Introduction
Major depressive disorder (MDD) is prevalent psychiatric disorder that causes substantial distress and socioeconomic burden (Thalamuthu, Mills, Berger et al., Reference Thalamuthu, Mills, Berger, Minnerup, Grotegerd, Dannlowski and Baune2022). Despite extensive research spanning genetics, immunology, inflammation, and neuroimaging, the fundamental pathophysiological mechanisms of MDD remain unclear. Identifying reliable biomarkers for the diagnosis and treatment of MDD is a critical and ongoing challenge in the field (Drysdale, Grosenick, Downar et al., Reference Drysdale, Grosenick, Downar, Dunlop, Mansouri, Meng and Liston2017; McCarron, Shapiro, Rawles, & Luo, Reference McCarron, Shapiro, Rawles and Luo2021).
Functional magnetic resonance imaging (fMRI), a noninvasive method for assessing brain structure and function, is a powerful tool for identifying psychiatric biomarkers and clarifying pathophysiological mechanisms (Chen et al., Reference Chen, Wang, Gong, Qi, Fu, Tang and Wang2022; Guo et al., Reference Guo, Tang, Xiao, Yan, Sun, Yang and Wang2024; Yang et al., Reference Yang, Xiao, Su, Gong, Qi, Chen and Wang2024). Recent meta-analyses of resting-state brain activity in unmedicated MDD patients compared to healthy controls (HCs) have consistently identified abnormal functional activity in brain regions such as the superior frontal gyrus (SFG), striatum, cerebellum, precuneus, occipital cortex, postcentral gyrus, and supramarginal gyrus. These findings suggest the potential of these regions as neuroimaging biomarkers for MDD diagnosis (Hao et al., Reference Hao, Chen, Mao, Zhong and Dai2019; Iwabuchi et al., Reference Iwabuchi, Krishnadas, Li, Auer, Radua and Palaniyappan2015; Wang, Zhao, Hu et al., Reference Wang, Zhao, Hu, Huang, Kuang, Lui and Gong2017; Yuan, Yu, Yu et al., Reference Yuan, Yu, Yu, Liang, Huang, He and Xiang2022).
Further research has investigated the changes in resting-state brain function in MDD patients following antidepressant treatment compared to HCs. These studies have revealed that altered brain function post-treatment is predominantly observed in the frontal and temporal lobes, cingulate gyrus, and hippocampus – regions that are crucial for attention, emotion processing, reward, and cognitive control (Guo, Liu, Xue et al., Reference Guo, Liu, Xue, Yu, Ma, Tan and Zhao2011; Wu, Li, Kuang et al., Reference Wu, Li, Kuang, Zhang, Lui, Huang and Gong2011; Wang, Li et al. Reference Wang, Li, Zhang, Zeng, Dai, Su and Si2014; Guo et al., Reference Guo, Cui, Liu, Chen, Xie, Wu and Zhao2018). However, findings across studies remain inconsistent, reflecting substantial heterogeneity in methodology and sample characteristics. These inconsistencies may be partly attributed to differences in the types of antidepressants used across studies (e.g. selective serotonin reuptake inhibitors, serotonin-norepinephrine reuptake inhibitors, or noradrenergic and specific serotonergic antidepressants) (Harmer, Duman, & Cowen, Reference Harmer, Duman and Cowen2017). Each class acts on distinct neurotransmitter systems and may therefore induce differential patterns of neural activation (Harmer et al., Reference Harmer, Duman and Cowen2017). Furthermore, even within the same pharmacological class, variability in dosage, treatment duration, or clinical response can lead to divergent activation results (Harmer et al., Reference Harmer, Duman and Cowen2017). For instance, Wang et al. reported a decrease in the amplitude of low-frequency fluctuation (ALFF) in the right middle temporal gyrus following treatment (Wang, Kuang et al. (Reference Wang, Kuang, Xu, Lei and Yang2014)), while Liu et al. reported an increase (Liu, Ma, Song et al., Reference Liu, Ma, Song, Tang, Jing, Zhang and Wang2015). Additionally, Wang et al. noted elevated regional homogeneity (ReHo) in the cerebellum of treated MDD patients compared to HCs (Wang, Li et al. (Reference Wang, Li, Zhang, Zeng, Dai, Su and Si2014)). Most of these studies compared medicated MDD patients with healthy controls, leaving it unclear whether the observed changes reflected the illness itself or medication effects. Because longitudinal within-subject fMRI studies remain scarce, the available evidence, including that synthesized in the present study, is largely based on cross-sectional comparisons between medicated and unmedicated patients. Such an approach can provide only indirect insight into treatment-related neural changes. Direct comparisons of brain function before and after antidepressant treatment in MDD patients are relatively rare. For instance, Wang et al. found decreased ReHo values in the left dorsal medial prefrontal gyrus and increased values in the right superior frontal gyrus following escitalopram treatment (Wang, Li et al. Reference Wang, Li, Zhang, Zeng, Dai, Su and Si2014). In a subgroup analysis of their meta-analysis, Zhou et al. found reduced fractional ALFF (fALFF) in the left middle occipital gyrus (MOG) and left posterior central gyrus in medicated MDD patients compared to their unmedicated counterparts; however, this comparison was based on only five studies (Zhou, Hu, Lu et al., Reference Zhou, Hu, Lu, Zhang, Chen, Gong and Huang2017). The limited number and variability of studies directly comparing brain function in MDD patients before and after treatment underscore the need for more comprehensive analyses.
Both functional activation and connectivity analyses can reveal the neural effects of antidepressant treatment, but they capture different aspects of brain function (Biswal & Uddin, Reference Biswal and Uddin2025). Connectivity describes large-scale network coordination, whereas activation indices such as ALFF and ReHo quantify the intensity and synchrony of intrinsic local neural activity (Biswal & Uddin, Reference Biswal and Uddin2025). Because antidepressant pharmacodynamics primarily modulate neurochemical processes at the regional level (Yamamura, Okamoto, Okada et al., Reference Yamamura, Okamoto, Okada, Takaishi, Takamura, Mantani and Yamawaki2016), the present study focused on activation-based alterations to characterize direct neurophysiological changes associated with medication use. To bridge these knowledge gaps, we conducted a coordinate-based meta-analysis of resting-state fMRI studies to compare brain function between medicated and unmedicated MDD patients. Our objectives were to explore brain function changes potentially influenced by antidepressant medications, identify possible therapeutic targets and neurophysiological mechanisms, and search for diagnostic biomarkers of MDD through a conjunction analysis that included comparisons between unmedicated/medicated MDD patients and HCs. To further characterize shared and distinct alterations, we implemented a series of bioinformatics strategies to differentiate the identified regions at the behavioral, genetic, and neurotransmitter levels. Figure 1A shows the workflow of all the main analyses that were conducted.
A. Work flow of main analyses in the current study. B. Diagram of the preferred reporting items for systematic review and meta-analysis (PRISMA). Abbreviations: ROI, region of interest; FC, functional connectivity; ReHo, regional homogeneity; ALFF, amplitude of low-frequency fluctuation; MDD, major depressive disorder, SDM-PSI, Seed-based d-mapping software; BAT, Brain Annotation Toolbox.
Methods
Literature search and study selection
On March 19, 2024, we conducted a comprehensive literature search across the PubMed, Embase, and Web of Science to identify studies examining resting-state fMRI in both medicated and unmedicated MDD patients. The search was restricted to English-language publications, with no limitations on geographical origin or article type. To ensure comprehensive coverage, we also screened the reference lists of retrieved articles for additional relevant studies. The search strategy incorporated the following keywords: (‘MDD’ OR ‘major depressive disorder’ OR ‘unipolar depression’ OR ‘depressive disorder’ OR ‘depression’ OR ‘depressed’) AND (‘amplitude of low frequency fluctuation’ OR ‘ALFF’ OR ‘low frequency fluctuation’ OR ‘LFF’ OR ‘amplitude of low frequency oscillation’ OR ‘LFO’ OR ‘regional homogeneity’ OR ‘ReHo’). Resting-state fMRI studies commonly employ amplitude-based indices such as ALFF/fALFF, which reflect the magnitude of spontaneous low-frequency oscillations, and ReHo, which quantifies local synchrony of BOLD signals. These measures provide complementary perspectives on intrinsic brain activity. Following previous meta-analyses in MDD (Liu et al., Reference Liu, Tu, Zhang, Yang, Liu, Lei and Zhang2021; Yuan et al., Reference Yuan, Yu, Yu, Liang, Huang, He and Xiang2022), we included both ALFF/fALFF and ReHo, as well as dynamic ALFF (dALFF). SDM-PSI integrates these indices using standardized effect sizes (z-scores), allowing for direct comparability across methods. This study is registered with PROSPERO (CRD42024538753) and conducted in accordance with Preferred Reporting Items for Meta-Analysis (PRISMA) guidelines (http://www.prisma-statement.org).
Inclusion criteria encompassed (1) comparative analysis of resting-state fMRI data between unmedicated/medicated MDD patients and HCs; (2) evaluation of brain activity using ALFF, fALFF, ReHo, or dALFF; (3) reporting of results in Montreal Neurological Institute (MNI) or Talairach coordinates; (4) whole-brain analyses; (5) baseline results in longitudinal studies; and (6) selection of studies with the largest sample size in cases of overlapping samples. Studies were excluded if they (1) did not explicitly differentiate between strictly unmedicated or medicated MDD patients; (2) lacked peak coordinate data; or (3) included MDD patients with severe somatic conditions.
Data extraction
Data extraction began on March 28, 2024. Relevant variables extracted included the first author’s last name, year of publication, sample sizes of MDD patients and HCs, number of female participants, mean age and standard deviation, type and scores of depression assessment scales (DAS), first-episode status, MRI scan intensity, analytical method (ALFF/ReHo/fALFF/ dALFF), software used, statistical thresholds, stereotactic coordinates, and effect size (t-statistic, z-score, or p-value). For the medicated MDD group, additional data were collected on medication type, duration of use, and pre- and post-treatment DAS scores. The literature search, study selection, and data extraction were conducted independently by two researchers, with discrepancies resolved through consultation with a third investigator.
Voxel-wise meta-analysis
Voxel-wise meta-analysis was performed using seed-based d-mapping with permutation of subject images (SDM-PSI) (version 6.23, http://www.sdmproject.com/) (Radua et al., Reference Radua, Mataix-Cols, Phillips, El-Hage, Kronhaus, Cardoner and Surguladze2012) to compare unmedicated and medicated MDD patients with HCs. The analysis followed SDM-PSI guidelines and employed a constrained maximum likelihood estimation method to balance bias and efficiency (Alegria, Radua, & Rubia, Reference Alegria, Radua and Rubia2016; Radua & Mataix-Cols, Reference Radua and Mataix-Cols2009). Initially, we extracted differences in brain activation from each study and a Gaussian kernel with a full width at half maximum (FWHM) of 20 mm was used within the gray matter mask. Subsequently, a composite map was constructed by averaging individual dataset maps, weighted by the square root of sample sizes to give greater influence to larger studies. Statistical significance was determined using a voxel-level threshold of p < 0.0025, an SDM z-score ≥ 1, and a minimum cluster extent of 10 voxels (Müller et al., Reference Müller, Cieslik, Laird, Fox, Radua, Mataix-Cols and Eickhoff2018). This threshold has been widely applied in recent SDM-PSI meta-analyses (Bore et al., Reference Bore, Liu, Huang, Kendrick, Zhou, Zhang and Becker2024; Chavanne & Robinson, Reference Chavanne and Robinson2021; Liu et al., Reference Liu, Klugah-Brown, Zhang, Chen, Zhang and Becker2022; Radua, Romeo, Mataix-Cols, & Fusar-Poli, Reference Radua, Romeo, Mataix-Cols and Fusar-Poli2013) because it provides a reasonable balance between Type I and Type II error rates, ensuring rigorous control of false positives. Data visualization was conducted using MRIcron (http://people.cas.sc.edu/rorden/mricron/).
A linear meta-regression model was applied to indirectly compare medicated and unmedicated MDD patients, based on contrasts between medicated MDD versus HCs and unmedicated MDD versus HCs. Only results exceeding a voxel-level p < 0.0025 and a cluster-level extent of ≥10 voxels were reported, corresponding to multiple comparison-corrected p < 0.05.
To examine the convergence of ALFF/ReHo changes between unmedicated and medicated MDD patients, we combined thresholded meta-analysis maps using the SDM-PSI conjunction analysis function (Radua et al., Reference Radua, Romeo, Mataix-Cols and Fusar-Poli2013; Radua, van den Heuvel, Surguladze, & Mataix-Cols, Reference Radua, van den Heuvel, Surguladze and Mataix-Cols2010). Statistical significance was determined using a voxel-level, p < 0.0025.
Functional annotation, genetic analysis
The behavioral and genetic characteristics of shared and distinct brain regions between medicated and unmedicated MDD patients were analyzed using the Brain Annotation Toolbox (BAT) (Liu, Rolls, Liu et al., Reference Liu, Rolls, Liu, Zhang, Yang, Du and Feng2019). BAT integrates activation maps from Neurosynth (Yarkoni et al., Reference Yarkoni, Poldrack, Nichols, Van Essen and Wager2011) and gene expression data from the Allen Human Brain Atlas (AHBA) (Shen, Overly, & Jones, Reference Shen, Overly and Jones2012). The Neurosynth decoder (https://neurosynth.org/decode/) was used to perform reverse inference decoding, linking brain regions with psychological processes. This decoder combines brain imaging data, text mining techniques, and machine-learning methods to calculate probability mappings between brain regions and psychological terms (Yarkoni et al., Reference Yarkoni, Poldrack, Nichols, Van Essen and Wager2011). Additionally, the AHBA offers a comprehensive ‘all genes-all structures’ profile of the human brain, enabling detailed analysis of gene expression patterns (Shen et al., Reference Shen, Overly and Jones2012).
Following the steps outlined in the BAT user manual, we conducted functional annotation and genetic characterization analyses. The top 10 functional terms associated with each brain region were extracted and plotted according to their correlation strength. Statistical significance was assessed using a permutation test conducted 100 times, and a significance threshold of p-values <0.05 was applied. For genetic characterization, the top 10 genes with the highest expression levels in each region were identified and plotted. The significance of these gene expression profiles was also evaluated using a permutation test (p < 0.05).
Neurotransmitter analysis
To explore the topographic association between shared and differential activation patterns in MDD (medicated versus unmedicated) and the distribution of key neurotransmitter systems, we used JuSpace v1.5 (https://github.com/juryxy/JuSpace) (Dukart, Holiga, Rullmann et al., Reference Dukart, Holiga, Rullmann, Lanzenberger, Hawkins, Mehta and Eickhoff2021). JuSpace enables spatial correlation analysis between MRI-derived brain activation patterns and PET- or SPECT-derived maps of neurotransmitter systems, such as dopamine, serotonin, and noradrenaline (Dukart et al., Reference Dukart, Holiga, Rullmann, Lanzenberger, Hawkins, Mehta and Eickhoff2021).
In this study, correlation analyses were restricted to the significant clusters identified by the SDM-PSI meta-analysis, rather than the entire brain. This approach was chosen to enhance the specificity of the molecular interpretation by focusing on brain regions that showed robust and replicable alterations in MDD. We focused on key neurotransmitter systems implicated in depression, particularly the 5-hydroxytryptamine (5-HT1a, 5-HT1b, and 5-HT2a) receptors (Savli, Bauer, Mitterhauser et al., Reference Savli, Bauer, Mitterhauser, Ding, Hahn, Kroll and Lanzenberger2012) and dopamine (D1 and D2) receptors (Alakurtti, Johansson, Joutsa et al., Reference Alakurtti, Johansson, Joutsa, Laine, Bäckman, Nyberg and Rinne2015; Kaller, Rullmann, Patt et al., Reference Kaller, Rullmann, Patt, Becker, Luthardt, Girbardt and Sabri2017), as well as dopamine (Dukart et al., Reference Dukart, Holiga, Rullmann, Lanzenberger, Hawkins, Mehta and Eickhoff2021) and noradrenaline transporters (Hesse, Becker, Rullmann et al., Reference Hesse, Becker, Rullmann, Bresch, Luthardt, Hankir and Sabri2017). These systems have been closely associated with MDD. We calculated Pearson’s correlation coefficients to test the spatial correlation between changes in brain activation after taking antidepressants and receptor/transporter distributions. Statistical significance was assessed via a permutation test (10,000 permutations) and FDR correction, with significance set at p < 0.05 (Dukart et al., Reference Dukart, Holiga, Rullmann, Lanzenberger, Hawkins, Mehta and Eickhoff2021). This approach offers a biologically meaningful framework for linking neuroimaging data with underlying neurotransmitter information, providing insights into the neurobiological mechanisms involved in MDD and its treatment.
Heterogeneity, publication bias and reliability analysis
Inter-study heterogeneity was assessed using the I2 statistical within a random-effects model. SDM-PSI was applied using standard kernel size and thresholds (FWHM = 20 mm, p = 0.005 uncorrected, FDR correction, peak height z = 1, minimum cluster size = 10 voxels) (Amad, Radua, Vaiva, Williams, & Fovet, Reference Amad, Radua, Vaiva, Williams and Fovet2019; Gong, Wang, Qiu et al., Reference Gong, Wang, Qiu, Chen, Luo, Wang and Wang2020). Heterogeneity was classified as low (I2 = 0–25%), moderate (I2 = 25–50%), large (I2 = 50–75%), or extreme (I2 = 75–100%). Publication bias was assessed using Egger’s test, with p < 0.05 indicating significant bias (Egger et al., Reference Egger, Davey Smith, Schneider and Minder1997).
These analyses provided insights into brain development, depression risk factors, and potential therapeutic targets. To evaluate the robustness of the primary meta-analysis results, we conducted a leave-one-out sensitivity analysis, sequentially omitting individual datasets to evaluate the consistency of the meta-analysis results (Radua & Mataix-Cols, Reference Radua and Mataix-Cols2009). To account for the imbalance in study numbers, we additionally performed a random sampling sensitivity analysis. Specifically, 21 datasets were repeatedly drawn at random from the 59 unmedicated datasets to match the 21 medicated datasets (derived from 15 papers), and both comparison and conjunction analyses were re-run 100 times. The coordinates, t-value files, and results of the 100 random runs can be accessed digitally at the following link: https://osf.io/49zgy/. The results of these 100 iterations, including key comparisons and conjunction analyses, are summarized in Supplementary Table S1.
To further assess the potential confounding effects of previous medication exposure, we conducted supplementary analyses restricted to medication-naïve patients (i.e. those who had never received antidepressant treatment). Activation differences between medicated and medication-naïve groups, as well as their conjunction maps, were also calculated to examine shared and distinct brain alterations. These analyses were performed following the same SDM pipeline as the main meta-analysis, and full results are provided in the Supplementary Figure S1 and Table S2 and are publicly available at https://osf.io/49zgy/files.
Results
Characteristics of the included studies and experiments
Unmedicated MDD
A total of 59 studies on unmedicated MDD and 15 studies involving 21 datasets on medicated MDD were included in the meta-analysis (Figure 1B). The unmedicated MDD group comprised 2,618 patients and 2,486 HCs. Among these studies, 26 applied ALFF as the primary measure, 32 used ReHo, and 13 used fALFF. In addition, 8 studies combined ALFF and ReHo, 2 combined ALFF and fALFF, 2 combined ReHo and fALFF, and 2 applied all three methods. Comprehensive data on participants’ age and gender were available for all studies. The mean age of unmedicated MDD patients was 31.22 years, with a female proportion of 61.4%. For HCs, the mean age was 31.26 years, with a female accounting for 58.2%. Additionally, 51 studies focused on first-episode MDD, while 8 included patients who had been medication-free for at least 2 weeks. Detailed demographic characteristics are provided in Supplementary Table S3.
Medicated MDD
This meta-analysis included 21 datasets from 15 studies, comprising 541 medicated MDD patients and 483 HCs. Of these studies, 7 used ALFF, 8 used ReHo, 2 used fALFF, 1 combined ALFF and ReHo, and 1 applied all three measures. The mean age of medicated MDD patients was 35.31 years, with a female ratio of 65.95%, whereas HCs had a mean age of 34.38 years, with 59.2% female representation. Medication duration ranged from 2 to 8 weeks, though three studies did not report this information (Liu et al., Reference Liu, Ma, Song, Tang, Jing, Zhang and Wang2015; Sun, Ma, Chen et al., Reference Sun, Ma, Chen, Wang, Guo, Luo and Xiao2022; Sun, Ma, Du et al., Reference Sun, Ma, Du, Wang, Guo, Luo and Hou2022; Wang, Wu et al. Reference Wang, Wu, Wang, Wang, Wang, Jin and Li2023), and four studies did not specify the type of antidepressant used (Fang et al., Reference Fang, Mao, Jiang, Li, Wang and Wang2015; Sun, Ma, Chen et al., Reference Sun, Ma, Chen, Wang, Guo, Luo and Xiao2022; Sun, Ma, Du et al., Reference Sun, Ma, Du, Wang, Guo, Luo and Hou2022; Wang, Wu et al. Reference Wang, Wu, Wang, Wang, Wang, Jin and Li2023). Detailed demographic and medication data can be found in Supplementary Table S4. We compared demographic and clinical variables between medicated and unmedicated MDD samples. No significant group differences were found in age, gender, or symptom severity (all p > 0.05 Supplementary Table S5).
Neural activation differences in the main meta-analysis
Medicated MDD versus HCs
SDM analysis revealed increased activation in the left striatum and decreased activation in the right precentral gyrus in medicated MDD patients compared to HCs (Table 1 and Figure 2A). Heterogeneity in these regions was minimal and no significant publication bias was found (p-values >0.05). Jackknife sensitivity analyses confirmed reproducibility of these findings across at least 18 of the 21 experiments (Supplementary Table S6).
Clusters showing differences among unmedicated MDD, medicated MDD, and HCs
Abbreviations: MDD, major depressive disorder; HCs, healthy controls; BA, Brodman areas; MNI, Montreal neurological institute, SDM, seed-based d-mapping.
Brain regions showed significant resting-state-related neural activation differences between groups. Meta-analyses results regarding (A) differences in resting-state-related neural activation between Medicated MDD and HCs, (B) differences in resting-state-related neural activation between Unmedicated MDD and HCs, (C) differences in resting-state-related neural activation between Unmedicated MDD and Medicated MDD, as well as (D) conjunction of Unmedicated MDD and Medicated MDD (versus HCs). Areas with decreased resting neural activation values are shown in blue, and areas with increased resting neural activation values are shown in red. The color bar indicates the maximum and minimum SDM-Z values. Abbreviations: HCs, healthy controls; MDD, major depressive disorder; SDM seed-based d-mapping.
Unmedicated MDD versus HCs
As shown in Table 1 and Figure 2B, the meta-analysis indicated that activation in the MOG, left precuneus, left lingual gyrus, and right fusiform gyrus was decreased in patients with drug-naïve MDD compared with HCs. No regions of increased activation were identified. No significant heterogeneity or publication bias were observed (Egger’s test, p > 0.05). Jackknife sensitivity analyses showed consistent reduced activity in the left MOG, left precuneus, and left lingual gyrus across all 59 analyses, while reduced activation in the right MOG and right fusiform gyrus were significant in at least 56 of 59 analyses (Supplementary Table S7).
Unmedicated MDD versus medicated MDD
As shown in Table 1 and Figure 2C, the meta-analysis revealed increased activation of the left striatum in medicated MDD patients compared to unmedicated MDD patients. No significant was detected, and Egger’s test indicated no publication bias (p-values >0.05).
(Medicated MDD versus HCs) and (unmedicated MDD versus HCs) conjunction
Conjunction analysis revealed overlapping reductions in neural activation in both medicated and unmedicated MDD patients relative to HCs. Specifically, significant reductions were observed in the left MOG (peak MNI = −40, −84, 4; p < 0.001; 75 voxels) and left calcarine fissure/surrounding cortex (peak MNI = −12, −48, 6; p < 0.001; 60 voxels) (Figure 2D).
Behavioral characterization and genetic expression
When we input the left MOG into the BAT, we found that it was functionally enriched in terms related to ‘recognition’, ‘face processing’, ‘reading’, and ‘motion’ (Figure 3A). For the left calcarine fissure/surrounding cortex, the highest correlations were found with terms, such as ‘memory’, ‘attention’, ‘perception’, and ‘semantic processing’ (Figure 3B). For the left striatum, the highest correlations were found with terms such as ‘memory’, ‘perception’, ‘semantic processing’, and ‘language’ (Figure 3C). Genetic expression analyses revealed Transcription Factor AP-2 Beta (TFAP2B) Prolactin (PRL), Solute Carrier Family 5 Member 7 (SLC5A7) and Iroquois Homeobox 2 (IRX2) were among the top 10 most expressed genes in the left MOG and left calcarine fissure/surrounding cortex (Figure 3D,E). However, SLC5A7, PRL, IRX2, and diamine oxidase were among the 10 genes with the highest expression levels in the left striatum (Figure 3F). The full names of all associated genes, along with the proteins they encode and their functions, are provided in Supplemental Table S8–S10.
Functional annotation results for the left middle occipital gyrus (A), left calcarine fissure/surrounding cortex (B), and left striatum (C); Genetic analysis of the top 10 genes identified in these key regions: left middle occipital gyrus (D), left calcarine fissure/surrounding cortex (E), and left striatum (F).
Association between the distribution of neurotransmitter
We investigated the relationship between significantly shared activation patterns in medicated and unmedicated MDD and neurotransmitter distribution. A significant positive correlation was observed between these activation patterns and the distribution of 5-HT2a receptors (Fisher’s z = 1.14, FDR-corrected p < 0.01), while a significant negative correlation was found with noradrenaline transporters (Fisher’s z = −1.10, FDR-corrected p < 0.01). No significant correlations were found with the distributions of 5-HT1a, 5-HT1b, dopamine (D1 and D2) receptors, or dopamine transporters (Figure 4A). Additionally, we examined the relationship between significantly different activation patterns in medicated versus unmedicated MDD and neurotransmitter distribution. A significant negative correlation was identified with 5-HT1a (Fisher’s z = −0.23, FDR-corrected p < 0.01) and 5-HT2a receptors (Fisher’s z = −0.22, FDR-corrected p = 0.02). In contrast, significant positive correlations were found with D1 receptors (Fisher’s z = 0.46, FDR-corrected p < 0.01), D2 receptors (Fisher’s z = 0.54, FDR-corrected p < 0.01), and dopamine transporters (Fisher’s z = 0.46, FDR-corrected p < 0.01) (Figure 4B).
Spatial correlation analysis between brain regions exhibiting co-variation in medication use and non-use in major depressive disorder (A), and brain regions showing significant increases post-medication compared to pre-medication in major depressive disorder (B), along with their relationship to neurotransmitters. Abbreviations: 5-HT, 5-Hydroxytryptamine; DAT, dopamine transporters; NAT, noradrenaline transporters.
Discussion
To our knowledge, this is the first meta-analysis to systematically compare resting-state neural activity between medicated and unmedicated patients with MDD. Indirect comparisons between unmedicated and medicated MDD patients revealed heightened activity in the left striatum following antidepressant treatment. Conjunction analysis demonstrated overlapping reductions in activation within the left MOG and left calcarine fissure across both unmedicated and medicated MDD groups. Additionally, functional characterization indicated that these affected regions are primarily involved in visual processing, attention, and semantic processing, while genetic expression analyses highlighted associations with metabolic regulation, endocrine regulation, neural development, and neurotransmitter-related genes. Finally, neurotransmitter mapping revealed significant correlations between MDD-related neural alterations and the distribution of serotonergic and dopaminergic receptors. However, as this comparison was cross-sectional rather than longitudinal, the observed differences should be interpreted as associative rather than causal.
Comparison with previous resting-state meta-analyses in MDD
Although the present study specifically examined local spontaneous neural activity (ALFF and ReHo), it is informative to situate our findings within the broader literature on resting-state abnormalities in MDD. Previous coordinate-based meta-analyses of resting-state functional connectivity (rsFC) have consistently revealed large-scale network dysregulation, particularly involving the default mode network (DMN), salience network, and fronto-limbic circuitry (Kaiser, Andrews-Hanna, Wager, & Pizzagalli, Reference Kaiser, Andrews-Hanna, Wager and Pizzagalli2015; Zhang, Zhang, Wang et al., Reference Zhang, Zhang, Wang, Lei, Jiang, Xiong and Liu2025). These network-level abnormalities are functionally related to emotion regulation, self-referential processing, and cognitive control – domains that overlap with the frontal and cingulate regions identified in the present activation-based meta-analysis (Kaiser et al., Reference Kaiser, Andrews-Hanna, Wager and Pizzagalli2015; Zhang et al., Reference Zhang, Zhang, Wang, Lei, Jiang, Xiong and Liu2025). Such convergence suggests that antidepressant treatment may modulate both local neural activity and distributed network coordination within emotion-regulation systems (Whitfield-Gabrieli & Ford, Reference Whitfield-Gabrieli and Ford2012).
In contrast to rsFC studies, which primarily quantify the strength of inter-regional coupling, activation-based measures capture the amplitude and synchrony of intrinsic neural signals within each region (Zang, He, Zhu et al., Reference Zang, He, Zhu, Cao, Sui, Liang and Wang2007; Zuo, Kelly, Di Martino et al., Reference Zuo, Kelly, Di Martino, Mennes, Margulies, Bangaru and Milham2010). Given that antidepressant pharmacodynamics act directly on neurochemical processes at the local level (Johansen, Armand, Plavén-Sigray et al., Reference Johansen, Armand, Plavén-Sigray, Nasser, Ozenne, Petersen and Knudsen2023), our activation-based approach provides complementary insight into the neurophysiological substrates of treatment response. The consistency between our regional activation results and prior rsFC findings strengthens the reliability and interpretability of the present results, indicating that both methodological perspectives converge on similar cortical–subcortical systems implicated in the pathophysiology and treatment of MDD (Gong et al., Reference Gong, Wang, Qiu, Chen, Luo, Wang and Wang2020; Hao et al., Reference Hao, Chen, Mao, Zhong and Dai2019; Iwabuchi et al., Reference Iwabuchi, Krishnadas, Li, Auer, Radua and Palaniyappan2015; Yuan et al., Reference Yuan, Yu, Yu, Liang, Huang, He and Xiang2022).
Differences in resting-state brain activity with medication MDD compared to unmedicated MDD
Our findings suggest that medicated MDD patients exhibit increased bilateral striatal activation compared to unmedicated patients. The striatum, a core component of the basal ganglia, is involved in cortico-striatal-thalamo-cortical circuits related to memory, learning, task performance, and reward – all of which are implicated in the therapeutic effects and responses in MDD (Rupprechter, Romaniuk, Series et al., Reference Rupprechter, Romaniuk, Series, Hirose, Hawkins, Sandu and Steele2020; Sequeira, Silk, Ladouceur et al., Reference Sequeira, Silk, Ladouceur, Hanson, Ryan, Morgan and Forbes2021). Striatal activation and functional connectivity during reward processing are known to change following antidepressant treatment (Admon, Kaiser, Dillon et al., Reference Admon, Kaiser, Dillon, Beltzer, Goer, Olson and Pizzagalli2017; Heller, Johnstone, Light et al., Reference Heller, Johnstone, Light, Peterson, Kolden, Kalin and Davidson2013; Wang, An, Gao et al., Reference Wang, An, Gao, Zhang, Chen, Li and Si2019), and meta-analyses suggest that increased striatal activation correlates with MDD remission (Wang, He et al. Reference Wang, He, Yang, Wang, Zou, Xiao and Zhu2023). The striatum is also linked to higher-order cognitive control and emotional processing through its connectivity with the frontoparietal network and the DMN (Ho et al., Reference Ho, Connolly, Henje Blom, LeWinn, Strigo, Paulus and Yang2015; Kaiser et al., Reference Kaiser, Andrews-Hanna, Wager and Pizzagalli2015). Functional decoding further supports our findings by indicating that striatal function is related to memory and perception. Neurotransmitter analyses revealed a negative correlation between increased striatal activation and 5-HT1a, 5-HT2a receptors, and a positive correlation with D1, D2 receptors, and dopamine transporters. Previous studies show that long-term antidepressant use downregulates 5-HT2A receptor expression (Gray & Roth, Reference Gray and Roth2001), and antagonists of 5-HT2A receptors produce antidepressant effects in animal models (Guiard & Di Giovanni, Reference Guiard and Di Giovanni2015). Meanwhile, 5-HT2A receptors significantly affect the function of 5-HT1A receptors (Naumenko, Bazovkina, & Kondaurova, Reference Naumenko, Bazovkina and Kondaurova2015; Naumenko et al., Reference Naumenko, Bazovkina, Kondaurova, Zubkov and Kulikov2010). 5-HT1A receptors inhibit serotonin release from nerve endings and decrease serotonin concentration in the salient interstitial space (Giorgioni et al., Reference Giorgioni, Bonifazi, Botticelli, Cifani, Matteucci, Micioni Di Bonaventura and Del Bello2024). Serotonin controls the release of other neurotransmitters (including dopamine and glutamate) into the dorsal striatum while regulating striatal microcircuits and projection populations (Spring & Nautiyal, Reference Spring and Nautiyal2024). Antidepressant drugs may affect striatal regions via fiber connections with the prefrontal cortex, which has extensive serotonin distribution (Hori et al., Reference Hori, Mimura, Nagai, Hori, Kumata, Zhang and Minamimoto2024). Moreover, D1R activation in the striatum is associated with reinforcement of cortical neural patterns (Vendrell-Llopis et al., Reference Vendrell-Llopis, Read, Boggiano, Hetzler, Peitsinis, Stanley and Isacoff2025), and D1/D2 receptors are critical for DA signaling integration, with their absence impairing striatal function (Bonnavion et al., Reference Bonnavion, Varin, Fakhfouri, Martinez Olondo, De Groote, Cornil and Giros2024). Gene expression analysis also identified high expression of SLC5A7 in this region, which facilitates choline entry into cholinergic neurons (Gigout et al., Reference Gigout, Wierschke, Lehmann, Horn, Dehnicke and Deisz2012), essential for acetylcholine synthesis (Kenny, Scharenberg, Abu-Remaileh, & Birsoy, Reference Kenny, Scharenberg, Abu-Remaileh and Birsoy2025). Acetylcholine and dopamine systems interact in reward processing within the striatum (Krok et al., Reference Krok, Maltese, Mistry, Miao, Li and Tritsch2023). This interaction likely contributes to cognitive functions, such as learning, memory, and decision-making.
Thus, the functional modulation of striatal regions likely underlies the therapeutic effects of antidepressant medications, reflecting a combination of altered cortico-striatal connectivity, neurotransmitter regulation (especially serotonin and dopamine systems), and region-specific gene expression, which together improve reward processing, cognitive control, and emotional regulation.
Shared resting-state brain activity alterations in medicated and unmedicated MDD
Our conjunction analysis revealed reduced activation in the left MOG and the left calcarine fissure/surrounding cortex in both medicated and unmedicated MDD patients compared to HCs. This pattern was consistently observed across both groups, reinforcing its robustness. As key components of the occipital lobe, these regions are involved in visual processing and higher-order cognition, such as attention, recognition, and memory encoding – domains often impaired in MDD (Teng, Zhou, Ma et al., Reference Teng, Zhou, Ma, Tan, Wu, Guan and Zhang2018). Functional decoding indicated that these regions are involved in attention, recognition, and memory encoding – cognitive domains often impaired in MDD. Abnormalities in visual attention, such as increased sensitivity to negative stimuli and attentional bias, are commonly reported in depression and correlate with symptom severity and suicidality Abnormalities in visual attention, such as increased sensitivity to negative stimuli and attentional bias, are commonly reported in depression and correlate with symptom severity and suicidality (Villa et al., Reference Villa, Pinkham, Kaufmann, Granholm, Harvey and Depp2018; Wang, Guobule, Li, & Li, Reference Wang, Guobule, Li and Li2021). Together with prior structural imaging findings of occipital lobe abnormalities in MDD (Giakoumatos et al., Reference Giakoumatos, Tandon, Shah, Mathew, Brady, Clementz and Keshavan2013; Maller et al., Reference Maller, Thomson, Rosenfeld, Anderson, Daskalakis and Fitzgerald2014), these results suggest that both functional and anatomical alterations in visual regions may underlie core aspects of depressive symptomatology. In addition to our findings, a large number of previous resting-state fMRI studies have reported reduced MOG activation in MDD (Cerullo, Eliassen, Smith et al., Reference Cerullo, Eliassen, Smith, Fleck, Nelson, Strawn and Strakowski2014; Hao et al., Reference Hao, Chen, Mao, Zhong and Dai2019; Yuan et al., Reference Yuan, Yu, Yu, Liang, Huang, He and Xiang2022; Zhong, Pu, & Yao, Reference Zhong, Pu and Yao2016), further reinforcing the robustness of our results.
Gene expression analysis identified high levels of TFAP2B in these regions, suggesting its role in visual cortex-related neurotransmission. TFAP2B regulates monoamine system genes, and its dysregulation may impact serotonergic and noradrenergic pathways (Nilsson et al., Reference Nilsson, Sonnby, Nordquist, Comasco, Leppert, Oreland and Sjöberg2014; Nutt, Reference Nutt2008). Consistent with this, our neurotransmitter correlation analysis revealed a positive association between reduced activation in these regions and 5-HT2a receptor expression and a negative association with norepinephrine transporter availability. The influence of 5-HT2a receptors and norepinephrine transporters on the underlying neural mechanisms of depression is further emphasized (Nutt, Reference Nutt2008).
Additionally, alterations in occipital function may also reflect broader disruptions at the systems neuroscience level. Moreover, the reduced occipital activation observed in our study may reflect broader alterations in large-scale brain networks. Prior research has shown that hypoactivity in visual areas is often coupled with hyperactivity in the DMN (Runia et al., Reference Runia, Yücel, Lok, de Jong, Denys, van Wingen and Bergfeld2022), suggesting a potential imbalance between internally focused thought and external information processing in MDD. This imbalance may contribute to the excessive rumination and diminished cognitive flexibility commonly observed in depressive states.
These findings underscore the occipital cortex – not only as a region of consistent functional hypoactivity in MDD but also as a potential site where neurochemical, genetic, and cognitive mechanisms converge. The reproducible reduction in left MOG activation, regardless of medication status, highlights its potential as a stable neurophysiological marker for MDD diagnosis and subtyping.
Limitations
Our meta-analysis has several limitations. First, this study adopted a cross-sectional design comparing medicated and unmedicated MDD patients rather than a longitudinal follow-up of the same individuals. As such, the results provide indirect rather than causal evidence of antidepressant-related neural changes. Second, ‘medication’ includes diverse antidepressant agents with different mechanisms, which could influence outcomes. However, most antidepressants ultimately converge on modulating key neurotransmitter systems, such as serotonin, norepinephrine, and dopamine (Cheng & Bahar, Reference Cheng and Bahar2019). Grouping medicated MDD patients together, therefore, allows us to capture shared neurobiological effects of antidepressant treatment that may underlie their clinical efficacy. Previous studies have reported that various antidepressants produce comparable effects on resting-state brain networks, including the DMN and emotion-regulation circuits (Brakowski et al., Reference Brakowski, Spinelli, Dörig, Bosch, Manoliu, Holtforth and Seifritz2017). Nonetheless, residual effects of prior antidepressant use, especially long half-life agents, may still confound our findings. Third, as SDM relies on coordinate-based rather than image-based data, further validation with original statistical maps is warranted. Fourth, our analyses were limited to adults, and antidepressant effects in adolescents or older populations may differ (Lee et al., Reference Lee, Shin, Lee, Yoo, Kim and Brent2023). Fifth, although effect-size standardization, random-effects modeling, and sensitivity analyses were applied to mitigate methodological heterogeneity, residual variability related to MRI acquisition protocols (e.g. scanner type, field strength) and preprocessing pipelines cannot be fully excluded. Sixth, illness duration was not consistently reported across studies, preventing direct group comparisons and representing a potential confound. Finally, all included studies were conducted in China, reflecting the current distribution of available literature; however, this limits the generalizability of our findings, highlighting the need for future multicenter, cross-cultural investigations.
Conclusions
We found significantly increased activation in the left striatum in medicated MDD patients, suggesting it as a potential target for antidepressant treatment. Striatal activity was further linked to memory, perception, and serotonergic/dopaminergic systems, with associated genes like SLC5A7, PRL, and IRX2. In contrast, both medicated and unmedicated MDD patients showed reduced activation in the left MOG, which may serve as a stable neurobiological marker for MDD diagnosis due to its role in visual-cognitive processing. These findings highlight distinct yet complementary neural signatures underlying MDD, with implications for treatment and biomarker development.
Supplementary material
The supplementary material for this article can be found at http://doi.org/10.1017/S0033291725102699.
Author contribution
Contributors: RFS, ZXW, and YKD conceived and designed the study. ZXW, YHS, and YH supervised the study. RFS, MLL, and YKD performed the statistical analysis. CY, YQW, DTL, and DY carried out data cleaning and material support. RFS and ZXW drafted the manuscript.
Funding statement
This study was supported by the Sichuan Science and Technology Program (2024NSFSC1564), Postdoctoral Fellowship Program of CPSF (GZC2023180), the China Postdoctoral Science Foundation (2024 M762243), and the Postdoctoral Research Fund of West China Hospital, Sichuan University (2024HXBH046).
Competing interests
There is no conflict of interest to declare.
Open access
