Cognitive, neuroimaging, and genetic insights on the interthalamic adhesion from a large cohort study of 591 subjects

Abstract

Both thalami can be connected by the Interthalamic Adhesion (IA), a white matter tract that crosses the 3rd ventricle. Its presence varies among individuals and remains poorly understood. This study examines the IA’s prevalence, anatomical variations, genetic determinants, and cognitive associations. Data from 591 healthy subjects (25–35 years) from the Human Connectome Project were analyzed, and grouped into monozygotic (MZ) or dizygotic (DZ) twins, non-twin siblings, and unrelated individuals. MRI was used to characterize the IA, while neuropsychological assessments and Freesurfer parcellations were used to assess cognition and anatomical differences between subjects with or without an IA. The IA was absent in 12.7% of subjects, more commonly in males (20.0%) than females (6.3%). No significant differences in age, education, or cognition were found between individuals with or without an IA. IA absence was associated with increased cerebrospinal fluid volumes, enlarged third ventricles, and thinning in several cortical areas. Genetic analysis on the IA presence or absence revealed a heritability estimate of 34% with a higher concordance among MZ twins (96%) than in other groups. The remaining 4% discrepancy was observed in male pairs only. This study underscores the genetic basis of IA, highlighting sexual dimorphism and neuroanatomical differences associated with its absence, despite unaffected cognition in healthy individuals.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-20469-w.

Introduction

The thalamus is a bilateral nuclear complex located in the diencephalon¹. It plays a pivotal role in processing and transmitting sensory information to the cerebral cortex. It also contributes to higher-order functions such as attention, consciousness, and memory through cortico-cortical connections². An interthalamic adhesion (IA), located on the median border across the third ventricle, can connect both thalami^3–6.

IA studies are scarce, and much remains to be investigated to better understand its role. This structure is absent in approximately 20% of healthy subjects, more often in males (17%) than females (9%)^4,7–15. This variability is unusual for brain structures. Commonly presented as a single adhesion in the anterosuperior quadrant^3–6, the IA can also vary in shape. Broader forms occur in 18% of cases, more frequently in females, while duplications are reported in 2–10%, and rare forms such as bilobar, multiple, tubular, or rudimentary, in less than 3% of subjects^{1,2,4,14,16–19}. In 2–3% of cases, the “kissing thalami” phenomenon can obscure the IA on MRI^11,12,15,17.

The IA is a white matter tract with uncertain connectivity. Diffusion imaging suggests it associates with the anterior thalamic radiations, projecting to the orbitofrontal and medial frontal cortex^14,20,21. In its absence, fibers may reroute via the posterior commissure, suggesting compensatory pathways²¹. The IA also connects to the amygdala, hippocampus, entorhinal cortex, insula, pericalcarine cortex, cuneus, nucleus accumbens, caudate nucleus, lateral habenula, and posterior commissure^14,22, implying an eventual role in interhemispheric communication.

Some researchers consider the IA vestigial¹⁶, due to its inconsistent presence. However, studies suggest cognitive relevance, especially in attention^11,20, executive functions^11,23, and verbal memory²³. While no cognitive differences are seen in healthy individuals based on IA presence, thalamic stroke patients with an IA experience fewer deficits in verbal memory and executive functions, hinting at a protective or compensatory role¹⁵. The role of the IA in cognition necessitates further research.

The cause of IA absence remains debated, with potential age correlations. A 3D transvaginal neurosonography study on fetuses demonstrated an IA presence of 100%, and in an MRI study on teenagers, this prevalence was maintained^24,25. Accordingly, a progressive thinning and elongation of IA in an oldest cohort of patients has been noted in an MRI study²⁶, and broad IAs are more frequent in those under 40^19,23. Still, a meta-analysis found no age-related differences in IA prevalence or length¹⁸.

Different studies have suggested a relationship between IA absence, or shorter length, and ventricle enlargement^3,27–30. Subjects without an IA have larger third ventricles, in both healthy individuals and those with schizophrenia or bipolar disorder. A negative correlation also exists between third ventricle size and IA length³⁰. Given that IA and the ventricular system develop together^28,31,32, IA anomalies may indicate midline developmental abnormalities. These include conditions such as Chiari II malformation, hydrocephalus, Cornelia de Lange syndrome, and diencephalic-mesencephalic dysplasia^17,33–37. Among pediatric patients with midline abnormalities, IA absence is four times more likely compared to healthy subjects³⁸. Some studies suggest that elevated third ventricle pressure could compress or rupture the IA, as illustrated by a case where the IA disappeared after hydrocephalus developed³⁹. It remains unknown whether neurological diseases associated with global atrophy, such as Alzheimer’s disease and its impact on the thalamus, are associated with IA shrinkage or reduced prevalence. Similarly, the effect of multiple sclerosis on the IA remains to be investigated.

Many studies have reported increased rates of IA absence or shorter length in neurodevelopmental and neuropsychiatric disorders such as schizophrenia, borderline personality disorder, major depression, and bipolar disorder^{8,28,30,40,41}. Altered connections to the amygdala, frontal, and anterior cingulate cortex have been linked to these disorders and are all structures demonstrated to be connected to the IA^42–44. In summary, IA absence might be related to early neurodevelopmental abnormalities, particularly of midline structures, leading to abnormal neural networks, including the thalamus and related regions. Those abnormal neural networks could lead to functional and structural aberrations, such as deficits in the dopaminergic system. Indeed, a recent meta-analysis found that IA absence is twice as likely in schizophrenia spectrum disorders¹⁸, consistent with known thalamic and dopamine-related abnormalities^40,46–48. IA absence may thus signal risk for neuropsychiatric disorders^8,18.

Genetic factors potentially influence IA presence, but this remains unproven. However, many IA-related psychiatric conditions are highly heritable^49–51. A study on monozygotic and dizygotic twins estimated the heritability of bipolar disorder at 85%⁵². It is plausible that IA presence is under partial genetic control, linked to neurodevelopmental and psychiatric risks.

In summary, many intriguing questions about the IA remain unresolved. This study seeks to advance understanding by studying a large cohort of 591 healthy individuals, including monozygotic and dizygotic twins. Our primary hypothesis is that genetics play a significant role in influencing IA characteristics, as it is consistently linked to genetically related disorders. Additionally, we hypothesize that the absence of the IA should not lead to cognitive differences, given its absence in some individuals. To explore these hypotheses, MRI from healthy young adults participating in the Human Connectome Project (HCP) were analyzed to: (1) assess the presence or absence of the IA, (2) classify its variants, and (3) examine associations with anatomical and neuropsychological differences. Subjects were grouped as monozygotic (MZ) twins, dizygotic (DZ) twins, non-twin siblings, and unrelated individuals.

Materials and methods

Participants

Data were provided by the Human Connectome Project, WU-Minn Consortium (Principal Investigators: David Van Essen and Kamil Ugurbil; 1U54MH091657) funded by the 16 NIH Institutes and Centers that support the NIH Blueprint for Neuroscience Research; and by the McDonnell Center for Systems Neuroscience at Washington University⁵³. Participants were healthy individuals from families in Missouri, United States. Inclusion criteria required the absence of severe neurodevelopmental, neuropsychiatric, or neurological disorders, as well as diseases such as diabetes or hypertension. Twins born before 34 weeks of gestation and non-twins born before 37 weeks were excluded. Each subject underwent neuroimaging and neuropsychological examinations. More specifically, the HCP Young Adult sample, comprising 591 subjects selected from 1207 based on the availability of magnetic resonance imaging (MRI) data, including T1 and T2 sequences, was used. The sample included 306 females (51.8%) and 285 males (48.2%), with an average age of 28.5 years (SD = 3.6, range 22–35 years) and 15 years of education (SD = 1.8).

To investigate the genetic influences on the prevalence of the IA, subjects were divided into four groups: monozygotic twins (MZ), dizygotic twins (DZ), non-twin siblings, and unrelated individuals. The monozygotic and dizygotic relationships were genetically verified. For unrelated subjects, random pairings were created to form equivalent groups.

Neuropsychological assessment

Relevant psychological and neuropsychological data were extracted from the HCP database. This included measures of depression, anxiety, and attention/hyperactivity from the Statistical Manual (DSM)-Oriented Scales⁵⁴. Cognitive abilities were assessed using the NIH Toolbox Cognition Battery⁵⁵, which included tests for working memory (List Sorting Working Memory Test), language (Oral Reading Recognition Test), and episodic memory (Picture Sequence Memory Test for non-verbal episodic memory, and the Penn Word Memory Test for verbal episodic memory). Data for depression, anxiety, and attention tests were missing for six individuals.

MRI acquisition

MRI images were acquired using a Siemens 3T “Connectome Skyra” scanner. For T1 sequences, images were obtained with a 3T scanner using the following parameters: 3D MPRAGE sequences, voxel size = 0.7 × 0.7 × 0.7 mm, TE = 2.14 ms, TR = 2400 ms, TI = 1000 ms, flip angle = 8°, FOV = 224 × 224 mm, slice thickness = 0.7 mm isotropic. The parameters for the 3D T2-SPACE sequences were: voxel size = 0.7 × 0.7 × 0.7 mm, TE = 565 ms, TR = 3200 ms, FOV = 224 × 224 mm, slice thickness = 0.7 mm isotropic. Images were coregistered and reoriented to align the anterior and posterior commissure to standardize orientation and facilitate comparison.

IA and IA variants identification protocol

Two independent raters characterized the presence, absence, and type of IA following a previously established protocol¹⁵. An extended video protocol to characterize the IA is publicly available [DOI: 10.13140/RG.2.2.35022.47689], and a training dataset can be sent on request. MRI images were reviewed using MRIcron software, starting with axial slices and then sagittal and coronal slices to refine the characterization of the IA. Confirmation was done using T2-weighted images. An IA was identified if a structure connecting both thalami was observed on at least one slice between the anterior and posterior commissure. Variants were classified as standard, broad, bilobar, or double based on specific criteria as the size and shape¹⁵. More specifically, a standard form is characterized by a thin adhesion comprising less than one-third of the thalamus on the current axial and coronal slices, with a small, rounded shape on sagittal slices. A broad variant refers to a larger form of the IA, identified when comprising over one-third of the thalamus on the current axial and coronal slices. Using the size of the thalami as a reference allows for normalization of IA size across subjects. The bilobar form is distinguished on sagittal slices by a bilobed shape and on coronal slices two distinct adhesions can be visualized and must not be mistaken for a double variant. Finally, the double variant is identified by two independent IA, independently of the size, which needs to be confirmed on all slices. To improve the previous protocol and help better differentiate the standard form from a broad one, especially in the case of close thalami, it is possible to segment the IA on sagittal slices and identify the amount of shared matter between the two thalami.

Two raters (A.G. and M.S.R) conducted the identification process blindly to avoid biased sibling correlations. Cases with the “kissing thalami” phenomenon were excluded from statistical analyses. Consequently, the sibling or unrelated paired subject was also excluded from the genetic analyses, as our focus was on pairs. In cases of disagreement between the examiners, discussions were held to reach a consensus. A third expert opinion (J.P.V) was sought to resolve the differences if necessary. Inter-rater concordance was evaluated using Cohen’s Kappa coefficient for every 100 subjects.

Statistical analyses

Demographic data

The impact of handedness, biological sex, age, and education level on IA prevalence was examined using χ² tests for qualitative data and Student’s t-tests or Mann-Whitney tests for quantitative data. ANOVA or Kruskal-Wallis tests were used for variables with multiple categories.

Prevalence of the IA, brain volumes and cortical thickness

FreeSurfer parcellations were used to extract volumes and cortical thicknesses measurements for each subject, maintaining the original nomenclature for consistency. To compare anatomical differences between subjects with and without an IA, general linear models were employed including age, gender, and intracranial volume (ICV) as covariates. Both uncorrected and FDR-corrected p-values are reported, alongside with the partial eta squared (η²) as a measure of effect size.

For visual representation, cortical regions with a significant FDR-corrected p-values are displayed for each hemisphere in a histogram. Additionally, cortical regions by hemisphere with p-values below 0.005, indicating highly significant results, are visualized on a 3D brain model using FreeSurfer parecellations and MRIcroGL.

Prevalence of the IA and neuropsychological consequences

An independent t-test was conducted to study the variation in neuropsychological test scores based on IA presence or absence.

Prevalence of the IA and genetic causes

Logistic regressions were conducted to evaluate the association between the IA and genetics by comparing the four groups (MZ, DZ, non-twin, unrelated) and considering whether each paired subject possesses an IA or not. χ² tests with Bonferroni correction were used for pairwise comparisons. The same tests were performed for genetic analysis of IA variants, but only data related to the presence of IA in both paired subjects were retained.

Twin studies provide a valuable approach for estimating heritability, which corresponds to how much of the variation in a phenotypic trait is due to variation in genetic factors⁵⁶. Indeed, MZ twin pairs share 100% of their genetic material, while DZ twin pairs share, on average, 50%. Both types of twins, however, share a similar environment. Any phenotypic differences observed in MZ twins are generally attributed to unique environmental influences. By assuming that these unique environmental factors contribute equally to the variance in both MZ and DZ pairs, it is possible to estimate the genetic contribution to the observed trait. This is done using Falconer’s Formula for heritability: 2*(rMZ - rDZ). rMZ and rDZ represent the concordance rates for the trait (e.g., having the same IA or IA variant) in MZ and DZ twins, respectively. The difference between the concordance rates of MZ and DZ twins reflects the proportion of the phenotypic variance attributable to genetic factors. The result is then multiplied by 2 to account for the full genetic effect, given that MZ twins share 100% of their genes, while DZ twins share only 50%. However, it is important to note that twin studies do not account for the potential influence of epigenetics, which may lead to an overestimation of heritability by incorrectly attributing environmental contributions to genetic factors.

Finally, to anatomically and neuropsychologically compare siblings only differing by the presence or absence of the IA, the analyses were restricted to paired subjects when their adhesion status differed. In this aim, a paired Bayesian t-test was conducted to reduce bias due to low sample sizes, applied to volumes, cortical thickness, and psychological or cognitive test scores by group.

Results

Prevalence of the IA: demographics factors

The IA was characterized by two independent examiners with a mean concordance of 0.85 for the presence or absence and 0.79 for identifying anatomical variants using Cohen’s Kappa score. Out of the 591 subjects, 55 cases had kissing thalami, resulting in a final sample size of 536 subjects. Among these, 87.3% had an IA, while 12.7% did not. The absence of the IA was significantly higher in males (20%) compared to females (6.3%) (p < 0.001, χ² test = 22.6, df = 1). No significant differences between subjects with and without an IA were observed for age, education, or manual laterality (Mann-Whitney U; p-value > 0.05).

Prevalence of IA variants

Recent developments in the study of the IA have included the analysis of anatomical variants¹⁹. Following a previously established protocol¹⁵, we identified that among all subjects, 61% had a standard single IA, 25% had a broad variant, and less than 1% exhibited either a bilobar or double variant (Fig. 1).

Fig. 1.

Typical single or absent IA and its most frequent anatomical variants (broad, double, bilobar) illustrated on the same slice of T1- or T2-weighted MRI. Blue arrows indicate the IA location. Note the lower contrast on T1w compared to T2w MRI, especially on sagittal slices, preventing the accurate location of the first double variant.

Prevalence of the IA: association with brain structure

Based on the literature about IA structural connectivity, we first address the hypothesis that anatomical differences exist between individuals with and without an IA. Using FreeSurfer parcellations, we sought to identify volume and cortical thickness variations associated with the IA presence. All comparisons (mean ± SD by groups, p-values, and eta squared (η²)) are reported in Supplementary Table 1.

Subjects without an IA showed several differences compared to those with an IA. GLMs corrected for age, gender, and ICV, showed significantly increased cerebrospinal fluid volumes and increased third ventricle volume among subjects without an IA (Fig. 2A). This was associated with smaller corpus callosum (Fig. 2A). No differences could be demonstrated concerning bi-thalamic volume (Supp. Table 1). In addition, some brain regions showed cortical thinning, most notably the bilateral inferior and superior parietal, supramarginal, rostral middle frontal, precentral and paracentral, superior frontal, precuneus, and the banks of the superior temporal sulcus (Fig. 2B-C). While most differences were bilateral, some were unilateral. Figure 3 highlights the 9 brain regions with the most significant thinning in the absence of IA (p < 0.05) with a schematic representation of these regions.

Fig. 2.

Partial Eta Squared resulting from comparisons of brain volumes and cortical thicknesses depending on the presence (N = 468) or absence (N = 68) of an IA, using GLMs with age, gender, and ICV as covariates. The comparisons are presented for (A) midline or general regions, (B) left, or (C) right brain regions. Only thicknesses with significant increase in case of IA presence are displayed. Yellow bars represent regions with bilateral differences. The names of the brain areas are from the nomenclature of FreeSurfer. CC: corpus callosum. FDR-corrected p-values: p < 0.05 *, p < 0.01 **, p < 0.001 ***.

Table 1.

Demographic data by genetic groups.

	Monozygotic	Dizygotic	Not Twin Siblings	Unrelated
N	175	98	146	117
Female	57.0%	54.0%	47.3%	54.7%
Mean age ± SD	29.3 ± 3.3	28.6 ± 3.4	27.1 ± 3.7	29.1 ± 3.7
Mean Years of Education ± SD	15.1 ± 1.8	15.3 ± 1.6	14.8 ± 1.9	14.8 ± 1.8
IA presence (%)	89.7	82.7	88.4	86.3

Fig. 3.

Illustration of the 9 cortical regions, as extracted from Freesurfer, with the most significant thinning in subjects without an IA compared to those with an IA (p < 0.005, GLM analysis, Fig. 2) using MRIcroGL (V1.2.20220720; https://www.nitrc.org/projects/mricrogl). These regions demonstrate evidence of decreased cortical thickness. The figure provides a superior view (left), a left view (top), and a right view (bottom), with a cube indicating spatial orientation. The color legend identifies each region by name.

Prevalence of the IA and association with affective and cognitive abilities

Given that the IA is absent in a large portion of the population and has been associated with brain differences, we explored whether the presence or absence of IA affects affective or cognitive abilities. Independent t-tests revealed no significant differences between these groups in terms of depression, anxiety, attention, verbal and non-verbal episodic memory, language, or working memory (t-test, p-values > 0.1).

Genetic influence on the IA

To investigate the potential genetic influence on IA development, its prevalence was compared across different genetic groups: monozygotic twins (MZ), dizygotic twins (DZ), non-twin siblings, and unrelated individuals.

IA’s prevalence between groups

After excluding “kissing” thalami cases, the sample included 175 MZ twins, 98 DZ twins, 146 non-twin siblings, and 117 unrelated individuals. The exclusions of subjects paired with those having a kissing thalamus were not applied to retain a larger effective sample size for subsequent sociodemographic analyses, which did not rely on paired data. Biological sex and education levels did not differ significantly between the genetic groups (χ² test = 3.10, p-value = 0.38, df = 3; ANOVA, p-value > 0.05). However, non-twin siblings were slightly younger than MZ twins and unrelated subjects (27.1 vs. 29.3 years old; Dunn’s test with Bonferroni correction, p-value < 0.001). No significant variation in IA prevalence was observed between the groups (χ² test = 3.10, p-value = 0.38, df = 3) (Table 1). Notably, the non-twin siblings group included 41 mixed-gender pairs, the unrelated group included 23 mixed-gender pairs, while no mixed-gender pairs were present in the MZ and DZ twin groups.

Genetic effect on IA presence among groups

For the following analyses involving pairs of subjects, exclusions were made for related “kissing” cases (subjects paired with those having a kissing thalamus) and one unmatched unrelated subject, resulting in 164 monozygotic twins, 94 dizygotic twins, 136 non-twin siblings, and 116 unrelated subjects.

A logistic regression model was applied to determine whether each pair of subjects shared the same IA characterization by groups. The results revealed a significant difference between the genetic groups in predicting the presence or absence of IA between paired subjects (χ² test = 32, p-value < 0.001, df = 506). The model’s accuracy in predicting the dependent variable was 86%. The odds ratios were 7 for the MZ group (IC95: 2.7–18.5), 3.7 for the DZ group (IC95: 2.3–6.0), 0.35 for the Non-Twin group (IC95: 0.8–3.1), and 0.33 for the Unrelated group (IC95: 0.44–1.6). This indicates a strong effect in the MZ group on the probability of having the same IA presence/absence, while no such effect was observed in the other groups. MZ twins had a 96% concordance rate for IA presence/absence, significantly higher than DZ twins (79%), Non-Twin siblings (85%), and Unrelated individuals (75%) (χ² test = 28, p-value < 0.001, df = 3) (Fig. 4).

Fig. 4.

Probability to have the same IA between pairs and by genetic groups (MZ: monozygotic, DZ: dizygotic) ± 95% confidence interval. χ² p-values Bonferroni corrected: *** < 0.001, ** < 0.005.

Heritability of the IA and sexual dimorphism

Based on the computed concordance rates of 0.96 for MZ twins and 0.79 for DZ twins, the heritability estimate for the presence or absence of the IA was found to be 0.34, indicating that 34% of the phenotypic variance is attributable to genetic factors, while the remaining variance is likely due to other influences.

Interestingly, 100% of female MZ twins had the same IA characterization, while three pairs of male twins exhibited different IA characteristics (χ² test = 8.8, p-value = 0.004, df = 1). Among these male pairs, those without an IA showed an average increase in third ventricle volume of 122 mm³.

Genetic influence on the IA variants

Prevalence of ia’s anatomical variants between groups

We then investigated whether genetic factors influenced specific IA anatomical variants. Due to their low frequency, statistical analysis was not conducted for the “Bilobar” and “Double” variants. No effects of age, years of education, or manual laterality were found between groups (Kruskal-Wallis test; p-value > 0.5). However, biological sex significantly influenced IA type, with a higher prevalence of broad IA in females (χ² test, p-value < 0.001, effect size = 17.4, df = 3). No significant differences in the proportion of IA variants were observed between genetic groups (χ² test = 6.6, p-value > 0.5, df = 9) (Table 2).

Table 2.

Demographic data by IA anatomical variants.

i	One	Broad	Double	Bilobar	Absent
All subjects	328 (61%)	132 (25%)	4 (< 1%)	4 (< 1%)	68 (13%)
Mean age ± SD	28.5 ± 3.7	28.6 ± 3.5	25.8 ± 3.9	30.3 ± 4.2	28.6 ± 3.3
Mean years of education  ± SD	15 ± 1.7	14.9 ± 1.8	14.5 ± 1.3	16.3 ± 0.5	15 ± 1.7
Female (%)	51.2	72.0	50.0	75.0	26.5
MZ (%)	61.1	27.4	< 1	< 1	10.3
DZ (%)	56.1	24.5	1	1	17.4
Not Twin (%)	65.1	21.9	1.4	0	11.6
Unrelated (%)	60.7	23.9	0	1.7	13.7

Genetic effect on the adhesion type among groups

Genetics’ effect on the adhesion type was evaluated using paired data between siblings and unrelated subjects having an IA, through logistic regression. Results showed a significant difference between the genetic groups concerning the ability to predict having the same IA variants between paired subjects (χ² test = 21, p-value < 0.001, df = 449). The accuracy of the model indicating the ability to predict having the same IA variant was 68%. The odds Ratio for the MZ group was 3.3 (IC95: 1.8–6.0), 1.3 in DZ (IC95: 0.8–2.0), 1.4 for the Not Twin Group (IC95: 0.8–2.5), and 1.2 for the Unrelated group (IC95: 0.6–2.1). It indicates a slight effect of the group MZ on the probability of having the same IA, while no such effect was observed in other groups. Rates of pairs of subjects having the same IA in each group were 81% for MZ, 56% for DZ, 64% for Not Twin, and 60% for Unrelated, which highlights higher randomness among all groups except MZ (χ² test = 20.5, p-value < 0.001, df = 3) (Fig. 5). The MZ group had a significantly better prediction of having the same IA compared to the three other groups without differences between biological sex (χ² test = 0.002, p-value > 0.05, df = 1).

Fig. 5.

Probability of having the same IA between each sibling or randomly paired unrelated subjects by genetic groups (MZ: monozygotic, DZ: dizygotic) ± 95% confidence interval. χ² p-values corrected for Bonferroni corrections: *** < 0.001, ** < 0.005.

Heritability of IA anatomical variants

Based on the computed concordance rates of 0.81 for monozygotic (MZ) twins and 0.56 for dizygotic (DZ) twins, the heritability estimate for the trait was found to be 0.50, indicating that 50% of the phenotypic variance is attributable to genetic factors, while the remaining variance is likely due to other influences.

Discussion

Leveraging a large cohort of 591 subjects, we found that the IA was more prevalent in females (94%) than males (80%). These differences were associated with a larger third ventricle, smaller corpus callosum, and cortical thinning in several regions in subjects without an IA. Despite these anatomical variations, no significant neuropsychological difference was evidenced. This study also reveals, for the first time, genetic influences on both the presence and anatomical variations of IA, with notable sexual dimorphism. Specifically, females exhibited a higher prevalence of IA, more broad anatomical variants, and there was 100% concordance in IA presence among female MZ twins, highlighting a significant biological sex impact on the IA.

Prevalence of IA

Consistent with previous literature, the IA was present in about 87% of subjects, with a higher absence rate in males (20%) compared to females (6%)^{2,4,7,8,10–15}. No significant differences were found in age, education, or manual laterality. While our dataset of 591 subjects is the largest in IA literature involving both neuroimaging and neuropsychological assessments in healthy individuals (e.g¹¹.: 402²⁰;: 233), the age range of our participants is limited, spanning from 22 to 35 years. Despite this limitation, our findings align with Borghei et al.¹¹, who found no age effects. In contrast, Damle et al.²⁰, who included a broader age range of 8 to 68 years, could observe age-related influences on IA characteristics and cognitive performance. This highlights the need for future studies with a broader age range to understand these relationships better.

Anatomical differences

Structural analyses revealed significant differences between individuals with and without an IA. The absence of IA was associated with increased cerebrospinal fluid volumes and enlarged third ventricles in individuals without an IA. This finding aligns with earlier studies linking ventricle enlargement to the absence of IA, even more among patients with schizophrenia^29,30. For example, among the three pairs of male monozygotic twins with differing IA characteristics reported in our study, the sibling without an IA systematically had a larger third ventricle volume.

Additionally, individuals without IA exhibited reduced corpus callosum volumes alongside cortical thinning in several areas. These included bilateral supramarginal, superior parietal, inferior parietal, rostral middle frontal, superior frontal, post central and left cuneus. This suggests that those regions are cortical areas of IA projections via both thalami, aligning with previous tractography studies^{11,12,14,20,21}. However, other structures were also found to be structurally connected with the IA in the literature, such as the amygdala, hippocampus, accumbens, and caudate nucleus, which we did not identify in this study^{11,12,14,20,21}. In addition to the literature and among the most significant differences in the absence of an IA, we also identified thinning in the bilateral banks of the superior temporal sulcus. The left pars opercularis and right pars triangularis were also thinner in the absence of IA. Connecting those findings with atlas-guided tract reconstruction and suggestions that thalamic radiations are the areas through which IA fibers reach cortical areas^11,12,20,21, the pars opercularis, triangularis, and orbitalis are indeed known projection areas of the anterior thalamic radiations, while posterior thalamic radiations project to the parietal and occipital lobes. The superior thalamic radiations connect the thalamus to the precentral and paracentral gyri, and the inferior thalamic radiations to the insula, temporal, and frontal lobes⁵⁷.

Neuropsychological and affective impact of the presence or absence of the IA

Despite these anatomical differences, this study did not find significant neuropsychological or affective difference associated with IA presence/absence in this group of healthy individuals. The different studies that have tried to find neuropsychological differences between groups of healthy subjects with and without an IA have also usually failed to report any difference¹⁵, except marginally^11,20. It is possible that the psychological tests used in the HCP and the other studies may not be sensitive enough to detect subtle differences in healthy subjects, all the more so as there is yet no known specific function associated with the IA that could be tested. The HCP group of subjects used in this project was also young (22–35 years old) and homogeneous, further suggesting that subjects might have performed at the ceiling. It is possible that using more sensitive tests in groups of older subjects could reveal subtler differences.

In addition, it is possible that compensatory neural mechanisms might mitigate potential cognitive impacts of IA presence/absence, supporting the hypothesis that the brain can reorganize itself to maintain cognitive function in the absence of IA¹⁵. IA functions could also be taken over by other commissural pathways, such as the corpus callosum or the anterior or posterior commissure, in case of IA absence^21,22. Overall, associating the IA with specific cognitive functions might prove difficult.

Genetic influence on IA prevalence and anatomical variants

The results of this study highlight a significant genetic influence on the presence and anatomical variations of the IA. Monozygotic (MZ) twins exhibited a higher concordance rate of IA presence (96%), compared to dizygotic (DZ) twins (79%), non-twin siblings (85%), or unrelated subjects (75%). However, 4% of subjects in the MZ group did not share the same IA characterization as their siblings, potentially implicating biological sex effects since these discrepancies were only found among men. The heritability estimate indicated that 34% of the IA presence and 50% of its anatomical variance is attributable to genetic factors, while the remaining variance is likely due to other influences and require further study.

This analysis aligns with previous literature highlighting strong anatomical correlations between MZ twins compared to matched controls, underscoring the genetic contribution to both prenatal and postnatal brain development⁵⁸. However, MZ twins can exhibit differential diagnoses of neurodevelopmental disorders such as schizophrenia, bipolar disorder, and attention deficit hyperactivity disorder (ADHD). Notably, in schizophrenia, the affected twin more often shows third ventricle enlargement, indicating anatomical differences not solely explained by genetics⁵⁹. Among discordant twins for these conditions, various anatomical differences have been observed alongside epigenetic variations^60,61. For instance, differential X-chromosome inactivation has been suggested in MZ female twins discordant for bipolar disorder⁶², and epigenetic changes, such as alterations in DNA methylation of the dopamine D2 receptor gene or the catechol-O-methyltransferase gene, have been noted in MZ twins discordant for schizophrenia⁶³.

Biological sex differences and genetic factors modulating IA prevalence

Biological sex emerged as a significant factor of IA presence, with females exhibiting a higher prevalence of IA than males, in robust agreement with previous studies^{2,4,7,9,12,14,15,19,20}. Consistent with more recent findings, a higher rate of broad IA in females was also demonstrated in this study^5,6,19,23. In line with these particular features present only in females, our novel finding shows 100% concordance in IA presence among female MZ twins (compared to 96% for males). This genetic influence could arise from sex chromosome (e.g., X-linked or Y-linked genes such as SRY), autosomal loci with sex-biased expression, or hormonal pathways (e.g., androgen/estrogen effects),

Although males’ and females’ brains are very similar, differences may exist⁶⁴. However, a recent meta-synthesis demonstrated no reliable subcortical volumes or cortical thickness differences when adjusting for brain size, nor lateralization dimorphism⁶⁵ Structural connection studies also show discrepancies, but females tend to have a slightly larger corpus callosum and anterior commissure^66–68. These larger structures in females are demonstrated in the present study to be larger in individuals with IA, suggesting a need to refine further research to understand these differences. Currently, the only consistently reliable difference between males and females is the different IA prevalence and anatomical variant⁶⁵.

The higher IA absence in males aligns with their increased susceptibility to neurodevelopmental disorders (e.g., schizophrenia, autism)^69–73, which often involve thalamic and dopaminergic dysfunction^74–79. Preceding these results, Tresniak et al.¹⁸ suggested an IA role of the dopaminergic system in schizophrenia spectrum disorders etiopathology^40,47,48. However, direct evidence linking IA absence to dopaminergic pathways is limited. Instead, IA absence may serve as a marker of altered midline development, possibly reflected by enlarged third ventricles and increased CSF as suggested in our study.

Bridging the link between IA, midline development, the dopaminergic system, and sexual dimorphism, several hypotheses emerge for future investigations. First, estrogen’s neuroprotective role on dopaminergic neurons^80–82 may contribute to higher IA retention in females. Second, as SRY (Sex-determining Region on the Y chromosome) is expressed in dopamine-abundant brain regions, regulating dopamine biosynthesis in males⁸³, its dysregulation among males during early development could alter the dopaminergic system, leading to IA anomalies. Finally, epigenetic studies indicate sex-specific signals induced by both environmental and preprogrammed hormonal or genetic cues^84–87, causing differential gene expression between males and females during development and across the lifespan^87,88. These epigenetic phenomena can impact the brain even after the perinatal window closes⁸⁹.

Significance

Overall, our results support the hypothesis that IA presence is influenced by genetic predisposition. Since no cognitive impairments were observed, the absence of the IA may not have a direct pathological effect but could instead serve as a marker of abnormal neurodevelopment of midline structures (e.g., ventricular enlargement) involving mechanisms to which males are more vulnerable. This is consistent with reports linking IA absence to neurodevelopmental disorders^8,18, although whether this reflects shared underlying mechanisms requires further investigation. Speculatively, associated with other abnormalities, IA absence could possibly be related to structural and functional aberrations such as deficits in the dopaminergic system and could involve a higher risk of developing a neuropsychiatric disorder^8,18. Being in possession of an IA would then indicate the integrity of adjacent midline structures and the third ventricle, reflecting a greater probability of better outcomes. Given that the IA is typically easy to visualize on high-quality MRI, its presence or absence may hold promise as a practical and informative neurodevelopmental biomarker.

Limits

This article presents several limitations. The association between IA and age could not be identified due to the limited age distribution of subjects. Additionally, the low number of siblings with different IA characterizations in the MZ group prevented proper neuropsychological statistical analyses. For similar reasons, genetic analyses were also constrained. However, future studies with larger sample sizes and greater variance should consider employing more complex models, such as ACE models⁹⁰, to disentangle the variance in IA prevalence attributable to genetic factors (A), shared environment (C), and unique environment (E), while controlling for covariates. Finally, the DZ group did not include mixed biological sex sibling pairs for reasons remaining unclear to us. The Not-Twin group was the only one with mixed-biological sex sibling pairs.

Conclusion

This study provides robust evidence that genetic factors significantly influence IA prevalence and anatomical variants. While IA absence correlates with specific neuroanatomical differences, it does not appear to impact cognitive function in healthy individuals. Future research should focus on the clinical significance of IA absence, its potential role as a biomarker for neurodevelopmental abnormalities, as well as the possible mechanisms underlying compensatory neural pathways that preserve cognitive function in the absence of IA.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

J.P.V. and E.J.B. designed the study.J.P.V. performed the computations and analysis under the supervision of L.D. and E.J.B. andwrote the manuscript with their support. A.G and M.R did some computations under the supervision of J.P.V.J.P, P.P, J.P., L.D, E.J.B substantively revised the manuscript.

Data availability

All data are provided by the Human Connectome Project (HCP) and are either part of a publicdataset or shared on request by contacting J.P.V.

Declarations

Competing interests

The authors declare no competing interests.