hand preference found in 90% of the human population. An association with handedness, however, has so far remained elusive [Westerhausen et al., 2007]. We hypothesized that the CST should display a leftward asymmetry in right‐handers, and a rightward asymmetry in left‐handers.

Secondly, we hypothesized that the maturation of the CST during adolescence differs between both sexes, due to the influence of testosterone. This prediction is based on the previous observations of a faster growth of white matter in males during childhood and adolescence [De Bellis et al., 2001; Lenroot et al., 2007] and our recent finding that age‐related increase in the overall volume of white matter during male adolescence is mediated, in part, by testosterone [Perrin et al., 2008]. Based on the discrepancy between age‐related changes in volume and those in magnetization‐transfer ratio, an indirect index of myelination, we speculated that this global increase is related to an increase in axonal calibre [Perrin et al. 2008]. Given the known relationships between T1W intensity on one hand and, on the other hand, axonal calibre and/or axon density at the level of the CST [Yagishita et al., 1994], we predicted that the aGMd will increase with age in males but not in females and that this sex‐specific increase is explained by the rise of testosterone.

Thirdly, we predicted that variations in aGMd in the pCST reflect inter‐individual differences in manual skills. It is not known whether, within the human species, inter‐individual variations in manual skills relate to the morphology of the CST.

MATERIALS AND METHODS

Participants

Analyses were conducted on a set of 409 healthy French Canadian adolescents (age ranging between 12 and 18 years, 204 males, 205 females) from the Saguenay Youth Study dataset [SYS, detailed in Pausova et al., 2007]. In this sample, there were 40 self‐reported left‐handers (22 females). Self‐reported handedness was recorded by a psychometrician before the performance of motor tasks; this information was used to determine the ordering of the use of left and right hand when testing manual skills. Handedness was also documented using an 18‐item questionnaire, with scores ranging from 18 to 90 for extreme right‐handers and extreme left‐handers respectively [based on Crovitz and Zener, 1962] (Table I). We used the questionnaire to verify that the self‐reported handedness was consistent with the “writing hand” item from this questionnaire. In this report, handedness was based on the hand used for writing, as reported by the subject during neuropsychological testing and in the Crovitz & Zener questionnaire. We chose a simple binary classification based on the hand used for writing, a much practised and extremely lateralized task that requires an early and extensive learning [Perelle and Ehrman, 2005]. This definition has been used in large populations [Peters et al., 2006], and provides an unambiguous operational criterion of handedness [Perelle and Ehrman, 2005]. No handedness information was available for three participants; they were not included in the relevant analyses. There was no difference in terms of age or IQ (WISC III) between the four sex and handedness groups (Table I).

Table I.

Demographics

Sex handedness	Females		Males
	Left‐handers	Right‐handers	Left‐handers	Right‐handers
Age (months)	188.3 ± 25.5, n = 22	183.1 ± 23.2, n = 181	179.4 ± 23.4, n = 18	180.7 ± 22.4, n = 185
Bioavailable testosterone (nmol/l)	0.5 ± 0.38, n = 22	0.5 ± 0.37, n = 181	10.4 ± 06.6, n = 18	8.7 ± 05.4, n = 181
IQ	105.8 ± 11.5, n = 22	104.7 ± 12.1, n = 181	103.6 ± 13.9, n = 18	105.4 ± 12.8, n = 185
Handedness scores	74.0 ± 11.2, n = 22	27.2 ± 07.1, n = 181	60.8 ± 17.6, n = 18	26.1 ± 06.2, n = 185

Means ± SD for age, bioavailable testosterone levels, IQ (WISC III), and an 18 item handedness scale, ranging 18–90 (extreme right‐handers to extreme left‐handers, 54 being the center).

Fine Motor Skills

In the Grooved Pegboard task, participants were required to fit key‐shaped pegs into similarly shaped holes on a 4‐by‐4 inches board, beginning at the left side of the board with the right hand and at the right side with the left hand. Participants were instructed to complete the task as rapidly as possible. Two trials were performed with each hand, starting with the dominant hand (self‐report). The order was RLLR for right‐handers, LRRL for left‐handers. The average completion times were used as scores.

In the repetitive tapping task [Leonard et al., 1988], the subjects had to hit a target (metal plate) located on either the left or the right of a board, as fast as possible, with a stylus for a period of 15 s. One trial was performed with each hand, starting with the dominant hand. We recorded the number of taps.

Bioavailable Testosterone

Fasting blood samples were taken in the morning (between 8 and 9 am) and analysed via radioimmunoassay (Testosterone RIA DSL‐4000, Diagnostic Systems Laboratory, Inc., TX) to measure levels of testosterone (nmol/l) and sex‐hormone‐binding‐globulin (SHBG). The level of bioavailable testosterone (nmol/l) was then calculated [Södergård et al., 1982]. Testosterone levels were not different between the two handedness groups but, as expected, were very different between males and females (Table I).

Magnetic Resonance Imaging

For each participant, a high‐resolution T1W image of the brain was acquired on a Phillips 1.0‐T scanner using the following parameters: 3D RF‐spoiled gradient echo scan with 160 slices, 1‐mm isotropic resolution, TR = 25ms, TE = 5 ms and flip angle = 30°.

Image Processing

T1‐weighted images were first non‐uniformity corrected [Sled et al., 1998] and intensity normalized. They were then non‐linearly registered to the SYS333 study‐specific template, an average of 333 subjects of the SYS dataset registered to the MNI305 template. This step was achieved using the Image Registration Tool Kit software package [Rueckert et al., 1999], with a sampling distance of 5 mm. The registered brain images were then classified [Zijdenbos et al., 2002] into grey, white matter and cerebro‐spinal fluid (GM, WM, CSF images).

The average of the 409 grey matter images was then slightly smoothed using a 2 × 2 × 1 mm FWHM Gaussian filter. The pCST was clearly apparent on this mean image as a patch of apparent grey‐matter in the posterior part of the internal capsule, a location consistent with previous reports [Yagishita et al., 1994]. In order to segment the pCST patch, a 2‐D watershed transform [Vincent and Soille, 1991] was applied to 26 consecutive axial slices covering the internal capsule from z = −4 to z = 21 mm (see Fig. 2). We used the 8‐neighbour definition of connectivity, and pixels with grey‐matter probabilities below a threshold of 0.15 were excluded from the transform. In each hemisphere, the watershed basin corresponding to the pCST was first identified on the most ventral slice (z = −4). The criterion for this was the highest average probability of being included in the CST as defined by the myelin‐based stereotaxic probability map [Rademacher et al., 2001]. On subsequent slices, the watershed whose peak coordinates were closest to those of the pCST watershed basin of the previous slice (Euclidean distance) was selected. In each hemisphere and each slice, we then applied a 2‐D region‐growing method to add sequentially 30 neighbouring pixels with the highest probability of being classified as grey matter within the pCST. The number of 30 pixels gave the most reliable results. This rater‐independent strategy allowed for the creation of masks of similar volume in the left and right hemispheres, thus making the hemispheric comparisons unbiased [Hagmann et al., 2006]. The region was not allowed to grow into any adjacent watersheds (thalamus or putamen/globus pallidus); it remained free, however, to extend into the internal capsule. By combining the regions drawn on all slices, we obtained two volumes of interest of 806 and 804 voxels on the left and right side respectively (two voxels lacked at z = 6 mm on the right because the watershed was too constraining). These masks were used to measure the average aGMd in the pCST region in each individual grey‐matter map. Both the masks and the grey‐matter maps were projected back to native space, and a 2‐mm 3D Gaussian blur was applied to the GM maps.

Statistical Analysis

Hypothesis 1: We assessed the effects of hemisphere, sex and handedness on aGMd values using a repeated‐measures type II ANOVA (with hemisphere as a within‐subject factor, sex and handedness as between‐subjects factors).

Hypothesis 2: In order to assess the effects of sex and age on aGMd in the pCST and their interaction, we computed linear models that included sex and age as explanatory variables of the left or right aGMd and of the asymmetry index. To complement these analyses, the effects of testosterone were also tested by linear regression separately in males and females, with and without age correction.

Hypothesis 3: Both the unimanual performance on pegboard and tapping and the asymmetry between the left and right hands were analyzed in the same way as the aGMd data (repeated measures type II ANOVA, sex and handedness as between‐subjects factors and performing hand as within‐subjects factor). Subjects with scores distant by more than 3 standard deviations from the mean on the left or right hand or the asymmetry index were rejected from the analysis. We then correlated aGMd values with the hand skill variables in males and females separately, with and without age correction. When used, asymmetry indices were computed with the formula (L − R)/(L + R) for the pCST, and (R − L)/(L + R) for the manual performance variables.

Statistics were computed using R (http://www.r-project.org, “car” package), and power calculations with the G power software.

RESULTS

Measurements of aGMd in pCST

On 25 consecutive 1‐mm‐thick axial slices of a grey‐matter probability map obtained in the population of 409 adolescents (see Fig. 2), we were able to identify automatically the aGMd peaks corresponding to the left and right pCST (seed pixels). The aGMd of the seed pixels was higher in the left hemisphere at all levels except in the lowest, denser, slices (see Fig. 3). This effect was also present when considering the individual aGMd densities averaged over the entire left and right pCST region (Table II). The left and right aGMd values were highly correlated (r = 0.86).

Figure 3.

Seed pixel profiles for the two volumes of interest. The coordinates are given in the two first graphs (absolute x and y coordinate versus z level), the seed density in the last graph. The left hemisphere appears in closed circles, the right in open circles. The apparent grey matter density is higher on the left, and also increases dorso‐ventrally.

Table II.

Descriptive statistics

Sex handedness	Females				Males
	Left‐handers		Right‐handers		Left‐handers		Right‐handers
Side	L	R	L	R	L	R	L	R
aGMd	0.31 ± 0.18	0.27 ± 0.16	0.31 ± 0.15	0.26 ± 0.15	0.37 ± 0.17	0.38 ± 0.15	0.41 ± 0.17	0.36 ± 0.16
n	22		181		18		185
Pegboard (s)	12.2 ± 1.4	12.4 ± 1.7	13.1 ± 2.1	11.5 ± 1.6	12.9 ± 1.7	14.0 ± 2.2	13.8 ± 2.1	12.5 ± 1.6
n	21		181		18		182
Tapping	98.0 ± 8.1	93.1 ± 8.6	87.3 ± 9.4	101.2 ± 10.7	99.9 ± 11.6	99.3 ± 13.4	93.4 ± 11.9	107.6 ± 12.0
n	21		176		18		174

Descriptive statistics (mean ± SD) for the apparent grey matter density of the putative corticospinal tract, pegboard task (execution time in s) and tapping task (number of taps in 15 s). The data are sorted by handedness, sex and side (hemisphere or hand). The sample sizes (n) fluctuate due to missing values and outlier exclusion (3 standard deviations criterion).

Hypothesis 1: Hemispheric Asymmetry of the pCST, and Effects of Sex and Handedness

In the repeated measures ANOVA, with handedness and sex as between‐subject factors and hemisphere as within‐subject factor, the three‐way interaction was not significant (F _1,402 = 2.46, P = 0.117). We noticed, however, a two‐way interaction between handedness and hemisphere (F _1,402 = 5.85, P = 0.016) indicating that the inter‐hemispheric difference was different in the right and left‐handers irrespective of sex (mean difference in inter‐hemispheric difference between right‐handers and left‐handers = 0.034, Cohen's d = 0.40). The L > R asymmetry of the pCST was observed in right‐handers but was lower in left‐handers. The hemispheric asymmetry was not significant in the left‐handed group (mean inter‐hemispheric difference = 0.013, Cohen's d = 0.14; one‐sample t‐test, n = 40, P = 0.39). It is interesting to note that male left‐handers were on average symmetric, whereas all other groups showed a L > R asymmetry (see Fig. 4). Given the small number of left‐handed subjects in each sex group, we explored this trend with post‐hoc tests (Table III). A one‐sample t‐test, assessing the presence of a hemispheric asymmetry in the male left‐handed group, did not reach significance (Table III). In contrast, female left‐handers still displayed a trend for an L > R asymmetry (Table III, Fig. 4); a larger sample of n = 45 would have been needed for this asymmetry to become significant (alpha = 0.05 and power = 0.8). The L > R asymmetry was evident and highly significant in both (large sample‐size) right‐handed groups, with a “medium” effect size (d > 0.5, Table III). The effect of handedness on the asymmetry was significant only in males (two‐sample student's t‐test, males, P = 0.009; females, P = 0.4; Table III). Although the three‐way interaction was not significant, the two‐way interaction between hemisphere and handedness appears to be driven by males (see Fig. 4). In fact, the small number of left‐handed subjects makes it difficult to detect differences in inter‐hemispheric asymmetries between male and female left‐handers (Table III). The sex‐by‐hemisphere interaction was not significant (F _1,402 = 0.68, P = 0.4), indicating that sex did not influence the asymmetry in pCST aGMd. The sex‐by‐handedness interaction was not significant (F _1,402 = 0.07, P = 0.79), indicating that the effect of sex was similar in both handedness groups. The “main effect” of hemisphere, i.e. the L > R asymmetry in the whole population, was highly significant (F _1,402 = 104.87, P < 10⁻¹⁵). The main effect of sex was also very significant (F _1,402 = 42.30, P < 10⁻¹⁰). The aGMd values were overall greater in males as compared with females in both hemispheres. There was no significant main effect of handedness (F _1,402 < 0.01, P = 0.99).

Figure 4.

Boxplots of the inter‐hemispheric difference (L − R) in the apparent grey‐matter density (aGMd) of the putative corticospinal tract (pCST) at the level of the internal capsule, in the four sex and handedness groups. The small boxes represent the means.

Table III.

Asymmetry of the putative corticospinaltract

Test	Mean ± SD [95% CI]	Cohen's d	T	P value	Corr. P value	Detectable difference (d)
Male LHd (n = 18)	−0.013 ± 0.102 [−0.064 0.037]	0.13	−0.56	0.58	–	0.0715 (0.70)
Male RHd (n = 185)	0.045 ± 0.089 [0.032 0.058]	0.51	6.92	7 × 10−11	6 × 10−10	0.0185 (0.21)
Female LHd (n = 22)	0.034 ± 0.079 [–0.001 0.069]	0.43	2.02	0.056	0.45	0.0495 (0.62)
Female RHd (n = 181)	0.048 ± 0.080 [0.036 0.060]	0.60	8.08	9 × 10−14	7 × 10−13	0.0168 (0.21)
RHd–LHd, males (n = 203)	0.059 ± 0.090 [0.015 0.103]	0.65	2.63	0.009	0.072	0.063 (0.70)
RHd–LHd, females (n = 203)	0.014 ± 0.082 [−0.050 0.021]	0.17	0.77	0.44	–	0.052 (0.64)
Males–females, LHd (n = 40)	−0.047 ± 0.089 [−0.106 0.011]	0.53	−1.65	0.11	0.88	0.080 (0.91)
Males–females, RHd (n = 186)	−0.003 ± 0.088 [−0.020 0.015]	0.03	0.30	0.77	–	0.026 (0.29)

Post‐hoc comparisons for the hemispheric asymmetries in pCST aGMd (L–R). The presence of a significant asymmetry in the four sex and handedness groups and differences between groups were assessed with one‐ and two‐sample Student's t‐tests. Cohen's d is a measure of effect size (ratio of mean and standard deviation). Corrected P values (Bonferroni for 8 tests) are also given. Statistical power estimates are presented in the form of the detectable difference with alpha = 0.05 and power = 0.8 (mean difference and Cohen's d).

Hypothesis 2: Sex Difference in the Developmental Course of the pCST During Adolescence and the Influence of Testosterone

In both hemispheres, the aGMd in the pCST increased with age in males (P < 10⁻⁴; simple regression in males left: r = 0.35, right: r = 0.38, Fig. 5, Table IV) but not in females (left r = 0.06, right = −0.01, non significant). This sex difference in the effect of age on the aGMd in the pCST was statistically significant (linear models, interaction between age and sex, left: P = 6 × 10⁻³; right: P = 10⁻⁴). Thus, the observed sex effect was age dependent: the predicted means for the two groups at 12 years of age were not different between males and females, whereas they were very different at 18 years of age. The asymmetry index of the aGMd did not vary with age in either males or females.

Figure 5.

Effects of sex and age (in months) on the putative corticospinal tract (pCST) apparent grey‐matter density (aGMd), for the left and right hemispheres. Females appear in open circles and dashed lines, males in closed circles and solid lines. The regression over the whole group is given by the dotted line.

Table IV.

Effects of age, sex and testosterone on the pCST apparent grey matter density

	Intercept	Std err	P value	Slope	Std err	P value	R 2
Left pCST aGMd, linear model with age and sex
Males	0.306	0.021	–	0.0027	0.0005	5 × 10−8	0.146
Males–females	−0.012	0.030	NS	−0.0024	0.0007	5.7 × 10−4
Left pCST aGMd, linear regression with bioavailable testosterone (n = 404)
Total (n = 404)	0.305	0.010	–	0.0112	0.0013	2 × 10−15	0.145
Males (n = 200)	0.307	0.021	–	0.0111	0.0020	10−7	0.131
Right pCST aGMd, linear model with age and sex
Males	0.270	0.019	–	0.0027	0.0005	10−8	0.171
Males–females	−0.006	0.028	NS	−0.0027	0.0006	2.5 × 10−5
Right pCST aGMd, linear regression with bioavailable testosterone (n = 404)
Total (n = 404)	0.260	0.009	–	0.0115	0.0013	<2 × 10−16	0.167
Males (n = 200)	0.272	0.019	–	0.0105	0.0018	4 × 10−8	0.142
Testosterone by age
Males (n = 200)	3.12 nmol/l	0.57	10−7	0.160 nmol/l/mth	0.013	<2 × 10−16	0.417

Apparent grey‐matter density (aGMd) of the putative corticospinal tract (pCST) as explained by age and sex or testosterone, and relationship between age and testosterone. The intercept is placed at the lowest age in the study (145 months, 12 years old). When pCST aGMd is explained by sex and age, the first row gives the intercept and slope in males, whereas the second row gives the relative difference in intercept and slope between males and females. All these models were strongly significant.

Bio‐available testosterone shows a significant increase with age during adolescence in males only (r = 0.65, P < 10⁻⁴, Table IV). There was a significant relationship between aGMd in the pCST and bio‐available testosterone in males (left: r = 0.36, right: r = 0.38, both P < 10⁻⁴). Furthermore, this relationship between testosterone and aGMd in pCST in males survived a statistical correction for the effects of age, albeit explaining less variance attributable to testosterone above and beyond the effect of age (P = 5 × 10⁻³, R ² = 0.04, on both sides after removing the effects of age on both the dependent and independent variable, i.e. partial correlation). In the light of the slopes observed in males (Table IV), the plasma levels of testosterone in females (Table I) were not variable enough to allow a meaningful regression analysis. But computing the regression in males only or in males and females together did not affect the slope or intercept of the regression line (Fig. 6, Table IV) indicating that the age‐dependent sex difference could be mediated by the progressive increase in bio‐available testosterone in males during adolescence.

Figure 6.

Effect of testosterone on the putative corticospinal tract (pCST) apparent grey‐matter density (aGMd), for the left and right hemispheres. Females appear in open circles and dashed lines, males in closed circles and solid lines. The average regression is given by the dotted line.

Hypothesis 3: Development of Manual Skills During Adolescence and Maturation of the pCST

In both left and right‐handers, the time taken to insert all the pegs into the board was shorter with the dominant hand than with the non‐dominant (performing‐hand by handedness interaction, F _1,398 = 36.69, P = 3 × 10⁻⁹, Fig. 7, Table II). Similarly, the dominant hand was also able to hit the target a greater number of times in 15 s than the non‐dominant hand (F _1,385 = 123.57, P < 10⁻¹⁵). The left‐handers were more symmetric than the right‐handers at this task (Table II). Over the whole sample, with a majority of right‐handers, the right‐hand was better (main effect of performing hand for pegboard and tapping: F _1,398 = 148.48, F _1,385 = 733.5, both P < 10⁻¹⁵). Females outperformed males at the pegboard task (main effect of sex: F _1,398 = 29.62, P = 1 × 10⁻⁷) while the converse was true for the repetitive tapping task (F _1,385 = 35.15, P = 7 × 10⁻⁹).

Figure 7.

Evolution of manual skills with age (in months) at adolescence. First row: grooved pegboard execution time (s), second row: repetitive tapping (number of taps in 15 s). Third column: asymmetry indices (R − L)/(L + R). Males appear in triangles and solid lines, females in open circles and dashed lines. The average regression is given by the dotted line.

Completion times in the pegboard task improved significantly with age for both hands (see Fig. 7) in females (P < 10⁻⁴, left: r = −0.40, right: r = −0.27) but not in males (P > 0.2, left r = −0.09, right r = −0.006; linear models, sex by age interaction: left, P = 0.002; right, P = 0.01). The asymmetry in the performance of this task with the dominant and non‐dominant hands decreased significantly with age in females only (r = 0.16, P = 0.02 in females; r = 0.06, P = 0.38 in males but no significant slope difference between the two sexes). This was indeed due to the left hand improving with age faster than the right hand. As regards the tapping task, the number of taps improved significantly with age in both males and females, and for both hands (simple regression, all four P values < 3 × 10⁻³, r ranging 0.21−0.42). The asymmetry tended to decrease slightly with age in males only (r = −0.15, P = 0.04, but no significant slope difference between the two sexes).

Prior to the statistical correction for age, we observed modest correlations between increasing aGMd values and improvement in manual performance as measured with the repetitive tapping task in males (r ranging 0.16–0.26, P < 0.05) but not in females. Similar tests conducted with the pegboard data did not yield any significant result. The above correlation observed in males did not survive a statistical correction for the effects of age. In keeping with the effect of handedness observed in males, there was a slight but significant relationship between the degree of asymmetry in tapping and the asymmetry in the aGMd of the pCST in this group (r = 0.176, P = 0.01); no such relationship was observed in the female group. This effect was not dependent on age.

DISCUSSION

This study described the morphological variability of the pCST and its behavioural relevance in the largest sample of healthy volunteers to date. We confirmed two out of the three hypotheses.

Hypothesis 1: Hemispheric Asymmetry and Handedness

We found a significant left > right hemispheric asymmetry in the pCST in both males and females right‐handers, which was reduced in left‐handers, particularly males. This is to our knowledge the first report of an association between handedness and morphological features of the putative CST. The asymmetry of the pCST was, however, not reversed in left‐handers. The fact that we did not find such an asymmetry reversal is not necessarily surprising. At the behavioural, functional or anatomical levels, a smaller degree of asymmetry was often described in left‐handers as compared with right‐handers [Amunts et al., 2000; Kim et al., 1993; Schmidt et al., 2000a; Solodkin et al., 2001]. From what we inferred of the microstructural bases of the T1W intensity of the pCST (see below), we would hypothesize that the handedness‐related asymmetry stems from an inter‐hemispheric difference in the proportion of large CST axons. This asymmetry would thus be established early during brain development [Dubois et al., 2008], and remain essentially unchanged until the onset of the aging process. This would further explain the absence of the relationship between pCST and the improvement of manual skills during adolescence; the number of large fibres from the Betz cells could only decrease, leaving no room for further improvement. Theoretically, larger axons on the dominant side would allow for faster conduction times [Hartline and Colman, 2007; Rushton, 1951] and/or increased axonal transport capacities [but Inestrosa and Alvarez, 1988]; yet, an increase in the axon diameter throughout adolescence was not shown to have any influence on motor skills. If we assume that the fast, large fibres are indeed involved in fine motor control, a larger number of such fibres on the dominant side could afford the dominant M1 a relatively better control over the motoneurons of the hand. This frequent left hemispheric bias would facilitate the preferred use of the right hand during development. Theoretically, the subjects in whom this bias would be absent or reversed would in contrast be more likely to become left‐handed. The fact that the association between pCST asymmetry and hand preference was more complex than we had initially hypothesized, and the possible interaction with sex (evidenced by Amunts et al. [ 2000] at the level of the central sulcus), suggests that other factors than pCST asymmetry are involved in the early determination of hand preference. No relationship was found between levels of bio‐available testosterone and either handedness or asymmetry of the pCST, suggesting that testosterone does not play a role in the establishment of these particular behavioural and neuroanatomical asymmetries—as far as it can be observed during adolescence.

Hypothesis 2: Age and Sex

We observed a marked sex‐specific effect of age: the two sexes had the same average aGMd in the pCST at age 12 years, with an increase occurring only in males over the next 6 years resulting, in turn, in a large sex difference at the age of 18 years. Testosterone levels were different between the two sexes, with this divergence emerging during puberty. Testosterone alone was just as explanative as a model involving a sex‐by‐age interaction.

Increases in T1W signal in white matter typically accompany early stages of myelination [Barkovich, 2000], while decreases of the signal are associated with pathologies such as amyotrophic lateral sclerosis [Yagishita et al., 1994] or Wallerian degeneration [increase in T2 contrast after 14 weeks, Kuhn et al., 1989]. The fact that we saw an age‐related increase rather than decrease in aGMd in males argues against myelination being a relevant determinant of the observed changes in the pCST. According to Yagishita et al. [ 1994], the diameter of the largest fibers is the major determinant of the T1W hypointensity of the pCST. Both Yagishita et al. [ 1994] and Mirowitz et al. [ 1989] had observed a dramatic increase in the detection rates of the pCST after infancy, and they established a compelling parallel between this phenomenon and Lassek's early histometric data on the development of the pyramidal‐axon diameter during the first two years of life, and then at 22 and 80 years old [Lassek, 1942]. Lassek reported that while in neonates all fibers are about 1 μm or less in size, at 8–11 months a differentiation between large and thick myelinated axons and medium‐to‐small axons occurs. At 7 years of age, the size of the largest fiber is of 12 μm [Verhaart, 1950], half of the mature size, and the pCST is still seldom detected with MRI. After 13 years of age, the MRI detection rates become very high. Mirowitz's detection rates show a tendency to decrease with age throughout adulthood, which is consistent with the age‐related decline in the axonal calibre [Mirowitz et al., 1989; Terao et al., 1994]. It should be noted that in previous MR studies white‐matter density in the internal capsule increased with age during childhood and adolescence [Paus et al., 1999], and was higher in the left vs. right hemisphere [Herve et al., 2006]. In these two studies, it is possible that the pCST was not detected as apparent grey‐matter by the tissue classification procedure because of different image contrast and/or classification algorithms.

Our results suggest that the age‐dependent changes observed in males during adolescence are mediated by an effect of testosterone on axonal calibre. To affect axonal calibre, it would be necessary for testosterone to influence the number and/or properties of microtubules and/or neurofilaments [Hoffman et al., 1987; Marszalek et al., 1996]. This hypothesis is consistent, for example, with experimental findings of testosterone‐induced upregulation of β‐tubulin expression [Butler et al., 2001; Matsumoto et al., 1994], or androgen regulation of axonal growth in motoneurons [for a review, see Fargo et al., 2008]. Since the effect of testosterone during adolescence would obviously relate to the late maturation of axon calibre, this phenomenon would likely affect the whole fibre spectrum of the pCST, small or large, causing the decrease in aGMd.

Hypothesis 3: Manual Skills and pCST

The increase in performance and decrease in inter‐manual skill asymmetry with age are consistent with the literature on the development of manual skills, as were the observed sex differences [Bryden and Roy, 2005; Roy et al., 2003; Schmidt et al., 2000a, b]. It is interesting to note, however, that the difference between females and males in the pegboard task seems to emerge during adolescence while, in the tapping task, the sex difference was already present at 12 years of age. Yet, contrary to our third hypothesis, we found no reliable correlation between manual skills and the increase in aGMd in the pCST in males during adolescence. Overall, we conclude that the increase in manual performance during adolescence is more likely to be mediated by other structures than the CST. It has been suggested that the evolution of the manual skills asymmetry pattern could itself be related to the maturation of inter‐hemispheric interactions through the corpus callosum, or to an asynchronous maturation of the left and right hemispheres, the left starting earlier than the right [Roy et al., 2003].

The effect of age on aGMd of the pCST seen in males may be a local expression of the global maturation of WM during adolescence, bearing no specific relevance to hand skills. Although one would assume that such a global phenomenon would contribute to the establishment of cognitive differences seen between adult males and females [Gur et al., 1999], the functional role of this sex‐related change itself, if any, is unclear at present. Furthermore, whether or not it could be of importance for brain disorders remains to be assessed. Differential maturation of white matter could be a relevant factor for the emergence of psychopathology during adolescence [Paus et al., 2008], and also later in life [Cahill, 2006]. For instance, sex differences in recovery after stroke have been documented [Stein, 2007; Sue‐Min et al., 2005].

CONCLUSION

Here, we took advantage of the presence of a T1W hypo‐intensity in the internal capsule likely corresponding to the CST and tested three hypotheses about the relationship between hemispheric asymmetry in this anatomical phenomenon, handedness and sex‐specific brain maturation of pCST during adolescence. The relationship between the hemispheric asymmetry in aGMd in the pCST and handedness is best explained by an early appearing (constitutive) difference in the number of large fibres in the CST of the “hand‐dominant” hemisphere. The observed sexual dimorphism in the T1W hypointensity, established during adolescence, appears to be strongly associated with the progressive rise of testosterone during male adolescence. This latter finding parallels the larger increase in white matter volume seen in males over this period. We speculate that testosterone affects axonal cytoskeleton and, in turn, axonal calibre, rather than myelin‐generating oligodendrocytes.