Journal of Affective Disorders 256 (2019) 148–155
Contents lists available at ScienceDirect
Journal of Affective Disorders journal homepage: www.elsevier.com/locate/jad
Research paper
Hippocampus-driving progressive structural alterations in medication-naïve major depressive disorder
T
Yuting Lia,†, Chun Wanga,b,c,d,†, Changjun Tengc, Kaili Jiaoc, Xiu Songc, Yarong Tanc, ⁎ Chaoyong Xiaoc, Ning Zhangc,d,e, Yuan Zhonga,b, a
School of Psychology, Nanjing Normal University, Nanjing 210097, China Jiangsu Key Laboratory of Mental Health and Cognitive Science, Nanjing Normal University, Nanjing 210097, China c Nanjing Brain Hospital Affiliated to Nanjing Medical University, Nanjing 210029, China d Functional Brain Imaging Institute of Nanjing Medical University, Nanjing, 210029, China e Cognitive Behavioral Therapy Institute of Nanjing Medical University, Nanjing, 210029, China b
A R T I C LE I N FO
A B S T R A C T
Keywords: Hippocampus Major depressive disorder Causal network of structural covariance Morphometric MRI Granger causality
Background: Major depressive disorder (MDD) is associated with abnormalities in brain structure. However, structural abnormality findings have been inconsistent and how structural changes lead to progressive morphometric alterations in depressed brain regions remains unclear. Methods: High-resolution T1-weighted magnetic resonance images of first-episode medication-naïve MDD patients (20 men, 36 women) and healthy control participants (33 men, 23 women) were evaluated. Voxel-based morphometry analysis was conducted based on T1-weighted images. The causal network of structural covariance analysis (CaSCN) was accomplished by applying Granger causality analysis to the sequenced T1-weighted images in order to assess causal effect of structural changes. Results: When comparing MDD patients and healthy controls, gray matter was greater in the bilateral amygdala, the bilateral hippocampus, the left parahippocampus, and the right fusiform, while it was lessened in the bilateral brainstem, the bilateral pallidum, and the bilateral thalamus. Selecting the hippocampus as the seed region to run further CaSCN analysis revealed that the hippocampus is a prominent node that exerts a causal effect on the amygdala and regions of the default mode network. Limitations: Our sample size was small and the subjects groups’ ages were not well matched. We also recognize that the hippocampus is not necessarily the original source of brain network alteration in MDD. Conclusions: The CaSCN clarified the causal relationship between progressive gray matter alterations in the hippocampus and in other regions. Our work provided evidence of a network spread mechanism in terms of the causal influence of hippocampal alteration on progressive brain structural alterations in MDD.
1. Introduction Major depressive disorder (MDD) is associated with abnormal interactions between brain regions that regulate emotion and cognitive functions (Korgaonkar et al., 2014). Morphometric studies have found gray matter (GM) alteration in regions that regulate emotional and cognitive functions such as the amygdala, hippocampus, anterior cingulate cortex, insula, and thalamus in MDD (Czléh and Lucassen, 2007; Du et al., 2012; Frodl et al., 2003; Lai, 2013; Peng et al., 2011; Sacher et al., 2012). However, results concerning expanded or atrophied GM in
these key regions remain inconclusive. Some researchers believe MDD is a progressive brain illness and studies have focused on whether there exists a correlation between illness progression information (such as illness duration) and morphometric alteration (Chen et al., 2016; Mckinnon et al., 2009). Several studies have revealed that both the hippocampus and amygdala show different alteration patterns depending on the stage of MDD (Alexanderbloch et al., 2013; Eijndhoven et al., 2009; Frodl et al., 2003). Furthermore, Arnone et al. found that hippocampal abnormality was considered an important contributor to the pathogenesis of
Abbreviations: MDD, Major Depressive Disorder; GM, Gray Matter; SCN, Structural Covariance Network Analysis; GCA, Granger Causal Analysis; CaSCN, Causal Structural Covariance Network Analysis; GMV, Gray Matter Volume; HRSD, Hamilton Rating Scale for Depression, VBM, Voxel-based Morphometry ⁎ Corresponding author at: Nanjing Normal University, School of Psychology, 122 Ninghai Rd. Nanjing, 210097, China. E-mail address:
[email protected] (Y. Zhong). † Li Yuting and Wang Chun should be considered as joint first author. https://doi.org/10.1016/j.jad.2019.05.053 Received 7 July 2018; Received in revised form 5 May 2019; Accepted 27 May 2019 Available online 28 May 2019 0165-0327/ © 2019 Elsevier B.V. All rights reserved.
Journal of Affective Disorders 256 (2019) 148–155
Y. Li, et al.
Rating Scale for Depression (HRSD) and all scored less than 8. Anatomical images obtained by magnetic resonance imaging (MRI) were evaluated for both groups by an experienced radiologist to exclude participants with bad imaging quality or gross brain abnormalities. Patients were first sorted according to illness duration and HRSD score. Illness duration was defined as time span from the first episode of depression to the scan time. Then, patients were artificially divided into two equal groups according to the median illness duration (0.7 years). Written informed consent was obtained from all patients and healthy control participants after a complete description of the current study. This study was approved by the Medical Ethics Committee in Nanjing Brain Hospital.
depression and gray matter reduction in the hippocampus appeared to be specific to the depressed state. Taken together, this data suggests that GM alteration of the hippocampus was associated with illness duration, although a distinct correlation between illness duration and GM alteration in MDD has yet to be established. The relationship between GM alterations in MDD is rarely explored and remains uncertain despite the availability of many analysis techniques. One technique is Structural covariance network analysis (SCN), a powerful tool that can evaluate interregional gray matter relationships. SCN has previously been used to investigate structural connectivity in MDD, the results of which revealed that the structural covariance patterns in the emotion network (such as the amygdala, angular, and posterior cingulate cortex) were disrupted in MDD (Wu et al., 2016). Although SCN has identified synchronized GM alterations in MDD, it cannot reveal causal effect between regions as it neglects time information (Alexanderbloch et al., 2013). Another technique, Granger causal analysis (GCA), has been identified as a useful method for revealing causal effects among brain regions and has been used in functional time-series data analysis for MDD. Assuming that gray matter volume images can be accurately ordered according to progressive information such as illness duration, GCA can be applied to the sequenced data in order to construct a causal network of structural covariance (CaSCN) (Jiang et al., 2018; Zhang et al., 2016). Unlike SCN, which is merely able to measure the synchronization of interregional morphometric alterations, CaSCN is assumed to describe progressive network damage and assessing the causal relationships of morphometric alterations among brain regions. Thus, by sequencing cross-sectional imaging data according to illness progression information, GCA can be used to explore the causal relationship between structural alterations. In summary, the gray matter alteration pattern in MDD patients remains uncertain, the relationship between structural alterations and illness duration is still unclear, and the causal relationship between gray matter alterations is ambiguous. We chose CaSCN as the main method of analysis and included only first-episode medication-naïve patients with MDD in the current study. We hypothesized that MDD would exhibit altered gray matter in regions thought to be involved in MDD, such as the hippocampus and amygdala, and that gray matter alterations might be related to illness duration. Furthermore, we postulated that when morphometric MRI images are ranked according to illness duration, CaSCN is assumed to be used to characterize the progressive structural alterations of brain regions in first-episode medication-naïve patients with MDD.
2.2. MRI scans MRI images (3.0 Tesla Magneton Vision, Siemens, Erlangen, Germany) were obtained using a fast spoiled gradient-echo sequence (repetition time [TR] = 1900 ms; echo time [TE] = 2.48 ms; inversion time = 900 ms; total acquisition time 4.18 min; number of acquisitions = 32; flip angle = 9°; field of view [FOV] = 256 mm × 256 mm; matrix 256 × 256, slice thickness = 1 mm). A birdcage head coil fitted with foam padding and earplugs was used to reduce the head movement and noise. 2.3. Data preprocessing Voxel-based morphometry (VBM) analysis of high-resolution T1weighted images was performed with the Computational Anatomy Toolbox (CAT12; http://dbm.neuro.uni-jena.de/cat12/) implemented in statistical parametric mapping software (SPM12; http://www.fil.ion. ucl.ac.uk/spm/software/spm12). First, all images were checked for artifacts. Then, images were transformed into MNI space by normalizing to a symmetric template with a 12-parameter affine-only nonlinear transformation and resampled to 1.5 × 1.5 × 1.5 mm3. All images were segmented into three tissue classes representing gray matter, white matter, and cerebrospinal fluid. Finally, all the resultant probabilistic gray matter maps were smoothed with an 8-mm FWHM isotropic Gaussian Kernel for ensuing morphological analyses. The generated smoothed GM images were used for subsequent group comparisons. The statistical analysis pipeline can be clearly seen in Fig. 1. 2.4. Statistical analysis
2. Methods
2.4.1. VBM and correlation analysis To estimate the overall gray matter volume alteration in first-episode medication-naive MDD, the two-sample t-test implemented in SPM12 was applied to compare the gray matter images of first-episode medication-naive MDD patients and healthy controls (p < 0.01, FWE corrected). To find the network regions showing progressive gray matter volume alteration, correlation analysis (p < 0.05, FDR corrected) was performed between gray matter images and illness duration. Total intracranial volume, gender, and age were regressed as covariates in all above analyses.
2.1. Participants A total of 56 unmedicated first-episode major depressive disorder patients were recruited from outpatients of the Department of Medical Psychology of Affiliated Nanjing Brain Hospital. Psychiatric diagnoses were determined by the consensus of two expert psychiatrists who concurred on a diagnosis based on the Structural Clinical Interview for DSM-IV-TR Axis I/Patients (SCID-I/P). Inclusion criteria included: (i) first-episode of MDD, (ii) 20–50 years old, (iii) right-handedness. Exclusion criteria included: (i) history of serious physical or neurological illness, (ii) substance abuse or mental retardation, (iii) significant life changes within 6 months, (iv) contraindication to MRI scan, (v) personality disorder, (vi) other mental illnesses. None of the patients were ever treated with psychotropic medication. For comparison, 56 right-handed healthy control participants (HCs) aged 20–50 years old with no history of psychiatric disorders, neurological disorders, or serious physical disease were socially recruited. As healthy control participants were collected somewhat randomly, the ages of the patient and healthy control groups were not well matched. Nevertheless, participants in the HC group were evaluated with the 24-item Hamilton
2.4.2. Mapping stage-specific GM alteration patterns and synchronized GM alteration with the hippocampus In order to investigate stage-specific gray matter volume alterations in the MDD group, the two-sample t-test was performed. Given that stage-specific comparisons can reveal the dynamic gray matter alteration patterns in different depression states, in the current study we compared two subgroups and aimed to determine whether the gray matter alterations would be different as illness duration prolonged. In line with Zhang's work, patients with MDD were categorized into two subgroups according to illness duration and HRSD scores (the gray matter images were sequenced from low to high according to illness 149
Journal of Affective Disorders 256 (2019) 148–155
Y. Li, et al.
Fig. 1. Statistical analysis pipeline. Pipeline showing a step-by-step process of statistical analysis. (A) Map the overall and stage-specific GMV alteration. The twosample t-test was applied to map the overall and stage-specific GMV alteration. For the overall GMV alteration, the two-sample t-test was applied between the whole first-episode medication-naïve MDD group and the whole healthy control group. For the stage-specific comparison, the GM images of the first-episode medicationnaïve patients were firstly sequenced according to the illness duration & HRSD scores and then divided into two subgroups. Healthy control participants were also divided into two subgroups and the two-sample t-test was finally conducted between the patient subgroups and the health control subgroups. (B) Map the correlational relationship between GMV alteration and illness duration. The correlation test was applied between the GMV and illness duration. (C) Select the hippocampus according to the two-sample t-test and correlation test as the seed region in structural covariance network (SCN) and causal network of structural covariance analysis (CaSCN). (D) Map the hippocampus-associated SCN using the Functional Connectivity (FC). (E) Map the hippocampus-associated CaSCN using the Granger Causality Analysis (GCA). The GMV images were firstly sequenced according to the illness duration and HRSD scores, the progressing information (illness duration and HRSD scores) was seen as a “pseudo-time series”. Thus, coefficient-based CaSCN was voxel-wisely performed in the whole brain using GCA.
duration first, then the gray matter images with the same illness duration were sequenced from low to high according to HRSD score); each subgroup had 28 patients (Illness duration: stage I / II = 0.1–0.7 years / 0.7–8.0 years) (Zhang et al., 2016). Each subgroup's GM images were then compared to those belonging to healthy controls using the two-sample t-test (p < 0.01, FWE corrected). Total intracranial volume, gender, and age were all regressed as covariates in the analyses mentioned above. We chose the median duration as standard to separate the MDD group for two reasons: First, statistically, the median can divide a set of data into two equal subgroups (Walsh, 1949), and we wanted the two subgroups to be matched precisely. Second, Keller et al. reported that more than 50% of the MDD patients in his cohort recovered in 6 months (Keller et al., 1992, 1982) and the median duration of the patients is 0.7 years, which is close to 6 months. Illness duration in the current study can be seen as a no-treatment interval, which, according to Keller et al. (1986) and Scott et al. (1992), is highly interrelated with severity of illness, suggesting that the notreatment interval can reflect the severity of illness in MDD. In the current study, the illness duration was used to indicate the severity of the disease; we noticed that the longer the illness duration, the worse the disease. Based on results from the group two-sample t-test, the bilateral hippocampus was selected as a seed region for subsequent SCN analysis as it displayed significant GM expansions (t = 5.67, p = 5.8 × 10−8). The aim of SCN was to map synchronized GM alterations with the hippocampus. SCN was performed using multi-regression implemented in SPM12 (http://restfmri.net).
overall group two-sample t-test. CaSCN was performed using Granger Causality Analysis (GCA) implemented in REST1.8. The GM images of MDD patients were sequenced from short to long illness duration as a “pseudo-time series”. Voxel-wise coefficient-based CaSCN was conducted over the whole brain using GCA. By applying GCA to the artificial time series, CaSCN can evaluate the causal relationship between brain regions’ GM alterations. Also, by applying GCA to the pseudo-time series morphometric data through data sequence, CaSCN can estimate the causal effect of the structural alteration of one region on others. Assuming that the hippocampus is the primary region involved in GM alteration in MDD, we adopted a GC value of X to Y in the current research, and the original GC map was transformed to a Z score map. The results of CaSCN are presented at a threshold of p < 0.05, FDR corrected.
3. Results Demographic and clinical data of patients and healthy controls 56 first-episode medication-naïve MDD patients and 56 healthy control participants were evaluated. 56 patients (20 men and 36 women; mean ± standard deviation age: 35.1 ± 8.9 years; range: 21–50 years; mean ± standard deviation illness duration: 1.1 ± 1.3 years; range: 0.1–8.0 years; mean ± standard deviation HRSD score: 25.1 ± 5.1; range: 18–35) and 56 healthy controls (23 men and 33 women; mean ± standard deviation age: 30.3 ± 8.0 years; range:20–36 years) composed the final cohort. For the stage-specific comparison, 28 patients (10 men and 18 women; mean ± standard deviation age: 34.2 ± 9.7 years; range: 21–51 years; mean ± standard deviation illness duration: 0.4 ± 0.2 years; range: 0.1–0.7 years; mean ± standard deviation HRSD score: 24.0 ± 4.7; range: 18–34) with shorter illness duration & lower HRSD scores and 28 healthy control participants (15 men and 13 women; mean ± standard deviation age: 34.1 ± 6.4 years; range: 24–44 years) were included in the
2.4.3. Mapping causal effect of GM alteration patterns with seed-based CaSCN To map causal effect of hippocampal alteration on whole-brain gray matter volume alteration, CaSCN was conducted using the hippocampus as a seed region, which had previously been identified from the 150
Journal of Affective Disorders 256 (2019) 148–155
Y. Li, et al.
corrected). The seed region was selected from the hippocampus, which showed significantly expanded gray matter volume in the group twosample t-test. The bilateral insula, bilateral temporal lobe, anterior cingulate cortex, and posterior cingulate cortex all demonstrated synchronized GM alteration with the bilateral hippocampus in MDD (Fig. 4).
Table 1 Demographic and clinical data of patients and healthy controls. Group Overall group
Stage I
Stage II
Age Gender (male/female) Duration HRSD score Age Gender (male/female) Duration HRSD score Age Gender (male/female) Duration HRSD score
HC
MDD
t / x2
p
30.7 (8.0) (23/33) – – 34.1 (6.4) (13/15) – – 27.4 (8.1) (10/18) – –
35.1 (8.9) (20/36) 1.1(1.3) 25.1 (5.1) 34.2 (9.7) (10/18) 0.4 (0.2) 24.0 (4.7) 35.9 (8.1) (10/18) 1.8 (1.6) 26.2 (5.4)
2.70 .34 – – .05 1.29 – – 3.95 0.00 – –
< 0.01 .56 – – .96 .29 – – < 0.01 1.00 – –
3.4. Hippocampus-associated CaSCN in first-episode medication-naïve patients with MDD Causal structural covariance network analysis (CaSCN) can show causal effect of regional changes. Assuming that the hippocampus is the initial region of GM alteration in MDD, the GC value of X to Y was adopted in the current study. The hippocampus, which showed significantly expanded gray matter volume, was selected as the seed region. Positive granger causal values were distributed in the precuneus, the left middle temporal lobe, the right angular gyrus, the left inferior parietal lobe, the postcentral gyrus, and the right amygdala. Negative granger causal values were found in the left hippocampus (Fig. 5). The results are also summarized in Table 5.
Gender differences are reported as frequency (n) and were evaluated with Chisquare contingency test, the other demographics are reported as Means (SD) using two-sample t-test. Abbreviations: HC, healthy controls; MDD, major depressive disorder; HRSD, Hamilton Rating Scale for Depression.
comparison of stage I; 28 patients (10 men and 18 women; mean ± standard deviation age: 35.9 ± 8.1 years; range: 25–50 years; mean ± standard deviation illness duration: 1.8 ± 1.6 years; range: 0.7–8.0 years; mean ± standard deviation HRSD score: 26.2 ± 5.4; range: 18–35) with longer illness duration & higher HRSD scores and 28 healthy control participants (18 men and 10 women; mean ± standard deviation age: 27.4 ± 8.1 years; range: 20–46 years) were included in the comparison of stage II. More detailed demographic and clinical data of patients and healthy control participants are presented in Table 1.
4. Discussion The current study revealed gray matter abnormalities in first-episode medication-naïve patients with MDD relative to healthy controls; GM expansions were found in the bilateral amygdala, the bilateral hippocampus, the bilateral parahippocampus, and the right fusiform gyrus. Diminished GM was found in the bilateral brainstem, the bilateral pallidum, and the bilateral thalamus. Stage-specific alteration patterns were also observed as significant GM expansions were documented in the bilateral hippocampus in stage I compared to stage II. Correlation results showed that illness duration was associated with gray matter alteration of the hippocampus, the left inferior parietal lobe, the left inferior temporal lobe and the bilateral fusiform. Within the brain regions that displayed gray matter alterations, the hippocampus was the only region that showed a negative correlation between GM and illness duration. Thus, the hippocampus was selected as the seed region for further CaSCN analysis. Through the use of CaSCN analysis, the current study estimated the causal influence of a seed region, the hippocampus, on progressive structural alteration of depressive brain regions in first-episode medication-naïve patients with MDD. CaSCN analysis was performed in the current study by applying GCA to cross-sectional morphometric data. Compared to stage-specific and longitudinal studies, CaSCN has the advantage of describing network properties of involved regions (Bernhardt et al., 2013, 2009; Mcdonald et al., 2009; Zhang et al., 2016). Giving the cross sectional GM data an artificial time series by sequencing the illness progression information makes CaSCN a very innovative tool for analysis (Zhang et al., 2016). On the other hand, GCA connectivity indicates that one region's neural activity can precede and predict that of another region (Goebel et al., 2003; Hamilton et al., 2011; Jiao et al., 2011; Zhang et al., 2016). Thus, in line with GCA, CaSCN can reflect the structural alterations in seed regions that precede and predict the structural alterations that occur in other network regions in relation to illness duration or other illness progression information. The results of stagespecific comparisons and SCN analysis confirmed the results of the hippocampus-associated CaSCN. Although the stage-specific alteration patterns we observed in the hippocampus and amygdala cohered with previous research, some of our findings did not. For example, although we reported greater GM in the amygdala in MDD, many studies found it to be reduced (Bora et al., 2012; Buss et al., 2012; Eijndhoven et al., 2009; Romanczuk-Seiferth et al., 2014), a discrepancy that may be caused by the inclusion of different clinical features (Bora et al., 2012; Du et al., 2012; Mckinnon et al., 2009; Sacher et al., 2012). Also, the conclusion made by Arnone
3.1. Overall and stage-specific GM alteration patterns in MDD relative to healthy controls Group comparisons of GM between MDD and HCs can precisely show the pathological changes in first-episode medication-naive patients with major depressive disorder. Compared to the HC group and at a lower threshold (p < 0.01, FWE corrected), gray matter volume was greater in the bilateral amygdala, the bilateral hippocampus, the bilateral parahippocampus, and the right fusiform in MDD, while volume was atrophied in the bilateral brainstem, the bilateral pallidum, and the bilateral thalamus (Fig. 2A). The results can also be seen in Table 2. Stage-specific GM comparisons revealed dynamic morphometric alterations among network regions in first-episode medication-naive patients with major depressive disorder. For instance, we noted significant GM expansions in the bilateral hippocampus in stage I compared to stage II (Fig. 2B). The results can be also seen in Table 3. 3.2. Correlation analysis between GM and progressive factors of illness duration The correlation map showed dynamic GM alterations in network regions as illness duration prolonged. We discovered a positive correlation between illness duration and GM alteration in the left inferior parietal lobe, the left inferior temporal lobe and the bilateral fusiform while there was a negative correlation between illness duration and GM alteration in the bilateral hippocampus (Fig. 3). Of all the regions that displayed a correlation with illness duration, the hippocampus was the only region that also showed GM alteration, thus the hippocampus was selected as the region of interest. The results are also summarized in Table 4. 3.3. Hippocampus-associated SCN of first-episode medication-naïve patients with MDD Structural covariance network analysis was performed based on cross-sectional morphometric data of the MDD group (p < 0.05, FDR 151
Journal of Affective Disorders 256 (2019) 148–155
Y. Li, et al.
Fig. 2. (A) Gray matter volume alteration in first-episode Medication-naïve MDD Relative to HC. Group comparisons of GMV between first-episode medication-naive MDD and HC using the two-sample t-test (p < 0.01, FWE corrected) revealed expamded GMV in the bilateral amygdala, the bilateral hippocampus, the bilateral parahippocampus, and the right fusiform. Lessened GMV was found in the bilateral brainstem, the bilateral pallidum, and the bilateral thalamus. (B) Progressive patterns of stage-specific gray matter alteration in first-episode medication-naïve MDD relative to HC. Depression stages were categorized by illness duration and HRSD score: stage I = 0.1 – 0.7 years, stage II = 0.7 – 8.0 years. Compared with HC, GMV of the bilateral hippocampus and bilateral amygdala expanded in stage I in MDD. Combined with the results of the correlation test, GMV of the bilateral hippocampus in first-episode medication-naïve MDD was larger than that of healthy controls, but the bilateral hippocampus atrophied as illness duration (depression stage) prolonged.
expansions in our results were structurally connected to regions of the amygdala. What's more, GM of the amygdala was found to be expanded in MDD (Frodl et al., 2002, 2003). Secondly, structural brain changes seem to be state dependent in MDD (Arnone et al., 2013; Eijndhoven et al., 2009) and previous research has tended to mix patients with different states of depression in their studies. In the current study, we reported on a cohort of MDD patients that only included first-episode medication-naïve MDD patients. In the current study, CaSCN indicated the potential causal relationship between GM alteration of the hippocampus and regions of the default mode network (DMN), the left inferior parietal lobe, the postcentral gyrus, and the right amygdala. Meanwhile, SCN revealed that synchronized GM alterations were found between the hippocampus and regions of the DMN, the bilateral insula, and the anterior cingulate cortex, therefore the results of CaSCN could be partially proven by the results of SCN in the current study. Some studies have put forward that transneuronal spread could be used to explain the network-based atrophy observed in brain disorders, which means alteration may begin in an epicenter and then spread along axonal pathways in a “prion-like” manner (Ahmed et al., 2016; Zhou et al., 2012). In our study, we used CaSCN to reveal the existing transneurnoal spread and posited that there was causal effect between the GM alterations of the hippocampus and the crucial nodes of the DMN including the precuneus, the left middle temporal lobe and the right angular gyrus. Though we found GM expansions in the bilateral hippocampus in patients with MDD compared to HCs, GM of the bilateral hippocampus atrophied as illness duration prolonged. Additionally, the hippocampus exhibited positive causal effects on regions such as the precuneus, left middle temporal lobe, right angular cortex, left inferior parietal lobe,
Table 2 GM alterations in first-episode medication-naïve MDD relative to HC. Brain regions
Increased GM Amygdala (L/R) Hippocampus (L/R) ParaHippocampus (L/R) Fusiform (R) Decreased GM Brainstem (L/R) Pallidum (L/R) Thalamus (L/R)
MNI coordinates (x, y, z)
Peak t-value
Number of voxels
−21,0,−15/21,3,−15 −15,−9,−18/27,−6,−18 −12,0,−18/18,−6,−18
12.88/13.28 8.51/9.56 11.19/9.65
58/48 100/88 51/88
24,−39,−18
6.14
69
−3,−30,−12/6,−33,−12 −18,−6,3/18,6,3 −12,−6,0/18,−9,0
−7.51/−6.81 −13.45/−23.4 −12.6/−10.36
178/189 61/50 42/33
Abbreviations: MNI, Montreal Neurological Institute; L, left; R, right; GM, gray matter volume.
and colleagues that gray matter reduction in the hippocampus appears to be specific to the depressed state due to the observation of bilateral reductions in gray matter of the hippocampus in patients with current depression while untreated patients in stable remission had no changes (Arnone et al., 2013) directly opposes our results which detail hippocampal GM expansions in the early stage of first-episode medicationnaïve patients with MDD although hippocampal GM did atrophy as illness duration prolonged. To the best of our knowledge, few studies have found hippocampal GM expansions in first-episode medicationnaïve patients with MDD; therefore, we propose two possible explanations for the bilateral hippocampus GM expansions we found in the current study. Firstly, the regions of the hippocampus that showed GM 152
Journal of Affective Disorders 256 (2019) 148–155
Y. Li, et al.
Table 3 Stage-specific gray matter alterations in first-episode medication-naïve MDD relative to HC (Stage: I/II). Brain regions Increased GM Amygdala Hippocampus ParaHippocampus Decreased GM Thalamus Brainstem Pallidum
MNI coordinates (x, y, z) (Stage: I/II)
Peak t-value (Stage: I/II)
Number of voxels (Stage: I/II)
L R L R L R
−21,0,−18/21,6,−15 −21,0,−15/21,3,−15 −18,−9,−18/−18,−9,−15 21,−6,−15/21 −9 −15 −15,0,−18 18,−6,−18/18,−3,−18
10.36/7.70 10.31/9.51 8.11/6.27 8.50/6.40 8.70 7.72/6.46
55/38 40/33 55/16 55/20 17 37/20
L R L R L R
−21,−21,0/−12,−6,3 21,−24,0/18,−9,0 0,−17,−15/−9,−15,−3 9,−24,−12/12,−15,−3 −15,−6,3/−15,−3,3 21,−3,−3/18,−3,3
−6.25/−8.25 −7.64/−6.36 −8.17/−8.25 −8.02/−6.36 −10.55/−7.76 −8.80/−9.21
21/14 15/10 95/29 98/26 36/11 35/30
Abbreviations: L, left; R, right; MNI: Montreal Neurological Institute; Stage: I/II, Patients categorized arbitrarily into 2 subgroups (n = 28) from low to high stages of epilepsy duration. (Epilepsy duration: stage I/II=0.1–0.7 years /0.7–8.0 years.).
Fig. 3. Correlational relationship between gray matter volume alteration and illness duration. Negative correlations were found in the bilateral hippocampus (red arrows). Table 4 Correlational relationship between the GMV alterations and illness duration. Brain regions Positive correlation Inferior parietal lobe (L) Inferior temporal lobe (L) Fusiform (L/R) Frontal lobe (L/R) Negative correlation Hippocampus (L/R) Brainstem (L/R)
MNI coordinates (x, y, z)
Peak t-value
Number of voxels
−42,−42,36 −42,−33,−28.5 −42,−39,−22.5/39,−28.5,−30 −9,31.5,−10.5/16.5,33,−10.5
1.97 1.94 2.75/2.21 1.77/3.13
259 631 312/417 70/1001
−30,−22.5,−12/34.5,−33,−1.5 −9,−18,−13.5/12,−21,−18
−2.28/−1.86 −2.99/−2.40
477/48 266/177
Abbreviations: L, left; R, right; MNI: Montreal Neurological Institute.
Fig. 4. Patterns of hippocampus-associated structural covariance networks. Structural covariance networks were constructed based on cross-sectional morphometric data from a large cohort of first-episode medication-naïve MDD patients. The seed region, the bilateral hippocampus, was picked from regions showing significant gray matter volume increases. Synchronized gray matter volume alteration with the hippocampus were present in the bilateral insula, temporal lobes, and anterior cingulate cortex in MDD.
default mode network (DMN) (Fransson and Marrelec, 2008) and studies have found that abnormal functional connectivity between the hippocampus/parahippocampus, and DMN regions could help to classify depressed individuals from healthy controls (Zeng et al., 2012). In
postcentral gyrus, and right amygdala which is in line with previous MDD studies (Bora et al., 2012; Buss et al., 2012; Du et al., 2012; Sacher et al., 2012). Interestingly, some of these regions, the precuneus, temporal lobe, and angular cortex, make up core brain regions of the 153
Journal of Affective Disorders 256 (2019) 148–155
Y. Li, et al.
individual heterogeneity and environmental factors by using CaSCN. Finally, since the hippocampus is not necessarily the original source of brain network alterations in first-episode medication-naïve patients with MDD, confirming the original region implicated in structural alterations in MDD would allow us to precisely describe the dynamic structural alteration in MDD and obtain more information regarding how depression originates and changes. 5. Conclusion In the current study, we found that the hippocampus showed gray matter alteration and was the only brain region that showed a negative correlation with illness duration in first-episode medication-naïve patients with MDD. The hippocampus-associated CaSCN clarified the causal relationship between progressive gray matter alterations in the hippocampus and in other regions included in the DMN in first-episode medication-naïve patients with MDD. Our work provided evidence of the network spread mechanism in terms of the causal influence of hippocampal alteration on progressive brain structural alterations in first-episode medication-naïve patients with MDD. Contributors Zhong Yuan, Zhang Ning and Wang Chun completed the conception of the study. Teng Changjun, Jiao Kaili, Song Xiu, Tan Yarong and Xiao Chaoyong acquired the data. Li Yuting and Zhong Yuan completed data analysis. Li Yuting, Wang Chun and Zhong Yuan drafted and complished the final manuscript, which all authors reviewed.
Fig. 5. Hippocampus-associated causal networks of structural covariance in first-episode medication-naïve MDD. Granger causal analysis was applied to sequenced morphometric data according to illness duration and HRSD score. Positive granger causal values were distributed in the bilateral temporal lobe, angular, right amygdala, and right hippocampus. Negative granger causal values were found in the left amygdala and left hippocampus.
Declaration of Competing Interest None. Funding source
our study, the positive GC values may indicate that the hippocampal atrophy predicted the atrophy of the DMN regions and provided a new perspective on the relationship between these GM changes. Some aspects of this study could be improved. First, the sample size was small; only 56 depressed patients were included in our study. If confirmed in larger samples, our results might serve as a biomarker for the neuroimaging mechanism of MDD. Second, the onset of depression was difficult to define. Structured interviews should be conducted to identify progressive information in first-episode medication-naïve patients with MDD in the future. Third, CaSCN assumes that all the regions investigated have similar trajectory during years. However, due to neuroplasticity and compartmentation of the brain, different parts of the brain might follow different trajectories. Analysis of longitudinal data could help to clarify and confirm the results of CaSCN. Additionally, the relationship between illness duration and structural features could be impacted by the arbitrary distribution of illness,
This study was supported by National Natural Science Foundation of China (81571344, 81871344); Natural Science Foundation of Jiangsu Province (BK20161109); the Natural Science Foundation of the Higher Education Institutions of Jiangsu Province, China(18KJB190003); key research and development program (Social Development) project of Jiangsu province (BE20156092015); Nanjing Medical Science and Technique Development Foundation, Outstanding Youth Project (JQX14008); the Postdoctoral Science Foundation of China (2014M552700); "Six Talent Peak" High-Level Talent Selection and Training Plan of Jiangsu Province (2018-WSN-109). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Acknowledgments None.
Table 5 Causal effect of hippocampus GM alteration on whole brain GMV alteration with illness duration. Brain regions Positive GC Value Precuneus (L) Middle temporal lobe (L) Angular (L) Inferior parietal lobe (L) Postcentral (L) Amygdala (R) Inferior temporal lobe (L/R) Negative GC Value Hippocampus (L)
MNI coordinates (x, y, z)
PeakGC-value (Z-score)
Number of voxels
−1.5,−33,37.5 −57,−18,−24 −43.5,−55.5,24 −54,−40.5,46.5 −69,−19.5,30 21,0,−19.5 −52.5,−7.5,−31.5/54,0,−30.5
11.99 9.99 7.82 9.00 7.35 16.38 7.10/9.03
542 222 84 213 31 81 44/96
−19.5,−7.5,−19.5
−13.47
19
Abbreviations: L, left; R, right; MNI: Montreal Neurological Institute. 154
Journal of Affective Disorders 256 (2019) 148–155
Y. Li, et al.
Supplementary materials
Hamilton, J.P., Chen, G., Thomason, M.E., Schwartz, M.E., Gotlib, I.H., 2011. Investigating neural primacy in major depressive disorder: multivariate Granger causality analysis of resting-state fMRI time-series data. Mol. Psychiatry 16 (7), 763–772. Jiang, Y., Luo, C., Li, X., Duan, M., He, H., Chen, X., Woelfer, M., 2018. Progressive reduction in gray matter in patients with schizophrenia assessed with MR imaging by using causal network analysis. Radiology 287 (2), 171832. Jiao, Q., Lu, G., Zhang, Z., Yuan, Z., Wang, Z., Guo, Y., Liu, Y., 2011. Granger causal influence predicts BOLD activity levels in the default mode network. Hum. Brain Mapp. 32 (1), 154–161. https://doi.org/10.1002/hbm.21065. Keller, M.B., Lavori, P.W., Mueller, T.I., Endicott, J., Coryell, W., Hirschfeld, R.M., Shea, T., 1992. Time to recovery, chronicity, and levels of psychopathology in major depression. A 5-year prospective follow-up of 431 subjects. Arch. Gen. Psychiatry 49 (10), 809–816. Keller, M.B., Lavori, P.W., Rice, J., Coryell, W., Hirschfeld, R.M., 1986. The persistent risk of chronicity in recurrent episodes of nonbipolar major depressive disorder: a prospective follow-up. Am. J. Psychiatry 143 (1), 24–28. Keller, M.B., Shapiro, R.W., Lavori, P.W., Wolfe, N., 1982. Recovery in major depressive disorder: analysis with the life table and regression models. Arch. Gen. Psychiatry 39 (8), 905–910. Korgaonkar, M.S., Fornito, A., Williams, L.M., Grieve, S.M., 2014. Abnormal structural networks characterize major depressive disorder: a connectome analysis. Biol. Psychiatry 76 (7), 567–574. https://doi.org/10.1016/j.biopsych.2014.02.018. Lai, C.H., 2013. Gray matter volume in major depressive disorder: a meta-analysis of voxel-based morphometry studies. Psychiatry Res. Neuroimaging 211 (1), 37. https:// doi.org/10.1016/j.pscychresns.2012.06.006. Mcdonald, C.R., Mcevoy, L.K., Gharapetian, L., Fennemanotestine, C., Hagler, D.J., Holland, D., Dale, A.M., 2009. Regional rates of neocortical atrophy from normal aging to early Alzheimer disease. Neurology 73 (6), 457. https://doi.org/10.1212/ WNL.0b013e3181b16431. Mckinnon, M.C., Yucel, K., Nazarov, A., Macqueen, G.M., 2009. McKinnon MCM, Yucel K, Nazarov A, MacQueen GM. A meta-analysis examining clinical predictors of hippocampal volume in patients with major depressive disorder. J Psychiatry Neurosci 34: 41-54. J. Psychiatry Neurosci. Jpn 34 (1), 41–54. Peng, Liu, Nie, Li, Shan, Wang, Li., 2011. Cerebral and cerebellar gray matter reduction in first-episode patients with major depressive disorder: a voxel-based morphometry study. Eur. J. Radiol. 80 (2), 395–399. https://doi.org/10.1016/j.ejrad.2010.04.006. Romanczuk-Seiferth, N., Pöhland, L., Mohnke, S., Garbusow, M., Erk, S., Haddad, L., Walter, H., 2014. Larger amygdala volume in first-degree relatives of patients with major depression. Neuroimage Clin. 5 (9), 62–68. https://doi.org/10.1016/j.nicl. 2014.05.015. Sacher, J., Neumann, J., Fünfstück, T., 2012. Mapping the depressed brain: a meta-analysis of structural and functional alterations in major depressive disorder. J. Affect. Disord. 140 (2), 142–148. https://doi.org/10.1016/j.jad.2011.08.001. Scott, J., Eccleston, D., Boys, R., 1992. Can we predict the persistence of depression. Br. J. Psychiatry J. Ment. Sci. 161 (11), 633. Walsh, J., 1949. Applications of some significance tests for the median which are valid under very general conditions*. Publ. Am. Stat. Assoc. 44 (247), 342–355. Wu, H., Sun, H., Wang, C., Yu, L., Li, Y., Peng, H., Jiang, T., 2016. Abnormalities in the structural covariance of emotion regulation networks in major depressive disorder. J. Psychiatr. Res. 84, 237. https://doi.org/10.1016/j.jpsychires.2016.10.001. Zeng, L.L., Shen, H., Liu, L., Wang, L., Li, B., Fang, P., Hu, D., 2012. Identifying major depression using whole-brain functional connectivity: a multivariate pattern analysis. Brain A J. Neurol. 135 (Pt 5), 1498. Zhang, Z., Liao, W., Xu, Q., Wei, W., Zhou, H.J., Sun, K., . . ., Lu, G., 2016. Hippocampusassociated causal network of structural covariance measuring structural damage progression in temporal lobe epilepsy. Hum. Brain Mapp. 38 (2), 753. Zhou, J., Gennatas, E.D., Kramer, J.H., Miller, B.L., Seeley, W.W., 2012. Predicting regional neurodegeneration from the healthy brain functional connectome. Neuron 73 (6), 1216. https://doi.org/10.1016/j.neuron.2012.03.004.
Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.jad.2019.05.053. References Ahmed, R.M., Devenney, E.M., Irish, M., Ittner, A., Naismith, S., Ittner, L.M., Hodges, J.R., 2016. Neuronal network disintegration: common pathways linking neurodegenerative diseases. J. Neurol. Neurosurg. Psychiatry 87 (11), 1234–1241. https://doi.org/10. 1136/jnnp-2014-308350. Alexanderbloch, A., Giedd, J.N., Bullmore, E., 2013. Imaging structural co-variance between human brain regions. Nat. Rev. Neurosci. 14 (5), 322–336. Arnone, D., Mckie, S., Elliott, R., Juhasz, G., Thomas, E.J., Downey, D., Anderson, I.M., 2013. State-dependent changes in hippocampal grey matter in depression. Mol. Psychiatry 18 (12), 1265–1272. https://doi.org/10.1038/mp.2012.150. Bernhardt, B.C., Kim, H., Bernasconi, N., 2013. Patterns of subregional mesiotemporal disease progression in temporal lobe epilepsy. Neurology 81 (21), 1840–1847. https://doi.org/10.1212/01.wnl.0000436069.20513.92. Bernhardt, B.C., Worsley, K.J., Kim, H., Evans, A.C., Bernasconi, A., Bernasconi, 2009. Longitudinal and cross-sectional analysis of atrophy in pharmacoresistant temporal lobe epilepsy. Neurology 72 (20), 1747. https://doi.org/10.1212/01.wnl. 0000345969.57574.f5. Bora, E., Fornito, A., Pantelis, C., Yücel, M., 2012. Gray matter abnormalities in major depressive disorder: a meta-analysis of voxel based morphometry studies. J. Affect. Disord. 138 (1–2), 9. https://doi.org/10.1016/j.jad.2011.03.049. Buss, C., Pruessner, J.C., Mayberg, H., Mletzko, T., Nemeroff, C., Heim, C., 2012. Larger amygdala volumes after childhood trauma associated with depression and cortisol response to psychosocial stress in adulthood. Eur. J. Psychotraumatol. 3 (2), 75–85. https://doi.org/10.3402/ejpt.v3i0.19445. Chen, V.C., Shen, C.., Liang, S.H., Li, Z.H., Tyan, Y.S., Liao, Y.T., Weng, J.C., 2016. Assessment of abnormal brain structures and networks in major depressive disorder using morphometric and connectome analyses. J. Affect. Disord., 205, 103. https:// doi.org/10.1016/j.jad.2016.06.066. Czéh, B., Lucassen, P.J., 2007. What causes the hippocampal volume decrease in depression? Are neurogenesis, glial changes and apoptosis implicated? Eur. Arch. Psychiatry Clin. Neurosci. 257 (5), 250. Du, M.Y., Wu, Q.Z., Yue, Q., Li, J., Liao, Y., Kuang, W.H., Gong, Q.Y., 2012. Voxelwise meta-analysis of gray matter reduction in major depressive disorder. Prog. Neuropsychopharmacol. Biol. Psychiatry 36 (1), 11–16. https://doi.org/10.1016/j. pnpbp.2011.09.014. Eijndhoven, P.V., Wingen, G.V., Oijen, K.V., Rijpkema, M., Goraj, B., Verkes, R.J., Tendolkar, I., 2009. Amygdala volume marks the acute state in the early course of depression. Biol. Psychiatry 65 (9), 812–818. https://doi.org/10.1016/j.biopsych. 2008.10.027. Fransson, P., Marrelec, G., 2008. The precuneus/posterior cingulate cortex plays a pivotal role in the default mode network: evidence from a partial correlation network analysis. Neuroimage 42 (3), 1178. Frodl, T., Meisenzahl, E., Zetzsche, T., Bottlender, R., Born, C., Groll, C., Möller, H.J., 2002. Enlargement of the amygdala in patients with a first episode of major depression. Biol. Psychiatry 51 (9), 708–714. Frodl, T., Meisenzahl, E.M., Zetzsche, T., Born, C., Jager, M., Groll, C., Moller, H.J., 2003a. Larger amygdala volumes in first depressive episode as compared to recurrent major depression and healthy control subjects. Biol. Psychiatry 53 (4), 338–344. https://doi.org/10.1016/S0006-3223(02)01474-9. Pii S0006-3223(03)01474-9. Goebel, R., Roebroeck, A., Kim, D.S., Formisano, E., 2003. Investigating directed cortical interactions in time-resolved fMRI data using vector autoregressive modeling and Granger causality mapping. Magn. Reson. Imaging 21 (10), 1251–1261.
155