Introduction
Adolescent depression is increasing worldwide particularly in recent years (Lu et al., Reference Lu, Lin and Su2024, Zhou et al., Reference Zhou, Chen, Gu, Xiang, Hajcak and Wang2024), leading to impact on patients’ development, society, and the well-being of future generations. Depression typically onsets around the age of 10 years, followed by a considerable increase during adolescence (Uhlhaas et al., Reference Uhlhaas, Davey, Mehta, Shah, Torous, Allen, Avenevoli, Bella-Awusah, Chanen, Chen, Correll, Do, Fisher, Frangou, Hickie, Keshavan, Konrad, Lee, Liu and Wood2023). Despite of different treatments for depression, the outcomes have been overall unsatisfying (Murphy et al., Reference Murphy, Capitao, Giles, Cowen, Stringaris and Harmer2021, Walter et al., Reference Walter, Abright, Bukstein, Diamond, Keable, Ripperger-Suhler and Rockhill2023), potentially due to neurobiological complexity and heterogeneity. Brain network perspectives are promising to reveal the physiopathological mechanisms of depression and guide clinical treatment (Malgaroli et al., Reference Malgaroli, Calderon and Bonanno2021). Patients with major depressive disorder (MDD) have been found to exhibit extensive alterations in brain networks such as the default mode, attention, and reward networks (Lynch et al., Reference Lynch, Elbau, Ng, Ayaz, Zhu, Wolk, Manfredi, Johnson, Chang, Chou, Summerville, Ho, Lueckel, Bukhari, Buchanan, Victoria, Solomonov, Goldwaser, Moia and Liston2024, Tse et al., Reference Tse, Ratheesh, Tian, Connolly, Davey, Ganesan, Gotlib, Harrison, Han, Ho, Jamieson, Kirshenbaum, Liu, Ma, Ojha, Qiu, Sacchet, Schmaal, Simmons and Zalesky2024). To facilitate early identification and intervention, it is necessary to identify early-stage brain targets of depressed adolescents and deepen our understanding of mechanistic progression of brain network disruption.
Adolescent depression, as a brain network disorder, involves multiple brain structural abnormalities, especially within the cognitive control, reward, and self-referential processing networks. The study conducted by the Enhancing NeuroImaging Genetics through Meta-Analysis (ENIGMA) consortium found adolescents with MDD showed significant atrophies of cortical surface area in the ventromedial prefrontal cortex (vmPFC), frontoparietal, primary motor, and visual areas (Schmaal et al., Reference Schmaal, Hibar, Samann, Hall, Baune, Jahanshad, Cheung, van Erp, Bos, Ikram, Vernooij, Niessen, Tiemeier, Hofman, Wittfeld, Grabe, Janowitz, Bulow, Selonke and Veltman2017). Another large-scale sample study, the Adolescent Brain and Cognitive Development study, reported similar cortical atrophies in the frontal and temporal lobes, in addition to abnormalities in white matter integrity (Shen et al., Reference Shen, MacSweeney, Chan, Barbu, Adams, Lawrie, Romaniuk, McIntosh and Whalley2021). Reductions in subcortical volumes such as the amygdala, striatal, and hippocampal were also reported, which were involved in negative emotion processing and memory processing (Long et al., Reference Long, Li, Wang, Cao, Wu, Roberts, Gong, Kemp and Jia2023, Zhang et al., Reference Zhang, Liu, Yang, Zhang, He, Xu, Jin, Gao, Yao, Yu, Hommel, Zhu, Wang and Zhang2024). These findings indicated that adolescent depression involved disruptions across multiple brain regions. Some researchers have further attempted to examine brain changes in adolescent depression as the disease progresses. For example, Zhang et al. (Reference Zhang, Liu, Yang, Zhang, He, Xu, Jin, Gao, Yao, Yu, Hommel, Zhu, Wang and Zhang2024) found depressed adolescents with mean illness durations at 1.3 years showed volumetric reductions of hippocampus that were limited within the core cornu ammonis regions, while longer durations of depressive episodes were associated with more pronounced and extended hippocampal changes (Zhang et al., Reference Zhang, Liu, Yang, Zhang, He, Xu, Jin, Gao, Yao, Yu, Hommel, Zhu, Wang and Zhang2024). However, these studies did not explore the network disruptions triggered by brain abnormalities in the early stages of the disease course (Redlich et al., Reference Redlich, Opel, Burger, Dohm, Grotegerd, Forster, Zaremba, Meinert, Repple, Enneking, Leehr, Bohnlein, Winters, Frobose, Thrun, Emtmann, Heindel, Kugel, Arolt and Dannlowski2018, Tannous et al., Reference Tannous, Amaral-Silva, Cao, Wu, Zunta-Soares, Kazimi, Zeni, Mwangi and Soares2018, Wang et al., Reference Wang, Li, Wang, Hommel, Xia, Qiu, Fu and Zhou2023b, Zhang et al., Reference Zhang, Liu, Yang, Zhang, He, Xu, Jin, Gao, Yao, Yu, Hommel, Zhu, Wang and Zhang2024). In other words, they have not addressed whether and how abnormalities in an early region can lead to dysfunction in broader brain networks. Thus, investigating the initiating brain abnormalities at the early stage of adolescent depression and the subsequently triggered brain network abnormalities is necessary to understand the neurobiological pathways of depression.
Granger causality (GC) analysis is a commonly used approach to elucidate the potential disrupted networks/pathways by which initial structural alterations extend to wider regions as the illness progresses (Zhang et al., Reference Zhang, Liao, Xu, Wei, Zhou, Sun, Yang, Mantini, Ji and Lu2017). Via constructing causal structural covariance network, GC analysis estimates the interregional casual/directional relationships with the structural alterations proceed. It has been used to identify the primary hub of gray matter abnormalities in epilepsy, schizophrenia, and MDD, and demonstrated good reliability and stability (Han et al., Reference Han, Zheng, Li, Liu, Wang, Jiang, Wen, Zhou, Wei, Pang, Li, Zhang, Chen and Cheng2023a, Jiang et al., Reference Jiang, Luo, Li, Duan, He, Chen, Yang, Gong, Chang, Woelfer, Biswal and Yao2018, Zhang et al., Reference Zhang, Liao, Xu, Wei, Zhou, Sun, Yang, Mantini, Ji and Lu2017). For depression specifically, researchers have focused on adult population and applied GC analysis to examine MDD disease-stage-specific alterations by defining stage groups using illness duration (Han et al., Reference Han, Zheng, Li, Liu, Wang, Jiang, Wen, Zhou, Wei, Pang, Li, Zhang, Chen and Cheng2023a, Li et al., Reference Li, Wang, Teng, Jiao, Song, Tan, Xiao, Zhang and Zhong2019, Lu et al., Reference Lu, Cui, Chen, He, Sheng, Tang, Yang, Luo, Yu, Chen, Li, Deng, Zeng and Chen2023). Results showed gray matter atrophies in the insula and hippocampus, along with an increase in the vmPFC volume in the early stage; these structural changes might be the markers of early brain abnormalities of adult onset MDD. As the illness duration increases, these targets may trigger other regional disruptions in a chain manner, further impacting the cognitive, reward, and default networks (Yun and Kim, Reference Yun and Kim2021). On the other hand, limited research has investigated the progressive structural atrophy pattern in adolescent depression from the perspective of illness duration. The causal structural networks underlying brain volume atrophies in adolescents with MDD remain unclear.
This study aimed to estimate the initial gray matter alterations in adolescent depression and further investigate the directional influences of abnormalities within brain networks. 3D T1-weighted images were collected for first-episode adolescent patients with MDD and age- and sex-matched healthy controls. We first identified the disease-stage specific gray matter abnormalities using voxel-based morphometry. A GC analysis was then conducted to construct casual structural covariance network to estimate the causal relationship of structural alterations (Jiang et al., Reference Jiang, Luo, Li, Duan, He, Chen, Yang, Gong, Chang, Woelfer, Biswal and Yao2018, Xu et al., Reference Xu, Luo, Peng, Guo, Zhong, Liu, Weng, Ou, Yan, Wang, Zeng, Zhang, Hu and Liu2023). Finally, we identified pathways and conducted functional decoding for the potentially disrupted structural networks to explore their functional characterization. We hypothesized that adolescents with a short illness duration of MDD would exhibit gray matter atrophies in few localized regions, and those with longer durations would show more extensive alterations. We expected that, as the disease progressed, these early and localized atrophies would gradually extend to wider regions involving the default mode, frontoparietal, and reward networks.
Methods
Participants
We included a total of 162 adolescent participants from the ongoing Shandong Adolescent Neuroimaging of Depression (SAND) project (Wang et al., Reference Wang, He, Zhu, Hu, Yao, Hommel, Beste, Liu, Yang and Zhang2022, Zhang et al., Reference Zhang, Liu, Yang, Zhang, He, Xu, Jin, Gao, Yao, Yu, Hommel, Zhu, Wang and Zhang2024). Among them, 80 were diagnosed with MDD by two clinical psychiatrists from the Shandong Mental Health Center, based on the DSM-5 criteria. All MDD patients met the following inclusion criteria: (a) age ranging from 10 to 19 years; (b) experiencing their first episode of depression; (c) no presence of comorbidity with other psychiatric illnesses such as schizophrenia or autism; and (d) no substance abuse or mental retardation. The severity of depression was assessed using the Children’s Depression Inventory (CDI). Their anxiety and suicide risk were also assessed using Multidimensional Anxiety Scale for Children (MASC) and Nurses’ Global Assessment of Suicide Risk (NGASR). Another 82 participants were healthy controls. They were matched to MDD patients on age and sex, and were recruited locally through advertisements. Healthy controls had no current/historic psychiatric illness or family history of psychiatric disorders.
All these participants were eligible for MRI scans and had no history of serious physical or neurological diseases. Written informed consent was obtained from all participants and their parents. This study was approved by the local ethics committee of both Shandong Mental Health Center and Shandong Normal University.
Structural image acquisition
All 3D T1-weighted data were acquired from a 3T MRI scanner (SIMENS Verio) at the Qilu Hospital of Shandong University using a magnetization prepared rapid gradient-echo (MPRAGE) sequence. The acquisition parameters were as follows: repetition time = 2400 ms, echo time = 2.19 ms, inversion time = 1000 ms, flip angle = 8°, matrix = 256 × 256 mm, voxel = 0.9 × 0.9 × 0.9 mm3, and 208 sagittal slices.
Image preprocessing
We visually examined all structural images before preprocessing, to check for any brain malformations, motion artifacts, and failed whole-brain coverage. To prepare all T1 images for the voxel-based morphometry analysis, we used Computational Anatomy Toolbox (CAT12, http://dbm.neuro.uni-jena.de/cat12/) with the following sequential steps: (1) checking the artifacts and adjusting origins to the anterior commissure, (2) normalizing to MNI template and resampling at 1.5 mm, (3) segmenting into gray matter, white matter, and cerebrospinal fluid, (4) modulating the gray matter maps, and (5) smoothing the gray matter maps with an 8-mm full width at half-maximum Gaussian kernel. Additionally, the total intracranial volume (TIV) was calculated for each participant for subsequent statistical analysis. After preprocessing, any images with poor quality were excluded based on the CAT12 reports (Gaser et al., Reference Gaser, Dahnke, Thompson, Kurth, Luders and Neuroimaging2024).
Voxel-based morphometry analysis
Two-sample t tests were performed to reveal the overall alterations in gray matter volume by comparing the patients to healthy controls. To acquire the stage-specific gray matter volume alterations, we utilized two grouping strategies. The primary strategy divided the patients into two stages based on the median of illness duration (median: 15.67 months; range: 0.13–85.27 months; Stage 1: 0 to ≤15.67 months, N = 41; Stage 2: >15.67 to 86 months, N = 39; Supplementary Table S1) (Chen et al., Reference Chen, Cui, Fan, Guo, Tang, Sheng, Lei, Li, Lu, He, Yang, Hu, Deng and Chen2020, Zhang et al., Reference Zhang, Liu, Yang, Zhang, He, Xu, Jin, Gao, Yao, Yu, Hommel, Zhu, Wang and Zhang2024). The second strategy aimed to validate the primary one, which categorized the patients into three groups according to cut-offs in a previous study (Han et al., Reference Han, Zheng, Li, Liu, Wang, Jiang, Wen, Zhou, Wei, Pang, Li, Zhang, Chen and Cheng2023a) (Stage 1: <12 months, N = 27; Stage 2: 12–24 months, N = 33; Stage 3: ≥24 months, N = 20; Supplementary Table S2). For both strategies, patients in each stage were separately compared to healthy controls using the two-sample t test to obtain stage-specific volume alterations. Age, sex, and TIV were included as covariates in these analyses. The false discovery rate (FDR) method was used to correct multiple comparisons and the significance was set at p < 0.05.
To test the associations between gray matter alterations and clinical characteristics, we calculated partial correlations between mean gray matter volume of significant areas emerging from earlier analyses and scores on depressive severity, anxiety level, and suicide risk. Covariates included age, sex, and TIV. Results were corrected by the FDR method (p < 0.05). The mean gray matter volume of the significant regions from the overall comparison and stage-specific analyses were extracted with an 8-mm spherical radius.
Causal structural covariance network analysis
To explore the potential causal mechanisms of brain atrophy, we used seed-based causal structural covariance networks analysis based on previous research (Zhang et al., Reference Zhang, Liao, Xu, Wei, Zhou, Sun, Yang, Mantini, Ji and Lu2017). First, we arranged gray matter data of all patients in order of disease duration from shortest to longest. This sequencing provided time-series information for cross-sectional data that described the progressive characteristics of the patients. Next, causal structural covariance network analyses were performed to explore the pattern of progressive structural changes as disease duration increased. Significant regions observed in the comparisons between Stage 1 MDD patients and healthy controls were selected as seed regions (see results, i.e. vmPFC, dACC, MCC, insula, precentral, and superior frontal gyrus). We performed signed path coefficient GC analysis using the REST software (http://www.restfmri.net) and calculated the directional influences between seed regions and the whole brain at the voxel level. This approach integrated GC analysis with structural covariance network techniques to characterize causal effects of regional morphological changes during illness progression. Positive GC values represented alterations that were in the same direction (increase/decrease) as those in the seed region. We focused on positive GC values to reveal structural atrophy in other regions, as prior studies suggested (Xu et al., Reference Xu, Luo, Peng, Guo, Zhong, Liu, Weng, Ou, Yan, Wang, Zeng, Zhang, Hu and Liu2023). The positive GC maps were converted to Z scores, and corrected by FWE (family-wise error, p < 0.01, i.e. z > 5.8 and GC value >0.10). Same covariates as in the previous analyses were included.
From the positive GC maps, we identified nodes that appeared in at least four of the six maps. The nodes were extracted by calculating the mean gray matter volume for coordinates with a spherical radius of 8 mm. To further investigate the causal effects among the nodes, we performed a bivariate signed path coefficient GC analysis to construct a node-wise causal network. Similar to the voxel-wise analysis, a positive GC value >0.10 was set as the threshold. The binary out-degree and in-degree values were computed for each node to determine the causal target or causal source of the progressive structural changes. We also calculated partial correlations between each node and symptom severity.
As a supplement, we repeated the GC analysis on CDI, MASC, and NGASR scores to examine the causal relationship between symptom severity and morphological changes (Chen et al., Reference Chen, Cui, Fan, Guo, Tang, Sheng, Lei, Li, Lu, He, Yang, Hu, Deng and Chen2020). The CDI, MASC, and NGASR scores were ranked for the causal structural covariance network analysis and same analyses were performed at the voxel level. We also repeated the GC analysis using age sequencing to examine the potential influence of ongoing brain development on our findings.
Pathways identification and functional decoding
To explore the potential pathways between the initiating brain source (the region with largest out-degree value, i.e. vmPFC; see results) and the mostly affected target (the region with largest in-degree value, i.e. posterior cingulate cortex; see results), we selected connections with the highest GC values between nodes, starting from the vmPFC and ending at the posterior cingulate cortex. The brain regions connected along these paths were those with the strongest directional GC values. Using this approach, we identified six potential pathways with the strongest GC effects in the entire GC brain network. Furthermore, we examined the functional roles of overall potential pathways and each single pathway by referring to the BrainMap database. Functional roles were identified by examining the taxonomic labels, which consisted of five behavioral domains and 60 subdomains. Bonferroni correction was performed with a p < 0.05 (Lancaster et al., Reference Lancaster, Laird, Eickhoff, Martinez, Fox and Fox2012).
Results
Demographic information and clinical characteristics
There were no significant differences in age, sex, and TIV between MDD patients and healthy controls. MDD patients scored higher in assessments of depression, anxiety, and suicide risk (ps < 0.001; Table 1).
Table 1. Demographic and clinical characteristics of adolescent participants
Note: M (SD), mean (standard deviation); n (%), number (percentage); MDD, major depressive disorder; HCs, healthy controls; TIV, total intracranial volume; CDI, Children’s Depression Inventory; MASC, Multidimensional Anxiety Scale for Children; NGASR, the Nurses’ Global Assessment of Suicide Risk scale.
* p < 0.05.
As for Stages 1 and 2 MDD groups, neither of them differed from healthy controls in terms of age, sex, and TIV, while both had higher scores of depression, anxiety, and suicide risk (Supplementary Tables S1 and S2).
Gray matter volume alterations
Compared to healthy controls, the MDD group exhibited significant gray matter atrophies in the vmPFC, insula, dorsal ACC, subgenual ACC, MCC, posterior cingulate cortex, thalamus, superior frontal gyrus, hippocampus, inferior parietal lobule, precentral gyrus, amygdala, superior temporal gyrus, middle temporal gyrus, middle occipital gyrus, and cerebellum crus2 (Figure 1 and Supplementary Table S3). Partial correlation analyses between these regions and clinical characteristics including depression, anxiety, and suicide risk found no significant associations (Supplementary Table S4).
Figure 1. Overall and stage-specific gray matter alterations in adolescent patients with depression.
In terms of disease stages, Stage 1 MDD patients showed decreased volumes in the vmPFC, dACC, MCC, right insula, right precentral, and right superior frontal gyrus when compared to controls; Stage 2 MDD patients had more severe alterations in the vmPFC, dACC, MCC, right insula, and right precentral, in addition to atrophies in the subgenual ACC, middle frontal gyrus, hippocampus, amygdala, inferior parietal lobule, superior/middle temporal gyrus, middle occipital gyrus, and cerebellum crus2 (Figure 1 and Supplementary Table S5). Results from the second grouping strategy also showed similar trends toward gray matter atrophy in relation to increase in disease duration (Supplementary Table S6 and Supplementary Figure S1). Partial correlation analyses between these changes and clinical characteristics found no significant association (Supplementary Tables S7 and S8).
Causal effects of brain alterations
Via GC analysis for seed regions at the voxel level, we constructed causal structural covariance networks for the vmPFC, dACC, MCC, right insula, right precentral, and right superior frontal gyrus, respectively (Figure 2, Supplementary Tables S9–S14). The six seed regions showed positive directional effects on widespread cortical areas, primarily within the default and frontoparietal networks, as well as subcortical volumes including caudate and putamen, which were core structures in the reward circuit. GC maps from analysis for severity of symptoms showed similar trends (Supplementary Figures S2–S4). The results using age sequencing (Supplementary Figure S5) differed from our original findings on illness duration, indicating the specificity of MDD-related brain changes.
Figure 2. The causal effects of early gray matter atrophies in adolescent patients with depression.
We identified 32 nodes from the six positive GC maps. A directional causal network for these nodes was constructed to show the causal connectivity (Figure 3a). The vmPFC was the highest out-degree node, projecting to most of the 32 nodes. The posterior cingulate cortex was the highest in-degree node and received most projections from other regions. In addition, the dACC, MCC, and right superior frontal gyrus nodes had high out-degree values and did not receive causal projections from other regions (Figure 3b).
Figure 3. Bivariate signed path coefficient granger causality analysis shows directional influences among the brain regions. (a) A directional causal network for 32 nodes, including six early regions of alteration and 26 nodes exhibiting causal connectivity to the early alterations. (b) The binary out-degree and in-degree values of each node. Specifically, the binary out-degree value of a node represents the total number of paths projected to other nodes, while the binary in-degree value represents the total number of paths projected to that node. The abbreviations for brain regions are listed in Supplementary Table S15.
In the partial correlation analyses between nodes of GC maps and clinical characteristics, we found smaller left putamen and bilateral caudate were correlated with higher depression scores (p < 0.05; Supplementary Table S15 and Supplementary Figure S6).
Pathways identification and functional decoding
We identified six potential pathways from the vmPFC to the posterior cingulate cortex. The overall pathways primarily involved the right postcentral, bilateral posterior middle temporal gyrus, right dorsal lateral prefrontal cortex, left middle frontal gyrus, and right inferior parietal lobule (Supplementary Table S16). We performed functional decoding for these regions. They were primarily associated with cognition (attention, memory, language, reasoning, and social cognition), action (preparation, inhibition, execution, and observation), perception (pain, vision, and somesthesis), and emotion (anger and fear) (Figure 4). We subsequently conducted separate functional decoding for each single pathway. Our results showed that pathways II, III, and IV primarily contributed to the reported associations with cognition, action, and perception (Supplementary Table S17).
Figure 4. The pathways from the vmPFC to the posterior cingulate cortex in adolescent depression and functional decoding.
Discussion
This study focused on progressive gray matter atrophy in adolescents with MDD. First, a voxel-based morphometry analysis indicated that patients with shorter illness duration showed gray matter atrophy in localized brain regions including the vmPFC, dACC, MCC, right insula, right precentral, and right superior frontal gyrus. With a prolonged course of the illness, gray matter atrophy extended to more widespread areas. Subsequently, our GC analysis showed that the early abnormalities may causally induce disruptions in regions within the default mode, frontoparietal, and reward networks. Notably, the vmPFC demonstrated the highest out-degree value, possibly representing the initial source of brain abnormality in adolescent depression. The posterior cingulate cortex demonstrated the highest in-degree value, possibly representing the mostly affected area. Finally, we identified distinct structural pathways from the vmPFC to the posterior cingulate cortex, which suggested that cognitive function may be most susceptible, as depression progressed and disrupted brain pathways.
In line with previous findings, we found that MDD patients exhibited widespread gray matter atrophy compared to healthy controls (Chen et al., Reference Chen, Liu, Chen, Liu, Tang, Tian, Wang, Lu and Zhou2024, Shen et al., Reference Shen, MacSweeney, Chan, Barbu, Adams, Lawrie, Romaniuk, McIntosh and Whalley2021). Compared to patients with longer illness duration, those with shorter duration of illness (≤15.67 months) displayed localized gray matter atrophy in the vmPFC, dACC, MCC, insula, right precentral, and right superior frontal gyrus. Alterations in these regions have also been reported in previous studies (Long et al., Reference Long, Li, Wang, Cao, Wu, Roberts, Gong, Kemp and Jia2023, Pannekoek et al., Reference Pannekoek, van der Werff, van den Bulk, van Lang, Rombouts, van Buchem, Vermeiren and van der Wee2014, Vulser et al., Reference Vulser, Lemaitre, Artiges, Miranda, Penttila, Struve, Fadai, Kappel, Grimmer, Goodman, Stringaris, Poustka, Conrod, Frouin, Banaschewski, Barker, Bokde, Bromberg, Buchel and Paillère-Martinot2015, Webb et al., Reference Webb, Weber, Mundy and Killgore2014) and were considered as origins of structural alterations (Han et al., Reference Han, Zheng, Li, Liu, Wang, Jiang, Wen, Zhou, Wei, Pang, Li, Zhang, Chen and Cheng2023a, Lu et al., Reference Lu, Cui, Chen, He, Sheng, Tang, Yang, Luo, Yu, Chen, Li, Deng, Zeng and Chen2023). Importantly, our findings further revealed that during a longer illness duration broader atrophy may arise in the subgenual ACC, middle frontal gyrus, hippocampus, amygdala, inferior parietal lobule, superior temporal gyrus, and cerebellum crus II (Gray et al., Reference Gray, Muller, Eickhoff and Fox2020, Long et al., Reference Long, Qin, Pan, Fan and Li2024, Wang et al., Reference Wang, Hu, Yan, Li, Wu, Qiu and Zhu2023a). Previous research has also indicated that patients with a longer duration of depression exhibited more atrophied regions (Lu et al., Reference Lu, Cui, Chen, He, Sheng, Tang, Yang, Luo, Yu, Chen, Li, Deng, Zeng and Chen2023). A prolonged illness duration was significantly correlated with structural atrophies in multiple brain regions (Lemke et al., Reference Lemke, Romankiewicz, Forster, Meinert, Waltemate, Fingas, Grotegerd, Redlich, Dohm, Leehr, Thiel, Enneking, Brosch, Meller, Ringwald, Schmitt, Stein, Steinstrater, Bauer and Dannlowski2022). These results suggested that gray matter atrophy may expand from localized regions to more extensive areas as the duration of illness increased, representing progressive brain alterations in adolescent depression.
Our GC analysis revealed that early gray matter atrophy exerted directional influences on broader brain networks. All the six initial atrophic regions exhibited positive causal effects on the default mode and frontoparietal networks, and subcortical regions within the reward circuits. These dysfunctional networks were closely related to emotional dysregulation, cognitive impairments, and self-processing in depression (Berman et al., Reference Berman, Peltier, Nee, Kross, Deldin and Jonides2011, Jamieson et al., Reference Jamieson, Harrison, Razi and Davey2022, Pan et al., Reference Pan, Xu, Zhou, Chen, Wei, Lu, Shang, Wang and Huang2020, Piguet et al., Reference Piguet, Cojan, Sterpenich, Desseilles, Bertschy and Vuilleumier2016). Our findings suggested that early atrophy regions may influence the cognitive and emotional functions through shared network mechanisms (Davey and Harrison, Reference Davey and Harrison2022, Tse et al., Reference Tse, Ratheesh, Tian, Connolly, Davey, Ganesan, Gotlib, Harrison, Han, Ho, Jamieson, Kirshenbaum, Liu, Ma, Ojha, Qiu, Sacchet, Schmaal, Simmons and Zalesky2024). Additionally, the subcortical regions within the reward circuits were closely linked to reward-related emotional and cognitive processing (Luking et al., Reference Luking, Pagliaccio, Luby and Barch2016, Lynch et al., Reference Lynch, Elbau, Ng, Ayaz, Zhu, Wolk, Manfredi, Johnson, Chang, Chou, Summerville, Ho, Lueckel, Bukhari, Buchanan, Victoria, Solomonov, Goldwaser, Moia and Liston2024, Rappaport et al., Reference Rappaport, Kandala, Luby and Barch2020). Taken together, the causal targets induced by initial abnormalities were not confined to a single network but spread across multiple brain networks.
Our results further suggested that the atrophy in the vmPFC may occur in a temporally leading position, where gray matter atrophy occurred first and triggered widespread disruptions in brain networks. The vmPFC may represent an early biomarker and a treatment target for adolescent depression. Applying transcranial direct current stimulation to the vmPFC has been shown to effectively enhance emotional regulation in depressed patients (Konicar et al., Reference Konicar, Prillinger, Klobl, Lanzenberger, Antal and Plener2022). Improvement of MDD following repetitive transcranial magnetic stimulation was also associated with increased baseline resting state connectivity of the vmPFC (Long et al., Reference Long, Du, Zhao, Wu, Zheng and Lei2020). Taken together with our findings, these evidence supported the vmPFC as a promising target for noninvasive neuromodulation therapies in depressed adolescents. Utilizing an H7 deep coil for precisely targeted deep transcranial magnetic stimulation with the vmPFC as the target may offer an effective treatment approach for adolescent depression (Di Passa et al., Reference Di Passa, Prokop-Millar, Yaya, Dabir, McIntyre-Wood, Fein, MacKillop, MacKillop and Duarte2024, Regenold et al., Reference Regenold, Deng and Lisanby2022), particularly for those first-episode and at early-stage cases.
We identified six pathways from the vmPFC to the posterior cingulate cortex. These pathways further supported the complexity and heterogeneity of the underlying mechanisms in adolescent depression (Zhang et al., Reference Zhang, Peng, Sweeney, Jia and Gong2018). Each pathway represents interactions among different brain regions and may involve distinct manifestations related to emotional regulation, cognitive function, and perceptual processing, according to the functional decoding results (Herrman et al., Reference Herrman, Patel, Kieling, Berk, Buchweitz, Cuijpers, Furukawa, Kessler, Kohrt, Maj, McGorry, Reynolds, Weissman, Chibanda, Dowrick, Howard, Hoven, Knapp, Mayberg and Wolpert2022). Damage in these pathways may underlie the diversity of clinical symptomatology in patients with depression (Ho et al., Reference Ho, Connolly, Henje Blom, LeWinn, Strigo, Paulus, Frank, Max, Wu, Chan, Tapert, Simmons and Yang2015, Luo et al., Reference Luo, Li, Hu, Lu, Wang, Lan, Mai, Liu, Zhang, Chen, You, Zeng, Chen, Liang, Chen, Zhou and Ning2024). Future research should explore the associations between clinical symptoms and specific pathways from the perspective of individualized brain networks.
Several limitations should be mentioned. First, the causal structural network was based on cross-sectional data, which could not establish true causality and directly reflect the real temporal sequence of illness progression. Longitudinal studies are needed to elucidate the causal effect of structural alterations. Second, only group-level causal structural networks can be obtained due to methodology limitations. Future studies could construct individual-level structural networks to more precisely investigate the progressive brain changes in adolescent depression (Han et al., Reference Han, Zheng, Li, Zhou, Jiang, Fang, Wei, Pang, Li, Zhang, Chen and Cheng2023b). Finally, our study only utilized illness duration and severity indicators of representative symptoms (depression, anxiety, and suicide risk). Future research should investigate more other clinical characteristics and manifestations to provide a deeper understanding of the neurobiological progression mechanisms in depression and guide treatment strategies.
In summary, this study found that depressed adolescents exhibited gray matter atrophies, which were initially localized and gradually extended to regions within the default mode, frontoparietal networks, and reward circuits as the illness progressed. Our study demonstrated that the progression of adolescent depression was associated with the expansion of gray matter atrophies, with the vmPFC potentially severing as the initial source of brain abnormality. These findings could deepen our understanding of the progression of structural atrophy in adolescent depression and underscore the importance of early diagnosis and intervention.
Supplementary material
The supplementary material for this articlecan be found at http://doi.org/10.1017/S0033291725101542.
Data availability statement
The data that support the findings of this study are available on request from the corresponding author, Kangcheng Wang.
Author contribution
Jiahui Chen: conceptualization, study design, data collection, formal analysis, and writing. Xinjuan Jin, Junqi Gao, and Ying Yang: data collection and study design. Yihao Zhang, Yixin Zhang, Changlin Bai, Feiyu Xu, Yuan Yao, and Wenxin Zhang: data collection. Xingxing Zhu: study design and writing. Kangcheng Wang: conceptualization, study design, methodology, formal analysis, and writing.
Funding statement
This research was supported by the National Natural Science Foundation of China (32000760), the China Postdoctoral Science Foundation Funded Project (2019M662433, 2023T160397), the Postdoctoral Innovation Project in Shandong Province (239735), the Youth Innovation Team in Universities of Shandong Province (2022KJ252), and the Shandong Provincial Natural Science Foundation (ZR2024QH085).
Competing interests
The authors declare none.
Open access
