Biological factors contributing to divergent cortical abnormalities in adolescent and adult psychiatric disorders

Abstract

Cortical morphological alteration patterns differ between adolescent and adult psychiatric disorders. However, the biological factors contributing to the divergence are unclear. Cortical thickness (CT) alterations in adolescents and adults with attention deficit hyperactivity disorder (ADHD), bipolar disorder (BD), major depressive disorder (MDD), and obsessive-compulsive disorder (OCD) were derived from the ENIGMA. We examined whether the structural connectome constrains disease-related CT alterations, followed by identifying likely epicenter regions and testing the hub vulnerability hypothesis. Using neurotransmitter, transcriptome, and mitochondria atlases, we furthermore investigated the neurochemical basis, genetic architecture, and molecular energetic landscape related to the CT alterations. Results showed that the structural connectome constrained CT alterations in adult psychiatric disorders rather than their adolescent counterparts. The epicenters were largely consistent in adolescents and adults for ADHD and MDD, while divergent for BD and OCD. The demonstration of CT alterations in adolescent BD, adult BD, and adult OCD as a function of connectome degree centrality was consistent with the hub vulnerability hypothesis. We also found distinct neurotransmitter systems linked to CT alterations in psychiatric adolescents and adults. Transcriptomic contextualization showed that CT alterations in adult ADHD, adult MDD, and adolescent OCD were related to genes involving essential components of the cerebral cortex, signal pathway, and nervous system development, while those in adolescent and adult BD to synapse and catabolic process. Additionally, mitochondrial features were associated with CT alterations in almost all conditions. Our findings may elucidate the biological factors associated with the differential cortical abnormalities between psychiatric adolescents and adults.

Introduction

Psychiatric disorders have been recognized as significant contributors to the global burden of disease [1]. Continuing improvements in the precision of in vivo neuroimaging techniques, and magnetic resonance imaging (MRI) in particular, have made it increasingly feasible to decode psychiatric neuropathology. Moreover, big-data neuroscience initiatives, such as the Enhancing NeuroImaging Genetics through Meta-Analysis (ENIGMA) consortium, have aggregated brain MRI data in thousands of healthy and psychiatric individuals across multiple sites from around the world [2]. In this context, dedicated ENIGMA working groups have revealed cortical morphological abnormalities in adolescents and adults with psychiatric disorders including attention deficit hyperactivity disorder (ADHD) [3], bipolar disorder (BD) [4], major depressive disorder (MDD) [5], and obsessive-compulsive disorder (OCD) [6]. Notably, the location and nature of these cortical structural alterations differ between adolescent and adult patients. For example, adult BD patients showed more widespread cortical thinning, mainly involving the bilateral frontal, temporal, and parietal regions, whereas abnormalities in adolescent BD were more circumscribed and were primarily characterized by thinning of the right supramarginal gyrus [4]. Although both adult and adolescent OCD patients showed parietal cortical thinning, adult OCD was mainly characterized by thinning of the bilateral inferior parietal cortex, whereas adolescent OCD showed thinning in the bilateral superior parietal cortex, left inferior parietal cortex, and left lateral occipital cortex [6]. However, the biological factors contributing to these divergent patterns are poorly understood, which hampers the development of effective individualized treatments for psychiatric disorders.

The human brain is a highly intricate network composed of continuously interacting neuronal populations anatomically connected by a complex web of axonal fibers, referred to as the structural connectome [7, 8]. The structural connectome enables inter-regional communication and integration, giving rise to synchronized activity patterns and co-activation across brain regions [9, 10]. In addition to facilitating transport of molecules and organelles essential for neurotransmission and metabolism [11], axonal fibers can also serve as conduits of pathological spread [12, 13]. In this framework, previous research has demonstrated that the spatial patterns of cortical morphological abnormalities are constrained by the brain’s connectome architecture in many neurodegenerative and psychiatric diseases [14–17]. Following this, an epicenter mapping approach is typically employed to further determine the origin of disease propagation, i.e., identifying the likely regional epicenters [14–18]. Additionally, hub regions in the structural connectome, which have more connections to other regions, are assumed to be more vulnerable to disease and thus tend to show more pronounced structural abnormalities, a concept termed the nodal stress or hub vulnerability hypothesis [12, 19]. This hypothesis has been confirmed in multiple neuropsychiatric disorders [16, 18, 20]. Collectively, investigation of the relationships with aspects of the structural connectome may help elucidate the underlying neurobiological processes that shape the spatial patterning of cortical morphological alterations in psychiatric disorders.

Advances in nuclear imaging techniques and tracers have afforded reliable and accurate quantification of a rich range of neurotransmitter receptors and transporters across the whole brain [21]. These neurotransmitter maps have offered researchers sufficient material to investigate the neurochemical basis of brain structure and function [22–24]. In parallel, with the recent introduction of comprehensive, brain-wide gene expression atlases, such as the Allen Human Brain Atlas (AHBA) [25, 26], comes the emergent field of imaging transcriptomics, whose goal is to uncover the genetic architecture underlying neuroimaging findings by identifying genes with spatial profiles of regional expression that track anatomical variations in a variety of neuroimaging phenotypes [27–30], including structural and functional abnormalities in psychiatric disorders [31–34]. In addition, given that mitochondria power brain activity and mitochondrial defects are linked to brain disorders, a comprehensive brain atlas of mitochondrial characteristics has recently been developed to provide a foundation for exploring the molecular energetic landscape that enables normal brain function and underlies abnormal neural processes relevant to brain disorders [35]. Combined, the current availability of these state-of-the-art atlases could open new opportunities for biologically contextualizing disease-related cortical morphological alterations.

In this study, we derived cortical thickness (CT) alterations in large cohorts of adolescent and adult patients with ADHD, BD, MDD, and OCD from the ENIGMA. Using a normative structural connectome constructed based on diffusion MRI data from the Human Connectome Project (HCP), we examined whether the structural connectome constrains disease-related CT alteration patterns, followed by identifying likely epicenter regions and testing the hub vulnerability hypothesis. In conjunction with the neurotransmitter, transcriptome, and mitochondria atlases, we furthermore investigated the neurochemical basis, genetic architecture, and molecular energetic landscape associated with the CT alterations. A systematic flowchart of the study design is shown in Fig. 1.

Fig. 1. A systematic flowchart of the study design.

A We derived CT alterations in large cohorts of adolescent and adult patients with ADHD, BD, MDD, and OCD from ENIGMA neuroimaging summary statistics. B Using a normative structural connectome constructed based on diffusion MRI data from the HCP, we examined whether the structural connectome constrains disease-related CT alteration patterns, followed by identifying likely epicenter regions and testing the hub vulnerability hypothesis. C In conjunction with the publicly available neurotransmitter, transcriptome, and mitochondria atlases, we furthermore investigated the neurochemical basis, genetic architecture, and molecular energetic landscape related to the CT alterations. ADHD attention deficit hyperactivity disorder, BD bipolar disorder, CT cortical thickness, ENIGMA Enhancing Neuroimaging Genetics through Meta Analysis, HCP Human Connectome Project, MDD major depressive disorder, MRI magnetic resonance imaging, OCD obsessive-compulsive disorder.

Materials and methods

ENIGMA neuroimaging summary statistics

For our analyses, we used multi-site neuroimaging summary statistics published by the ENIGMA Consortium and publicly available within the ENIGMA Toolbox (https://github.com/MICA-MNI/ENIGMA) [36]. Included psychiatric disorders comprised ADHD [3], BD [4], MDD [5], and OCD [6], which were investigated in adolescent and adult samples respectively. Sample sizes and age ranges of the ENIGMA cohorts are summarized in Table S1. We focused on case-control differences in CT by performing cross-site random-effects meta-analyses of Cohen’s d values with adjustment for covariates including age, sex, and site. In the resultant Cohen’s d maps, positive/negative effects indicated increases/decreases in CT in patients relative to controls. Prior to the computation of summary statistics, structural MRI data were preprocessed, segmented and parcellated based on the Desikan-Killiany atlas [37] in FreeSurfer [38] at each site and according to standard ENIGMA quality control protocols (http://enigma.ini.usc.edu/protocols/imaging-protocols). Individual cohort investigators obtained approval from local institutional ethics boards, and informed consent was obtained from study participants or their guardians.

Normative structural connectome

To assess the connectome constraints on disease-related CT alterations and to identify likely epicenter regions, a normative structural connectome was constructed based on diffusion MRI data of 207 unrelated healthy adults (83 males, mean age ± standard deviation = 28.73 ± 3.73 years, age range = 22–36 years) from the HCP [39]. Details on diffusion MRI data processing and structural connectome construction are provided in Supplementary method. After these procedures, a group-level 68 × 68 structural connectivity matrix based on the Desikan-Killiany atlas was generated.

Connectome constraint model

A connectome constraint model was used to examine the connectome constraint effects on CT alterations (Cohen’s d maps) in adolescent and adult psychiatric disorders (ADHD, BD, MDD, and OCD) separately. To achieve this goal, we initially obtained the estimated CT alteration of node i (${\hat{d}}_i$) based on its connected neighbor nodes using the connectome constraint model [14, 15, 40] defined as: ${\hat{d}}_i=\frac{1}{N_i}{\sum }_{j=1,j\ne i}^{N_i}{d}_j{A}_{\mathit{ij}}$, where N_i is the number of structurally connected neighbors of node i and d_j is the CT alteration observed in the j-th neighbor of node i, and A_ij defines the structural connectivity between nodes i and j. The group-level structural connectome was used to define the connected neighbors of each node. Then, we calculated the spatial correlation between the empirical and estimated CT alterations across 68 cortical nodes. We tested the spatial correlation against two null models to examine the statistical significance. In the first null model, we used the spin permutation test to investigate whether the observed correlation was specific to the actual CT alteration pattern rather than due to the spatial autocorrelation [41]. Specifically, we first recorded the spherical coordinates of centroids for each node in the Desikan-Killiany atlas. Then, we randomly rotated the nodes while maintaining spatial autocorrelation and reassigned node values to the nearest nodes. This procedure was repeated 1000 times to create surrogate brain maps. These surrogate brain maps were used to calculate the estimated CT alterations and repeat the above correlation analysis. The second rewired null model was used to evaluate whether the observed correlation is determined by the empirical structural connectome topology rather than the basic spatial embedding of the structural connectome (such as the distribution of node degree and edge length) [42]. Specifically, we first divided edges into different bins according to their Euclidean distance. To preserve the degree sequence and approximate edge length distribution of the empirical structural connectome, edge pairs were randomly swapped within each bin. Finally, 1000 rewired connectomes were generated by repeating this procedure using a combination of 10 distance bins and 50,000 edge swaps [15]. For the above two null models, P values were calculated as the fractions of correlations in null models that exceeded the observed correlation.

Epicenter mapping

The connectome constraint model evaluates the degree to which structural connectome constrains the spatial patterns of disease-related CT alterations, yet it is unable to identify regions from which the pathological spread may initiate. Thus, we employed an epicenter mapping approach [16, 18] to further investigate whether certain brain nodes function as likely epicenter regions. Specifically, we used cross-region Pearson correlations to assess the spatial similarity of each node’s normative whole-brain structural connectivity profile with the whole-brain patterns of CT alterations (Cohen’s d maps) in adolescent and adult psychiatric disorders (ADHD, BD, MDD, and OCD) separately. A brain region can be considered an epicenter regardless of its CT alteration magnitude, as long as it is strongly connected to regions with greater CT alterations and weakly connected to regions with smaller CT alterations. Statistical significance was assessed using a spin permutation test, where 1000 spatially constrained surrogate maps were generated to derive a null distribution and P value was calculated as the fraction of correlations in the null distribution that exceeded the observed correlation. Brain regions with spin-test P values < 0.05 were considered epicenters.

Relation to connectome degree centrality

To test the hub vulnerability hypothesis that hub regions are more vulnerable to disease and tend to show more pronounced structural abnormalities, we performed spatial correlations between disease-related CT alteration patterns (Cohen’s d maps) and the weighted degree centrality derived from the normative structural connectome. Weighted degree centrality of a given region was defined by the sum of its all weighted structural connections. Brain regions with higher degree centrality are considered hub regions. Statistical significance was also examined using the above-described spin permutation test.

Relation to neurotransmitters

To explore the neurochemical basis of CT alterations in adolescent and adult psychiatric disorders, we examined their spatial correlations with 19 maps of neurotransmitter receptors and transporters [21] involving serotonin, acetylcholine, cannabinoid, dopamine, GABA, histamine, glutamate, opioid, and norepinephrine (Table S2). Specifically, Pearson correlation coefficients between the Cohen’s d maps and these neurotransmitter maps were calculated across 68 cortical regions. Likewise, we used the spin permutation test to assess statistical significance and P values were adjusted using Bonferroni method.

Relation to transcriptome

We adopted neuroimaging-transcriptome association analysis to investigate the genetic architecture associated with disease-related CT alterations, using brain-wide gene expression data provided by the AHBA dataset (http://www.brainmap.org) [43]. This dataset was derived from six adult postmortem human donors (Table S3). Microarray expression data were processed in abagen [44], including intensity-based filtering, normalization and aggregation within Desikan-Killiany parcels and across donors. Only genes with a similarity of r > 0.2 across donors were included, resulting in 12,668 genes for the analysis [36]. Then, the spatial associations between gene expression and disease-related CT alteration patterns (Cohen’s d maps) were examined using partial least squares (PLS) regression, a multivariate statistical technique that can predict a set of dependent variables from a set of independent variables [45]. PLS regression is technically well suited to the highly collinear gene expression data and has been extensively applied to neuroimaging-transcriptome association research [29, 30, 46]. In our analysis, the gene expression data were defined as the independent variables, and the Cohen’s d values of CT alterations were designated as the dependent variables. PLS regression identified the first PLS component where the weighted sum of gene expression (gene score) was most strongly correlated with the Cohen’s d values. Pearson correlation coefficient between the gene score and the Cohen’s d values was calculated, and its statistical significance was assessed using the spin permutation test. Briefly, we created 1000 surrogate CT alteration maps that maintained spatial autocorrelation and then used these surrogate maps to repeat the above PLS regression analysis to produce a null distribution. P value was determined by calculating the frequency with which the true correlation coefficient exceeded those in the null distribution.

Gene contribution was estimated with gene loading, that is, Pearson correlation coefficient between an individual gene’s expression pattern and the PLS-derived gene score across brain regions. Then, genes were segregated based on the sign of their loadings, where positive genes refer to genes with positive loadings and negative genes refer to those with negative loadings. Strongly contributing genes were defined as those among the top 10% of the positive and negative genes. To better understand the biological significance of all strongly contributing genes, we conducted functional enrichment analyses with the ToppGene portal (https://toppgene.cchmc.org/) [47]. Gene ontology (GO) was utilized to determine the biological functions including biological process (BP), cellular component (CC) and molecular function (MF). Fisher’s exact tests were used to evaluate the statistical significance. Multiple testing was corrected using the false discovery rate (FDR) method with a corrected P value of 0.05.

Relation to brain mitochondria

To understand the molecular energy landscape that underlies disease-related CT alterations, we examined their spatial correlations with a comprehensive brain atlas of mitochondrial features. These mitochondrial features were derived from a coronal slab of the right hemisphere of a 54-year-old neurotypical male, including oxidative phosphorylation (OXPHOS) enzyme activities (CI, CII and CIV), mitochondrial density (MitoD), mitochondrial respiratory capacity (MRC), and tissue respiratory capacity (TRC) [35]. Statistical significance of the spatial correlations was determined by the spin permutation test.

Robustness analyses

To assess the robustness of the connectome constraint and hub vulnerability findings, we performed leave-one-region-out analyses. For the connectome constraint analysis, with one cortical region removed each time, the estimated CT alteration map was recalculated for the remaining regions, and its spatial correlation with the empirical map was recomputed. For the hub vulnerability analysis, with one cortical region removed each time, weighted degree centrality of the normative structural connectome was recalculated, and its spatial correlations with disease-related CT alteration patterns were recomputed across the remaining regions.

Results

Structural connectome constraints

Whole-brain CT alteration patterns (Cohen’s d maps) in adolescent and adult psychiatric disorders (ADHD, BD, MDD, and OCD) from the ENIGMA cohorts are shown in Fig. 1A. To determine whether structural connectome constrains the CT alteration patterns, we constructed a normative structural connectome based on diffusion MRI data collected from an independent cohort of 207 healthy adults from the HCP dataset. Based on the normative structural connectome, we obtained the estimated CT alteration of each node according to its connected neighbor nodes using the connectome constraint model. Significant spatial correlations were found between the empirical and estimated CT alterations across 68 cortical nodes for adult ADHD (r = 0.648, P_spin = 0.002, P_rewire = 0.030), adult BD (r = 0.625, P_spin < 0.001, P_rewire < 0.001), adult MDD (r = 0.491, P_spin = 0.011, P_rewire = 0.014), and adult OCD (r = 0.164, P_spin = 0.012, P_rewire < 0.001) (Fig. 2A), suggesting structural connectome constraints on CT alteration patterns in these psychiatric conditions. However, no significant correlations were observed for adolescent ADHD, adolescent BD, adolescent MDD or adolescent OCD (Fig. 2A).

Fig. 2. Structural connectome constraints and epicenter mapping.

A Scatterplots of the spatial correlations between the empirical and estimated CT alterations across 68 cortical nodes. Asterisks represent significant correlations, suggesting structural connectome constraints on disease-related CT alteration patterns. B Brain maps showing the regional epicenters whose structural connectivity profiles significantly correlated with disease-related CT alteration patterns. ADHD attention deficit hyperactivity disorder, BD bipolar disorder, CT cortical thickness, MDD major depressive disorder, OCD obsessive-compulsive disorder.

Epicenter regions

After determining the extent to which disease-related CT alterations were constrained by the structural connectome, we employed an epicenter mapping approach to further investigate whether certain brain nodes function as likely epicenter regions, by comparing each node’s structural connectivity profile with the whole-brain CT alteration patterns. Results showed the bilateral superior frontal gyrus as the shared epicenters of both adolescent and adult ADHD, with the bilateral anterior cingulate cortex as the epicenters specific to adult ADHD. The epicenters of adolescent BD were located in the frontal lobe (bilateral superior frontal, bilateral caudal middle frontal, and right precentral gyrus) and parietal lobe (left inferior parietal and bilateral postcentral gyrus). The epicenters of adult BD were localized in the frontal lobe (bilateral middle frontal and bilateral inferior frontal gyrus), left temporal lobe (left inferior temporal and left middle temporal gyrus), and bilateral inferior parietal gyrus. With respect to MDD, the common epicenters included the bilateral medial orbitofrontal gyrus, bilateral frontal pole, and bilateral anterior cingulate cortex, with the left superior frontal gyrus specific to adolescent MDD and the right temporal pole specific to adult MDD. The epicenters of adolescent OCD were located in the bilateral superior parietal, right inferior parietal, and right lateral occipital cortex. The epicenters of adult OCD were localized in the frontal lobe (right superior frontal, bilateral rostral middle frontal, and bilateral inferior frontal gyrus), left inferior parietal gyrus, right temporal lobe (right inferior temporal and right superior temporal gyrus), right lateral occipital cortex, and right anterior cingulate cortex. All the epicenters of adolescent and adult psychiatric disorders (ADHD, BD, MDD, and OCD) are illustrated in Fig. 2B (P_spin < 0.05).

Hub vulnerability

To test the hub vulnerability hypothesis, spatial correlations between disease-related CT alteration patterns and nodal degree centrality of the structural connectome were conducted. We found significant negative associations of degree centrality with CT alterations in both adolescent (r = −0.336, P_spin = 0.003) and adult (r = −0.354, P_spin = 0.006) BD as well as in adult OCD (r = −0.390, P_spin = 0.001) (Fig. 3A), indicating that hub regions with higher degree centrality tend to present more pronounced structural abnormalities. No significant associations were observed for ADHD, MDD, or adolescent OCD.

Fig. 3. Hub vulnerability and neurochemical basis.

A Scatterplots of the spatial correlations between disease-related CT alteration patterns and nodal degree centrality of the structural connectome. Asterisks represent significant correlations, indicating that hub regions with higher degree centrality tend to present more pronounced structural abnormalities. B Heatmap showing the spatial correlations between 19 neurotransmitter maps and disease-related CT alteration patterns. *P_spin < 0.05, Bonferroni corrected. 5HT 5-hydroxytryptamine, 5HTT 5-hydroxytryptamine transporter, ADHD attention deficit hyperactivity disorder, BD bipolar disorder, CB cannabinoid type, CT cortical thickness, D dopamine, DAT dopamine transporter, GABAa gamma-aminobutyric acid a, H histamine, M muscarinic, MDD major depressive disorder, mGluR metabotropic glutamate type, MOR mu opioid receptor, NET norepinephrine transporter, NMDA N-methyl-D-aspartate receptor, OCD obsessive-compulsive disorder, VAChT vesicular acetylcholine transporter.

Neurochemical basis

We performed spatial correlations between 19 neurotransmitter maps and disease-related CT alteration patterns to explore their neurochemical basis. The significant associations are shown in Fig. 3B and Supplementary File S1 (P < 0.05, Bonferroni corrected). Specifically, CT alterations in adult ADHD showed significant negative correlations with H3 (r = −0.582, P_spin = 0.010), MOR (r = −0.707, P_spin = 0.029), and VAChT (r = −0.612, P_spin < 0.001). CT alterations in adolescent BD showed a significant negative correlation with mGluR5 (r = −0.442, P_spin = 0.029). CT alterations in adult BD showed a significant positive correlation with 5HTT (r = 0.557, P_spin < 0.001) as well as negative correlations with CB1 (r = −0.478, P_spin = 0.019), 5HT2a (r = −0.513, P_spin < 0.001), and 5HT4 (r = −0.474, P_spin = 0.021). CT alterations in adult MDD showed significant negative correlations with CB1 (r = −0.513, P_spin = 0.010), D1 (r = −0.508, P_spin = 0.048), D2 (r = −0.461, P_spin = 0.038), MOR (r = −0.685, P_spin = 0.010), and 5HT1a (r = −0.534, P_spin < 0.001). CT alterations in adult OCD showed a significant positive correlation with 5HTT (r = 0.367, P_spin = 0.019). No significant correlations were found for adolescent ADHD, adolescent MDD, or adolescent OCD.

Genetic architecture

With use of the AHBA transcriptomic dataset, we investigated the associations between gene expression and disease-related CT alteration patterns to elucidate their genetic architecture. PLS regression identified the first PLS components where the weighted sum of gene expression (gene score) was most strongly correlated with CT alterations in adult ADHD (r = 0.618, P_spin < 0.001), adolescent BD (r = 0.405, P_spin = 0.019), adult BD (r = 0.420, P_spin = 0.014), adult MDD (r = 0.644, P_spin = 0.001), and adolescent OCD (r = 0.542, P_spin = 0.020) (Fig. 4A). All the 12,668 genes were ranked by their contributions to the correlations (i.e., gene loadings), with 1267 genes (the top 10% of genes with positive loadings and negative loadings) defined as strongly contributing genes. To better understand the biological significance of the strongly contributing genes, we conducted functional enrichment analyses. The results of functional enrichment are listed in Supplementary File S2 and are depicted in Fig. 4B (P < 0.05, FDR corrected). Specifically, CT alterations in adult ADHD, adult MDD, and adolescent OCD were linked to gene categories primarily involving essential components of the cerebral cortex (axon, dendrite and synapse), signal pathway, and nervous system development. CT alterations in adolescent and adult BD were linked to gene categories primarily involving synapse and catabolic process.

Fig. 4. Genetic architecture.

A Scatterplots of the overall associations between gene expression and disease-related CT alteration patterns identified by PLS regression. *P_spin < 0.05. B Functional enrichment of the strongly contributing genes (the top 10% of genes with positive loadings and negative loadings). For each bubble chart, x-axis denotes rich factor and y-axis denotes items from the GO. Rich factor refers to the ratio of the number of strongly contributing genes annotated to the item to the number of all genes annotated to the item. Bubble size represents the number of genes overlapping with those belonging to each item, and bubble color represents -Log₁₀(P) with P values corrected by the FDR method. ADHD attention deficit hyperactivity disorder, BD bipolar disorder, BP biological process, CC cellular component, CT cortical thickness, FDR false discovery rate, GO gene ontology, MDD major depressive disorder, MF molecular function, OCD obsessive-compulsive disorder, PLS partial least squares.

Molecular energetic landscape

Spatial correlations between mitochondrial features (CI, CII, CIV, MitoD, MRC, and TRC) and disease-related CT alteration patterns were examined to understand the molecular energy landscape related to the CT alterations. We found that CT alterations in both adolescent and adult ADHD were negatively correlated with MitoD (adolescent ADHD: r = −0.378, P_spin = 0.016; adult ADHD: r = −0.383, P_spin = 0.012). CT alterations in adult BD were negatively correlated with CI (r = −0.272, P_spin = 0.035), CIV (r = −0.297, P_spin = 0.027), and TRC (r = −0.293, P_spin = 0.028). CT alterations in both adolescent and adult MDD were positively correlated with CII (adolescent MDD: r = 0.247, P_spin = 0.033; adult MDD: r = 0.392, P_spin = 0.003). CT alterations in adolescent OCD were positively correlated with MitoD (r = 0.452, P_spin = 0.011) (Fig. 5). No significant spatial correlations were found for adolescent BD or adult OCD.

Fig. 5. Molecular energetic landscape.

Spatial correlations between mitochondrial features and disease-related CT alteration patterns were examined to understand the molecular energy landscape related to the CT alterations. The resultant Pearson correlation coefficients were assessed against a spatial autocorrelation preserving null model (spin permutation test). Points represent the empirical Pearson correlation coefficients. In the boxplots of the null distributions derived from 1000 permutations, the ends of the boxes represent the first (25%) and third (75%) quartiles, the center line represents the median, the whiskers represent the non-outlier end-points, and the solid circles represent outliers. ADHD attention deficit hyperactivity disorder, BD bipolar disorder, CI, NADH-ubiquinone oxidoreductase; CII; succinate dehydrogenase; CIV cytochrome c oxidase, MDD major depressive disorder, MitoD mitochondrial density, MRC mitochondrial respiratory capacity, OCD obsessive-compulsive disorder, TRC tissue respiratory capacity.

Robustness analyses

Leave-one-region-out analyses further supported the robustness of our findings. For the connectome constraint analysis, the same patterns of structural connectome constraints on disease-related CT alterations were observed with one cortical region removed each time (Supplementary File S3). For the hub vulnerability analysis, we found similar negative associations of connectome degree centrality with CT alterations in adolescent BD, adult BD, and adult OCD using leave-one-region-out analyses (Supplementary File S4).

Discussion

Based on ENIGMA neuroimaging summary statistics from large cohorts of psychiatric adolescents and adults, the current study revealed that the structural connectome constrained CT alteration patterns in adult ADHD, BD, MDD, and OCD rather than their adolescent counterparts. The epicenter regions were largely consistent in adolescents and adults for ADHD and MDD, while divergent for BD and OCD. Concurrently, the demonstration of CT alterations in adolescent BD, adult BD, and adult OCD as a function of connectome degree centrality was consistent with the hub vulnerability hypothesis in these conditions. We also found distinct neurotransmitter systems linked to CT alterations in adolescent and adult psychiatric disorders. Transcriptomic contextualization showed that CT alterations in adult ADHD, adult MDD, and adolescent OCD were related to gene categories primarily involving essential components of the cerebral cortex, signal pathway, and nervous system development, while those in adolescent and adult BD to synapse and catabolic process. Additionally, mitochondrial features were found to associate with CT alterations in almost all psychiatric disorders except for adolescent BD and adult OCD.

In line with previous studies reporting constraint effects of the brain connectome on cortical morphological alterations in neurodegenerative and psychiatric disorders [14–16], our findings in adult psychiatric disorders confirm a connectome-constrained spatial pattern of cortical thickness alterations. Notably, given the rather small correlation for adult OCD, the connectome constraint effect in this condition should be interpreted with caution. Conversely, the constraint effects were not observed in adolescent psychiatric disorders. This suggests that in adults with psychiatric disorders, the structural connectome may play a more prominent role in shaping cortical morphological alteration patterns. While the exact mechanisms remain unclear, the spatial patterning of cortical morphological alterations based on the structural connectome may be explained by several factors. First, cortical regions structurally connected by axonal projections often exhibit comparable cytoarchitectonic features, such as neuronal density and laminar differentiation [48–51]. Second, the structural connectome facilitates interregional communication and spontaneous neuronal activity synchronization among connected regions [52], in alignment with Hebbian learning principles where synchronously firing neurons are more likely to form or strengthen connections [53]. Third, the structural connectome acts as transport channels for substances, sustaining neurophysiological homeostasis through axonal transport of nutrient factors and metabolic substrates [54, 55]. Likewise, the structural connectome can also act as a conduit for the spread of pathology, such that illness processes originating in one area can propagate to affect distributed systems via multiple mechanisms [12, 56]. In neurodegenerative diseases such as dementia and Alzheimer’s disease, pathological proteins have been shown to spread along axons, and disease-related cortical morphological change patterns are indeed constrained by the structural connectome [14, 57]. Although psychiatric disorders show limited evidence of visible pathological protein aggregates, subtle alterations in protein homeostasis may occur and propagate through synaptically connected brain regions [58], and the structural connectome-constrained cortical morphological alterations have been demonstrated in psychiatric disorders [15, 17]. These factors may render psychopathological changes to spread along structurally connected brain regions. In addition, research has shown that synapse density peaks before adolescence, subsequently decreasing during the adolescent period and then remaining relatively stable until later in life [59]. This decrease in synaptic density during adolescence is primarily due to synaptic pruning, a process that has been shown to be exacerbated by psychiatric disorders [60, 61]. Thus, the neurodevelopmental processes occurring during adolescence, such as synaptic pruning and axonal recombination, interact with emerging psychopathology, which may partially explain the lack of connectome constraints in psychiatric adolescents.

In addition to identifying differential epicenter regions across psychiatric disorders, we found that the epicenter regions were largely consistent in adolescents and adults for ADHD and MDD, while divergent for BD and OCD. The current observation of the bilateral superior frontal gyrus as the shared epicenters of both adolescent and adult ADHD may reflect the core pathological mechanisms underlying ADHD, particularly in terms of attention regulation and executive function [62]. Cortical thinning in frontal regions is associated with the delayed maturation hypothesis [63, 64]. Shaw et al. reported that children with ADHD reached peak CT in the prefrontal cortex much later [65]. Furthermore, we found the bilateral anterior cingulate cortex as the additional epicenters in adult ADHD. Given that the anterior cingulate cortex has a tight structural connection to the superior frontal gyrus [66], one may speculate that the epicenters of adult ADHD may extend from the superior frontal gyrus to the anterior cingulate cortex. For both adolescent and adult MDD, the bilateral orbitofrontal cortex and anterior cingulate cortex constitute the common epicenters. Previous meta-analyses have reported reductions in volume and CT in these regions in MDD [5, 67]. It is also evident that structural abnormalities in these regions occur early in the course of MDD [68] and might play a key role in the dysregulation of negative emotion, value representation, and self-referential processing [69]. For BD and OCD, the epicenters were widely distributed across the frontal, parietal, temporal, and occipital regions, which may account for the diversity of their clinical manifestations and pathophysiological characteristics [62, 70, 71]. Moreover, the identified epicenter regions differed between adolescent and adult patients, which could have important clinical implications for developing more effective targets for therapeutic intervention in different conditions.

Our data demonstrated CT alterations in adolescent BD, adult BD, and adult OCD as a function of connectome degree centrality. Albeit trend-level significance, this was also the case for adolescent OCD. These findings support the hub vulnerability hypothesis in BD and OCD irrespective of patient populations. Several biological mechanisms could lead to a marked susceptibility of hub regions. According to the stress-threshold model, centrally located neuronal populations are especially vulnerable to pathophysiological perturbations due to their higher neuronal activity and concomitant metabolic demands [72, 73]. In addition, hub regions are more likely to be part of the association cortices which, partly due to their prolonged window of developmental plasticity, have been shown to be especially vulnerable to developmental insults [74].

We found distinct neurotransmitter systems linked to CT alterations in adolescent and adult psychiatric disorders. In adult ADHD, CT alterations were correlated with H3, MOR, and VAChT. Previous literature has shown close associations between these neurotransmitter receptors/transporters and ADHD [75–78]. For instance, blockade of centrally localized H3 receptors by selective H3 receptor antagonists has been shown to enhance the release of neurotransmitters such as histamine and acetycholine, which play important roles in cognitive processes of ADHD [75]. Histamine H3 receptor antagonists/inverse agonists possess potential to treat deficits of some aspects of cognitive and motor functions in ADHD [76]. CT alterations in adolescent BD were correlated with mGluR5. Glutamate is the central mediator of excitatory synaptic transmission in the human brain, facilitating learning, memory, and synaptic plasticity [79]. Previous studies have also found that glutamate concentrations in adolescent BD patients are reduced and affect the patients’ symptoms [80, 81]. However, CT alterations in adult BD were associated with CB1 and serotonergic system. The endocannabinoid system is a widespread neuromodulatory system that plays crucial roles in central nervous system development, synaptic plasticity, and the response to endogenous and environmental insults [82]. Serotonergic transmissions are critically involved in the pathophysiology of neuropsychiatric disorders and is modulated by the endocannabinoid system [83]. Similarly, adult MDD demonstrated comparable associations with CB1 and serotonergic system. These findings jointly highlight the intricate interplay between the endocannabinoid and serotonergic systems in mood disorder pathophysiology, suggesting that their combined modulation may offer novel therapeutic targets for BD and MDD. Furthermore, our results also showed that adult MDD was associated with dopamine receptors. Dopamine dysregulation has been hypothesized to underlie some core depressive symptoms, such as anhedonia and psychomotor disturbance [84, 85].

Our neuroimaging-transcriptome association analysis revealed that CT alterations in adult ADHD, adolescent BD, adult BD, adult MDD, and adolescent OCD were spatially associated with transcriptomic profiles. Specifically, CT alterations in adult ADHD, adult MDD, and adolescent OCD were related to gene categories primarily involving essential components of the cerebral cortex (axon, dendrite and synapse), signal pathway, and nervous system development, while those in adolescent and adult BD to synapse and catabolic process. It is largely known that normal communication between brain regions relies on the integrity of neuron, axon, dendrite, and synapse. Indeed, information exchange between neurons primarily takes place at synapses of axons and dendrites [86, 87]. With respect to signal transduction, synapses are the locus where information is transferred from a pre- to a postsynaptic neuron, which is largely mediated by neurotransmitters that are released by the presynaptic axon terminals and then bind to receptors on the postsynaptic dendritic spines [88]. Thus, abnormality in synaptic connections results in impaired processing of information, which is the common pathological basis of many psychiatric disorders [89, 90]. Abnormal folding and accumulation of proteins as a result of defective protein catabolic process may contribute to aggregation events. A recent review has emphasized that abnormal folding and accumulation of intracellular proteins may be a potential pathological mechanism of psychiatric disorders [91].

Contextualization of molecular energetic landscape showed that mitochondrial features were associated with CT alterations in almost all psychiatric disorders except for adolescent BD and adult OCD. It is generally accepted that many neuronal activities are energy-taxing and human nervous system consumes a great deal of energy. Mitochondria are assumed to be the major energy source providing ATP through oxidative phosphorylation to maintain the normal neuronal homeostasis and function [92]. As one of the high-energy consuming organs, the brain is vulnerable to impaired energy metabolism such that even mild changes in energy metabolism in the human brain are associated with disturbance in neural function. Previous literature has already discovered that impaired energy metabolism is one of the consistent characteristics in psychiatric disorders [93]. For example, a decreased ratio of phosphomonoesters to phosphodiesters in subjects with BD was consistent with differences in membrane turnover, suggesting alterations in phospholipid metabolism and mitochondrial function in the brains of subjects with BD [94]. In addition, alterations in high-energy phosphate metabolism and regulation of oxidative phosphorylation have been found in the brains of subjects with MDD [95]. Collectively, psychiatric disorders are closely related to abnormal mitochondrial energy metabolism, which may provide a new perspective for understanding the pathogenesis of psychiatric disorders. In the present study, however, the associations between CT alterations and mitochondrial features were not observed for adolescent BD or adult OCD. In adolescent BD, previous studies suggest that metabolic abnormalities may be relatively localized, manifesting as blood flow-oxygen metabolism mismatch [96], frontal high-energy phosphate alterations [97], and hemisphere-specific abnormalities [98]. In adult OCD, metabolic findings also appear relatively circumscribed [99], and evidence for mitochondria-targeted interventions in OCD remains inconclusive [100]. Therefore, brain energetic abnormalities may be present in these conditions, but they are not necessarily captured by the current spatial correlation analyses.

Several limitations need to be mentioned. First, we analyzed neuroimaging summary statistics from the ENIGMA datasets rather than the original data, so we cannot determine whether the possible use of controls shared among studies affected our results. Second, the cortical morphological alteration data from the ENIGMA datasets are only available for the Desikan-Killiany parcellation. It would thus be relevant to re-evaluate our results based on higher-resolution cortical parcellations, and/or to assess vertex-level data in future efforts. Third, we focused our analyses solely on the cerebral cortex. Analysis of subcortical structures will be part of our future investigations. Fourth, the data of disease-related CT alterations, structural connectome, neurotransmitters, transcriptome, and mitochondrial features were derived from different samples. In particular, the structural connectome, neurotransmitter, transcriptome, and mitochondrial data were obtained from adults, at odds with the ENIGMA neuroimaging data from adolescents. This developmental mismatch may bias our results. Future studies using sample- or age-matched datasets are needed to validate our findings. Finally, for the ENIGMA datasets, the sample sizes of the adolescent cohorts are smaller than the adult cohorts, which could lead to reduced accuracy of the cortical thickness alteration maps for adolescent psychiatric disorders and may ultimately result in biased findings of the spatial correlation analyses.

In conclusion, based on ENIGMA neuroimaging summary statistics and HCP diffusion MRI data, we investigated connectome architecture shaping the divergent CT alteration patterns in adolescent and adult psychiatric disorders. Combined with the publicly available neurotransmitter, transcriptome, and mitochondria atlases, we further examined the neurochemical basis, genetic architecture, and molecular energetic landscape related to the divergence. Our findings may advance the understanding of the biological factors contributing to the differential cortical abnormalities between psychiatric adolescents and adults, which could inform future personalized precision medicine in psychiatry.

Author contributions

SC and CT performed the data processing, statistical analyses, and visualization. SX, XH, and KL contributed to public data acquisition. SC and JZ drafted and revised the manuscript. JZ and YY acquired funding, provided guidance, and reviewed the manuscript. All authors read and approved the final manuscript.

Funding

The study was supported by the National Natural Science Foundation of China (grant numbers: 82471952 and 82371928), the Anhui Provincial Natural Science Foundation (grant number: 2308085MH277), and the Scientific Research Foundation of Anhui Medical University (grant number: 2022xkj143).

Data availability

All data analyzed in this paper were obtained from open-access sources. Disease-related Cohen’s d maps derived from the ENIGMA meta-analyses were accessed via the ENIGMA Toolbox (https://enigma-toolbox.readthedocs.io/en/latest/). Through the toolbox, we also accessed normative structural connectome data from a Human Connectome Project young adult sample (HCP; http://www.humanconnectome.org/) and gene expression data from the Allen Human Brain Atlas (https://human.brain-map.org/) as accessible through abagen.

Code availability

The code used to perform the analyses in this study is available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Ethics approval

The present study only utilized publicly available or published summary-level data and did not involve individual-level data from participants. Therefore, additional ethics approval was not required for the present secondary analysis. For the original ENIGMA studies, individual cohort investigators obtained approval from their local institutional review boards or ethics committees, and informed consent was obtained from all participants or their legal guardians.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41398-026-04247-4.