Human language's complexity distinguishes us from other primates, characterized by significant left-hemisphere dominance. This lateralization likely stems from structural and functional brain asymmetries developing prenatally and in infancy. While adult-like language network lateralization emerges by age 4, approximately 85% of adults exhibit left-hemisphere dominance, with the remaining individuals showing no clear dominance or, rarely, rightward dominance. The left-hemisphere language network encompasses various regions, including hubs in the inferior frontal gyrus and superior temporal sulcus. Right hemisphere homologues also participate, particularly in comprehension. Language-related cognitive performance is highly heritable, with genetic factors influencing language-related neurodevelopmental disorders like dyslexia. Neonatal asymmetries suggest an early developmental basis influenced by genetic programs. Genome-wide association studies (GWAS) have identified loci associated with language/reading performance, dyslexia, and handedness, many expressed in the embryonic/fetal brain, indicating early developmental impact. However, these studies only explain a small portion of heritable variance. This study uses a novel approach: analyzing resting-state functional connectivity within the language network in a large sample to perform a well-powered GWAS. Resting-state fMRI, being task-free, avoids task design biases affecting lateralization estimates and offers a holistic view of the language network. Previous studies on language network connectivity focused on a limited number of regions without considering hemispheric differences in connectivity, prompting this study's focus on left-right hemisphere homotopies.

Previous research has established the heritability of language abilities and the role of genetics in language disorders like dyslexia. Studies using GWAS have identified specific genomic loci associated with language-related traits, but these have only accounted for a small percentage of the heritable variance. Previous work by Mekki et al. (2022) identified 20 loci associated with functional language network connectivity using resting-state fMRI but did not fully consider hemispheric differences. This study addresses these gaps by employing a more comprehensive analysis of both connectivity and hemispheric differences across a larger sample.

This study utilized resting-state fMRI and genetic data from 29,681 participants of white British ancestry from the UK Biobank aged 45-82. The language network was defined using the AICHA atlas, encompassing 18 left-hemisphere regions and their right-hemisphere homotopes, resulting in 630 intra- and interhemispheric connectivity measures and 153 hemispheric differences. After quality control, multivariate genome-wide association studies (mvGWAS) were performed using MOSTest, accounting for potential confounders (age, sex, etc.). Heritability analysis using GCTA identified 629 heritable edges and 103 heritable hemispheric differences. For common genetic variants (MAF ≥1%), mvGWAS identified 14 independent genomic loci associated with language network edges, three of which were also associated with hemispheric differences. Gene mapping using FUMA and MAGMA revealed associated genes preferentially expressed prenatally in the brain. The direction and magnitude of effects for lead SNPs were examined. Polygenic scores for language abilities, dyslexia, and left-handedness, calculated using PRS-CS, were analyzed using canonical correlation analysis (CCA) to assess their association with language network connectivity and asymmetry. Exome-wide association studies (using REGENIE and SKAT-O) investigated the contribution of rare, protein-altering variants (MAF <1%) to functional connectivity, using both broad and strict filters for variant impact. Burden analysis further examined the aggregate effect of rare variants on connectivity measures.

The mvGWAS identified 14 independent genomic loci significantly associated with language network edges (p ≤ 5 × 10⁻⁸), with three loci also significantly associated with hemispheric differences. The most significant locus was on chromosome 3, near the *EPHA3* gene, associated with generally reduced connectivity in minor allele carriers. Other significant loci were near *ZIC4* (increased connectivity) and *PLCCE1:PLCE1 AS1* (stronger increase in intrahemispheric connectivity). Analysis of Brainspan gene expression data showed preferential expression of associated genes during prenatal brain development. Polygenic scores for higher language abilities showed a significant association with stronger left-hemisphere connectivity and a leftward shift in hemispheric differences. Conversely, polygenic scores for dyslexia and left-handedness were associated with rightward shifts in connectivity asymmetry, particularly impacting interhemispheric connectivity for dyslexia and right intrahemispheric connectivity for left-handedness. Exome-wide analysis identified five genes (*NIBAN1, MANEAL, SLC25A48, DUSP29, TRIP11*) associated with language network edges and two genes (*WDCP, DDX25*) associated with hemispheric differences based on rare, protein-altering variants. Burden analysis revealed that increased numbers of rare variants in *MANEAL* were associated with decreased overall connectivity, while increased *DDX25* variants were linked to a rightward shift in intrahemispheric connectivity asymmetry.

This study's findings provide new insights into the genetic architecture of the language network. The strong association with *EPHA3*, involved in neurogenesis and axon guidance, highlights its role in establishing left-right brain asymmetries supporting language specialization. The association with *TBC1D5*, involved in autophagy and receptor metabolism, further implicates this gene in the development of the brain's language network. The observed leftward shift in asymmetry for individuals with higher polygenic scores for language abilities aligns with the notion that this asymmetry reflects optimal language processing. The rightward shift in individuals with higher polygenic risk for dyslexia and left-handedness is consistent with previous research suggesting reduced left-hemisphere dominance in these conditions. The identification of novel genes through exome-wide analysis offers promising avenues for future research investigating the biological mechanisms underlying language network development and function. The study's use of resting-state fMRI avoids task-related biases in the assessment of connectivity.

This large-scale imaging genetics study identified 14 genomic loci associated with language network connectivity and its hemispheric differences using common genetic variants. Polygenic scores demonstrated a link between reduced leftward asymmetry and lower language abilities, dyslexia, and left-handedness. Exome-wide analysis implicated seven additional genes based on rare variants. These findings advance our understanding of the genetic contributions to language network organization and related behavioral traits, opening avenues for future research focusing on the identified genes and their functional roles in brain development.

The study's cross-sectional design limits causal inferences. The UK Biobank sample, while large, overrepresents healthy participants and may not fully represent the general population. The use of resting-state fMRI does not directly measure language lateralization, and the parcel-based analysis may not capture all functionally relevant aspects of brain network architecture. The exome-wide findings, based on a marginal significance threshold, require replication in independent datasets.