SORL1 rs1699102 Moderates the Effect of Sex on Language Network

Abstract

Background:

Language ability differs between the sexes. However, it is unclear how this sex difference is moderated by genetic factors and how the brain interacts with genetics to support this specific language capacity. Previous studies have demonstrated that the sorting protein-related receptor (SORL1) polymorphism influences cognitive function and brain structure differently in males and females and is associated with Alzheimer’s disease risk.

Objective:

The aim of this study was to investigate the effects of sex and the SORL1 rs1699102 (CC versus T carriers) genotype on language.

Methods:

103 non-demented Chinese older adults from Beijing Aging Brain Rejuvenation Initiative (BABRI) database were included in this study. Participants completed language tests, T1-weighted structural magnetic resonance imaging (MRI) and resting-state functional MRI. Language test performance, gray matter volume, and network connections were compared between genotype and sex groups.

Results:

The rs1699102 polymorphism moderated the effects of sex on language performance, with the female having reversed language advantages in T carriers. The T allele carriers had lower gray matter volume in the left precentral gyrus. The effect of sex on language network connections was moderated by rs1699102; male CC homozygotes and female T carriers had higher internetwork connections, which were negatively correlated with language performance.

Conclusion:

These results suggest that SORL1 moderates the effects of sex on language, with T being a risk allele, especially in females. Our findings underscore the importance of considering the influence of genetic factors when examining sex effects.

Keywords

Alzheimer’s disease functional connectivity language network rs1699102 sex

INTRODUCTION

Language ability is one of the most comprehensive and high-level brain functions for humans and plays an important role in daily life. Previous studies have identified that language ability differs between sexes [1–3], with females outperforming males in verbal fluency, speech articulation, and grammatical skills [4]. Like other cognitive functions, language ability has been found to have a moderate to high level of heritability [5], with genetic factors accounting for 54% to 79% of its variance even in older adults [6]. It has been suggested that genetic factors contribute to differences in the brain and behavior between males and females [7, 8]. Previous studies have also found that sex and genetics interact to influence cognition [9–11]. However, it is unclear how genetic factors influence both sex effects and brain functions/structures, especially concerning language ability.

The sorting protein-related receptor (SORL1) gene is involved in amyloid precursor protein (APP) processing and trafficking, amyloid-β (Aβ) degradation, and lipid metabolism [12, 13]. The SORL1 gene polymorphism has been linked to language ability [10, 14] and other cognitive functions [11, 15–17] in elderly individuals. However, these SORL1 effects may interact with sex, as gene expression in the brain and the risk of Alzheimer’s disease (AD) differ between sexes in elderly individuals [18, 19]. Additionally, SORL1 effects on cognitive performance differ cross-sectionally and longitudinally in males and females [10, 11]. Furthermore, sex and SORL1 interaction effects on whole brain and regional hippocampus (Hip) atrophy [20] and white matter integrity [11] have been discovered. The single-nucleotide polymorphism (SNP) rs1699102 is located in the 3’ region of the SORL1 gene [21]. The SNP rs1699102 and haplotypes that encompass this SNP have been linked to the risk of AD in various ethnic groups [22–24]. Thus, this SNP may also affect brain cognition, structure, and function in a sex-specific manner.

Previous research on the effect of sex on brain activation for language has yielded mixed results. Some studies have shown different brain activation patterns for language in males and females [25, 26], while other studies did not [27]. However, meta-analyses reported a small or no sex effect on brain activation for language [28–30], suggesting that males and females have more similarities than differences in functional brain activation patterns for language. The mixed findings between studies may be due to sample size, task and stimulus heterogeneity across studies, but they could also be attributed to a failure to consider genetic factors. We raised this possibility, as language ability is known to be influenced by genetics, and genetic factors have been found to be related to gray matter volume (GMV) in cortical regions known to be involved in language processing [31, 32]. Furthermore, language processing relies on multiple regions and the connections between them, which compose the language network [33, 34]. Therefore, it is important to examine brain functional connections to understand how the brain organizes itself with respect to language. A recent study found sex differences in brain connectivity among language regions, as well as in brain network segregation and integration during language processing [35]. Additionally, language-related network connectivity was found to be influenced by genetic factors [36]. However, few studies have examined how the interaction between genetics and sex affects language ability. For example, it is largely unknown whether the interactive effects of genetic factors and sex on language manifest across the brain network.

In the present study, we used data from 103 non-demented elderly individuals who completed language ability test, high-resolution structural magnetic resonance imaging (MRI) and resting-state functional MRI, to examine the effects of SORL1 rs1699102 and sex on language performance, language-related regional GMV and language network connections. A voxel-based morphometry analysis was used to assess the regional GMV. Modularity analysis was used to divide the language-related regions into subnetworks based on their resting-state connectivity, and the connection attributes were calculated based on the networks. This study aims to examine the hypothesis that the effects of sex on language networks are moderated by the SORL1 rs1699102 polymorphism.

MATERIALS AND METHODS

Participants

This study included 103 participants from the Beijing Aging Brain Rejuvenation Initiative (BABRI) [37], which is an ongoing community-based cohort study focusing on cognitive, neuroimaging and sociodemographic factors related to aging and dementia. Details of participant selection have been previously described in our previous paper [37] and a brief description can also be seen in the Supplementary Material. The BABRI project was approved by the Ethics Committee of the State Key Laboratory of Cognitive Neuroscience and Learning and the Institutional Review Board of the Imaging Center for Brain Research at Beijing Normal University, and conformed to the principles of the Declaration of Helsinki.

The inclusion criteria of this study were as follows: (a) native Chinese speakers; (b) 50 years of age or older; (c) at least 6 years of education; (d) a score of 24 or higher on the Chinese version of the Mini-Mental State Examination [38]; and (e) no history of any neurologic, psychiatric, or systemic illnesses known to influence cognitive function. To evaluate language function, all participants took the Category Verbal Fluency Test (CVFT) [39]. Written informed consent was obtained from each participant. Blood samples were collected from all participants and used for genetic analysis.

Genotyping

All participants were prescreened for the rs1699102 genotype using TaqMan allele-specific assays on the 7900HT Fast Real-Time PCR System (Applied Biosystems, Foster City, CA). Another two SNPs, rs429358 and rs7412, were also genotyped to determine the APOE genotype (ɛ4 carriers or non-carriers). The sample success rates for all three SNPs were 100% (i.e., no failures across the participants to ‘call’ the polymorphisms), and the reproducibility of the genotyping was 100% according to duplication of at least 10% of the genotypes. Based on rs1699102 genotyping, participants were divided into two groups: 57 CC homozygotes and 46 T carriers (including 2 homozygotes).

MRI data acquisition

Participants were scanned with MRI within six months of the neuropsychological testing, and the mean interval time in this study was 2.16 months. Participants were scanned on a 3.0 T Siemens Trio scanner in the Imaging Center for Brain Research at Beijing Normal University. High-resolution T1-weighted structural images were collected from all participants using the sagittal 3D magnetization-prepared rapid gradient-echo (MP-RAGE) sequence (acquisition parameters: 176 sagittal slices, TR = 1900 ms, TE = 3.44 ms, slice thickness = 1 mm, flip angle = 9°, inversion time = 900 ms, FOV = 256×256 mm², acquisition matrix = 256×256). Resting-state functional MRI data were collected from 52 CC homozygotes and 44 T carriers for 8 minutes using an echo-planar imaging sequence (acquisition parameters: 33 axial slices, TR = 2000 ms, TE = 30 ms, slice thickness = 3.5 mm, flip angle = 90°, acquisition matrix = 64×64, FOV = 200×200 mm²).

Data processing

Structural image processing

The structural images were preprocessed using the Computational Anatomy Toolbox (CAT) of the Statistical Parametric Mapping (SPM, http://www.neuro.uni-jena.de/cat/) software. The structural images were segmented into gray matter (GM), white matter, and cerebrospinal fluid, and the individual GM map was spatially normalized to MNI (Montreal Neurological Institute) space and smoothed with an 8-mm full-width half-maximum (FWHM) Gaussian kernel. A general linear model (GLM) was constructed using individual GM images and covariates such as age, education, APOE genotype and total intracranial volume (TIV) to test the effect of genotype and sex. Because we focused on language ability, a mask of the language-related regions was used. Regions of interest (ROIs) were selected from language-related regions based on a previous study [40]. In brief, regions involved in language processing, as well as their contralateral homologues, were selected. A total of 58 ROIs (29 ROIs for each hemisphere) were obtained (Fig. 1) from the Harvard-Oxford probabilistic atlas (probability threshold at 25%) to create the mask. The GM results were corrected for multiple comparisons using a Gaussian random field (GRF) with a voxel significance of p < 0.001 and a cluster significance of p < 0.05.

Fig. 1

Region of interests involved in this study. FP, Frontal Pole; IC, Insular Cortex; SFG, Superior Frontal Gyrus; MidFG, Middle Frontal Gyrus; IFGtri, Inferior Frontal Gyrus, pars triangularis; IFGoper, Inferior Frontal Gyrus, pars opercularis; PreCG, Precentral Gyrus; aSTG, Superior Temporal Gyrus, anterior division; pSTG, Superior Temporal Gyrus, posterior division; aMTG, Middle Temporal Gyrus, anterior division; pMTG, Middle Temporal Gyrus, posterior division; toMTG, Middle Temporal Gyrus, temporooccipital part; aITG, Inferior Temporal Gyrus, anterior division; pITG, Inferior Temporal Gyrus, posterior division; toITG, Inferior Temporal Gyrus, temporooccipital part; aSMG, Supramarginal Gyrus, anterior division; pSMG, Supramarginal Gyrus, posterior division; AG, Angular Gyrus; iLOC, Lateral Occipital Cortex, inferior division; FOrb, Frontal Orbital Cortex; aPaHC, Parahippocampal Gyrus, anterior division; pPaHC, Parahippocampal Gyrus, posterior division; aTFusC, Temporal Fusiform Cortex, anterior division; pTFusC, Temporal Fusiform Cortex, posterior division; TOFusC, Temporal Occipital Fusiform Cortex; FO, Frontal Operculum Cortex; HG, Heschl’s Gyrus (includes H1 and H2); PT, Planum Temporale; Hip, Hippocampus.

Functional image processing

Resting-state functional MRI data were preprocessed using Data Processing & Analysis of Brain Imaging (DPABI, https://rfmri.org/DPABI) software [41]. The preprocessing of functional images included removing the first 10 time points, slice timing, realignment, nuisance covariate regression, bandpass filtering between 0.01 and 0.1 Hz, coregistration to high-resolution structural images, normalization to MNI space, resampling to 3×3×3 mm³, and smoothing with a 6-mm FWHM Gaussian kernel. Four participants with excessive head movements (translation > 3 mm or rotation > 3 degrees in any direction) were excluded (1 CC homozygote and 3 T carriers), leaving 92 participants in the functional image analysis.

ROI-to-ROI functional connectivity was obtained using the Graph Theoretical Network Analysis Toolbox (GRETNA, https://www.nitrc.org/projects/gretna/) [42] based on the language regions mentioned above using Pearson correlations converted to z scores using Fisher’s Z transformation. We obtained 58 by 58 correlation metrics for each participant, and only positive connectivity was used for further analysis. To obtain different language subnetwork modular structures, a data-driven modularity analysis was performed based on the whole group mean connectivity matrix [43]. The module identification algorithm optimizes the total number of modules and the associated module affiliation of regions by maximizing the modularity Q, defined as follows: $Q = \sum_{S = 1}^{N_{M}} [\frac{l_{s}}{L} - {(\frac{d_{s}}{2 L})}^{2}]$ where N_M is the number of modules (subnetworks), l_s is the total number of edges within module s, L is the total number of edges in the whole network, d_s is the total number of degrees at each region in the module s. We used the Louvain algorithm to detect the modules in this study [44]. The Louvain algorithm was carried out 10 times,10 Q values were obtained, and the modular partition corresponding to the max Q value was taken as the final partition. As a result, the language regions were divided into 4 modules, and each module was treated as a subnetwork for further analysis. To calculate graph measures, a 0.25 sparsity threshold was used on the participant’s positive connectivity matrix (i.e., only the strongest 25% of positive correlations were maintained for each participant). As there is no consensus or strong precedent on which specific sparsity threshold is best, we chose a value in the middle of a range of values that have commonly been used in related work. Additionally, to evaluate the reliability of our results, sparsity values of 0.2 and 0.3 were also used, and the results did not change (see Supplementary Material).

Then, three connectivity measures were used to compare sex and genotype groups: a) the network connectivity density, which was calculated as the average connectivity strength of all connections within the same subnetwork or between different subnetworks; b) the modularity interaction, which was calculated as the total number of connections within the same subnetwork or between different subnetworks; and c) the participation coefficient (PC) for each region, which reflected the ability of a region to maintain communication between its own subnetwork and other subnetworks, which was calculated as follows: $P C_{I} = 1 - \sum_{S = 1}^{N_{M}} {(\frac{K_{IS}}{K_{I}})}^{2}$ where N_M is the number of networks (4 in this study), K_IS is the number of connections between region I and regions in the subnetwork S, and K_I is the total number of region I connections in all subnetworks. The PC of region I ranged from 0 to 1, with higher values indicating more connections spread across multiple subnetworks.

Statistical analyses

PLINK software was used to assess the Hardy-Weinberg equilibrium. SPSS 20.0 software was used to conduct a two-way analysis of variance (ANOVA) with the rs1699102 genotype and sex as grouping factors to examine the demographic variables (age and years of education). A two-way analysis of covariance (ANCOVA) was conducted to examine neuropsychological test performance, network connectivity density, modularity interaction and the PC, with genotype and sex as grouping factors and age, years of education, and APOE ɛ4 status as covariates. Partial correlations were obtained to investigate the relationships between language performance and the significant group effect measures, with age, years of education, and APOE ɛ4 status as covariates. A statistical significance level of p < 0.05 was used in these analyses.

RESULTS

Demographics and cognitive performance

SORL1 rs1699102 genotype frequencies did not deviate from Hardy-Weinberg equilibrium (p > 0.05). For age, education, and APOE ɛ4 status, there were no significant main or interaction effects between sex and the rs1699102 genotype (Table 1). All of the subsequent analyses were adjusted for age, education, and APOE ɛ4 status. For the neuropsychological tests, no sex or rs1699102 genotype main effects were found, but significant interaction effects (Supplementary Figure 1) were found for the CVLT (F = 5.57, p = 0.020). The post hoc analysis revealed that in CC homozygotes, there was a marginally significant difference between females and males, with females outperforming males (p = 0.060). For T carriers, however, this pattern flipped, and males tended to perform better than females, suggesting that T may be a risk allele for females.

Table 1

Demographics and cognitive performance of all participants

	CC (N = 57)		CT/TT (N = 46)		rs1699102	Sex	rs1699102×Sex
	Male (N = 27)	Female (N = 30)	Male (N = 21)	Female (N = 25)	F/χ² (p) ^a	F/χ² (p) ^a	F/χ² (p) ^a
Age, y	69.93±6.86	70.80±6.87	71.14±7.50	68.84±5.50	0.08 (0.777)	0.30 (0.586)	1.48 (0.227)
Age range, y	56-82	56-81	56-85	58-80
Education, y	11.74±3.53	11.13±3.12	11.95±3.40	11.76±3.32	0.40 (0.527)	0.37 (0.546)	0.10 (0.754)
APOE ɛ4 carriers	10 (37%)	9 (30%)	6 (29%)	4 (16%)	1.61 (0.205)	0.57 (0.449)	2.26 (0.520)
MMSE	27.41±2.00	27.53±1.96	27.81±1.54	27.40±1.76	0.01 (0.907)	0.18 (0.671)	1.21 (0.274)
CVFT	41.85±8.90	45.07±9.13	46.15±7.68	43.00±7.58	0.03 (0.862)	0.19 (0.663)	5.57 (0.020)

Values are presented as the mean±SD. MMSE, Mini-Mental State Examination; CVFT, Category Verbal Fluency Test. ^a Comparisons between groups were performed using the Wald χ²-test for APOE ɛ4. Two-way analysis of covariance (two-way ANCOVA) was used to determine the main effect of genotype and sex on cognitive performance, as well as the interaction between them (with age, education, and APOE ɛ4 as covariates).

Effects of rs1699102 genotype and sex on gray matter volume

Table 2 shows the sex and rs1699102 genotype differences in the GMV of language regions (GRF p < 0.05 corrected). Significant sex effects were identified mainly in the left superior temporal gyrus (STG), with females having a lower GMV than males. Significant genotype effects were identified in the left precentral gyrus (PreCG), with higher volumes in CC homozygotes. We did not find any interaction effects between the rs1699102 genotype and sex in the GMV.

Table 2

Regions with significant rs1699102 genotype and sex effects on gray matter volume

Brain region	MNI coordinates			Cluster size	T value
	X	Y	Z
Gene effect CC > CTTT
Left Precentral Gyrus	–43.5	–3	57	1486	4.912
Sex effect male > female
Left Superior Temporal Gyrus	–63.5	–39	12	791	5.049

Voxel p < 0.001, Cluster p < 0.05, GRF corrected.

Effects of rs1699102 genotype and sex on language network connectivity

Network-wise changes

The mean connectivity matrix of all participants was subjected to a modularity analysis to determine whether groups of regions were highly connected with one another but less connected with other regions. Four modules were identified, and the modularity Q was 0.36. Based on their spatial distribution and function in language processing [33], we divided the modules into 4 subnetworks: the integration, semantics, articulatory and auditory networks. The details for the regions within each subnetwork are shown in Fig. 2A and Supplementary Table 1.

Fig. 2

A) Language subnetwork region distribution; as the entire network is bilaterally symmetrical, only one hemisphere is shown. B) Significant sex and rs1699102 interaction effects on internetwork connections between the integration and articulatory networks. C) Significant sex and rs1699102 interaction effects on the participation coefficient. FP, Frontal Pole; IC, Insular Cortex; SFG, Superior Frontal Gyrus; MidFG, Middle Frontal Gyrus; IFGtri, Inferior Frontal Gyrus, pars triangularis; IFGoper, Inferior Frontal Gyrus, pars opercularis; PreCG, Precentral Gyrus; aSTG, Superior Temporal Gyrus, anterior division; pSTG, Superior Temporal Gyrus, posterior division; aMTG, Middle Temporal Gyrus, anterior division; pMTG, Middle Temporal Gyrus, posterior division; toMTG, Middle Temporal Gyrus, temporooccipital part; aITG, Inferior Temporal Gyrus, anterior division; pITG, Inferior Temporal Gyrus, posterior division; toITG, Inferior Temporal Gyrus, temporooccipital part; aSMG, Supramarginal Gyrus, anterior division; pSMG, Supramarginal Gyrus, posterior division; AG, Angular Gyrus; iLOC, Lateral Occipital Cortex, inferior division; FOrb, Frontal Orbital Cortex; aPaHC, Parahippocampal Gyrus, anterior division; pPaHC, Parahippocampal Gyrus, posterior division; aTFusC, Temporal Fusiform Cortex, anterior division; pTFusC, Temporal Fusiform Cortex, posterior division; TOFusC, Temporal Occipital Fusiform Cortex; FO, Frontal Operculum Cortex; HG, Heschl’s Gyrus (includes H1 and H2); PT, Planum Temporale; Hip, Hippocampus. Values are expressed as the mean±SEM.; *p< 0.05; **p < 0.01; ***p < 0.001.

The connectivity differences among the 4 subnetworks were calculated from each participant’s connectivity matrix with a sparsity threshold (0.25). For the network connectivity density, a significant sex main effect was found for within-network connectivity, including in the semantics (F = 6.19, p = 0.015) and articulatory networks (F = 6.31, p = 0.014); a significant sex main effect was also found for the between-network connectivity of the integration and articulatory networks (F = 4.12, p = 0.045), with males having higher connectivity within and between networks than females. In addition, an interaction effect was found for the between-network connectivity of the integration and articulatory networks (F = 6.76, p = 0.011, Fig. 2B). In CC homozygotes, the post hoc results indicated that males had significantly higher connectivity than females (p < 0.001), while in T carriers, this difference diminished, and the sexes had equal connectivity density.

For the modularity interaction, males were found to have more connections in the semantics network than females (F = 6.63, p = 0.014). Interaction effects were also found in the internetwork connections between the integration and articulatory networks (F = 5.66, p = 0.020, Fig. 2B). Similar to the connectivity density results, post hoc tests indicated that male CC homozygotes exhibited more connections than female CC homozygotes (p = 0.008).

Region-wise changes

Sex effects were found in the PCs of the left posterior division of the parahippocampal gyrus (pPaHC, F = 4.18, p = 0.044), right posterior division of the temporal fusiform cortex (pTFusC, F = 5.86, p = 0.018), left temporal occipital fusiform cortex (TOFusC, F = 6.10, p = 0.016), left Hip (F = 3.99, p = 0.049), right Hip (F = 4.37, p = 0.037), right insular cortex (IC, F = 4.49, p = 0.037), and right Heschl’s gyrus (HG, F = 12.47, p < 0.001), with females having higher PCs, indicating that these regions were more connected to subnetworks other than their own in females than in males.

Additionally, some regions had interaction effects (Fig. 2C), including the left posterior division of the inferior temporal gyrus (pITG, F = 7.84, p = 0.006), right angular gyrus (AG, F = 7.29, p = 0.008), left pPaHC (F = 5.65, p = 0.020), and right pars opercularis of the inferior frontal gyrus (IFGoper, F = 6.17, p = 0.015). Additionally, post hoc tests indicated that the PCs were lower in males than in females only in T carriers (pITG, p = 0.047; AG, p = 0.041; pPaHC, p = 0.004; IFGoper, p = 0.037). Notably, the differences in network connectivity and PCs between sex and genotype groups were similar to the different sparsity thresholds of 0.2 and 0.3 used to define connections (see Supplementary Material).

Correlations between language and network connectivity

Partial correlation analyses were conducted for each group between each significant network measure and language (Fig. 3). For the network-wise measure, significant correlations were found between the CVFT and the internetwork connectivity of the integration and articulatory networks in male CC homozygotes (r=-0.51 p = 0.013), and the strength of this association was significantly stronger than that in male T carriers (z = 1.81, p = 0.035, one-tailed) and marginally stronger than that in female CC homozygotes (z = 1.43, p = 0.076, one-tailed). A similar result was found for the modularity interaction between these two networks, and a significant correlation with language was also found in male CC homozygotes (r=-0.49, p = 0.017). The difference in correlation strength was significant between male CC homozygotes and female CC homozygotes (Z = 1.67, p = 0.048, one-tailed), and it was marginally significant between male CC homozygotes and male T carriers (Z = 1.39, p = 0.082, one-tailed). For the regional measure, significant correlations between language and the PC were found in female T carriers only in the right AG (r=-0.54 p = 0.020). The strength of the correlation for female T carriers was significantly stronger than that for female CC homozygotes (Z = 2.09, p = 0.018, one-tailed).

Fig. 3

Correlations between language performance and A) internetwork connections between the integration and articulatory networks and B) participation coefficient. CVFT, Category Verbal Fluency Test; AG, Angular Gyrus; *p<0.05.

DISCUSSION

In this study, we examined the effects of sex and SORL1 rs1699102 polymorphism on language performance and language network connections in both the GMV and intrinsic connections of non-demented elderly individuals. We found that 1) Behaviorally, the rs1699102 polymorphism moderated the effects of sex on language performance, with females tending to perform better than males in language in CC homozygotes, while in T carriers, this trend reversed; 2) The T allele carriers had lower GMV in the left precentral gyrus which related to language; 3) Interaction effect of sex and genotype was found on language network connections, with female T carriers having higher internetwork connections, which were negatively correlated with language performance. These results suggest that SORL1 moderates the effects of sex on language, with T being a risk allele, especially in females.

In terms of these effects on behavioral results, we did not find a main effect for sex and genotype. A significant interaction effect on cognitive performance was found in the language test, and rs1699102 moderated the effects of sex on language performance. Females tended to perform better than males in language, as previously reported [1–4]; however, we only found this trend in CC homozygotes, while in T carriers, this trend reversed, suggesting that the T allele may be deleterious in females. The SORL1 protein is involved in APP processing and trafficking, Aβ degradation, and could interact with ApoE and tau protein [12, 13]. The reduced expression of SORL1 was found in both AD and mild cognitive impairment and correlated with cognitive performance [45]. The SNP rs1699102 could significantly regulate SORL1 expression in human brain tissues, with the T allele associated with decreased expression [46]. And the T allele has also been found to be associated with a steeper age-related decline in episodic memory and processing speed [15]. Furthermore, previous studies suggest that SORL1 effect may interact with sex. For example, Cellini and colleagues have found a sex effect for SORL1 gene, with the association between genotype and AD mainly detected in females [18]. This sex bias is further explored by animal experiments, with Lee and colleagues finding that sex-specific alteration of SORL1 expression transpired during the aging process in females [19]. As for the cognitive function, Liang and colleagues found that the SORL1 SNP polymorphism interacts with sex in executive function [11]. A longitudinal study also found that several SORL1 SNPs were associated with cognitive change trajectories in older adults across multiple domains, including language ability; furthermore, these effects differed between sexes [10]. Taken together, our results, consistent with previous findings, suggest that SORL1 may influence cognitive function in a sex-specific manner. And the rs1699102 T allele may be a risk allele for language ability, especially in females; thus, it is crucial to consider sex and genetic factors at the same time.

Previous studies have identified that global and regional GM are related to general intelligence and cognitive function in a variety of areas, including language ability [47, 48]. In the current study, we examined the effects of sex and the rs1699102 polymorphism on the GMV of regions involved in language processing [40], and the main effects on these regions were found. The effects of sex on GMV have long been debated, with some studies finding little evidence [49], and others finding this effect, although sex may affect regions differently [50]. In our study, it was found that males had a larger temporal region volume than females. This advantage for males is consistent with previous findings that males had larger brain volumes and regional GM volumes distributed in the temporal lobe, occipital lobe, brain stem, and cerebellum [51]. With a large sample size (N = 40,028) from the UK Biobank, Williams and colleagues also confirmed the sex effect on brain structure and found that males had a larger volume in the left temporal pole [50]. Regarding the main genetic effects, larger volumes were found in CC homozygotes than in T carriers. This finding suggests that the T allele is associated with a reduced GMV, which is consistent with the findings of Li and colleagues [15]. They found that T allele carriers had faster volume reduction than CC homozygotes in several regions, including the frontal, temporal and limbic regions. Taken together, these results further confirm that the rs1699102 T allele may be a risk allele for language-related region GM atrophy.

Our study also found genotype and sex effects on the intrinsic connections between language networks. Brain networks are composed of multiple coactivating brain regions and are thought to be functionally specialized due to interregional connectivity [52]. Language processing involves multiple regions that comprise different networks. Here, we divided the language regions into different language subnetworks based on their connections using a data-driven method, and the regions that were connected closely formed a module. We named the four subnetworks based on their assumed functions from a previous study [33, 34]. The first module mainly consisted of temporal lobe regions including the Hip, fusiform cortex (FusC), parahippocampal gyrus (PaHC), middle temporal gyrus (MTG), and inferior temporal gyrus (ITG). These regions are involved in semantic functions such as semantic association and semantic memory [33, 53, 54]; thus, we defined it as the semantics network. The second module was primarily located in the auditory cortex, HG, STG, and PT; thus, we defined it as the auditory network. Another module consisted of the PreCG, inferior frontal gyrus, FO and supramarginal gyrus and was involved in orofacial motor activity, sublexical and semantic reading, and articulatory planning [33, 55]. Therefore, we defined this module as the articulatory network. The last module was composed of distinct regions from the frontal, parietal and temporal lobes, including the superior frontal gyrus, frontal pole, MidFG, AG, MTG, and ITG. These regions are involved in multiple language processes, such as word selection, semantics to phonology, visual word-form processing, integration and prediction in speech comprehension, speech initiation and spontaneity; as a result, the last module was defined as the integration network [33, 56]. It should be noted that our language network did not show lateralization in elderly individuals, and the regions in the network were bilaterally symmetrical. Language ability is left-lateralized, as the left hemisphere is dominant in language processing in most people [57]. However, hemispheric asymmetry decreases with age (also known as the HAROLD effect), which reflects a compensatory function or a dedifferentiation process [58]. Previous studies have also found that language lateralization is age-dependent [27] and reduced in elderly individuals [59, 60]. For example, Szaflarski and colleagues found that language lateralization slowly decreases between 25 and 70 years [59]. Thus, the bilateral symmetrical language network in this study may also reflect the effect of aging on brain function.

The language network connections also exhibited sex differences, as measured by connectivity density, modularity interaction and PC for regions in this study. Our results are consistent with previous findings that functional connections among distinct brain regions for language may differ between males and females [35]. Furthermore, the effects of sex on language network connections were moderated by SORL1. The interaction effect between sex and rs1699102 was found in the internetwork connections between the integration and articulatory networks, with more connections found in male CC homozygotes but not in male T carriers. However, this strong connection was negatively correlated with language performance in male CC homozygotes. Similarly, female T carriers had PCs that showed interaction effects (left pITG, right AG, left pPaHC, right IFGoper), with more connections distributed across different networks, while a high PC was also negatively correlated with language performance, suggesting that a high PC may not benefit female T carriers. This finding is consistent with previous lesion studies which found that the frontal and temporal cortices play important roles in fluency tests [61, 62]. The increased activation in the right inferior frontal gyrus during the verbal fluency test was negatively correlated with performance in older adults [63], as was the HAROLD phenomenon. The right AG was more activated during the cognitive control of language processes [64]. Thus, these elevated PCs during rest may indicate brain network dysfunction. In addition, because our language networks are data-driven and based on the mean connectivity of all participants, and both the internetwork connections and PCs reflect communication between different subnetworks, our language networks may reflect the degree to which the network pattern deviates from the mean connectivity pattern. Thus, in our study, internetwork connections during rest may be inefficient and incapable of contributing to language performance.

Several limitations of this study should be acknowledged. First, the language test we used was based on word fluency, which is influenced by both genetics and age. The language process includes multiple processes, and further research may consider other language tests to examine the genetic and sex effects on language. Second, our networks during rest were based on language regions from previous studies, and how brain connectivity during language processing is affected by sex and genetics remains unknown. Third, only the SNP rs1699102 was chosen, with no other SNPs discussed. Finally, our small sample size may also limit our findings. Future large sample study considering other SNPs or GWAS may give an improved and detailed depiction of the effects of sex and genetic factors on the language.

Footnotes

ACKNOWLEDGMENTS

The authors have no acknowledgments to report.

FUNDING

This work was supported by State Key Program of National Natural Science of China (grant number 82130118), Funds for International Cooperation and Exchange of the National Natural Science Foundation of China (grant number 81820108034), the National Natural Science Foundation of China (grant number 82071205).

CONFLICT OF INTEREST

Zhanjun Zhang is an Editorial Board Member of this journal but was not involved in the peer-review process nor had access to any information regarding its peer-review.

All other authors have no conflict of interest to report.

DATA AVAILABILITY

The data supporting the findings of this study are available from the corresponding author on reasonable request.

The supplementary material is available in the electronic version of this article: .

References

Wirth

, Horn

, König

, Stein

, Federspiel

, Meier

, Michel

, Strik

(2007) Sex differences in semantic processing: Event-related brain potentials distinguish between lower and higher order semantic analysis during word reading. Cereb Cortex 17, 1987–1997.

Moscoso del Prado

Martín F

(2017) Vocabulary, grammar, sex, and aging. Cogn Sci 41, 950–975.

Snitz

, Unverzagt

, Chang

, Bilt

, Gao

, Saxton

, Hall

, Ganguli

(2009) Effects of age, gender, education and race on two tests of language ability in community-based older adults. Int Psychogeriatr 21, 1051–1062.

Weiss

, Kemmler

, Deisenhammer

, Fleischhacker

, Delazer

(2003) Sex differences in cognitive functions. Pers Individ Dif 35, 863–875.

Lee

, Thalamuthu

, Henry

, Trollor

, Ames

, Wright

, Sachdev

, Team

(2018) Genetic and environmental influences on language ability in older adults: Findings from the Older Australian Twins Study. Behav Genet 48, 187–197.

Lee

, Henry

, Trollor

, Sachdev

(2010) Genetic influences on cognitive functions in the elderly: A selective review of twin studies. Brain Res Rev 64, 1–13.

Ngun

, Ghahramani

, Sanchez

, Bocklandt

, Vilain

(2011) The genetics of sex differences in brain and behavior. Front Neuroendocrinol 32, 227–246.

Tower

, Arbeitman

(2009) The genetics of gender and life span. J Biol 8, 38.

Mortensen

, Hogh

(2001) A gender difference in the association between APOE genotype and age-related cognitive decline. Neurology 57, 89–95.

10.

Reynolds

, Zavala

, Gatz

, Vie

, Johansson

, Malmberg

, Ingelsson

, Prince

, Pedersen

(2013) SORL1 predicts longitudinal cognitive change. Neurobiol Aging 34, 1710.e11–8.

11.

Liang

, Li

, Lv

, Shu

, Chen

, Li

, Zhang

, Hu

, Zhang

(2015) Sex moderates the effects of the Sorl1 gene rs2070045 polymorphism on cognitive impairment and disruption of the cingulum integrity in healthy elderly. Neuropsychopharmacology 40, 1519–1527.

12.

Andersen

, Rudolph

, Willnow

(2016) Risk factor SORL1: Fromgenetic association to functional validation in Alzheimer’s disease. Acta Neuropathol 132, 653–665.

13.

Yin

, Yu

, Tan

(2015) The role of SORL1 in Alzheimer’s disease. Mol Neurobiol 51, 909–918.

14.

Andrews

, Das

, Cherbuin

, Anstey

, Easteal

(2016) Association of genetic risk factors with cognitive decline: The PATH through life project. Neurobiol Aging 41, 150–158.

15.

, Lv

, Yang

, Wei

, Chen

, Li

, Zhang

(2017) SORL1 rs1699102 polymorphism modulates age-related cognitive decline andgray matter volume reduction in non-demented individuals. Eur JNeurol 24, 187–194.

16.

Seshadri

, DeStefano

, Au

, Massaro

, Beiser

, Kelly-HayesM , Kase

, D’Agostino

, Sr. , Decarli

, Atwood

, Wolf

(2007) Genetic correlates of brain aging on MRI and cognitive testmeasures: A genome-wide association and linkage analysis in theFramingham Study. BMC Med Genet 8 Suppl 1, S15.

17.

Cruz-Sanabria

, Bonilla-Vargas

, Estrada

, Mancera

, Vega

, Guerrero

, Ortega-Rojas

, María

, Romero

, Montañás

(2020) Analysis of cognitive performance and polymorphisms of SORL1, PVRL2, CR1, TOMM40, APOE,PICALM, GWAS_14q, CLU, and BIN1 in patients with mild cognitive impairment and cognitively healthy controls. Neurologia (Engl Ed) 36, 681–691.

18.

Cellini

, Tedde

, Bagnoli

, Pradella

, Piacentini

, Sorbi

, Nacmias

(2009) Implication of sex and SORL1 variants in Italian patients with Alzheimer disease. Arch Neurol 66, 1260–1266.

19.

Lee

, Peterson

, Freeman

(2017) Sex-specific characterizationand evaluation of the Alzheimer’s disease genetic risk factor sorl1in zebrafish during aging and in the adult brain following a 100ppbembryonic lead exposure. J Appl Toxicol 37, 400–407.

20.

Assareh

, Piguet

, Lye

, Mather

, Broe

, Schofield

, Sachdev

, Kwok

(2014) Association of SORL1 gene variants with hippocampal and cerebral atrophy and Alzheimer’s disease. Curr Alzheimer Res 11, 558–563.

21.

Caglayan

, Bauerfeind

, Schmidt

, Carlo

, Prabakaran

, Hubner

, Willnow

(2012) Identification of Alzheimer disease risk genotype that predicts efficiency of SORL1 expression in the brain. Arch Neurol 69, 373–379.

22.

Rogaeva

, Meng

, Lee

, Gu

, Kawarai

, Zou

, Katayama

, Baldwin

, Cheng

, Hasegawa

, Chen

, Shibata

, Lunetta

, Pardossi-Piquard

, Bohm

, Wakutani

, Cupples

, Cuenco

, Green

, Pinessi

, Rainero

, Sorbi

, Bruni

, Duara

, Friedland

, Inzelberg

, Hampe

, Bujo

, Song

, Andersen

, Willnow

, Graff-Radford

, Petersen

, Dickson

, Der

, Fraser

, Schmitt-Ulms

, Younkin

, Mayeux

, Farrer

, St

George-Hyslop P

(2007) The neuronal sortilin-related receptor SORL1 is geneticallyassociated with Alzheimer disease. Nat Genet 39, 168–177.

23.

Tan

, Lee

, Chen

, Teo

, Zhao

, Lee

(2009) SORL1 haplotypes modulate risk of Alzheimer’s disease in Chinese. Neurobiol Aging 30, 1048–1051.

24.

Wang

, Lei

, Zheng

, Li

, Cui

, Hao

(2016) Meta-analysis of the Association between Alzheimer Disease and Variants in GAB2, PICALM, and SORL1. Mol Neurobiol 53, 6501–6510.

25.

Baxter

, Saykin

, Flashman

, Johnson

, Guerin

, Babcock

, Wishart

(2003) Sex differences in semantic language processing: A functional MRI study. Brain Lang 84, 264–272.

26.

Garn

, Allen

, Larsen

(2009) An fMRI study of sex differences in brain activation during object naming. Cortex 45, 610–618.

27.

Hirnstein

, Westerhausen

, Korsnes

, Hugdahl

(2013) Sex differences in language asymmetry are age-dependent and small: A large-scale, consonant–vowel dichotic listening study with behavioral and fMRI data. Cortex 49, 1910–1921.

28.

Sommer

, Aleman

, Bouma

, Kahn

(2004) Do women really have more bilateral language representation than men? A meta-analysis of functional imaging studies. Brain 127, 1845–1852.

29.

Sommer

, Aleman

, Somers

, Boks

, Kahn

(2008) Sex differences in handedness, asymmetry of the planum temporale and functional language lateralization. Brain Res 1206, 76–88.

30.

Sato

(2020) The neurobiology of sex differences during language processing in healthy adults: A systematic review and a meta-analysis. Neuropsychologia 140, 107404.

31.

Jamadar

, Powers

, Meda

, Gelernter

, Gruen

, Pearlson

(2011) Genetic influences of cortical gray matter in language-related regions in healthy controls and schizophrenia. Schizophr Res 129, 141–148.

32.

Meda

, Gelernter

, Gruen

, Calhoun

, Meng

, Cope

, Pearlson

(2008) Polymorphism of DCDC2 reveals differences in cortical morphology of healthy individuals-a preliminary voxel based morphometry study. Brain Imaging Behav 2, 21–26.

33.

Price

(2012) A review and synthesis of the first 20 years of PET and fMRI studies of heard speech, spoken language and reading. Neuroimage 62, 816–847.

34.

Friederici

, Gierhan

SME

(2013) The language network. Curr Opin Neurobiol 23, 250–254.

35.

, Liang

, Ou

, Li

, Luo

, Tan

(2020) Sex differences in functional brain networks for language. Cereb Cortex 30, 1528–1537.

36.

Jamadar

, Powers

, Meda

, Calhoun

, Gelernter

, Gruen

, Pearlson

(2013) Genetic influences of resting state fMRI activity in language-related brain regions in healthy controls and schizophrenia patients: A pilot study. Brain Imaging Behav 7, 15–27.

37.

Yang

, Li

, Zhang

, Chen

, Li

, Wei

, Lu

, Liang

, Liu

, Shu

, Wang

, Guan

, Tao

, Wang

, Jia

, Ai

, Cui

, Wang

, Peng

, Zhang

, Chen

, Wang

, Zhao

, Wang

, Dong

, Wang

, ZhangZ; Beijing Aging Brain Rejuvenation Initiative workgroup (2021) Early prevention of cognitive impairment in the community population: The Beijing Aging Brain Rejuvenation Initiative. Alzheimers Dement 17, 1610–1618.

38.

Katzman

, Zhang

, Ouang

Ya Q

, Wang

, Liu

, Yu

, Wong

, Salmon

, Grant

(1988) A Chinese version of the Mini-Mental State Examination; impact of illiteracy in a Shanghai dementia survey. J Clin Epidemiol 41, 971–978.

39.

Mok

, Lam

, Chiu

(2004) Category verbal fluency test performance in chinese elderly with Alzheimer’s disease. Dement Geriatr Cogn Disord 18, 120–124.

40.

Bartha-Doering

, Kollndorfer

, Schwartz

, Fischmeister

FPS

, Alexopoulos

, Langs

, Prayer

, Kasprian

, Seidl

(2021) The role of the corpus callosum in language network connectivity in children. Dev Sci 24, e13031.

41.

Yan

, Wang

, Zuo

, Zang

(2016) DPABI: Data processing & analysis for (resting-state) brain imaging. Neuroinformatics 14, 339–351.

42.

Wang

, Wang

, Xia

, Liao

, Evans

, He

(2015) GRETNA: A graph theoretical network analysis toolbox for imaging connectomics. Front Hum Neurosci 9, 386.

43.

Newman

(2006) Modularity and community structure in networks. Proc Natl Acad Sci U S A 103, 8577–8582.

44.

Liang

, Zou

, He

, Yang

(2016) Topologically reorganized connectivity architecture of default-mode, executive-control, and salience networks across working memory task loads. Cereb Cortex 26, 1501–1511.

45.

Sager

, Wuu

, Leurgans

, Rees

, Gearing

, Mufson

, Levey

, Lah

(2007) Neuronal LR11/sorLA expression is reduced in mild cognitive impairment. Ann Neurol 62, 640–647.

46.

Liu

, Sun

J-Y

, Xu

, Yang

X-Y

, Sun

(2017) SORL1 variants showdifferent association with early-onset and late-onset Alzheimer’sdisease risk. J Alzheimers Dis 58, 1121–1128.

47.

Deary

, Penke

, Johnson

(2010) The neuroscience of human intelligence differences. Nat Rev Neurosci 11, 201–211.

48.

Hoffman

, Cox

, Dykiert

, Munoz

Maniega S

, Valdes

Hernandez MC

, Bastin

, Wardlaw

, Deary

(2017) Brain grey and white matter predictors of verbal ability traits in older age: The Lothian Birth Cohort 1936. Neuroimage 156, 394–402.

49.

Eliot

, Ahmed

, Khan

, Patel

(2021) Dump the “dimorphism”: Comprehensive synthesis of human brain studies reveals few male-female differences beyond size. Neurosci Biobehav Rev 125, 667–697.

50.

Williams

, Peyre

, Toro

, Ramus

(2021) Neuroanatomical norms in the UK Biobank: The impact of allometric scaling, sex, and age. Hum Brain Mapp 42, 4623–4642.

51.

Chen

, Sachdev

, Wen

, Anstey

(2007) Sex differences in regional gray matter in healthy individuals aged 44-48 years: A voxel-based morphometric study. Neuroimage 36, 691–699.

52.

Shafto

, Tyler

(2014) Language in the aging brain: The network dynamics of cognitive decline and preservation. Science 346, 583–587.

53.

Bartha-Doering

, Novak

, Kollndorfer

, Kasprian

, Schuler

, Berl

, Fischmeister

FPS

, Gaillard

, Alexopoulos

, Prayer

, Seidl

(2018) When two are better than one: Bilateral mesial temporal lobe contributions associated with better vocabulary skills in children and adolescents. Brain Lang 184, 1–10.

54.

Forseth

, Kadipasaoglu

, Conner

, Hickok

, Knight

, Tandon

(2018) A lexical semantic hub for heteromodal naming in middle fusiform gyrus. Brain 141, 2112–2126.

55.

Pulvermuller

, Huss

, Kherif

, Moscoso del Prado

Martin F

, Hauk

, Shtyrov

(2006) Motor cortex maps articulatory features of speech sounds. Proc Natl Acad Sci U S A 103, 7865–7870.

56.

Fujii

, Maesawa

, Motomura

, Futamura

, Hayashi

, Koba

, Wakabayashi

(2015) Intraoperative subcortical mapping of a language-associated deep frontal tract connecting the superior frontal gyrus to Broca’s area in the dominant hemisphere of patients with glioma. J Neurosurg 122, 1390–1396.

57.

Vigneau

, Beaucousin

, Herve

, Jobard

, Petit

, Crivello

, Mellet

, Zago

, Mazoyer

, Tzourio-Mazoyer

(2011) What is right-hemisphere contribution to phonological, lexico-semantic, and sentence processing? Insights from a meta-analysis. Neuroimage 54, 577–593.

58.

Cabeza

(2002) Hemispheric asymmetry reduction in older adults: The HAROLD model. Psychol Aging 17, 85–100.

59.

Szaflarski

, Holland

, Schmithorst

, Byars

(2006) fMRI study of language lateralization in children and adults. Hum Brain Mapp 27, 202–212.

60.

Nenert

, Allendorfer

, Martin

, Banks

, Vannest

, Holland

, Szaflarski

(2017) Age-related language lateralization assessed by fMRI: The effects of sex and handedness. Brain Res 1674, 20–35.

61.

Robinson

, Shallice

, Bozzali

, Cipolotti

(2012) The differing roles of the frontal cortex in fluency tests. Brain 135, 2202–2214.

62.

Baldo

, Schwartz

, Wilkins

, Dronkers

(2006) Role of frontal versus temporal cortex in verbal fluency as revealed by voxel-based lesion symptom mapping. J Int Neuropsychol Soc 12, 896–900.

63.

Meinzer

, Seeds

, Flaisch

, Harnish

, Cohen

, McGregor

, Conway

, Benjamin

, Crosson

(2012) Impact of changed positive and negative task-related brain activity on word-retrieval in aging. Neurobiol Aging 33, 656–669.

64.

Halsband

(2006) Bilingual and multilingual language processing. J Physiol Paris 99, 355–369.