Sex-Based Differences in Cortical and Subcortical Development in 436 Individuals Aged 4–54 Years

Abstract

Sex-based differences in brain development have long been established in ex vivo studies. Recent in vivo studies using magnetic resonance imaging (MRI) have offered considerable insight into sex-based variations in brain maturation. However, reports of sex-based differences in cortical volumes and thickness are inconsistent. We examined brain maturation in a cross-sectional, single-site cohort of 436 individuals (201 [46%] males) aged 4–54 years (median = 16 years). Cortical thickness, cortical surface area, subcortical surface area, volumes of the cerebral cortex, white matter (WM), cortical and subcortical gray matter (GM), including the thalamic subnuclei, basal ganglia, and hippocampi were calculated using automatic segmentation pipelines. Subcortical structures demonstrated distinct curvilinear trajectories from the cortex, in both volumetric maturation and surface-area expansion in relation to age. Surface-area analysis indicated that dorsal regions of the thalamus, globus pallidus and striatum, regions demonstrating structural connectivity with frontoparietal cortices, exhibited extensive expansion with age, and were inversely related to changes seen in cortical maturation, which contracted with age. Furthermore, surface-area expansion was more robust in males in comparison to females. Age- and sex-related maturational changes may reflect alterations in dendritic and synaptic architecture known to occur during development from early childhood through to mid-adulthood.

Introduction

Sex-based differences in human cortical and subcortical development have been established through ex vivo histological examination (Fernández-Guasti et al. 2000; Kruijver et al. 2001) and more recently through in vivo magnetic resonance imaging (MRI) studies, which have provided considerable insight into the development of the brain (Sowell et al. 2007; Neufang et al. 2009; Luders and Toga 2010; Koolschijn and Crone 2013; Mutlu et al. 2013).

Infant MRI studies have revealed that within the first 2 years of life, sex-based differences are evident, with males having up to 13% greater total cerebral volume in comparison to females (Knickmeyer et al. 2008). Studies examining sex differences in child and adult populations (4–40 years) with MRI have noted that males have greater total cerebral volumes (TCV) on the order of ~ 8–10% compared to females (Caviness et al. 1996; Giedd et al. 1996; Goldstein et al. 2001), consistent with larger head sizes. However, the volumes of GM and white matter (WM) differ between males and females (aged 3–30 years), with males having greater total WM and females having greater total GM (Caviness et al. 1996; Luders et al. 2005; Lenroot et al. 2007; Koolschijn and Crone 2013). Yet, several studies have reported inconsistent results regarding male and female differences in brain morphometry. In particular, findings from cortical thickness studies have reported discordant results with one study reporting increased thickness in males aged 18–93 years (Salat et al. 2004), and another study reported increased thickness in temporal and parietal cortices in females relative to males who were aged 7–87 years (Sowell et al. 2007). Also in aging studies results are inconsistent; Király et al. found that subcortical volumes showed greater decreases with age in males, while Li et al. found the opposite pattern (Li et al. 2014; Király et al. 2016). However, the discrepancy in the findings is likely associated with the age of the participants as Király et al. studied adults in their early thirties, while Li and colleagues included young to older adults in their study (19–70 years).

With age, the cortex undergoes exponential development, with a global increase of cortical thickness, peaking at around 10–12 years coinciding with pubertal changes. Cortical thickness then gradually decreases in late adolescence into adulthood (Shaw et al. 2008; Giedd and Rapoport 2010; Gogtay and Thompson 2010; Raznahan et al. 2011; Taki and Kawashima 2012). Although more recent studies report variable decreases in cortical thickness from early childhood (Mills et al. 2016; Zielinski et al. 2014; Tamnes et al. 2017), this may be due to increasing myelination rather than cortical thinning (Natu et al. 2019). Variations in pubertal status, with girls reaching puberty 1–2 years earlier than boys has thought to underlie alterations in peak cortical thickness seen between sexes. Frontal and parietal GM peaks at 10.5 years in girls and 14.5 years in boys (Lenroot et al. 2007). However, a study that included boys and girls with a narrow age range (10–14 years) who were matched for pubertal status reported that no sex differences in cortical thickness were evident (Bramen et al. 2012). Sex differences were evident when using testosterone levels as predictors of cortical thickness.

Regional variations in cortical thickness and volumetric development can provide insight into the maturation of brain regions and the relation to the onset of cognitive processes. Furthermore, trajectories of brain development can provide insight into when interventions may be most effective during sensitive windows of development. Thus, the aim of the current study was to examine age- and sex-related maturation of cortical and subcortical areas in a large single-site cohort of typically developing children, adolescents, and adults aged 4–54 years. We used T1-weighted anatomical scans acquired at 1.5 T. The entire cerebral cortex, GM and WM, and subcortical GM were segmented using automatic segmentation pipelines. The volumes of the GM and WM, cortical thickness, and surface area as well as the subcortical surface area of the basal ganglia and thalamus were extracted to examine age- and sex-related differences in a large cohort of males and females aged 4–54 years.

Materials and Methods

Participants

A total of 514 individuals were scanned with MRI for this study. A total of 269 (53%) females and 245 (47%) males participated. Participants’ age range was 4–54 years (median age = 15 years 9 months; interquartile [IQR] range: 10 years and 9 months—23 years and 2 months).

All subjects were recruited from the community through advertisements at the Hospital for Sick Children, Toronto, Ontario. Exclusion criteria were history of concussion, brain injury, preterm birth, learning disabilities, seizures, or psychiatric illness. Informed written consent was obtained from all adult participants and the parents of the children, and children provided informed verbal assent.

MRI Data Acquisition

Participants underwent a standardized neuroimaging protocol at the research-dedicated MRI suite at the Hospital for Sick Children. Participants were scanned on a 1.5 T MRI scanner (GE Signa Excite, Waukesha, USA) using an 8-channel array head coil. The participants were scanned using a 3D-FSPGR sequence, producing volumes of T1-weighted axial slices. The majority of participants (n = 400, 78%) were scanned on the same sequence (flip angle = 15, number of averages = 2, voxels = 0.94 mm × 0.94 mm, slice thickness of 1.5 mm, repetition time [TR] = 9 ms). The remaining participants (n = 114, 22%) were scanned using identical parameters, except the TR ranged from 7.41–11.29 ms. No statistically significant differences in TR were evident (t = −0.93, P = 0.4).

Cerebral Volume, Surface Area and Cortical Thickness Measurements and Analysis

For all MRI scans, total cerebral volume, cortical surface area and cortical thickness were calculated using the corticometric iterative vertex-based estimation of thickness image-processing pipeline (Lerch and Evans 2005). Preprocessing steps included a nonuniformity correction (Sled et al. 1998), skull stripping (Smith 2002) and alignment to the common image space of the Montreal Neurological Institute (Collins et al. 1994). All T1-weighted images from children and adults were registered to the common template in MNI space. Transforming the child and adult scans in the same template space permitted averaging the data and examining the effects of age and biological sex on cortical thickness and surface area. Brain tissue was classified into WM, GM and cerebrospinal fluid (CSF) (Zijdenbos et al. 2002; Tohka et al. 2004). The GM, WM, and CSF volumes were calculated for the whole brain and the sum of these values was used to calculate total cerebral volumes.

Age-related and sex-based differences in GM and WM and total brain volume were assessed in a linear model. The between-group differences in GM and WM volumes were also assessed controlling for TCV.

For the cortical thickness analysis, the inner and outer cortical surfaces with 81 924 vertices each were extracted using partial-volume-effect classification (Kim et al. 2005). Cortical thickness was determined in native space using the distance between the inner and pial surfaces at each vertex. Images were smoothed (20 mm (Chung et al. 2003)) and non-linearly registered to a template surface (Lyttelton et al. 2007; Boucher et al. 2009).

Subcortical Measurements and Analysis

The whole thalamus and its subnuclei, according to the nomenclature of Hirai and Jones (Hirai and Jones 1989) were segmented automatically using a high-resolution subcortical atlas derived from serial histological data (Chakravarty et al. 2006) using a region of interest nonlinear registration procedure (Chakravarty et al. 2008). The subcortical segmentation of the thalamic nuclei has been validated using three approaches: (1) manual segmentations; (2) other automatic segmentation methods, and (3) intraoperative electrophysiological recordings in humans who were undergoing stereotactic neurosurgery for movement disorders (Chakravarty et al. 2008; Chakravarty et al. 2009).

The entire hippocampus was segmented using the MAGeT Brain (Multiple Automatically Generated Templates) pipeline using a previously validated segmentation algorithm (Pipitone et al. 2014).

Subcortical surface area was determined by creating surface-based representations of the subcortical structures (thalamus, globus pallidus, striatum) defined on an input atlas estimated using a marching cubes algorithm (Lerch et al. 2008). The surface-based representations were then smoothed. The nonlinear transformations, which mapped each subject’s image to 30 input templates, were concatenated and averaged across the template library. This step was performed to minimize the effects associated with registration errors and to increase precision and accuracy. The surface-based representations were warped to fit each template. The resulting surface representations were merged by estimating the median coordinate representation. In the last step, the surface-area values were blurred with a surface-based diffusion-smoothing kernel (striatum/thalamus = 5 mm, globus pallidus = 3 mm). This processing pipeline yielded a surface area with a total of 21 156 vertices across the subcortical structures.

Statistical Analysis

For the volume data, all analyses were performed with SPSS v25 software (Chicago, IL). For the thalami, basal ganglia, and hippocampi, the volumes of all structures described were averaged across the hemispheres and were used as the dependent variables in all subsequent analyses.

To investigate patterns of linearity and non-linearity in the relations among volumes and age, we performed a curve fitting analysis on the TCV, GM, WM, thalamic, basal ganglia, and hippocampal volumes. In the curve fitting analysis, the goodness of fit of first, second, and third order polynomial expansions was used to assess linear and non-linear associations with age. In the analysis, the volumes were entered as dependent variables and age (continuous variable) was the independent variable. The analyses were performed on each volume measure for the group data as well as for males and females separately.

To examine sex-based differences in volumes, in separate generalized linear models, the dependent variables were the volumes of the total cerebral cortex, GM, WM, the thalamus, basal ganglia, and hippocampal volumes, sex was included as a grouping variable and all analyses were adjusted using age as a continuous variable. All brain volumes (GM, WM, thalami, basal ganglia, hippocampi) were corrected for TCV by including this variable as a covariate in the analysis. Adjustment for TCV is to correct for males having generally larger head sizes in comparison to females (Paus 2010). For each model examining the association of sex with brain volumes, if significant main effects were evident for both sex and age then a subsequent interaction analysis was performed, adjusting for TCV.

Cortical thickness, cortical surface area and subcortical surface area analyses were performed using the Matlab-based program SurfStat (Worsley et al. 2009). A series of linear and non-linear models were applied to assess the association of cortical thickness and surface area to age-related changes. Model fits were calculated to examine developmental trajectories of cortical thickness as well as cortical and subcortical surface area measures. Curve estimation analysis was used to determine whether a linear, quadratic, or cubic model best fit the data.

Sex-based differences were subsequently examined in relation to cortical thickness and surface area and subcortical measures of surface area in linear models, adjusting for age.

If both age and sex effects were significant then the interaction between age and sex was performed using linear models. A correction for multiple comparisons was applied using Random Field Theory (Worsley et al. 1999). The significance level was set at P < 0.05.

Results

Participants

A total of 436 participants (201 males [46%], median age = 16 years, interquartile range [IQR] = 11–23.75 years; age range 4–54 years) are included in the final analysis. Table 1 contains the demographic information for the full sample.

Table 1.

Participant characteristics

Age groups	Females (n)	Males (n)	Total (n)	P value
4–11	63	48	111
Age [IQR]	8 [7–10]	8 [7–10]		0.1
12–20	91	85	176
Age [IQR]	16 [14–18]	15 [13–17]		0.1
21–34	72	59	131
Age [IQR]	26 [23–28]	26 [23–28]		0.1
35–54	9	9	18
Age [IQR]	42 [37.5–49.5]	38 [35.5–39.5]		<0.001

Probability values provide results using a generalized liner model comparing age between females and males. Abbreviations, IQR, interquartile range.

Total Cerebral Volume

A curve fitting analysis was applied to the TCV to examine linear and nonlinear associations with age. TCV peaked between ages 10–12 years of age followed by a decline in volume at older ages. The data were best fit by a quadratic model (P = 0.02, Table 2). Similar trends emerged when examining the data for males and females separately; however, the TCV data only in females displayed a significant quadratic fit with age (Fig. 1). The TCV data for males showed trends toward a linear or quadratic fit but neither were significant.

Table 2.

TCV, GM and WM: association with age in males and females

	Linear		Quadratic		Cubic
Brain Area	R 2	P value	R 2	P value	R 2	P value
Total cerebral volume	0.008	0.06	0.2	0.02	0.2	0.06
males	0.002	0.6	0.008	0.5	0.009	0.6
females	0.02	0.02	0.03	0.03	0.03	0.08
Total GM volume	0.3	<0.001	0.3	<0.001	0.3	<0.001
males	0.31	<0.001	0.31	<0.001	0.31	<0.001
females	0.35	<0.001	0.37	<0.001	0.37	<0.001
Total WM	0.14	<0.001	0.18	<0.001	0.19	<0.001
males	0.2	<0.001	0.25	<0.001	0.25	<0.001
females	0.15	<0.001	0.2	<0.001	0.2	<0.001

Probability values provide results using curve estimation analyses examining the association between volumes and age in the full sample as well as in males and females separately.

Figure 1.

Trajectories of TCV, GM and WM volumes in males (blue lines) and females (red lines) aged 4–54 years.

In a linear model, TCV was examined in relation to biological sex. TCV was greater in males than females (β = 127 247, P < 0.001), adjusting for age. Age significantly predicted TCV in the model (β = −1162, P < 0.001). In turn, a subsequent interaction analysis to examine the effects of age and sex on TCV was performed. The interaction between age and sex was significant (P < 0.001).

In a separate generalized linear model, an age by sex interaction was tested. Consistent with the curve fitting analysis, females demonstrated a slight negative association between TCV and age (β = −3671, P < 0.001), while males demonstrated a slight positive correlation with age (β = 2324, P < 0.001), reflected in the increase in TCV till the second decade.

Total GM

Total GM, adjusting for TCV demonstrated a significant decrease with age (Fig. 1, Table 2) starting in 4 year-old children and the findings persisted in 30 year-olds. From the third until the fifth decade of life GM volumes remained relatively unchanged. For the full cohort, aged 4–54 years, the GM volume data in relation to age were best fit by linear models (P < 0.001).

We examined sex-based differences in total GM volume. Adjusting for age at scan and TCV, there were no differences between males and females in GM (β = 1990, P = 0.5).

As no differences in GM volumes between males and females were evident after adjusting for TCV, the subsequent analysis of GM structures included this variable as a confounding factor.

Total WM

WM volumes, adjusting for TCV, significantly increased at older ages and reached their peak volume in the third decade (Fig. 1, Table 2). The data were fit by a linear model (P < 0.001). Males had greater WM volumes, (β = 59 647, P < 0.001) compared to females, when adjusting for age. However, when adjusting for TCV, no significant differences between the sexes were evident (P = 0.8).

Thalamic Volumes

The volumes of the entire thalamus peaked at 10–12 years of age and maintained similar volumes throughout adolescence and into the third decade followed by a steady decline in volume. The volumes of the entire thalamus were best fit by a quadratic model (P < 0.001, Supplementary Table 1–1). Sex differences were not significant when volumes were adjusted for TCV (β = 43.6, P = 0.5).

The association of individual thalamic nuclei volumes with age was assessed, adjusting for TCV. Several of the individual thalamic nuclei demonstrated similar age-related patterns to that seen in the total thalamic volumes including the pulvinar, ventrolateral nucleus, the central nucleus, the medial geniculate nucleus, and the medial dorsal nucleus. The volumes of the anterior nucleus demonstrated a positive linear relation with age until the fourth decade (P < 0.001), and then declined. The volumes of the anterior nucleus were also fit by a quadratic model (P < 0.001). The volumes of the ventral anterior, ventral posterior, lateral dorsal, lateral geniculate nucleus, and lateral posterior nuclei also demonstrated positive linear associations with age (all P < 0.01). Volumes of individual thalamic nuclei in males and females exhibited similar trends to that seen in the total volumes, except for the central and medial geniculate nuclei in males that demonstrated positive linear associations with age (P < 0.02, Supplementary Table 1–1).

Sex-related differences in thalamic nuclei were evident in the medial geniculate nuclei (males, (β = 17.2, P < 0.001), central nucleus (males, β = 11.5, P = 0.003), ventrolateral nucleus (females, β = 34.6, P = 0.008), adjusting for TCV and age. Age effects were significant in all the models (all, P < 0.005).

Subsequent interaction analyses examining the effects of sex and age on volumetric development, indicated that for the medial geniculate nucleus, only males demonstrated significantly positive linear changes with age (β = 1.2, P < 0.001). Similarly, for the central nucleus, only males demonstrated positive linear changes with age for the central nucleus (β = 0.95, P < 0.001). However, for the ventrolateral nucleus, females demonstrated a positive linear association with age (β = 2.65, P < 0.001).

No other thalamic nuclei demonstrated significant sex-based differences (P > 0.05), adjusting for TCV age.

Basal Ganglia Volumes

The volumes of the striatum were best fit by a quadratic model (P = 0.006, Table 3). The volumes peaked at 10–12 years followed by a slight decline after age 35. Males and females showed trends toward similar associations with age; however, the findings were not significant. No significant differences in volumes were evident between males and females when adjusting for TCV and age (P = 0.1).

Table 3.

Basal ganglia volumes: association with age in males and females

	Linear		Quadratic		Cubic
Brain area	R 2	P value	R 2	P value	R 2	P value
Striatum	0.002	0.4	0.02	0.006	0.02	0.02
males	0.002	0.5	0.03	0.06	0.03	0.1
females	0.01	0.1	0.02	0.07	0.02	0.1
Globus pallidus	0.004	0.2	0.008	0.2	0.01	0.2
males	0.001	0.7	0.004	0.7	0.006	0.8
females	0.01	0.4	0.01	0.3	0.01	0.4

Probability values provide results using curve estimation analyses examining the association between volumes and age in the full sample as well as in males and females separately.

The globus pallidus demonstrated a different pattern in relation to age in comparison to other subcortical structures in that no significant association with age was evident in the curve estimation analysis (all P > 0.05, Table 3). No association between globus pallidus volumes and age were evident in males and females. Females had smaller volumes of the globus pallidus relative to males (β = −45.3, P < 0.001), adjusting for TCV and age.

Hippocampal Volumes

Hippocampal volumes were best fit by a linear model (P < 0.001, Table 4). The peak volume of the hippocampus was seen at ~ 20 years of age. Hippocampal volumes in males and females exhibited similar associations with age. No differences in hippocampal volumes were evident between males and females (β = −83.1, P = 0.09), when adjusting for TCV and age.

Table 4.

Hippocampal volumes: association with age in males and females

	Linear		Quadratic		Cubic
Brain area	R 2	P value	R 2	P value	R 2	P value
Total hippocampal volume	0.06	<0.001	0.12	<0.001	0.13	<0.001
males	0.08	<0.001	0.2	<0.001	0.2	<0.001
females	0.07	<0.001	0.08	<0.001	0.08	<0.001

Probability values provide results using curve estimation analyses examining the association between volumes and age in the full sample as well as in males and females separately.

Cortical Thickness

Linear models assessing the effects of age on cortical thickness revealed that extensive regions of the cortex exhibited age-related cortical thinning (Supplementary Table 1–2). In particular, the bilateral insular cortices and the right medial prefrontal cortex demonstrated a decrease of > 4.3 millimeters in older participants (Supplementary Fig. 1–1).

Cortical thickness data also demonstrated significant quadratic associations with age (Fig. 2, Supplementary Table 1–3); however, the association was inverted. Similar regions in the cortex were found to show cubic relationships with age (Supplementary Table 1–4). Males, in comparison to females, showed increased cortical thickness in various frontal, parietal, temporal, and occipital regions (Table 5) adjusting for age.

Figure 2.

Vertex-based cortical thickness analysis (corrected P map): corrected P values for cortical brain regions demonstrating significant quadratic age effects (inverted), representing a slow decrease in cortical thickness values in childhood followed by a steady decline in values until the 3rd and 4th decades and leveling out at the 5th decade, controlling for sex and corrected for multiple comparisons (P < 0.05). The color bar indicates corrected P values for significant peaks at the cluster (left; light blue–dark blue) and vertex-levels (right; red– orange).

Table 5.

Cortical thickness: sex-based differences controlling for age

Region	BA	Hemisphere	x	y	z	t value	P values
Anterior cingulate gyrus	25	Right	1	14	−1	7.38	<0.0001
Anterior cingulate gyrus	24	Right	3	31	−4	7.20	<0.0001
Cingulate gyrus	31	Right	4	−42	27	6.63	<0.0001
Posterior cingulate gyrus	29	Right	9	−47	6	6.25	<0.0001
Anterior cingulate gyrus	32	Right	5	24	−11	6.16	<0.0001
Cingulate gyrus	24	Left	−3	−8	33	5.98	<0.0001
Precentral gyrus	4	Right	39	−16	46	5.49	<0.0001
Cingulate gyrus	23	Right	7	−32	30	5.43	<0.0001
Cuneus	18	Right	16	−93	20	5.43	<0.0001
Cingulate gyrus	23	Left	1	−23	27	5.38	<0.0001
Cuneus	19	Right	9	−91	25	5.27	0.001
Anterior cingulate gyrus	25	Left	−4	13	−9	5.25	0.001
Inferior parietal lobule	40	Right	51	−30	32	5.03	0.002
Parahippocampal gyrus	28	Right	25	−18	−18	4.99	0.002
Inferior frontal gyrus	47	Right	27	30	−14	4.96	0.003
Superior temporal gyrus	22	Right	48	−19	−9	4.94	0.003
Precuneus	7	Left	−13	−71	43	4.83	0.005
Anterior cingulate gyrus	32	Left	−6	40	5	4.80	0.005
Parahippocampal gyrus	28	Left	−19	−15	−24	4.80	0.005
Middle temporal gyrus	21	Right	56	−22	−7	4.77	0.006
Precuneus	7	Right	6	−34	45	4.75	0.007
Precuneus	2	Right	8	−37	46	4.65	0.01
Precentral gyrus	44	Right	47	0	12	4.59	0.01
Medial frontal gyrus	11	Left	−8	29	−12	4.54	0.02
Fusiform gyrus	20	Right	57	−39	−21	4.46	0.022
Parahippocampal gyrus	34	Right	16	−3	−13	4.44	0.024
Inferior temporal gyrus	20	Right	63	−30	−15	4.44	0.024
Cingulate gyrus	31	Left	−3	−43	28	4.41	0.027
Precentral gyrus	6	Left	−48	−4	9	4.38	0.031
Medial frontal gyrus	6	Right	7	3	57	4.37	0.032
Posterior cingulate gyrus	29	Left	−9	−43	7	4.32	0.038
Inferior frontal gyrus	13	Right	39	25	7	4.32	0.038
Postcentral gyrus	5	Right	24	−37	64	4.31	0.04
Middle frontal gyrus	46	Right	44	30	17	4.27	0.048
Anterior cingulate gyrus	33	Left	−1	12	25	4.26	0.048

Probability values provide results demonstrating regions of increased cortical thickness in males relative to females, adjusting for age, in the full sample; BA, Brodmann Area.

Examination of the interaction between sex and age revealed that older ages males had decreased thickness in the postcentral gyrus (x = 54, y = −13, z = 56, t = 7.7), the supplementary motor area (x = −3, y = −34, z = 55, t = 8.5), and the medial temporal lobe in the territory of the parahippocampal gyrus (x = 34, y = −13, z = −17, t = 4.3). Females had decreased thickness at older ages relative to males, in the temporal lobe in the fusiform gyrus (x = 48, y = −65, z = 14, t = 4.9), in the occipital lobe (x = 4, y = −66, z = 32, t = 6.1), the inferior parietal lobule (x = 52, y = −43, z = 44, t = 5.0), the precentral gyrus (x = 27, y = 10, z = 63, t = 4.7), medial frontal gyrus (x = 7, y = 38, z = 28, t = 4.7), the middle temporal gyrus (x = 55, y = 17, z = −12, t = 5.4), and the inferior frontal gyrus (x = 53, y = 24, z = 17, t = 4.4).

Cortical Surface Area

Cortical surface area showed only a few age-related expansions with age (Supplementary Table 1–5) that were localized to the cingulate gyrus. Age-related contractions in cortical surface were seen in many regions throughout the brain (Table 6, Fig. 3) with the peak decreases seen in the temporal and parietal cortices. However, few brain regions demonstrated quadratic or cubic effects of age. The surface area in bilateral insula was associated with quadratic age effects (Supplementary Table 1–6), while only the surface areas of the superior temporal gyri demonstrated cubic effects of age (Supplementary Table 1–7). All analyses were adjusted for sex.

Table 6.

Cortical surface area: age-related regional contractions of cortical surface area

Brain region	Hemisphere	BA	x	y	z	t	P value
Superior temporal lobe	Right	22	47	−4	−7	10.70	<0.001
Superior temporal lobe	Left	22	−45	−8	−7	10.40	<0.001
Superior frontal gyrus	Right	10	8	67	14	6.3	<0.001
Superior frontal gyrus	Left	10	−12	65	16	5.04	0.0003
Middle frontal gyrus	Right	9	45	29	27	6.20	<0.001
Subcallosal gyrus	Right	34	14	4	−16	6.09	<0.001
Superior occipital gyrus	Left	19	−34	−86	32	5.57	<0.001
Superior temporal gyrus	Left	38	−33	7	−19	5.30	<0.001
Postcentral gyrus	Left	5	−3	−43	69	4.70	0.0002
Interior frontal gyrus	Left	45	−55	21	17	4.70	0.0002
Precuneus	Left	7	−5	−72	37	4.20	0.001
Postcentral gyrus	Left	2	−59	−21	32	3.98	0.003

Probability values provide results demonstrating regional contractions in surface area related to age, adjusting for sex, in the full sample; BA, Brodmann Area.

Figure 3.

Cortical (left), thalamic (middle) and striatal (right) surface area: association with age. Left: vertex-based surface analysis of corrected P values of age-related regional contractions in cortical surface area. Middle and right: corrected P values of age-related subcortical surface area expansion. All analyses were adjusted for sex and corrected for multiple comparisons (P < 0.05). The color bar indicates corrected P values for significant peaks at the vertex level (left) as well as at the cluster level (right).

When adjusting for age, we examined sex-specific effects on cortical surface area (Table 7). Males had greater cortical surface area expansion in relation to females in a number of brain regions particularly in frontoparietal brain regions as well as temporal regions.

Table 7.

Cortical surface area: sex-based differences

Brain region	BA	Hemisphere	x	y	z	t	P value
Posterior cingulate gyrus	29	Right	8	−40	8	11.34	<0.0001
Middle frontal gyrus	11	Right	−23	31	−15	10.68	<0.0001
	11	Left	−10	37	−10	8.84	<0.0001
	6	Left	−37	5	61	7.45	<0.0001
Parahippocampal gyrus	36	Right	35	−22	−27	10.2	<0.0001
		Right	24	−55	−5	8.61	<0.0001
Orbitofrontal gyrus	47	Right	14	27	−24	9.37	<0.0001
Cingulate gyrus	31	Left	−2	−33	36	9.34	<0.0001
Inferior parietal lobule	40	Left	−54	−19	23	9.2	<0.0001
Superior parietal lobule	5	Right	13	−29	47	9	<0.0001
Cingulate gyrus	31	Right	9	−25	41	8.95	<0.0001
Cuneus	18	Right	15	−73	13	8.72	<0.0001
Paracentral Lobule	6	Left	−4	−27	52	8.7	<0.0001
Inferior occipital gyrus	18	Left	−39	−81	−14	8.02	<0.0001
Medial frontal gyrus	6	Left	−4	38	37	6.74	<0.0001
Inferior temporal gyrus	20	Left	−40	0	−44	8.14	<0.0001
Middle temporal gyrus	37	Left	−45	−69	6	7.7	<0.0001
	39	Right	41	−61	19	7.7	<0.0001
Precentral gyrus	4	Left	−53	−7	44	7.31	<0.0001
	4	Right	57	−8	45	7.65	<0.0001
	44	Right	54	10	6	4.88	<0.0001
Middle temporal gyrus	21	Left	−63	−25	−13	7.7	<0.0001
Anterior cingulate gyrus	24	Right	1	24	17	6.9	<0.0001
Medial frontal gyrus	9	Right	15	52	38	6.25	<0.0001
Superior frontal gyrus	10	Right	21	66	11	7.92	<0.0001
Inferior frontal gyrus	47	Right	38	25	−17	8.33	<0.0001
Postcentral gyrus	3	Right	59	−15	23	7.9	<0.0001
Superior temporal gyrus	22	Right	59	−8	4	6.59	<0.0001
Middle temporal gyrus	21	Right	61	2	−27	5.8	<0.0001
	22	Right	62	−39	13	8.14	<0.0001

Probability values provide results demonstrating cortical surface area expansions in males compared to females, adjusting for age, in the full sample; BA, Brodmann Area.

The interaction analysis between age and sex revealed that males exhibited surface area expansion in the occipital lobe (x = −1, y = −86, z = −1.4, t = 4.1) at older ages. Males also demonstrated surface-area contraction at older ages in the superior temporal gyrus (x = −47, y = −5, z = −4, t = 4.3). Females demonstrated surface area contraction at older ages in the superior frontal gyrus (x = 7, y = 56, z = 22, t = 4.4), the inferior frontal gyrus (x = 55, y = 24, z = 5, t = 4), the inferior parietal lobule (x = 46, y = −39, z = 47, t = 4), and the superior frontal gyrus (x = 36, y = 56, z = 13, t = 4).

Thalamic Surface Area

The dorsal and ventral thalami demonstrated positive linear expansion in surface area with age (P < 0.001, Fig. 3). The surface area in the territories of the lateral and posterior thalami were best fit by quadratic models (P < 0.001). These same regions also showed a significant association with age in a cubic model (P < 0.001).

Sex-based differences in the surface area of the thalami encompassed the majority of this deep GM structure, with males demonstrating increased surface area relative to females (P < 0.001).

Sex and age interaction effects revealed that males had greater surface area expansion relative to females in bilateral regions of the superior, lateral, and inferior thalami at older ages (all, P < 0.01). Females relative to males demonstrated surface area expansion in bilateral ventral thalami at older ages (both, P < 0.001).

Striatum Surface Area

The striatum exhibited significant linear changes in surface area in relation to age (P < 0.05; Fig. 3), adjusting for sex. In particular, the body and the tail of the caudate nucleus demonstrated surface area expansion with age. In addition, bilateral medial putamen as well as the left lateral putamen demonstrated significant expansion in surface area at older ages. The surface area data were fit against cubic and quadratic models. Only a small section of the medial putamen was shown to have a quadratic association with age (P < 0.05).

Males demonstrated greater surface area expansion of the striatum relative to females when adjusting for age. Findings were significant across the surface areas of the body and tail of the caudate and the putamen (P < 0.001).

An interaction analysis examined the association of striatal surface area in males and females in relation to age. Males demonstrated surface area expansion at older ages in the medial and lateral right putamen (both, P < 0.004), and the medial left putamen (P < 0.002). Additionally, the tails of the caudate nuclei demonstrated surface area expansion in males at older ages (both, P < 0.01). Females did not demonstrate age-related surface area changes relative to males.

Globus Pallidus Surface Area

The surface area of the globus pallidus increased linearly with age in or the superior, inferior and posterior regions (Supplementary Fig. 1–2). Few linear age-related changes were evident in anterior regions of the globus pallidus. The age-effects of the surface area in posterior regions of the globus pallidus were best fit by quadratic and linear models. The peak surface area in these posterior regions occurred at ~ 10 years of age and decreased thereafter.

Sex-based differences in the surface area of the basal ganglia revealed that the near entirety of the surface area of the basal ganglia exhibited greater expansion in males compared to females (P < 0.001).

A sex by age interaction analysis demonstrated that males showed greater surface area expansion relative to females at older ages in the posterior portion of the globus pallidus, bilaterally (both, P < 0.005). Females demonstrated a small area of surface area expansion at older ages relative to males in the left superior globus pallidus (P < 0.01).

Discussion

Understanding structural changes in the brain across development is challenging, due to the changing physiology and structural trajectories of different cortical and subcortical regions; yet this is critical to provide a window into brain and behavior associations that can be used to identify clinically relevant biomarkers. Considerable variability in the findings of sex differences in brain structure are likely due to many studies using smaller samples, data from multi-site studies or focusing on a single brain region. Thus, we examined age-related and sex-based differences in both cortical and subcortical maturation from a single site with a, large cohort of participants aged 4–54 years. We examined macrostructural alterations in the cortex and subcortical areas using volumetric and surface area measurements in relation to age in males and females. Our results indicate that across brain morphometric measures that the age of peak maturation coincided with puberty-associated growth curves. TCV, WM, thalamic, and basal ganglia volumes largely peaked in early adolescence at 10–12 years of age, while the hippocampus volumes reached peak volume later in the second decade of life. Few sex-based differences in volumes were evident when adjusting for TCV, except in some thalamic nuclei. Males and females demonstrated similar maturational growth curves except in the sensorimotor nuclei in the deep GM. Surface area expansion in subcortical structures in the thalamus, striatum and globus pallidus paralleled age-related maturational decreases in cortical thickness and to some extent cortical surface area. Subcortical surface area expansion was predominantly seen in dorsal regions, in territories demonstrating connectivity between frontoparietal association cortices.

Gross and Regional Volumetric Maturation

Reports of age-related variations in the development of subcortical regions have been found inconsistently. Age-related increases in thalamic and hippocampal volumes, adjusted for TCV, have been reported (Ostby et al. 2009), which paralleled changes in WM development. However, inverse relations with age were reported in the basal ganglia, adjusting for TCV (Sowell et al. 2002; Sowell et al. 2004; Ostby et al. 2009). The morphological changes such as expansion or contraction during development that underlie the volumetric changes remain largely unexplored.

Our findings demonstrating larger TCV and inverted age-related relations between GM and WM are largely in agreement with previous findings. The decrease in GM seen in the developing brain is thought to reflect a number of both progressive and regressive events impacting neuronal and glial processes. In early development, losses of GM are thought to reflect synaptic pruning along with increases in WM reflective of changes in myelination (Natu et al. 2019). Age-related decreases in GM volume in the aging brain are thought to reflect neuronal shrinkage or loss of neurons and/or dendrites (Terry et al. 1987; Jacobs et al. 1997).

Sex-Based Differences in Brain Morphometry

We demonstrated subtle differences in subcortical volumes in the thalamus, striatum, globus pallidus, and hippocampus between males and females when adjusting for total cerebral volume. Our findings are in agreement with previous research demonstrating that without adjusting for TCV that sex-based differences are evident in many subcortical brain regions; however, when the adjustment is made for TCV, few sex differences in subcortical maturation remain (Paus 2010).

Examination of age-related maturation of subcortical structures showed variations between males and females primarily in the thalamic sensorimotor nuclei. Differences in male and female cortical and subcortical development has been attributed to variations in the distribution of androgens, sex steroid receptors. In particular, brain regions with high densities of androgens including the basal ganglia, amygdalae, hippocampi, and cerebellum are noted to demonstrate MRI-based structural differences between males and females (Sarkey et al. 2008). In the current work, our findings demonstrate peak subcortical volumes at ~ 20 years of age in the hippocampi, which are expected to occur several years after pubertal changes. These findings may be a reflection of the wide age range of our sample, as a previous study found that the hippocampi demonstrated a steady increase in volume, but in a sample with both a younger and narrower age range (1 month—25 years (Uematsu et al. 2012)).

Additionally, variations in sex hormone levels have been associated with volumetric differences in the amygdalae and hippocampi (Bramen et al. 2011), whereby testosterone levels in males and females predicted GM volume in these structures (Neufang et al. 2009; Bramen et al. 2012). Androgen receptors have been identified in the ventrolateral thalamic nuclei based on findings from immunohistochemistry (Kritzer 2004). Our results demonstrating sex differences in the volumes of the ventrolateral nucleus of the thalamus may be a result of underlying alterations in androgen receptors.

The molecular mechanisms underlying sexual dimorphism in the human brain are still under active investigation (McCarthy and Arnold 2011). In rodent models, genetic and hormonal factors contribute to sexual differentiation (Arnold 2009). Epigenetic regulation of steroid receptors plays an important role in sex-based differences in the brain (Murray et al. 2009). For instance, the volume of subcortical structures can be modified by altering key epigenetic processes. The impact of epigenetic factors on sex-based differences in human brain development have yet to be elucidated but may be a contributing factor to volumetric differences.

Surface Area

A key finding of the current work was that surface areas of the thalamus, striatum, and globus pallidus underwent significant expansions up until the fifth decade. Additionally, the surface area expansions were largely linear. We also noted that males exhibited greater surface expansions in subcortical regions in comparison to females. Findings of greater surface area expansions with age are in line with a previous report that examined a large sample (n = 1172) of typically developing children and young adults aged 5–25 years (Raznahan et al. 2014). In that previous study, they reported surface area expansions also in the dorsal surfaces, particularly in the body and the tail of the caudate nucleus. In our sample that contains a wider age range, we demonstrate similar findings. Only older adults (mean age 66 years) demonstrated contraction in subcortical surface area (Tullo et al. 2019). Additionally, our results of surface area expansion in subcortical brain regions are complemented by cortical surface changes in the same cohort. The subcortical surface areas showed expansion in dorsal regions, regions that show connectivity with frontoparietal regions; however, we see inverse relations with cortical thickness and cortical surface area, which both demonstrated significant linear decreases with age. Subcortical expansion coupled with cortical thickness decreases and surface area reductions may reflect a number of maturational processes including alterations in dendritic arborization, WM architecture, and/or synaptic pruning. As subcortical surface area expanded with age, this may reflect more specialized and efficient WM microstructural connections with cortical areas. In the cortex, age-related changes are reflected in alterations in dendritic arborization rather than neuronal losses and are partially explained by the underlying alterations in WM (Raz et al. 1997; Freeman et al. 2008; Paus 2010). Future studies combing both human MRI studies and immunohistochemistry in animals would aid in determining the mechanisms of subcortical and parallel maturational changes in the cortex.

Relevance to Neurodevelopmental Disorders

Our findings in relation to cortical thickness and surface area would indicate that sex-based differences are prevalent in several cortical and subcortical areas. Findings are relevant to the study of neurodevelopmental disorders, whereby boys carry an increased risk for the development for common childhood psychiatric and neurodevelopmental disorders. For instance, autism spectrum disorder primarily affects males; the risk for autism spectrum disorder is 3–4 times that for boys relative to girls. Similarly, attention deficit hyperactivity disorder and oppositional defiant disorder are much more commonly seen in boys relative to girls (Loeber et al. 2000; Scahill and Schwab-Stone 2000; Munkvold et al. 2011). In contrast, self-injury including eating disorders, anxiety, and depression are more commonly seen in females compared to males. Understanding the trajectories of male and female brain maturation in typical development could provide key insight into the mechanisms of neuropsychiatric and neurodevelopmental disorders to identify sex-specific biomarkers that can later be used to assess the downstream effects of behavioral and/or pharmaceutical interventions. Future large-scale neuroimaging studies including participants with a wide age range would benefit from obtaining endocrine data. Relating structural neuroimaging findings to endocrine data would shed light on the association among pubertal changes, circulating gonadal steroids and behaviors that are mediated by subcortical structures (Forbes and Dahl 2010; Op de Macks et al. 2016).

Conclusions

In a large, single-site, cross-sectional cohort of 436 individuals aged 4 to 54 years, we examined global and regional volumetric cortical and subcortical maturation in males and females scanned with structural MRI. Through examination of volumetric differences, we report few sex-based differences when adjusting for TCV except in select thalamic nuclei. Sex-based differences were evident, however, in cortical measures of thickness and surface area in frontoparietal regions. We demonstrated that surface area expansion of subcortical structures is inversely related to cortical maturation, predominantly in subcortical structures known to show connectivity with frontoparietal cortical regions. This suggests that subcortical surface area expansion is associated with age-related maturational changes in dendritic and synaptic architecture underling the cortical changes in both white and GM during childhood, adolescence, and adulthood.

Our findings are based on the cross-sectional study of brain morphological differences in children, adolescents, and adults. To examine age and sex effects in brain morphology, however, we normalized all T1-weighted images to a common image space. Spatial normalization issues may arise, particularly for MRI scans acquired in young children (4–5 years), which may be of concern for studies with small sample sizes, where increased variability in the data may unduly influence the results. Confirmation of our findings in future longitudinal cohorts are needed to fully quantify age- and sex-based differences in the trajectories of brain development across the lifespan.

Funding

This work was supported by the Canadian Institutes of Health Research.