A systematic review of structural and functional magnetic resonance imaging studies on the neurobiology of depressive symptoms in schizophrenia spectrum disorders

Overview of study characteristics

Our initial search identified 5,765 potentially relevant studies, excluding duplicates. Following a review of titles and abstracts, 5570 studies were excluded, leaving 195 for full-text screening. After this stage, 162 studies were excluded, resulting in a final selection of 33 studies. Among these, 14 studies used T1-weighted structural MRI (sMRI) or dMRI, 18 used fMRI, and one study used both fMRI and sMRI (Fig. 2). Tables 1 and 2 present the characteristics of the structural and functional neuroimaging studies, respectively, including the number of participants, sex distribution, mean age, and quality assessment scores. Across included studies, sample sizes tended to be small (a group of less than 30 participants in 15 out of 33 studies), yet exhibited a wide variation (overall sample sizes ranged from 15 to 312 participants). Most study participants were male (1293 of 2007; 64%) and had an average age of 33.7 土 11.2. Antipsychotic medication use was reported in 21 studies, with 19 using chlorpromazine, one using olanzapine, and one using haloperidol equivalents; two studies specified that participants were drug or neuroleptic naive. Additionally, six studies reported the use of antidepressant medication. All studies used either DSM-III, DSM-IV, or DSM-5 for SSDs diagnosis (details provided in Tables 1–2). Thirty-one studies recruited patients with schizophrenia. Six studies included patients with schizoaffective disorder and one with schizophreniform disorder. Four studies designated patients as first-episode schizophrenia or psychosis. None of the studies included individuals with comorbid depressive disorders.

DSM Diagnostic and Statistical Manual of Mental Disorders, dMRI diffusion MRI, fMRI functional magnetic resonance imaging, rs-fMRI resting-state fMRI, sMRI structural MRI, SSDs schizophrenia spectrum disorders.

Quality assessment

One study received a low score (<4 points)³⁴, 15 studies received a moderate score (4–5 points)^{35,36,37,38,39,40,41,42,43,44,45,46,47,48,49}, and 17 studies received a high score (6+ points)^{27,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65} on the modified NOS. Validation of diagnoses by independent sources was frequently unreported, and additional points were lost relatively uniformly across the other evaluation criteria.

Structural studies

Table 1 provides an overview of the included structural studies, encompassing key details such as the main clinical measure of depressive symptoms, the neuroimaging metric used, and a summary of the study’s findings. Figure 3 summarizes frequently implicated brain regions and white matter tracts.

The color scale corresponds to the frequency of the region or tract reported. A A schematic illustration of regions implicated in depressive symptoms in SSDs. Subcortical regions are shown through a glass brain, and cortical regions are displayed on the cerebral cortex, as per the Desikan-Killiany Cortical Atlas parcellation. Top regions include the bilateral hippocampus, as well as the right frontal areas. Implicated regions are further subdivided into positive (B) and negative (C) associations with depressive symptoms. D A schematic illustration of tracts implicated in depressive symptoms in SSDs. Tracts are shown overlaid on a glass brain, as per the O’Donnell Research Group Fiber Clustering White Matter Atlas parcellation. CC Corpus Callosum, CR Corona Radiata, sFOF Superior Fronto-Occipital Fasciculus, SLF Superior Longitudinal Fasciculus and ThalR Thalamic Radiation. Top tracts include CR, ThalR, and SLF. Note that association and projection tracts are displayed in the left hemisphere, and only the genus of the CC is shown for clarity. Both dMRI studies found positive correlations between white matter tract integrity and depressive symptoms.

Depressive symptom measures

The association between structural neuroimaging metrics with depressive symptoms was assessed using scales such as the CDSS (n = 4)^27,35,39,64, the depression-anxiety subscale or depressive factor score of the Positive and Negative Syndrome Scale; PANSS (n = 3)^36,38,63, the depression-anxiety subscale or affect factor score of the Brief Psychiatric Rating Scale; BPRS (n = 3)^37,59,62, the HAMD (n = 3)^60,61,65, and the Maryland Trait and State Depression Scale; MTSD (n = 1)⁵⁸.

sMRI studies

Of the 14 structural studies, 10 employed metrics derived from T1-weighted sMRI such as morphology measurements related to volume, surface area, thickness, and size^{27,35,36,37,60,61,62,63,64,65}.

Six of these studies associated depressive symptoms with brain morphology using a regional-specific approach^{27,36,37,61,62,63}. Regions of interest (ROIs) included areas within the prefrontal cortex (namely, the dorsolateral prefrontal cortex (DLPFC) and orbitofrontal cortex)^27,61, the hippocampus^36,63, the amygdala⁶², and the cerebellum³⁷. A negative correlation emerged between the severity of depressive symptoms and both volume⁶¹ and thickness²⁷ within the prefrontal cortex. In the hippocampus, Bossù et al. found a negative association between the severity of depressive symptoms and volume⁶³, while Smith et al. reported a positive correlation between depression scores and fissure size; a measure suggestive of abnormal neurodevelopment³⁶. In the remaining studies, depression scores were positively associated with amygdalar volume⁶², and negatively associated with cerebellar volume³⁷.

Four sMRI studies investigated the relationship between depressive symptoms and brain morphology using a whole-brain approach^35,60,64,65. Kohler et al. reported increased left temporal lobe volume in patients with high depressive symptoms compared to those with low depressive symptoms⁶⁰, whereas an association of lower volume in the superior frontal and orbitofrontal gyrus with higher depression scores was identified by Siddi et al.⁶⁴. In a multivariate brain-behavior analysis, Buck et al. found specific patterns in females with SSDs, where fewer depressive symptoms were associated with changes in hippocampal subfields and varying thickness in specific cortical regions; such as lower thickness in the right superior temporal gyrus, entorhinal cortex, pars orbitalis, medial orbitofrontal gyrus and cingulate cortex, and high thickness in the left precentral gyrus, paracentral gyrus, cuneus, and lingual gyrus³⁵. Notably, this brain-behavior pattern also correlated with fewer negative symptoms, though to a lesser extent. Finally, Wei et al. found that individuals with comorbid depressive symptoms had significantly greater gray matter volume in the left isthmus cingulate and posterior cingulate cortex, as well as increased surface area in the left isthmus cingulate, left superior parietal gyrus, and right cuneus compared to those without depressive symptoms⁶⁵.

dMRI studies

Four studies used dMRI to assess the relationships between white matter tract integrity measures (i.e., FA, MD, radial diffusivity (RD), or white matter connectivity) and depressive symptoms in SSDs. Analytical methods across studies were highly heterogeneous. Chiappelli et al. used voxel-wise tract-based spatial statistics (TBSS)⁵⁸, Amodio et al. used probabilistic tractography³⁸, Long et al. used both voxel-wise TBSS and ROI probabilistic tractography³⁹ and Joo et al. used whole-brain tractography⁵⁹. Two of the four studies that used tractography did not find significant associations between alterations in white matter integrity and depressive symptoms in SSDs^38,59. However, Chiappelli et al. found that greater experience of depression, termed ‘trait depression’, was positively linked to both the overall average FA values throughout the brain and FA values specific to four white matter pathways: the corona radiata, thalamic radiation, superior longitudinal fasciculus, and superior frontal-occipital tract⁵⁸. Similarly, Long et al. found that patients with suicidal ideation exhibited elevated FA in several white matter tracts, including the corpus callosum, left anterior corona radiata, left superior corona radiata, and bilateral posterior corona radiata, as well as decreased MD in the splenium of the corpus callosum, bilateral posterior corona radiata, left posterior thalamic radiation and left superior longitudinal fasciculus³⁹. However, this finding should be interpreted with caution, as suicidal ideation in psychosis could have multiple etiologies (i.e., delusion content, auditory verbal hallucination) despite being measured using the CDSS.

Functional studies

Table 2 provides an overview of the included functional studies, encompassing key details such as the main clinical measure of depressive symptoms, the neuroimaging metric used, and a summary of the study’s findings. Figure 4 summarizes frequently implicated brain regions and networks, while Supplementary Fig. S5 provides a breakdown based on whether the findings are from resting-state or task-based analyses.

The color scale corresponds to the frequency of the region or network reported. A A schematic illustration of regions implicated in depressive symptoms in SSDs. Subcortical regions are shown through a glass brain, and cortical regions are displayed on the cerebral cortex, as per the Surface-Based Multimodal parcellation. Top regions include the left caudate and putamen and bilateral frontal area. Note that some studies investigated specific regions of interest (ROIs), and did not use a whole-brain approach. Implicated regions are further subdivided into positive (B) and negative (C) associations with depressive symptoms. D A schematic illustration of networks implicated in depressive symptoms in SSDs. Networks are displayed on the cerebral cortex, as per the Cole-Anticevic Brain-wide Network Partition, AUD Auditory Network, DMN Default Mode Network, FPN Frontoparietal Network, LAN Language Network, SMN Somatomotor Network, SN Salience Network. Networks were reported bilaterally but are displayed on the left hemisphere for clarity. Top network connections include within- DMN, FPN, and SN. All studies found negative correlations between network-based functional connectivity and depressive symptoms. One study identified both negative and positive associations with depressive symptoms (positive association found within-FPN, and between DMN-SMN).

Depressive symptom measures

The association between functional neuroimaging metrics with depressive symptoms was assessed using scales such as the depression-anxiety subscale or depressive factor score of PANSS (n = 6)^{34,41,51,52,53,57}, the CDSS (n = 5)^{40,43,45,47,48}, the depression-anxiety subscale or affect factor of BPRS (n = 3)^54,55,56, the Beck’s Depression Inventory; BDI/BDI-II (n = 3)^44,46,50, the MTSD (n = 1)⁴².

rs-fMRI studies

Nine fMRI studies utilized metrics derived from resting-state fMRI (rs-fMRI) data, such as functional connectivity^{34,40,52,53,54,55,56}, amplitude of low-frequency fluctuations (ALFF)⁵⁷, and global/network efficiency⁴¹.

Five of these studies investigated associations of depressive symptoms with brain function using a specific seed- or a-priori network-based approach. In a lower-quality ROI-based analysis of resting state functional connectivity, Xu et al. found no significant correlation between depressive symptoms and the substantia nigra/ventral tegmental area³⁴. However, in analyses of resting state functional connectivity based on specific networks of interest, depressive symptoms were linked to the default mode network (DMN)⁵⁵, salience network^40,52, and frontoparietal network (FPN)⁵³ (often synonymous with the central executive network; CEN).

The remaining four studies used a whole-brain regional or network-level approach^41,54,56,57. Analytical methods and findings across studies were variable. Li et al. demonstrated that an increase in ALFF, which quantifies the strength of low-frequency brain activity fluctuations, in the dorsolateral region of the superior frontal gyrus was significantly linked to a greater reduction in depression scores⁵⁷. Doucet et al. showed a robust pattern of functional network connectivity strongly correlated with improvements in depressive symptoms, with higher within-DMN connectivity being a significant positive predictor, while reduced within-CEN and diminished connectivity between DMN and sensorimotor networks acted as important negative predictors⁵⁶. Notably, this connectivity pattern also correlated with improvements in positive symptoms. Moreover, Lee et al. found the variance in depressive symptom severity can be explained by within-network connectivity of the salience network and connectivity between salience-language networks and somatomotor-auditory networks⁵⁴. Lastly, Su et al. used graph theoretical analysis of networks to show depression symptoms were significantly correlated with the overall efficiency of brain network information processing⁴¹.

task-fMRI studies

Nine studies employed task-based fMRI to evaluate the relationship between functional brain activity and depressive symptoms in SSDs^{42,43,44,45,46,47,48,50,51}; three of which investigated specific ROIs. Significant positive associations were found between functional activity in the ventral striatum during a monetary incentive delay task measuring reward processing^43,48, and visual-related regions during an object perception task⁴⁵ with depressive symptoms.

The remaining six studies used a whole-brain approach^{42,44,46,47,50,51}. Two of the studies did not find any significant associations between brain activation and depressive symptoms^50,51. Conversely, Lee et al. found that activity in the left posterior cingulate cortex was inversely correlated with overall depression scores⁴⁷. Arrondo et al. demonstrated a negative correlation between depression severity and ventral striatum activity during reward anticipation⁴⁶. Kumari et al. highlighted significant positive correlations between depression scores and brain activity in several regions while processing fearful expressions, including the left thalamus, para post-pre-central gyrus, putamen-globus pallidus, supramarginal gyrus, insula, inferior-middle frontal gyrus, and right superior frontal gyrus, extending to other frontal and cingulate gyri⁴⁴. Moreover, higher activity was noted in thalamic and superior frontal gyrus clusters among patients with moderate-to-severe depression compared to those with milder levels of depression. Lastly, Kvarta et al. found a significant inverse correlation between anticipatory threat-induced ventral anterior cingulate cortex cluster activation and trait depression⁴².

Multimodal study

A study with the largest sample size (n = 312) by Liang et al., employed a multimodal approach investigating both whole-brain fractional ALFF (fALFF) and gray matter volume in relation to depressive symptoms, assessed using the Montgomery-Asberg Depression Rating Scale (MADRS) (42). The authors investigated associations in schizophrenia and schizoaffective disorder groups separately and identified distinctions. In schizophrenia, elevated depression scores were linked to increased fALFF in the thalamus and hippocampus, as well as heightened gray matter volume in the insula and inferior frontal cortex. In schizoaffective disorder, higher depression scores were associated with increased fALFF and greater gray matter volume in the lingual and frontal gyrus.