Hyperconnectivity between the posterior cingulate and middle frontal and temporal gyrus in depression: Based on functional connectivity meta-analyses

Disrupted whole-brain resting-state functional connectivity (RSFC) of the posterior cingulate (PCC) has been highlighted to associate with cognitive and affective dysfunction in major depressive disorder (MDD). However, prior findings showed certain inconsistency about the RSFC of the PCC in MDD. This study aims to investigate the aberrant RSFC of the PCC in MDD using anisotropic effect-size version of seed-based d mapping (AES-SDM). Web of Science and PubMed were searched for studies investigating PCC-based RSFC in MDD. A total of 17 studies, involving 804 patients and 724 healthy controls (HCs), fit our selection criteria. Additionally, to seek for the link between functional and structural differences, we did a meta-analysis on the studies in conjunction with voxel-based morphology (VBM) analysis. The PCC showed higher RSFC with the left middle temporal gyrus (MTG) and the right middle frontal gyrus (MFG), and lower RSFC with the left superior frontal gyrus (SFG) and the left precuneus in patients with MDD than HCs. Moreover, the meta-regression analysis revealed a negative correlation between the FC alteration of the right MFG with the PCC and depression severity. Notably, the left MTG and the left MFG demonstrated gray matter deviations in conjunction analysis. Our results indicated that the aberrant RSFC between the PCC and brain regions sub-serving cognitive control and emotional regulation in patients with MDD. And such functional alterations may have structural basis. These findings may underlie the mechanisms of deficits in cognitive control and emotional regulation of MDD.


Introduction
Major depressive disorder (MDD), one of the most prevalent mental disorders, is characterized by a lack of interest, low self-worth, feeling of sadness, as well as cognitive and physical symptoms that disturb the normal life and work (Ferrari et al., 2013;Janca & Hiller, 1996). It has brought significant psychological burden for the patients as well as socioeconomic burden for their family and the whole society, and is the largest factor contributing to global disability (Liu et al., 2020). Because the pathogenesis of depression is not yet fully understood, the guideline for treatment mainly depends on clinical observations of depressive symptom severity (Köhler-Forsberg et al., 2020). With the advancing neuroimaging technology, we can now investigate the neuropathological mechanisms of MDD and identify potential neural biomarkers of MDD that could inform treatment efficacy (Drysdale et al., 2017).
Resting-state functional magnetic resonance imaging (rs-fMRI) is one of the commonly used neuroimaging approaches to characterize the spontaneous blood oxygen level dependent (BOLD) fluctuation, explicitly during task-free conditions. Many rs-fMRI studies have showed its validity of exploring the neurobiological mechanism of MDD (Iwabuchi et al., 2015;Wang et al., 2019;Zhou et al., 2020). The rs-fMRI has a number of advantages including reproducibility, non-invasiveness and higher sensitivity in detecting illness-related brain functional alterations (Fox & Raichle, 2007). Resting-state functional connectivity (RSFC) can measure the properties of intrinsic brain functional networks that can be attributed to clinical variables (Castellanos et al., 2013). Previous studies showed that the aberrant RSFC within default mode network (DMN) subnetworks could be a network-based biological marker for disease etiology and an indicator for clinical symptoms of MDD (Liu et al., 2018).
The posterior cingulate cortex (PCC), the core region of the DMN, plays a pivotal role in self-referential processing (Greicius et al., 2004;Johnson et al., 2006). One important function of the PCC is the facilitation of constructing mental models of personally significant events with two subsystems within the DMN, namely the medial temporal lobe (MTL) subsystem and the dorsal medial prefrontal cortex (dmPFC) subsystem (Andrews-Hanna et al., 2010). Empirical studies indicated that the dmPFC subsystem is activated in response to self-relevant mental simulations such as mentalizing scenarios and meta-cognitive processes of reflecting or deducing upon the present mental states of one's self and others (Frith & Frith, 2003), while the MTL subsystem involves in mnemonic scene construction, an important component process of future thinking (Andrews-Hanna et al., 2010;Hearne et al., 2015;Spalding et al., 2018). Taken together, sub-networks of DMN may contribute to rumination and its sub-component process in MDD, for its intrinsic relation with the psychological processes of self-referential processing and autobiographical memory (Zhou et al., 2020).
Interestingly, deviant RSFC of the PCC have been found to be correlated with remarkable clinical manifestations of depression, such as depressive severity and medication status, relying on distributed brain regions spanning multiple lobes. Independent component analysis (ICA), Zhu et al. (2012) found that functional connectivity between the PCC and angular gyrus involves in overgeneral autobiographical memory phenomena of MDD patients. Later advance demonstrated that the connection between the PCC and right superior/ middle frontal gyrus was inversely correlated with rumination among remitted MDD youths (Jacobs et al., 2014). Fresco et al. (2017) posited that connectivity between the PCC and anterior insula would be associated with improvements in decentering followed by emotion regulation therapy. Importantly, the PCC is a hub of the DMN that interacts with other predominant neurocognitive networks such as the salient network (SN) and central executive network (CEN). Menon (2011) proposed a unifying triple network model that focuses on these three core networks and postulated the within-and across-network interactions may characterize the cognitive and affective dysfunction in multiple psychiatric and neurological disorders. For example, a recent study using dynamic causal modelling (DCM) observed attenuated effective connectivity (EC) from the SN to DMN network, which may induce altered self-referential mental activity (i.e., excessive rumination) in MDD (Li, Liu, et al., 2020;Li, Zhao, et al., 2020). Previous studies have found that the RSFC between the DMN and CEN may be a response specific to electroconvulsive therapy (ECT) or antidepression treatment (e.g., sertraline) (Brakowski et al., 2017;Yang et al., 2016). Additionally, Kaiser et al. (2015) revealed hyperconnectivity between frontoparietal systems and regions of the DMN, which may associate with the presence of psychotic symptoms in severe late-life depression (Oudega et al., 2019). That is, exploring the potential moderators of RSFC may open new avenues for facilitating precise diagnosis and guiding new treatment strategies.
Evidence suggests striking differences in PCC connectivity with whole brain between patients with MDD and healthy control subjects (HCs) (Berman et al., 2014;Bluhm et al., 2009;Yang et al., 2016). However, these results are inconsistent across studies. For instance, on the one hand, some studies reported increased RSFC in depressed participants compared with HCs between the region of interest (ROI) at the PCC with several cortical and subcortical midline brain regions, such as bilateral precuneus, anterior cingulate cortex, medial prefrontal cortex (Alexopoulos et al., 2012;Andreescu et al., 2013). On the other hand, some studies reported decreased RSFC of the PCC with the bilateral caudate, thalamus, amygdala and the temporal cortex (Bluhm et al., 2009;Chase et al., 2014). This discrepancy could be attributed to several reasons, such as variations in the definition of the PCC seed, small sample size, subject heterogeneity, and different analytical methods across studies. To address this issue, a quantitative meta-analysis could help identify the most prominent and consistent PCC-based RSFC alterations in MDD by controlling the above-mentioned confounding factors. Importantly, to interpret whether BOLD-related differences reflect underlying morphological differences or whether they express pure hemodynamic differences, conducting conjunction of PCC-based VBM analysis and the RSFC analysis may be of necessity. Spontaneous alter in both neural activity and anatomy of certain regions may provide more stringent evidence for the underlying biomarkers for MDD .
The present study aimed to conduct a meta-analysis to unify findings of RSFC studies and identify consistent PCC-based RSFC changes in MDD. In addition, to explore the potential moderators moderating the PCC-based RSFC patterns, we performed an exploratory meta-regression analysis through the association between the PCC-based RSFC and the available clinical variables (e.g., depressive symptoms). Anisotropic effect-size version of seed-based d mapping (AES-SDM), a powerful tool based on wellestablished statistical measures for meta-analytic studies on differences in brain activity and spatial correlations between voxels (Radua & Mataix-Cols, 2012), was employed in the present study to integrate the results of the PCC-based RSFC abnormalities and delineate between-study heterogeneities. Moreover, we also set out to probe the structural basis of brain functional alterations by combining VBM and RSFC analysis. Considering previous observations of deviant RSFC between the PCC and voxels throughout the whole brain, we hypothesized that: (i) PCC would show consistent abnormal RSFC to the cortical and subcortical regions in MDD; and (ii) the identified altered RSFC might be moderated by clinical variables (e.g., depressive severity and antidepressant status).

Literature search and selection
A comprehensive literature search was conducted in Web of Science and PubMed for articles up to January 30, 2020, using the keywords "rest*(-ing)", "connect*(-ivity)", "depress*(-ion, -ive)" and "posterior cingulate". The reference section of the studies that fulfilled the inclusion criteria was manually screened to identify additional studies. For the study that the seed ROI or peak effect coordinates were not provided, authors were contacted for these information. Under this strategy, 471 papers were identified in total. The inclusion criteria were as follows: (i) articles were published in peer-reviewed English language journals; (ii) reported group comparison between MDD patients and healthy controls; (iii) examined the whole-brain RSFC of a seed in the PCC; (iv) reported coordinates in a standard stereotaxic space such as Talairach or Montreal Neurological Institute (MNI); and (v) reported rs-fMRI scan results including treatment studies with pretreatment (baseline) assessment. Studies of brain injury or other organic lesion were excluded. Exclusion criteria for patients were the presence of axis-I disorders other than MDD. The screening procedure for the current meta-analysis was prepared according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statements (Moher et al., 2009). The detailed screening protocol is illustrated in Fig. 1.

Quality assessment and data extraction
To achieve high standard of analysis, we assessed the quality of included studies by using a 12-points checklist containing the demographical and clinical characteristics of subjects as well as methods implemented during the imaging procedure (Li, Liu, et al., 2020;Li, Zhao, et al., 2020) (see Table S1 in Supplementary). Each point was scored as 0, Fig. 1 Flow chart of the identification of articles. BD, bipolar disorder; PCC, posterior cingulate; NH, network homogeneity; ReHo, regional homogeneity; ICA, independent component analysis; VBM, voxel-based morphometry; ALFF, amplitude low-frequency fluctuation 0.5 or 1 if the criteria were unfulfilled, partially met or completely met, respectively. All studies that met the inclusion criteria should pass the fractional line of quality assessment set ahead, scoring > 6.0. Specific information, including the demographic and clinical features (sample size, mean age, gender) and illness-related variables (duration of illness, mean age of first onset, medication status, score of severity) as well as scanning parameters (field strength of scanner, repetition/echo time) were retrieved as the basic data (see Table 1 and Table S2 in Supplementary) and were used to perform quality assessment in accordance with the checklist. A higher score represents a higher quality of original articles. Additionally, the current meta-analysis was based on coordinates showing the locations of significant group differences in RSFC. Therefore, coordinates of the PCC-seed ROI, and the corresponding peak of each significant between-group effect were extracted from each study. Following Kaiser et al., we categorized the disorder related effects into hypo-connectivity or hyper-connectivity in MDD group compared with HCs of each study (Kaiser et al., 2015). Hypo-connectivity (i.e., MDD < controls) was defined as reduced positive or increased negative RSFC in MDD versus controls; hyper-connectivity (i.e., MDD > controls) was defined as reduced negative or increased positive RSFC in MDD versus controls.

SDM meta-analysis
AES-SDM (version 6.12, http:// www. sdmpr oject. com) was employed in a coordinate-based meta-analysis. According to the guidelines of AES-SDM, we retrieved the reported peak coordinates and their corresponding t values and saved them into a text file (Radua et al., 2014). All z values were straightforwardly transformed into t values for data analysis using the SDM online converter. In case of studies not reporting any measure related to effect size (t score, z score, p value, etc.), "n" for hypo-connectivity or "p" for hyperconnectivity was written for the SDM software to conduct a pre-analysis to provide an effect size for these peaks. Data extracted from the 17 studies also included sample sizes (for both MDD and HC groups), software package for data analyses (FSL, SPM, or other), and coordinate space (Talairach or MNI space). The t-value of each peak was then calculated and converted to Hedges' g, a measure of standardized mean difference appropriate for small sample sizes (Emch et al., 2019). Once the datasets were prepared, a mean effect-size brain map was recreated in the preprocessing of the present meta-analysis, combing the data of each original study and demonstrating both negative and positive differences in the same map. The map was restricted in brain gray matter with a setting of anisotropy = 1, isotropic full-width at halfmaximum (FWHM) = 20 mm, and p = 0.005. Meanwhile, to evaluate the impact of anxiety symptoms on the RSFC of MDD patients, 11 studies were defined as subgroup by discarding 6 studies containing MDD patients with anxiety disorder diagnosis (Berman et al., 2011;Bluhm et al., 2009;Gaffrey et al., 2012;Jacobs et al., 2016;Peters et al., 2016;Satyshur et al., 2018). SDM meta-analytic study was also conducted for the subgroup, for comparing with the RSFC results of all included studies.

Voxel-based morphology analysis (VBM) analysis
To compare the structural differences between MDD patients and controls, AES-SDM was also used in combination with Sleuth software (version 3.0.4, http:// www. brain map. org/ sleuth/) for the VBM analysis. Search selections were performed in the VBM database to filter articles which: (i) included patients diagnosed with MDD; (ii) reported gray matter as experiments contrast; and (iii) reported MRI as imaging modality. To further determine qualification, studies were excluded if they: (i) had no HC group; (ii) included MDD patients with comorbid conditions except anxiety disorders; and (iii) reported no group comparison between MDD patients and healthy controls. After that, the qualified foci were exported in form of MNI coordinates and pooled to conduct a meta-analysis of VBM. Results were thresholded using FWHM = 20 mm, p = 0.005 and cluster extent = 10 voxels. To examine the functional alterations of PCC in MDD, the regions showing consistent abnormal gray matter density in MDD were overlapped with the PCC based FC alteration, which were then coded as the conjunction output. The overlapped regions might indicate the structural basis of the functional deviations of PCC in MDD.

Sensitivity analyses
To test the robustness of the main results, we conducted Jackknife sensitivity analysis or "leave-one-out cross validation" (LOOCV) analysis which consists of repeating the mean analysis by systematically excluding a different study each time and repeating the analysis (Radua & Mataix-Cols, 2009). If a previously significant brain region remains significant in most or even all of the LOOCV analyses, it can be concluded that this result has a high replicability (Duko et al., 2020). Any result presenting voxel threshold p < 0.005 with a peak Z > 1 and cluster extent > 10 voxels was regarded to have significantly elevated heterogeneity.

Meta-regression
Meta-regression, a kind of linear model analysis, was applied to explore potential sources of heterogeneity among studies. Variables explored by regression were the depressive symptom severity (the score of the Hamilton Rating Scale for Depression (HAMD)) of MDD patients and percentage of medication-naïve patients. According to the recommendation that meta-regression requires at least 9 studies (Radua & Mataix-Cols, 2009), we excluded the following variables: mean age of first onset, illness duration, and score of severity taped except HAMD. A stringent voxel threshold of p < 0.005, a peak of Z > 1 and a cluster extent threshold of 10 voxels was applied to minimize the risk of false-positive findings. In our result, the regression line (meta-regression SDM slope) was presented as a straight line and the SDM value derived from the proportion of studies that reported RSFC changes near the voxel, representing the difference of RSFC strength between MDD patients and HCs.

Included studies and sample characteristics
Seventeen studies met the inclusion criteria of the present study, with a total of 804 MDD patients and 724 healthy individuals. One of the studies did not report the sex ratio of participants. Patients with major depression from included studies were diagnosed using the Diagnostic and Statistical Manual of Mental Disorder-IV criteria (DSM-IV; 16 studies) or the Kiddie Schedule for Affective Disorders and Schizophrenia Present and Lifetime Version (KSADS-PL; one study) (Kaufman et al., 1997). Furthermore, 5 studies recruited drug-naïve patients, while 1 study included off medication but previously treated patient, and 8 studies reported samples combining offmedication and drug-naïve patients. Notably, the minimum time off medication of patients in the included studies was at least 2 weeks (2 weeks, 2 studies; 3 months, 1 study; 6 months, 1 study) according to the four studies that reported the duration of the medication-free period prior to the scan Liu et al., 2018;Satyshur et al., 2018;Zhang et al., 2015). All of these patients were also defined as "drug-naïve" in the present study, while in the rest three of the included studies patients were defined as medication state (Andreescu et al., 2013;Chase et al., 2014;Peters et al., 2016). Ultimately, this meta-analysis incorporated 83 peak coordinates extracted from all the included studies, 50 of which reported increased PCCbased RSFC in patients versus healthy controls and 33 revealed decreased RSFC. The mean quality score of these studies was 10.2 (total score was 12), indicating high quality. The detailed demographic and clinical characteristics of included studies and methodological details are listed Tables 1 and Table S2 in Supplementary. cortex and its relation to functional specialization

VBM results and conjunction of functional and structural alterations in MDD
By using Sleuth, 45 papers with 144 experiments were identified. After filtering these articles with exclusion criteria, 38 studies with 447 foci were included in the present VBM study. Patients with MDD exhibited consistently decreased gray mater density mainly in the right inferior frontal gyrus extending to the insular cortex compared with healthy controls. In contrast, the region that showed significant increased gray mater density was in the right cuneus cortex (Fig. 3b and Table S3 in supplementary). Importantly, the left MFG and the left MTG consistently overlapped in the VBM results and the functional connectivity alterations in MDD (Fig. 3c).

Sensitivity test for the main finding
Jackknife sensitivity analysis was conducted to test the robustness of the results. Table S4 lists the increased RSFC in the patients with PCC-seed in the right MFG across all studies, indicating a good reproducibility of these outcomes. Meanwhile, hypoconnectivity between the PCC and the left SFG remained significant in all but one combination of 17 datasets. The left MTG was stably hyperconnected with the PCC in MDD patients, which were preserved in 15/17 of datasets. The result in the left precuneus remained significant in all but three combinations. Of note, hypoconnectivity between the PCC and the right precuneus failed to emerge in most of the study combinations (13 of 17), indicating that this result lacks sufficient robustness and is required to be ruled out in discussion (Table S4 in supplementary).

Meta-regression
The HAMD score of patients was found to be negatively correlated with the extent RSFC changes between the right MFG and the PCC (Z = -3.186, p < 0.001; Fig. 4a and Table 3). In addition, the extent of RSFC changes between PCC and the right supplementary motor area (SMA) is associated with the percentage of drug-naïve of MDD patients of included studies (Z = 3.686, p < 0.001; Fig. 4b and Table 3).

Discussion
The current meta-analysis reviewed 17 studies reporting whole-brain RSFC of the PCC seed and revealed the altered RSFC of the PCC in MDD compared with controls. Specifically, we identified significant elevated RSFC of the PCC with the left MTG extending to the left inferior temporal gyrus as well as the right MFG in MDD. In addition, the PCC exhibited reduced RSFC with the left SFG and left precuneus. Furthermore, regions with aberrant connectivity had a partial overlap with regions that showed decreased gray matter density, namely the left MTG as well as the left MFG. These results remained largely unchanged in the whole-brain Jackknife sensitivity analysis.

Increased RSFC between MTG and PCC in depression
Hyperconnectivity was found in individuals with MDD between the PCC and left MTG which is an essential part of the dorsal attention system (DAN). The MTG is closely related to the successful decision making in the presence of a stimulus conflict and participates in emotional processing, selective attention and working memory (Corbetta & Shulman, 2002;Fox et al., 2006). Gaffrey et al. (2012) suggested that children with a history of preschool depression demonstrated increased RSFC between the MTG and PCC and such altered RSFC was associated with clinically relevant behavior such as decreased use of proactive emotion regulation strategies. Also, emerging evidence has indicated that abnormal MTG-DMN RSFC may contribute to negative thinking mode and pessimistic emotional experience in MDD patients, triggering self-referential processing and rumination (Ma et al., 2012). Thus, the current meta-analysis provides further evidence to support the involvement of the PCC and MTG in the underlying pathophysiology of MDD.

Altered RSFC between the dlPFC and PCC in depression
Compared with controls, the RSFC strength increased between the PCC and right MFG, and decreased between the PCC and left SFG in MDD. The SFG and MFG are the sub-regions of the dlPFC which is a core part of the CEN.
The CEN participates in a variety of executive functions, such as response inhibition, working memory and emotional regulation (Habas et al., 2009;Levy & Goldman-Rakic, 2000;Zhang et al., 2017). It is worth noting that, the SFG is critical for mental manipulation and monitoring information within higher levels of working memory processing (du Boisgueheneuc et al., 2006;Petrides, 2000). Evidence from neuroimaging studies suggested that the CCN is linked to memory impairment and attentional bias experienced by depressed patients (Rogers et al., 2004;Sheline et al., 2010). Hence, the failure of control and regulation from the CCN to the PCC may result in impaired cognitive control over negative emotions, ability to synthesize memories and decision making observed in depression. Fig. 4 Results of meta-regression. a meta-regression with the depressive severity (HAMD score); (b) meta-regression with the percentage of drug naïve patients. SMA, supplementary motor area The current study also found an increased RSFC between the PCC and the MFG in patients. The MFG involves in bottom-up/stimulus-driven orienting of attention and usually has an antagonistic relationship with the DMN and the attentional control originated from the middle frontal gyrus which helps to regulate the DMN activity (Fox et al., 2006;Kucyi et al., 2012). In our previous work, we have suggested that the enhanced RSFC between prefrontal cortex and DMN could be an adaptive process in response to disorder, and then it is conceivable that the adaption process is signaled by depression symptom severity (Zhang et al., 2020). Moreover, the regulation from the MFG to the DMN activity may also function as the disruption of depressive rumination by driving the attention away from current event, which might therefore alleviate the sad mood and negative cognitive deviation of a patient, at least temporarily (Andrews & Thomson, 2009;Morrow & Nolen-Hoeksema, 1990). It is in line with our finding of exploratory meta-regression that the extent of RSFC changes between PCC and MFG is negatively correlated with the HAMD score of depressed patients. Hence, this may help explain why depressed patients are able to control over depressive symptoms such as rumination and negative internalizing states during remitted period . Likewise, the deviations of connection between the PCC and SMA may also facilitate the medication-based alleviation in MDD symptoms, especially in psychomotor disturbance. The SMA, a crucial area for self-initiated movement, is expected to involve in high-order executive function that links to motor planning and initiation (Cunnington et al., 2002;Kohn et al., 2014). Consequently, future studies with longitudinal evidence may provide biomarkers for the use of motor training for MDD patients (Sarkheil et el., 2020).

Decreased RSFC between the PCUN and PCC in depression
The present study also revealed hypoconnectivity in MDD between the PCC and left precuneus, an important subregion of the DMN. Precuneus plays a particular role in cause-and-effect judgements, which may lead to the selfreferential processing ruminations (Jones & Bhattacharya, 2014;Lai, 2018). As suggested in a previous study, the connection between the precuneus and PCC involves in selfrelevant cognitive activities, such as the self-reflection and the maintenance of a sense of self-consciousness (Fransson & Marrelec, 2008). In line with our finding of decreased PCC-precuneus RSFC in patients with MDD, another fMRI study also reported decreased RSFC of the PCC-precuneus in the late-onset depression group (Liu et al., 2018) and such alteration can be interpreted as failure to attenuate selfreferential activity to a situationally appropriate manner in depression (Sheline et al., 2009).

Conjunction of functional and structural abnormalities
Structural basis of aberrant PCC-based RSFC results in MDD was assessed via conjunction of functional and structural deviations. Regions with altered RSFC had a significant overlap with regions that revealed decreased gray matter density, which were mainly located in the left MTG as well as the left MFG. Generally speaking, structural deficits may represent more stable and long-standing changes, while functional alterations measured by fMRI relate to acute illness stage and are more sensitive to drug utilization (Ren et al., 2013), which could be a possible explanation for those regions that only exhibited RSFC deficits. Several studies have detected the prominent loss of the gray matter volume in clusters encompassing the MFG and MTG, which may weaken the capacity of completing complex cognitive tasks of emotional significance in depression (Arnone et al., 2016;Kandilarova et al., 2019;Lai & Wu, 2014). It may be noted, another previous study has indicated significantly smaller gray matter volume in the right MTG in MDD patients with suicide history than in those who had not (Peng et al., 2014). Hence, it is suggested that the left MTG and the left MFG may be crucial to the underlying pathological mechanism of MDD, but future investigations are still needed to further confirm.

Limitations
Several limitations should be acknowledged in this study. First, the statistical analytic methods implemented in the included studies were limited to seed-based RSFC for the current meta-analysis, without using ICA or seed-derived network studies. Such strict strategy may cause certain omission but indeed provide a clear presentation of the altered connectivity pattern of the PCC. Secondly, we acknowledge that the limited number of included studies may impact the reproducibility of the results. Eickhoff et al. (2016) recommended to include at least 20 experiments in all types of coordinate-based meta-analyses to derive sufficient power for moderate effects. 17 experiments are enough to control the top-contribution of the most dominant experiments (Eickhoff et al., 2016). Thirdly, exploratory analysis of age-related differences in MDD patients was not perform due to inadequate longitudinal empirical data, despite of the wide age range in the included studies. Even though meta-regression analysis revealed a significant association between age and the aberrant PCC-based RSFC, the potential impact of age on human brain's RSFC is remained to be explored. Fourthly, the studies we included were cross-sectional, so the inference of a causal link between the brain dysfunction and clinical features still require more valid supportive evidence. Thus, future studies should consider employing Granger causality (GC) or DCM method that may confirm the direction of the connections in current result. Last but not least, the size and location inconsistency of the seed ROI definitions were observed among the included studies, which may limit the accuracy of our results. Future investigations such as overlap rate analysis on the PCC seed definitions will be necessary to more fully address this question.

Conclusion
The present meta-analysis identified significant and consistent PCC seed-to-whole-brain RSFC alterations in MDD, including hyperconnectivity of the PCC with the middle temporal gyrus and the middle frontal gyrus and hypoconnectivity with the superior frontal gyrus and the precuneus. And such functional deviations may have structural basis, especially in the left middle temporal gyrus and the left middle frontal gyrus. These functional abnormalities suggested the disrupted executive control in MDD, which may account for the core symptom of rumination. The findings may shed light on the neuropathology mechanisms of depression.