Molecular Psychiatry (2020) 25:1511–1525 https://doi.org/10.1038/s41380-019-0477-2 ARTICLE White matter disturbances in major depressive disorder: a coordinated analysis across 20 international cohorts in the ENIGMA MDD working group Laura S. van Velzen1,2,3 Sinead Kelly4,5 Dmitry Isaev6 Andre Aleman7 Lyubomir I. Aftanas8,9 Jochen Bauer10 ● ● ● ● ● ● Bernhard T. Baune 11,12,13 Ivan V. Brak 8,14 Angela Carballedo15,16 Colm G. Connolly 17,18 ● ● ● ● Baptiste Couvy-Duchesne19 Kathryn R. Cullen20 Konstantin V. Danilenko8 Udo Dannlowski11 Verena Enneking11 ● ● ● ● ● Elena Filimonova8 Katharina Förster11 Thomas Frodl15,21,22 Ian H. Gotlib23 Nynke A. Groenewold24,25 ● ● ● ● ● Dominik Grotegerd11 Mathew A. Harris 26 Sean N. Hatton27 Emma L. Hawkins26 Ian B. Hickie27 ● ● ● ● ● Tiffany C. Ho17,23,28 Andreas Jansen29 Tilo Kircher29 Bonnie Klimes-Dougan30 Peter Kochunov31 Axel Krug29 ● ● ● ● ● ● Jim Lagopoulos32 Renick Lee33 Tristram A. Lett 34 Meng Li35 Frank P. MacMaster36,37 Nicholas G. Martin 38 ● ● ● ● ● ● Andrew M. McIntosh 26,39 Quinn McLellan 40,41 Susanne Meinert11 Igor Nenadić29 Evgeny Osipov14 ● ● ● ● ● Brenda W. J. H. Penninx1 Maria J. Portella42,43,44 Jonathan Repple11 Annerine Roos45 Matthew D. Sacchet46 ● ● ● ● ● Philipp G. Sämann47 Knut Schnell48 Xueyi Shen26 Kang Sim49,50,51 Dan J. Stein 52 Marie-Jose van Tol 7 ● ● ● ● ● ● Alexander S. Tomyshev53 Leonardo Tozzi21,28 Ilya M. Veer34 Robert Vermeiren54,55,56 Yolanda Vives-Gilabert57 1234567890();,: ● ● ● ● ● 1234567890();,: Henrik Walter 34 Martin Walter58 Nic J. A. van der Wee55,59 Steven J. A. van der Werff55,59 ● ● ● ● Melinda Westlund Schreiner 30 Heather C. Whalley 26 Margaret J. Wright 60,61 Tony T. Yang17 Alyssa Zhu6 ● ● ● ● ● Dick J. Veltman1,62 Paul M. Thompson6 Neda Jahanshad 6 Lianne Schmaal2,3 ● ● ● Received: 23 August 2018 / Revised: 15 April 2019 / Accepted: 10 May 2019 / Published online: 30 August 2019 © The Author(s) 2019. This article is published with open access Abstract Alterations in white matter (WM) microstructure have been implicated in the pathophysiology of major depressive disorder (MDD). However, previous findings have been inconsistent, partially due to low statistical power and the heterogeneity of depression. In the largest multi-site study to date, we examined WM anisotropy and diffusivity in 1305 MDD patients and 1602 healthy controls (age range 12–88 years) from 20 samples worldwide, which included both adults and adolescents, within the MDD Working Group of the Enhancing Neuroimaging Genetics through Meta-Analysis (ENIGMA) consortium. Processing of diffusion tensor imaging (DTI) data and statistical analyses were harmonized across sites and effects were meta-analyzed across studies. We observed subtle, but widespread, lower fractional anisotropy (FA) in adult MDD patients compared with controls in 16 out of 25 WM tracts of interest (Cohen’s d between 0.12 and 0.26). The largest differences were observed in the corpus callosum and corona radiata. Widespread higher radial diffusivity (RD) was also observed (all Cohen’s d between 0.12 and 0.18). Findings appeared to be driven by patients with recurrent MDD and an adult age of onset of depression. White matter microstructural differences in a smaller sample of adolescent MDD patients and controls did not survive correction for multiple testing. In this coordinated and harmonized multisite DTI study, we showed subtle, but widespread differences in WM microstructure in adult MDD, which may suggest structural disconnectivity in MDD. These authors contributed equally: Laura S. van Velzen, Sinead Kelly Introduction These authors contributed equally: Neda Jahanshad, Lianne Schmaal Major depressive disorder (MDD) is a debilitating and Supplementary information The online version of this article (https:// highly prevalent psychiatric disorder, characterized by doi.org/10.1038/s41380-019-0477-2) contains supplementary depressed mood and loss of interest in daily activities [1]. material, which is available to authorized users. Although MDD is one of the leading causes of disability * Laura S. van Velzen worldwide [2], our understanding of the pathophysiological laurasvanvelzen@gmail.com basis of the disorder remains incomplete. In recent years, Extended author information available on the last page of the article neuroimaging analyses have helped to characterize the 1512 L. S. van Velzen et al. neuroanatomical basis of MDD; however, consistent pat- thalamic radiation [5] and the superior longitudinal fasci- terns of brain alterations have been difficult to identify due culus [17], while both higher and lower FA have been to limited power in previous studies and heterogeneity in reported in the fronto-occipital fasciculus [5, 17]. One major data analysis. To address this issue, the MDD working limitation of this form of literature-based meta-analysis is its group within the Enhancing Neuro Imaging Genetics dependency on published data and therefore susceptibility through Meta-Analysis (ENIGMA) consortium (http:// to publication bias. Moreover, results of the different ret- enigma.usc.edu/) initiated the largest coordinated meta- rospective meta-analyses highlight different WM tracts and analyses of brain structure in MDD to date to investigate the do not confirm findings from other meta-analyses, perhaps robustness or consistency of neuroimaging findings across as they include studies with different processing protocols many samples worldwide. Our recent studies revealed lower and different statistical analyses. hippocampal volume and altered cortical structure in MDD In the DTI project of the ENIGMA-MDD Working patients [3, 4]. Lower integrity of white matter (WM) tracts Group, we aimed to address these methodological issues connecting these cortical and subcortical regions may and increase statistical power by initiating a worldwide suggest a ‘disconnection-syndrome’ in MDD [5]. Identify- effort to perform the largest coordinated multi-cohort ana- ing patterns of alterations in WM in MDD could lead to the lysis on WM alterations in MDD to date. Standardized discovery of pathogenic processes and thereby guide protocols for image processing, quality assurance, and sta- development of new treatment targets for MDD; it could tistical analyses were applied using the ENIGMA-DTI also help provide ways of monitoring or predicting response protocols for multi-site DTI harmonization [20–22], and to currently available treatments. distributed to sites around the world. Harmonized effect- Diffusion tensor imaging (DTI) characterizes the direc- size estimates calculated across sites were then meta- tionality of water diffusion in the brain and allows for the analyzed. in vivo study of WM microstructural properties that cannot Our primary goal was to identify and rank the most be measured with standard anatomical MRI. Fractional robust associations between MDD diagnosis and WM anisotropy (FA) is a common measure derived from DTI microstructure in a large sample of 1305 MDD patients and ranging from 0 to 1, where higher values typically represent 1602 healthy controls across 20 samples from North directionally constrained diffusion within the WM, likely America, Europe, Asia, and Australia. In contrast to most due to more intact myelin, and greater uniformity and previous studies and meta-analyses that only examined FA, compactness of fiber bundles. MDD patients have been we also characterized axial diffusivity (AD), which is reported to show lower FA, on average, in numerous WM considered to represent a measure of axonal number, cali- tracts, including callosal, association, and commissural ber, and organization, radial diffusivity (RD), which may fibers; yet the pattern of deficits, and the degree of disrup- give more insight into myelination, and mean diffusivity tion is highly variable across studies [6–10]. Some studies (MD), often considered a measure of membrane density report lower FA in the uncinate fasciculus [7] and internal [23]. Due to the continued maturation of WM tracts capsule [8, 9], yet others did not replicate these findings [6]. throughout adolescence, we analyzed adolescent (age ≤ 21 Inconsistent findings across studies may be due, in part, to years) and adult (age > 21 years) patients and controls limited statistical power related to small sample sizes, as separately. Furthermore, we explored the modulating effects well as differences in analytical techniques between studies. of clinical characteristics of MDD including age of onset, Variations in demographic (e.g. age [11]) or clinical char- recurrence of depression, antidepressant use, severity of acteristics of participants (e.g. age of onset [12, 13] and depressive symptoms, and number of episodes. illness duration [14]) across studies may also contribute to conflicting results in the literature as increasing age, longer illness duration and early age of onset have been associated Materials and methods with lower FA [13–15]. Of note, adolescent MDD may show distinct patterns of WM changes compared to adult Study sample MDD, as childhood and adolescence is a peak period for WM maturation [16]. The ENIGMA-MDD DTI Working Group consists of 20 To date, four retrospective meta-analyses of DTI studies cohorts from 11 different countries and includes DTI scans have been performed in 2011 [17], 2013 [5] and 2016 from 1602 healthy controls and 1305 adults and adoles- [18, 19]. These have all reported lower FA in the corpus cents). Demographic and clinical characteristics for each callosum in adult MDD patients. Other meta-analytic results sample are presented in Tables 1 and 2. Diagnostic included lower FA in the anterior limb of the internal assessment measures and exclusion criteria for every site are capsule [19], inferior longitudinal fasciculus, posterior presented in Supplementary Table S1. All study participants Table 1 Demographic data for all samples included in the ENIGMA-MDD-DTI project Sample Adult samples (age > 21) Adolescent samples (age ≤ 21) Age Age % Female %Female MDD Total no. Total MDD Age Age % Female %Female MDD Total no. Total controls MDD controls of controls MDD controls of no. of MDD (mean ± (mean ± controls (mean ± (mean ± controls SD) SD) SD) SD) 1 Barcelona 46.41 ± 47.29 ± 72.40 78.20 29 55 7.82 7.88 2 Bipolar family study 24.70 ± 25.22 ± 53.40 55.60 73 18 1.88 2.21 3 CODE 38.45 ± 35.41 ± 50.00 59.30 22 27 13.29 11.76 4 DIP 44.50 ± 43.38 ± 65.00 62.50 20 16 14.71 15.75 5 Sexpect 33.75 ± 39.21 ± 15.00 42.10 20 19 7.20 11.11 6 EPISCA 14.60 ± 15.53 ± 85.00 82.40 20 17 1.64 1.55 7 MOTAR 34.65 ± 33.77 ± 57.70 61.30 26 31 12.14 8.53 8 MPIP 52.01 ± 50.98 ± 63.80 56.90 130 137 12.38 12.86 9 Munster 38.37 ± 38.82 ± 57.30 56.80 386 125 19.68 ± 18.92 ± 59.50 69.20 37 13 neuroimaging cohort 11.03 11.36 1.67 1.44 10 NESDA 54.91 ± 48.37 ± 45.50 69.50 22 57 8.59 10.05 11 Novosibirsk 40.13 ± 46.33 ± 46.70 77.10 30 48 8.82 12.23 12 QTIM 24.82 ± 25.15 ± 71.40 41.43 84 13 20.14 ± 19.71 ± 62.07 84.61 87 13 White matter disturbances in major depressive disorder: a coordinated analysis across 20 international. . . 1.89 2.22 1.16 1.29 13 Sydney 25.86 ± 24.87 ± 54.90 55.20 71 67 19.00 ± 17.73 ± 75.90 67.70 29 155 3.37 3.94 2.67 2.10 14 Imaging Genetics Dublin 38.26 ± 41.81 ± 52.20 62.30 46 53 12.40 10.76 15 UCSF 15.38 ± 15.74 ± 57.77 60.66 45 61 1.35 1.28 16 Child and Adolescent 19.40 ± 17.80 ± 80.00 55.00 10 20 Imaging Research 1.51 1.44 Calgary 17 FOR2017 34.73 ± 39.37 ± 60.90 59.60 274 218 19.54 ± 19.76 ± 73.10 64.00 26 25 12.27 12.92 1.36 1.05 1513 1514 L. S. van Velzen et al. provided written informed consent and the local institu- no. of MDD tional review boards and ethics committees approved each included cohort study. Total 68 Image processing and analysis % Female %Female MDD Total no. controls 36 Scanner and acquisition parameters for all sites are provided of in Supplementary Table S2. Preprocessing of diffusion weighted images, including eddy current correction, echo- planar imaging (EPI)-induced distortion correction, and tensor fitting, were performed at each site. After tensor 76.50 estimation, DTI images were processed using the ENIGMA-DTI protocols (see Supplemental Note 1). Pro- controls tocols with image processing as well as quality control 66.70 Adolescent samples (age ≤ 21) procedures are freely available as part of the ENIGMA-DTI webpage (http://enigma.ini.usc.edu/ongoing/dti-working- group/) and NITRC (https://www.nitrc.org/projects/ (mean ± 15.40 ± MDD enigma_dti/). Measures of FA, MD, RD, and AD were 1.82 Age SD) obtained for 25 regions of interest (please see Supplemen- tary Table S3 for a list of and description of the regions of (mean ± controls 15.67 ± interest). In all analyses, we combined ROIs across both 2.07 Total MDD Age SD) hemispheres by taking the mean of the left and right hemisphere regions weighted by the number of voxels; as we did not hypothesize any lateralized effect of the disorder, our primary analysis included bilaterally averaged Age (in years) and sex for patients and controls, presented separately for adult and adolescent samples measures to avoid potential issues of left/right flipping 14 23 between sites. % Female %Female MDD Total no. controls Statistical analysis 15 17 of First, differences between patients and controls in FA, MD, AD, and RD were examined within adult (age > 21 years) and adolescent (age ≤ 21 years) samples separately by linear 100.00 43.50 regression analysis, and Cohen’s d effect size estimates were calculated. This was done for all 25 regions of interest. controls All analyses were corrected for age and sex and linear 100.00 52.90 and nonlinear age and sex interactions (age-by-sex inter- action, age2 and age2-by-sex interaction). To examine Adult samples (age > 21) whether case-control effects were regional effects beyond (mean ± 35.00 ± 39.30 ± MDD a global effect, these analyses were performed again while 8.49 8.34 Age SD) including average FA, MD, RD, or AD as an additional covariate. (mean ± controls 32.00 ± 38.53 ± We further performed diagnosis-by-sex and diagnosis- 10.13 4.64 Age SD) by-age interaction analyses. Case-control differences were also examined in different age categories (10–<20 years; 18 University of Minnesota 20–<30 years; 30–<40 years; 40–<50 years, and 50–<60 20 IMH MDD Singapore years; please see Supplemental Note 4 and Supplemental Tables). Separately within adult (age > 21) and adolescent Table 1 (continued) (≤21 years) samples, we performed stratified meta-analyses to compare age of onset (adolescent onset ≤ 21 years of age, 19 Stanford adult onset > 21 years; in adult samples only), anti- Sample depressant use at the time of scanning (antidepressant users and antidepressant non-users) and MDD stage (first and Table 2 Clinical data for all samples included in the ENIGMA-MDD-DTI project: percentage of MDD patients using antidepressants, percentage of acute versus remitted MDD, age of onset of MDD, percentage of MDD patients with a comorbid anxiety disorder and severity of MDD symptoms, presented separately for adult (age > 21) and adolescent (age ≤ 21) samples Study Sample % AD use % first episode/ % Acute/ Age of onset % comorbid HDRS-17 severity BDI-II severity recurrent MDD remitted MDD MDD (mean ± anxiety disorder MDD (mean ± SD)a MDD (mean ± SD)b SD) Adult samples 1 Barcelona 94.50 32.70/67.30 63.60/36.40 32.91 ± 11.45 0.0 13.51 ± 9.04 NA 2 Bipolar family study 16.70 NA NA 22.80 ± 2.70 NA 4.94 ± 4.72 NA 3 CODE 0.00 0.00/100.00 100.00/0.00 NA NA NA NA 4 DIP 50.00 31.25/68.75 100.00/0.00 26.00 ± 13.46 31.3 NA 25.94 ± 10.03 5 Sexpect 100.00 20.00/80.00 100.00/0.00 30.87 ± 11.24 0.1 11.90 ± 5.39 20.86 ± 12.40 6 MOTAR 0.00 71.00/29.00 100.00/0.00 24.52 ± 9.93 45.2 NA NA 7 MPIP NA 0.00/100.00 70.00/30.00 32.20 ± 12.76 17.9 NA 14.25 ± 10.87 8 FOR2017 60.60 22.50/75.70 73.90/21.10 27.34 ± 13.12 12.0 8.85 ± 8.00 17.58 ± 11.16 9 Munster neuroimaging cohort 88.80 22.40/77.60 100.00/0.00 30.15 ± 12.13 NA 23.09 ± 5.15 28.56 ± 9.16 10 NESDA 33.90 6.80/93.20 37.30/62.70 25.44 ± 10.60 23.7 NA NA 11 Novosibirsk 18.80 60.40/39.60 100.00/0.00 38.60 ± 14.16 0.0 18.41 ± 4.75 31.73 ± 10.87 12 QTIM NA NA NA 19.54 ± 4.74 30.8 NA NA 13 Stanford 21.40 0.00/100.00 100.00/0.00 16.79 ± 6.34 64.3 18.86 ± 4.22 29.71 ± 6.41 14 Sydney 40.30 31.30/31.30 40.30/35.80 14.11 ± 7.39 26.9 12.48 ± 7.26 4.00 ± 3.39 15 Imaging Genetics Dublin 75.50 0.00/100.00 100.00/0.00 25.26 ± 12.67 NA 23.53 ± 5.00 33.17 ± 11.60 16 IMH MDD Singapore 100.00 34.80/65.20 21.70/78.30 33.39 ± 9.83 NA 6.30 ± 6.23 NA Adolescent samples 1 Child and Adolescent Imaging 65.00 0.00/100.00 100.00/0.00 14.10 ± 2.29 79.3 20.40 ± 6.31 30.50 ± 10.51 Research Calgary 2 EPISCA 11.80 100.00/0.00 100.00/0.00 NA 29.4 NA NA 3 University of Minnesota 23.50 23.59/32.40 NA 12.46 ± 2.35 63.3 NA 26.35 ± 12.06 White matter disturbances in major depressive disorder: a coordinated analysis across 20 international. . . 4 FOR2017 60.00 24.00/76.00 80.00/20.00 14.21 ± 2.96 17.6 11.45 ± 9.17 19.13 ± 12.02 5 Munster neuroimaging cohort 84.60 53.80/46.20 100.00/0.00 16.77 ± 2.89 NA 22.38 ± 3.86 30.08 ± 10.40 6 QTIM NA NA NA 17.92 ± 2.10 46.2 NA NA 7 Sydney 41.90 29.00/32.30 46.50/31.00 11.57 ± 5.52 23.2 12.65 ± 7.86 2.31 ± 0.95 8 UCSF 0.00 35.09/64.91 92.99/7.01 13.40 ± 2.34 61.0 NA 28.33 ± 11.12 a Hamilton Depression Rating Scale (HDRS-17; range: 0–52) b Beck Depression Inventory (BDI-II; range: 0–63) 1515 1516 L. S. van Velzen et al. recurrent episode patients) and to investigate associations analysis were also performed including only subjects within with symptom severity (Beck Depression Inventory (BDI- the UK Biobank age range (43–78 years). II) and the 17-item Hamilton Depression Rating Scale (HDRS-17)) and the number of depressive episodes. For the latter two analyses with continuous measures and interac- Results tion analyses, we extracted the beta regression coefficient as the effect size measure. Adults Analyses scripts are available on the ENIGMA-GitHub (https://github.com/ENIGMA-git/). All regression models After FDR correction, significantly lower FA was observed and effect sizes were computed for each site separately and for adult MDD patients (N = 921; age range 22–88) com- regression outputs were then meta-analyzed across sites. pared to healthy controls (N = 1265) in 16 of the 25 ROIs, with the largest effects observed for the full WM skeleton, Meta-analysis followed by the anterior corona radiata (ACR), corona radiata (CR), corpus callosum (CC), genu of the corpus A random-effects inverse-variance weighted meta-analysis callosum (GCC), body of the corpus callosum (BCC) and was conducted in R (metaphor package, version 1.9–118) anterior limb of the internal capsule (ALIC). Significantly to combine effect sizes from all sites. Heterogeneity lower FA was also observed in the superior fronto-occipital scores (I2), indicating the percentage of total variance fasciculus (SFO), sagittal stratum (SS), internal capsule explained by heterogeneity of the effects alone, were (IC), posterior corona radiata (PCR), superior corona also computed for each test, with lower values indicating radiata (SCR), inferior fronto-occipital fasciculus (IFO), lower effect size estimate variance across sites. The false fornix/stria terminalis (FXST), external capsule (EC), discovery rate (FDR) multiple comparisons correction was and cingulate gyrus of the cingulum bundle (CGC) applied to correct for the number of WM tracts. In addition, (Fig. 1, Supplementary Table S4 and Supplementary in the supplemental tables, we show which findings survive Fig. 1). No significant effects were observed for AD or FDR-correction across all four diffusion measures and the MD differences in adults (Tables S4 and S6 and Supple- 25 WM tracts (100 measures; see supplements). All repor- mentary Fig. 1). Higher RD for the adult sample was ted p-values are FDR-corrected p-values. observed across seven ROIs, including the FXST, BCC, SCR, hippocampal part of the cingulum bundle (CGH), the Secondary analyses full WM skeleton, CR, and SFO (Fig. 2, Table S7 and Supplementary Fig. 1). Moderator analysis Adolescents We performed moderator analyses using meta-regression to examine whether characteristics of individual sites After FDR correction, no significant differences were explained variation in effect sizes across sites. Methods and observed for FA, AD, MD, or RD between adolescent results of these analyses are presented in Supplemental MDD patients (N = 372) and healthy controls (N = 290) Note 2. (Tables S8–11 and Supplementary Figures. 2 and 3). UK Biobank Correction for average anisotropy/diffusivity To examine whether our findings generalize to a population No significant differences between MDD patients and study, we analyzed a large population cohort of adults (age controls were observed after correcting for average aniso- range: 43–78 years) from the UK Biobank (1st and 2nd tropy and diffusivity across the WM skeleton in the adult release) a large-scale population health study of adults in the samples (Tables S12–15) and adolescent samples United Kingdom. 2096 individuals were identified to have (Tables S16–19). These results indicate that no region ‘probable lifetime MDD’, as previously described [24], showed additional effects beyond the global effect. based on hospital admissions data and self-report of depressive symptoms. DTI measures from patients were Diagnosis by sex interaction compared to DTI measures from 3275 controls without mental disorders. Image acquisition and processing of UK Adults Biobank data is described in Supplemental Note 3. In order to examine overlap between the UK Biobank sample and There were no significant diagnosis-by-sex interaction the meta-analysis, the case-control analyses in the meta- effects in adult subjects (Tables S20–23). White matter disturbances in major depressive disorder: a coordinated analysis across 20 international. . . 1517 Fig. 1 Cohen’s d effect sizes for case-control differences in fractional anisotropy, mean diffusivity, radial diffusivity, and axial diffusivity across adults and adolescents Adolescents post-hoc tests showed that case-control differences increased with age. There were no interaction effects for A diagnosis-by-sex interaction for adolescents was AD, MD, and RD (Tables S29–31). observed for RD in the uncinate fasciculus (UNC; Table S27), post-hoc tests showed that this was driven by First and recurrent episodes higher RD in patients compared to controls in males only (d = 0.760, p = 0.011). Adults Diagnosis by age interaction Lower FA was observed for recurrent MDD patients (N = 645) compared to controls (N = 1053) across 15 ROIs, with Full group the largest effects observed in the ACR, GCC, CR, ALIC, full WM skeleton, IC and the whole CC. Significant effects Diagnosis-by-age interaction analyses were performed were also observed in the BCC, SFO, SS, PCR, superior across adults and adolescents to obtain the largest age range longitudinal fasciculus (SLF), retrolenticular part of the possible. Interaction effects were observed for FA in the internal capsule (RLIC), PTR, and IFO (Table S32). No BCC, CC, FXST, GCC, SS, and average FA (Table S28), significant effects were observed for MD or AD 1518 L. S. van Velzen et al. Fig. 2 Regional overlap in case- control differences in white matter integrity across adults and adolescents (Tables S33 and S34). Significantly higher RD was Table S60). No significant differences were observed observed for MDD patients in the SCR (Table S35). No between adult onset (N = 334) and adolescent onset MDD significant differences were observed between first-episode (N = 197) (Tables S64–67). MDD patients (N = 169) and controls (N = 816) or between first-episode MDD patients and recurrent MDD patients Antidepressant use at the time of scanning (Tables S36–43). Adults Adolescents No differences were observed between antidepressant No significant differences were observed between first users (N = 406) and controls (N = 848) (Tables S68–71). episode adolescent patients (N = 98) and controls (N = 130) Significantly lower FA was observed for antidepressant or between recurrent adolescents (N = 148), and controls non-users (N = 288) compared to controls (N = 962) in (N = 146) (Tables S44-55). the FXST, the entire WM skeleton, FX, CGH, SS, and EC (Table S72). No significant effects were observed for AD Age of onset or MD (Tables S73 and S74). Antidepressant non-users had higher RD in the FXST compared to controls Adults (Table S75). No differences were observed between antidepressant users (N = 335) and non-users (N = 164) No significant differences were observed between ado- (Tables S76–79). lescent age of onset MDD (≤21 years (N = 251) and controls (n = 869) (Tables S56–59). MDD patients with Adolescents an adult age of onset (>21years) (N = 399) had sig- nificantly lower FA compared to controls (N = 853) No significant differences were observed between adoles- across 10 ROIs, with the largest effects observed for the cent antidepressant users (N = 128) and controls (N = 138), EC, SS, and the IFO. Significant effects were also antidepressant non-users (N = 188) and controls (N = 156) observed in the CC, GCC, CGC, PCR, CR, SLF, and the or between antidepressant non-users (N = 104) and anti- FA across the full WM skeleton (Supplementary depressant users (N = 112) (Tables S80–S91). White matter disturbances in major depressive disorder: a coordinated analysis across 20 international. . . 1519 WM associations with symptom severity significant associations were no longer significant after controlling for average/global FA. Adults The regions that most strongly contributed to the global effect of lower FA were regions of the corona radiata and No significant associations were observed between anisotropy corpus callosum. Lower FA in the CC is consistent with or diffusivity and symptom severity scores assessed with the previous meta-analyses of WM in MDD [5, 18, 19]. The CC BDI (N = 477) or HDRS (N = 603) (Tables S92–99). is the largest interhemispheric commissure in the human brain [25], and connects amongst others the anterior cin- Adolescents gulate cortex and orbitofrontal cortex in both hemispheres. These regions play an important role in mood regulation No significant associations were observed between aniso- [26], show cortical thinning in MDD [4] and show func- tropy or diffusivity and BDI (N = 168) or HDRS (N = 191) tional abnormalities in relation to cognitive control, work- scores (Tables S100–107). ing memory and emotion processing in MDD [27–29]. Changes in FA in the CR have been observed in single site Moderator analysis studies of adult and geriatric depression [30–32]. This tract is part of the limbic-thalamo-cortical circuitry, and contains Results of the moderator analyses are presented in Sup- thalamic projections to cortical regions and also plays an plemental Note 2. important role in emotion regulation [33, 34]. In adults, global lower FA in some WM tracts in MDD UK Biobank was accompanied by global higher RD, yet no differences found in MD or AD. Previous work suggests that changes Significantly lower FA in the SFO was observed in patients in RD reflect changes in myelination or morphology of glial with probable MDD (N = 2096) compared to healthy con- cells [35, 36]. In line with this, a prior quantitative MRI trols (N = 3275) in the UK Biobank sample (Table S108). study of individuals with MDD reported global reductions Higher AD in patients was observed in the SCR of R1, an MRI parameter that is thought to reflect myelin (Table S109). Higher MD in patients was observed in the content [37]. Furthermore, previous post-mortem studies ACR, ALIC, CR, SCR, and SFO (Table S110). Finally, RD have reported lower oligodendrocyte density in the amyg- was higher in patients in the ACR, ALIC, CR, SCR, and dala and prefrontal cortex in MDD patients [38–40] and SFO (Table S111). lower expression of genes related to oligodendrocyte function [41]. Overall, we provided evidence that adult MDD is associated with subtle widespread differences in Discussion WM microstructure that are robust across many samples worldwide. A coordinated approach was used to perform the largest WM abnormalities in adult MDD appear to be driven by multi-site DTI study in MDD to date. Differences in FA and patients with more than one episode of MDD. Abnormal- diffusivity measures (AD, RD, and MD) were examined ities in these WM tracts may be related to cumulative effects between depressed patients and healthy controls, as well as of stress on brain structure. Psychological stress is asso- associations with clinical characteristics of the disorder. We ciated with increased glucocorticoid release and stress- included DTI data from 20 cohorts, including a total of related neuroinflammation, which have been shown to 1305 patients with MDD and 1602 healthy controls. negatively impact WM microstructure [42, 43]. Our results Separate analyses were performed for adolescent and adult suggest that exposure to multiple episodes of MDD may MDD, and we found evidence for globally lower FA and have a neuroprogressive effect on WM microstructure. We higher RD across the full WM skeleton in adult MDD did not observe any associations between FA or diffusivity patients. These effects were mostly driven by patients with measures and the number of MDD episodes, suggesting that recurrent episodes, patients with an adult (older than 21 the relationship with the number of episodes may not be years) age of onset of depression, and patients not taking linear. It is, however, also possible that WM abnormalities antidepressants at the time of scanning. in these tracts are a risk factor for an unfavorable course of In adult patients with MDD, we found evidence of MDD or that the study is underpowered to detect effects in subtle, but widespread lower FA in 16 out of 25 tract-based first-episode patients. ROIs including the corpus callosum, internal capsule, WM differences were also primarily found in patients inferior fronto-occipital fasciculus, corona radiata, cingu- with an adult age of onset (>21 years). The findings may be lum, fornix, superior fronto-occipital fasciculus (SFO), and related to age, as the mean age at the time of scanning was sagittal stratum. These effects appear to be global, as these higher in adults with an adult-onset compared to adults with 1520 L. S. van Velzen et al. an adolescent-onset depression. We speculate that depres- meta-analysis were not significant in the UK Biobank sion may interact with the normal aging process, as results sample. Including only subjects in the meta-analysis within of diagnosis-by-age interaction analysis showed that MDD the age range of UK Biobank did not increase the con- is associated with increased decline in overall and inter- sistency in findings in the meta-analysis and UK Biobank hemispheric FA with age. These results suggest that MDD (see Tables S140–143). We speculate that this different may be associated with accelerated brain aging, which has pattern of results across regions and in effect sizes between been shown for regional gray matter volume [44, 45], the meta-analysis and UK Biobank may be related to the although longitudinal studies are required to confirm this fact that UK Biobank is a community cohort study (of hypothesis. A previous study did not observe a diagnosis- relatively well individuals) with a lifetime diagnosis of by-age interaction effect on WM, but this study was per- MDD, while most studies in the meta-analysis have speci- formed in a smaller sample of patients and controls and may fically recruited participants with a current diagnosis MDD have been underpowered to detect small effects [46]. and may have more severe symptoms at time of scanning. Identifying neuroimaging biomarkers related to anti- Finally, differences in the processing of images between the depressant treatment is important, as it not only increases ENIGMA-DTI protocol and the UK Biobank protocol may our understanding of the pathophysiology of MDD, but may be important, although most regions of interest show good also reveal neural mechanisms through which anti- agreement between protocols (please see supplements). depressants improve mood. In this study we observed lower In adolescents with MDD (21 years or younger), dif- FA in the cingulum, sagittal stratum, external capsule, and ferences in WM microstructure of the corpus callosum fornix in patients who were not taking antidepressant (i.e., lower FA and higher MD and RD) were also medication at the time of scanning. No such differences observed compared to age-matched healthy controls, were observed between patients who were taking anti- however, these findings did not survive FDR-correction depressants and healthy subjects, which could suggest that for multiple testing. The effect sizes of case-control WM deficits may normalize with antidepressant use. In line comparisons in adolescents were comparable to adults in with this, selective serotonin reuptake inhibitors (SSRIs) the corpus callosum, but were lower in other tracts. It is have previously been associated with increased myelina- possible that the effects are smaller in adolescent MDD tion, plasticity and myelin repair [47]. SSRIs also increase compared to adult MDD, due to the shorter disease release of brain-derived neurotrophic factor (BDNF), a duration and lower recurrence of MDD episodes in ado- neuroprotective protein that impacts oligodendrocytes and lescents. As the sample size in adolescent MDD was increases myelination in the central nervous system [48]. smaller than in adult MDD, we may not have been able to However, caution is warranted when interpreting these detect similar small differences in WM in adolescent findings as medication effects, as we have only examined MDD as we observed in adult MDD. The effect sizes for antidepressant use at the time of scanning and do not have differences in FA in the corpus callosum were similar to data on past antidepressant use. Furthermore, the cross- the effect sizes observed in adult MDD, suggesting that sectional nature of this study and possible confounding larger studies in adolescent MDD are needed to confirm effect of disease severity also limit our ability to conclude whether subtle differences in FA of the corpus callosum that these findings are medication effects. While we can also be detected in adolescent MDD patients. acknowledge these limitations, this study is the largest study The reported abnormalities in WM microstructure are to date to examine associations between current AD use and likely not specific to MDD, as previous work from the WM microstructure. Our findings are in line with a study by ENIGMA schizophrenia working group has also revealed Zeng et al. [49], which is to our knowledge the only study to lower global FA and higher RD in schizophrenia patients date that has examined alterations in WM between compared to healthy controls [50]. The strongest effects medication-naive MDD patients and MDD patients using were observed in the same regions, namely the corona antidepressants. Although their sample was small, the radiata and corpus callosum, although the effect sizes authors conclude that antidepressant use normalizes WM were higher in patients with schizophrenia (maximum deficits in MDD. Ultimately, to fully elucidate the direct Cohen’s d = 0.42). WM microstructure studies are cur- effect of antidepressant use on WM microstructure, DTI rently underway in other ENIGMA disease working measures should be incorporated in randomized long- groups, allowing future comparisons across multiple itudinal clinical trials with antidepressant medication in mental disorders. medication-naive MDD patients. We observed significant differences in adult MDD In secondary analyses, we ran the same analyses on a patients in WM tracts that lead to the hippocampus, large dataset from UK Biobank to examine how our find- including the fornix/stria terminalis and the hippocampal ings generalize to an adult population sample. Effect sizes portion of the cingulum bundle. Smaller hippocampal were generally lower and most significant findings in the volumes (on average) is one of the most consistently White matter disturbances in major depressive disorder: a coordinated analysis across 20 international. . . 1521 reported neuroimaging findings in the depression literature Conclusion (for meta-analyses please see [3, 51]). Similar mechanisms may underlie both findings. Stress-related decreases in In this first coordinated multi-site study of WM tract hippocampal gliogenesis, a reduction in the number of integrity in MDD, we provide evidence for subtle, global hippocampal oligodendrocytes or stress-related morpholo- differences in FA and RD in multiple WM tracts in adult gical changes to hippocampal neurons (i.e. decreased den- MDD, with the strongest regional effects being observed in dritic length and number of dendritic spines), could affect tracts that have been implicated in emotion regulation. both hippocampal volume and WM microstructure [52, 53]. Future studies with a larger sample of adolescents are It is also possible, however, that hippocampal volume and needed to examine whether more subtle WM alterations WM microstructure are independently impacted by stress- already exist at this age. Our results suggest that widespread related mechanisms including glutamate excitotoxicity, structural dysconnectivity may play a role in the patho- inflammation, oxidative stress or an increase in glucocorti- physiology of MDD and may become more pronounced coids. Recent imaging genetics studies also suggest that a with recurrent episodes of MDD. genetic predisposition to such neuroanatomical variations may also be related to the genetic underpinnings of mental Acknowledgements The ENIGMA-Major Depressive Disorder working group gratefully acknowledges support from the NIH Big illness [54]. Data to Knowledge (BD2K) award (U54 EB020403 to PMT) and NIH grant R01 MH116147 (PMT). LS is supported by an NHMRC MRFF Limitations Career Development Fellowship (APP1140764). We wish to acknowledge the patients and control subjects that have particiaped int the study. We thank Rosa Schirmer, Elke Schreiter, Reinhold While the sample size and use of harmonized protocols in Borschke and Ines Eidner for image acquisition and data preparation, this multi-site study are strengths, we also acknowledge and Anna Oliynyk for quality checks. We thank Dorothee P. Auer and certain limitations. First, we used a cutoff of 21 years for F. Holsboer for initiation of the RUD study. We wish to acknowledge adolescent and adult MDD and for categorizing adoles- the patients and control subjects that have particiaped int the study. We thank Rosa Schirmer, Elke Schreiter, Reinhold Borschke and Ines cent and adult age of onset. While this is consistent with Eidner for image acquisition and data preparation, and Anna Oliynyk our previous work [3, 4], alternative definitions may affect for quality checks. We thank Dorothee P. Auer and F. Holsboer for our findings. We also performed the case-control analysis initiation of the RUD study. NESDA: The infrastructure for the using a cutoff of 25 years, with adolescents being 25 years NESDA study (www.nesda.nl) is funded through the Geestkracht program of the Netherlands Organisation for Health Research and or younger and adults older than 25 years. Results in Development (Zon-Mw, grant number 10-000-1002) and is supported adults were very similar to the findings using the cutoff at by participating universities (VU University Medical Center, GGZ 21 and there were still no significant case-control differ- inGeest, Arkin, Leiden University Medical Center, GGZ Rivierduinen, ences in adolescents. Second, analyses of effects of age University Medical Center Groningen) and mental health care orga- nizations, see www.nesda.nl. M-JvT was supported by a VENI grant and age of onset on WM microstructure were performed at (NWO grant number 016.156.077). UCSF: This work was supported the site-level prior to meta-analysis, and are therefore by the Brain and Behavior Research Foundation (formerly NARSAD) hampered by the limited range of age and age of onset to TTY; the National Institute of Mental Health (R01MH085734 to within each site. Future studies harmonizing individual TTY; K01MH117442 to TCH) and by the American Foundation for Suicide Prevention (PDF-1-064-13) to TCH. Stanford: This work was level data across studies may allow for data to be pooled supported by NIMH Grants R01MH59259 and R37101495 to IHG. and for the effects of age and age of onset to be examined MS is partially supported by an award funded by the Phyllis and across the total age range across studies (12–88 years). Jerome Lyle Rappaport Foundation. Muenster: This work was funded Third, while several limitations of TBSS have been by the German Research Foundation (SFB-TRR58, Projects C09 and Z02 to UD) and the Interdisciplinary Center for Clinical Research recognized (please see [55]), it remains the most widely (IZKF) of the medical faculty of Münster (grant Dan3/012/17 to UD). used method to examine group differences in WM Marburg: This work was funded by the German Research Foundation microstructure. Fourth, while DTI analysis was harmo- (DFG, grant FOR2107 DA1151/5-1 and DA1151/5-2 to UD; KI 588/ nized using standardized protocols, DTI acquisition was 14-1, KI 588/14-2 to TK; KR 3822/7-1, KR 3822/7-2 to AK; JA 1890/ 7-1, JA 1890/7-2 to AJ). IMH-MDD: This work was supported by the not harmonized, which may have affected our findings. National Healthcare Group Research Grant (SIG/15012) awarded to Still, our moderator analyses showed no effect of DTI KS. Barcelona: This study was funded by two grants of the Fondo de acquisition parameters on effect sizes. Including DTI data Investigación Sanitaria (FIS: PI 10/00372; FIS: 13/1057)from the that was acquired with different acquisition parameters Instituto de Salud Carlos III, by the Centro de Investigación Biomédica en Red de Salud Mental (CIBERSAM). The author is funded through helps ensure that the findings are not specific to any ‘Miguel Servet’ research contract (CP16-0020), co-financed by the specific acquisition scheme. Finally, in this study we were European Regional Development Fund (ERDF) (2016–2019). QTIM: only able to examine antidepressant use at time of scan- We thank the twins and singleton siblings who gave generously of ning, but future studies may want to examine associations their time to participate in the QTIM study. We also thank the many research assistants, radiographers, and IT support staff for data between duration or (cumulative) dose of antidepressants acquisition and DNA sample preparation. This study was funded by and WM microstructure. 1522 L. S. van Velzen et al. the National Institute of Child Health & Human Development (RO1 2. Mathers CD, Loncar D. Projections of global mortality and burden HD050735); National Institute of Biomedical Imaging and Bioengi- of disease from 2002 to 2030. PLoS Med. 2006;3:e442. neering (Award 1U54EB020403-01, Subaward 56929223); National 3. Schmaal L, Veltman DJ, van Erp TGM, Sämann PG, Frodl T, Health and Medical Research Council, Australia (Project Grants Jahanshad N, et al. Subcortical brain alterations in major depres- 496682, 1009064). NIH ENIGMA-BD2K U54 EB020403 (Thomp- sive disorder: findings from the ENIGMA Major Depressive son); R01 MH117601 (Jahanshad/Schmaal). Magdeburg: M.L. and Disorder working group. Mol Psychiatry. 2016;21:806–12. M.W. are funded by SFB 779. Bipolar Family Study: This study has 4. Schmaal L, Hibar DP, Sämann PG, Hall GB, Baune BT, Jahan- received funding from the European Community's Seventh Framework shad N et al. Cortical abnormalities in adults and adolescents with Programme (FP7/2007-2013). This paper reflects only the author's major depression based on brain scans from 20 cohorts worldwide views and the European Union is not liable for any use that may be in the ENIGMA Major Depressive Disorder Working Group. Mol made of the information contained therein. This work was also sup- Psychiatry. 2017;22:900–9. ported by a Wellcome Trust Strategic Award (104036/Z/14/Z). Min- 5. Liao Y, Huang X, Wu Q, Yang C, Kuang W, Du M, et al. Is nesota Adolescent Depression Study: The study was funded by the depression a disconnection syndrome? Meta-analysis of diffusion National Institute of Mental Health (K23MH090421), the National tensor imaging studies in patients with MDD. J Psychiatry Neu- Alliance for Research on Schizophrenia and Depression, the Uni- rosci. 2013;38:49–56. versity of Minnesota Graduate School, the Minnesota Medical Foun- 6. Ota M, Noda T, Sato N, Hattori K, Hori H, Sasayama D, et al. dation, and the Biotechnology Research Center (P41 RR008079 to the White matter abnormalities in major depressive disorder with Center for Magnetic Resonance Research), University of Minnesota, melancholic and atypical features: a diffusion tensor imaging and the Deborah E. Powell Center for Women's Health Seed Grant, study. Psychiatry Clin Neurosci. 2015;69:360–8. University of Minnesota. Dublin: This study was supported by Science 7. LeWinn KZ, Connolly CG, Wu J, Drahos M, Hoeft F, Ho TC, Foundation Ireland through a Stokes Professorhip grant to TF. MPIP: et al. White matter correlates of adolescent depression: structural The MPIP Sample comprises patients included in the Recurrent Uni- evidence for frontolimbic disconnectivity. J Am Acad Child polar Depression (RUD) Case-Control study at the clinic of the Max Adolesc Psychiatry. 2014;53:899–909, 909–7. Planck Institute of Psychiatry, Munich, German. The RUD study was 8. Zou K, Huang X, Li T, Gong Q, Li Z, Ou-yang L et al. Alterations of supported by GlaxoSmithKline. white matter integrity in adults with major depressive disorder: a mag- netic resonance imaging study. J Psychiatry Neurosci. 2008;33:525–30. 9. Choi KS, Holtzheimer PE, Franco AR, Kelley ME, Dunlop BW, Compliance with ethical standards Hu XP, et al. Reconciling variable findings of white matter integrity in major depressive disorder. Neuropsychopharmacol- Conflict of interest IBH has previously led community-based and ogy. 2014;39:1332–9. pharmaceutical industry-supported (Wyeth, Eli Lily, Servier, Pfizer, 10. Olvet DM, Delaparte L, Yeh F-C, DeLorenzo C, McGrath PJ, AstraZeneca) projects. He was a member of the Medical Advisory Weissman MM, et al. A comprehensive examination of white Panel for Medibank Private, a Board Member of Psychosis Australia matter tracts and connectometry in major depressive disorder. Trust and a member of Veterans Mental Health Clinical Reference Depress Anxiety. 2016;33:56–65. group. He is the Chief Scientific Advisor to, and an equity shareholder 11. Westlye LT, Walhovd KB, Dale AM, Bjørnerud A, Due- in, Innowell. AMM has received research support from Janssen, Eli Tønnessen P, Engvig A, et al. Life-span changes of the human Lilly and The Sackler Trust. AMM has also received speaker's fees brain white matter: diffusion tensor imaging (DTI) and volumetry. from Illumina and Janssen. The remaining authors declare that they Cereb Cortex. 2010;20:2055–68. have no conflict of interest. 12. Cheng Y, Xu J, Yu H, Nie B, Li N, Luo C, et al. Delineation of early and later adult onset depression by diffusion tensor imaging. Publisher’s note: Springer Nature remains neutral with regard to PLoS ONE. 2014;9:e112307. jurisdictional claims in published maps and institutional affiliations. 13. Taylor WD, MacFall JR, Gerig G, Krishnan RR. Structural integrity of the uncinate fasciculus in geriatric depression: rela- Open Access This article is licensed under a Creative Commons tionship with age of onset. Neuropsychiatr Dis Treat. 2007;3: Attribution 4.0 International License, which permits use, sharing, 669–74. adaptation, distribution and reproduction in any medium or format, as 14. de Diego-Adeliño J, Pires P, Gómez-Ansón B, Serra-Blasco M, long as you give appropriate credit to the original author(s) and the Vives-Gilabert Y, Puigdemont D, et al. Microstructural white- source, provide a link to the Creative Commons license, and indicate if matter abnormalities associated with treatment resistance, severity changes were made. The images or other third party material in this and duration of illness in major depression. Psychol Med. article are included in the article’s Creative Commons license, unless 2014;44:1171–82. indicated otherwise in a credit line to the material. If material is not 15. Davis SW, Dennis NA, Buchler NG, White LE, Madden DJ, included in the article’s Creative Commons license and your intended Cabeza R. Assessing the effects of age on long white matter tracts use is not permitted by statutory regulation or exceeds the permitted using diffusion tensor tractography. Neuroimage. 2009;46: use, you will need to obtain permission directly from the copyright 530–41. holder. To view a copy of this license, visit http://creativecommons. 16. Paus T, Collins D, Evans A, Leonard G, Pike B, Zijdenbos A. org/licenses/by/4.0/. Maturation of white matter in the human brain: a review of magnetic resonance studies. Brain Res Bull. 2001;54:255–266. 17. Murphy ML, Frodl T. Meta-analysis of diffusion tensor imaging References studies shows altered fractional anisotropy occurring in distinct brain areas in association with depression. Biol Mood Anxiety 1. Kessler RC, Angermeyer M, Anthony JC, DE Graaf R, Demyt- Disord. 2011;1:3. tenaere K, Gasquet I, et al. Lifetime prevalence and age-of-onset 18. Wise T, Radua J, Nortje G, Cleare AJ, Young AH, Arnone D. distributions of mental disorders in the World Health Organiza- Voxel-based meta-analytical evidence of structural dis- tion’s World Mental Health Survey Initiative. World Psychiatry. connectivity in major depression and bipolar disorder. Biol Psy- 2007;6:168–76. chiatry. 2016;79:293–302. White matter disturbances in major depressive disorder: a coordinated analysis across 20 international. . . 1523 19. Chen G, Hu X, Li L, Huang X, Lui S, Kuang W, et al. Dis- 36. Song S-K, Sun S-W, Ramsbottom MJ, Chang C, Russell J, Cross organization of white matter architecture in major depressive AH. Dysmyelination revealed through MRI as increased radial disorder: a meta-analysis of diffusion tensor imaging with tract- (but unchanged axial) diffusion of water. Neuroimage. based spatial statistics. Sci Rep. 2016;6:21825. 2002;17:1429–36. 20. Jahanshad N, Kochunov PV, Sprooten E, Mandl RC, Nichols 37. Sacchet MD, Gotlib IH. Myelination of the brain in major TE, Almasy L, et al. Multi-site genetic analysis of diffusion depressive disorder: an in vivo quantitative magnetic resonance images and voxelwise heritability analysis: a pilot project of imaging study. Sci Rep. 2017;7:2200. the ENIGMA-DTI working group. Neuroimage. 2013;81: 38. Regenold WT, Phatak P, Marano CM, Gearhart L, Viens CH, 455–69. Hisley KC. Myelin staining of deep white matter in the dorso- 21. Kochunov P, Jahanshad N, Sprooten E, Nichols TE, Mandl RC, lateral prefrontal cortex in schizophrenia, bipolar disorder, and Almasy L, et al. Multi-site study of additive genetic effects on unipolar major depression. Psychiatry Res. 2007;151:179–88. fractional anisotropy of cerebral white matter: comparing meta 39. Hamidi M, Drevets WC, Price JL. Glial reduction in amygdala in and megaanalytical approaches for data pooling. Neuroimage. major depressive disorder is due to oligodendrocytes. Biol Psy- 2014;95:136–50. chiatry. 2004;55:563–9. 22. Kochunov P, Jahanshad N, Marcus D, Winkler A, Sprooten E, 40. Uranova NA, Vostrikov VM, Orlovskaya DD, Rachmanova VI. Nichols TE, et al. Heritability of fractional anisotropy in human Oligodendroglial density in the prefrontal cortex in schizophrenia white matter: a comparison of Human Connectome Project and and mood disorders: a study from the Stanley Neuropathology ENIGMA-DTI data. Neuroimage. 2015;111:300–11. Consortium. Schizophr Res. 2004;67:269–75. 23. Alexander AL, Hurley SA, Samsonov AA, Adluru N, Hosseinbor 41. Aston C, Jiang L, Sokolov BP. Transcriptional profiling reveals AP, Mossahebi P, et al. Characterization of cerebral white matter evidence for signaling and oligodendroglial abnormalities in the properties using quantitative magnetic resonance imaging stains. temporal cortex from patients with major depressive disorder. Mol Brain Connect. 2011;1:423–336. Psychiatry. 2005;10:309–22. 24. Smith DJ, Nicholl BI, Cullen B, Martin D, Ul-Haq Z, Evans J, 42. Howell BR, McCormack KM, Grand AP, Sawyer NT, Zhang X, et al. Prevalence and characteristics of probable major depression Maestripieri D, et al. Brain white matter microstructure alterations and bipolar disorder within UK biobank: cross-sectional study of in adolescent rhesus monkeys exposed to early life stress: asso- 172,751 participants. PLoS ONE. 2013;8:e75362. ciations with high cortisol during infancy. Biol Mood Anxiety 25. Tomasch J. Size, distribution, and number of fibres in the human Disord. 2013;3:21. corpus callosum. Anat Rec. 1954;119:119–35. 43. Prasad KM, Upton CH, Nimgaonkar VL, Keshavan MS. Differ- 26. Picó-Pérez M, Radua J, Steward T, Menchón JM, Soriano-Mas C. ential susceptibility of white matter tracts to inflammatory med- Emotion regulation in mood and anxiety disorders: a meta- iators in schizophrenia: an integrated DTI study. Schizophr Res. analysis of fMRI cognitive reappraisal studies. Prog Neu- 2015;161:119–25. ropsychopharmacol Biol Psychiatry. 2017;79:96–104. 44. Sacchet MD, Camacho MC, Livermore EE, Thomas EAC, Gotlib 27. Wagner G, Koch K, Schachtzabel C, Reichenbach JR, Sauer H, IH. Accelerated aging of the putamen in patients with major Schlösser RGM. Enhanced rostral anterior cingulate cortex acti- depressive disorder. J Psychiatry Neurosci. 2017;42:164–71. vation during cognitive control is related to orbitofrontal volume 45. Frodl TS, Koutsouleris N, Bottlender R, Born C, Jäger M, Scupin reduction in unipolar depression. J Psychiatry Neurosci. 2008; I, et al. Depression-related variation in brain morphology over 3 33:199–208. years. Arch Gen Psychiatry. 2008;65:1156. 28. Phan KL, Wager T, Taylor SF, Liberzon I. Functional neuroa- 46. Kochunov P, Glahn DC, Rowland LM, Olvera RL, Winkler A, natomy of emotion: a meta-analysis of emotion activation studies Yang Y-H, et al. Testing the hypothesis of accelerated cerebral in PET and fMRI. Neuroimage. 2002;16:331–48. white matter aging in schizophrenia and major depression. Biol 29. Matsuo K, Glahn DC, Peluso MAM, Hatch JP, Monkul ES, Najt Psychiatry. 2013;73:482–91. P, et al. Prefrontal hyperactivation during working memory task in 47. Bartzokis G. Neuroglialpharmacology: myelination as a shared untreated individuals with major depressive disorder. Mol Psy- mechanism of action of psychotropic treatments. Neuropharma- chiatry. 2007;12:158–66. cology. 2012;62:2137–53. 30. Guo W, Liu F, Xue Z, Gao K, Wu R, Ma C, et al. Altered white 48. Xiao J, Wong AW, Willingham MM, van den Buuse M, Kilpa- matter integrity in young adults with first-episode, treatment- trick TJ, Murray SS. Brain-derived neurotrophic factor promotes naive, and treatment-responsive depression. Neurosci Lett. central nervous system myelination via a direct effect upon oli- 2012;522:139–44. godendrocytes. Neurosignals. 2010;18:186–202. 31. Choi S, Han K-M, Won E, Yoon B-J, Lee M-S, Ham B-J. 49. Zeng L-L, Liu L, Liu Y, Shen H, Li Y, Hu D. Antidepressant Association of brain-derived neurotrophic factor DNA methyla- treatment normalizes white matter volume in patients with major tion and reduced white matter integrity in the anterior corona depression. PLoS ONE. 2012;7:e44248. radiata in major depression. J Affect Disord. 2015;172:74–80. 50. Kelly S, Jahanshad N, Zalesky A, Kochunov P, Agartz I, Alloza 32. Alexopoulos GS, Glatt CE, Hoptman MJ, Kanellopoulos D, C, et al. Widespread white matter microstructural differences in Murphy CF, Kelly RE, et al. BDNF Val66met polymorphism, schizophrenia across 4322 individuals: results from the ENIGMA white matter abnormalities and remission of geriatric depression. J Schizophrenia DTI Working Group. Mol Psychiatry. Affect Disord. 2010;125:262–8. 2018;23:1261–9. 33. Sanjuan PM, Thoma R, Claus ED, Mays N, Caprihan A. Reduced 51. Santos MAO, Bezerra LS, Carvalho ARMR, & Brainer-Lima white matter integrity in the cingulum and anterior corona radiata AM. Global hippocampal atrophy in major depressive disorder: a in posttraumatic stress disorder in male combat veterans: a dif- meta-analysis of magnetic resonance imaging studies. Trends fusion tensor imaging study. Psychiatry Res. 2013;214:260–8. Psychiatry Psychother. 2018;40:369–78. 34. Drevets WC, Price JL, Furey ML. Brain structural and functional 52. Boku S, Nakagawa S, Toda H, Hishimoto A. Neural basis of abnormalities in mood disorders: implications for neurocircuitry major depressive disorder: beyond monoamine hypothesis. Psy- models of depression. Brain Struct Funct. 2008;213:93–118. chiatry Clin Neurosci. 2018;72:3–12. 35. Song S-K, Yoshino J, Le TQ, Lin S-J, Sun S-W, Cross AH, et al. 53. Czéh B, Lucassen PJ. What causes the hippocampal volume Demyelination increases radial diffusivity in corpus callosum of decrease in depression? Eur Arch Psychiatry Clin Neurosci. mouse brain. Neuroimage. 2005;26:132–40. 2007;257:250–60. 1524 L. S. van Velzen et al. 54. Van der Meer D, Rokicki J, Kaufmann T, Córdova-Palomera A, 55. Bach M, Laun FB, Leemans A, Tax CM, Biessels GJ, Stieltjes B, Moberget T, Alnæs D, et al. Brain scans from 21,297 individuals et al. Methodological considerations on tract-based spatial statis- reveal the genetic architecture of hippocampal subfield volumes. tics (TBSS). Neuroimage. 2014;100:358–69. Mol Psychiatry, 2018. https://doi.org/10.1038/s41380-018-0262- 7. [Epub ahead of print]. Affiliations Laura S. van Velzen1,2,3 Sinead Kelly4,5 Dmitry Isaev6 Andre Aleman7 Lyubomir I. Aftanas8,9 Jochen Bauer10 ● ● ● ● ● ● Bernhard T. Baune 11,12,13 Ivan V. Brak 8,14 Angela Carballedo15,16 Colm G. Connolly 17,18 ● ● ● ● Baptiste Couvy-Duchesne19 Kathryn R. Cullen20 Konstantin V. Danilenko8 Udo Dannlowski11 Verena Enneking11 ● ● ● ● ● Elena Filimonova8 Katharina Förster11 Thomas Frodl15,21,22 Ian H. Gotlib23 Nynke A. Groenewold24,25 ● ● ● ● ● Dominik Grotegerd11 Mathew A. Harris 26 Sean N. Hatton27 Emma L. Hawkins26 Ian B. Hickie27 ● ● ● ● ● Tiffany C. Ho17,23,28 Andreas Jansen29 Tilo Kircher29 Bonnie Klimes-Dougan30 Peter Kochunov31 Axel Krug29 ● ● ● ● ● ● Jim Lagopoulos32 Renick Lee33 Tristram A. Lett 34 Meng Li35 Frank P. MacMaster36,37 Nicholas G. Martin 38 ● ● ● ● ● ● Andrew M. McIntosh 26,39 Quinn McLellan 40,41 Susanne Meinert11 Igor Nenadić29 Evgeny Osipov14 ● ● ● ● ● Brenda W. J. H. Penninx1 Maria J. Portella42,43,44 Jonathan Repple11 Annerine Roos45 Matthew D. Sacchet46 ● ● ● ● ● Philipp G. Sämann47 Knut Schnell48 Xueyi Shen26 Kang Sim49,50,51 Dan J. Stein 52 Marie-Jose van Tol 7 ● ● ● ● ● ● Alexander S. Tomyshev53 Leonardo Tozzi21,28 Ilya M. Veer34 Robert Vermeiren54,55,56 Yolanda Vives-Gilabert57 ● ● ● ● ● Henrik Walter 34 Martin Walter58 Nic J. A. van der Wee55,59 Steven J. A. van der Werff55,59 ● ● ● ● Melinda Westlund Schreiner 30 Heather C. Whalley 26 Margaret J. Wright 60,61 Tony T. Yang17 Alyssa Zhu6 ● ● ● ● ● Dick J. Veltman1,62 Paul M. Thompson6 Neda Jahanshad 6 Lianne Schmaal2,3 ● ● ● 1 Department of Psychiatry, Amsterdam UMC, The Netherlands 14 Lab. of Experimental & Translational Neuroscience, Novosibirsk 2 State University, Novosibirsk, Russia Orygen, The National Centre of Excellence in Youth Mental Health, Parkville, Australia 15 Department of Psychiatry and Trinity Institute of Neuroscience, 3 Trinity College Dublin, Dublin, Ireland Centre for Youth Mental Health, The University of Melbourne, Melbourne, Australia 16 North Dublin Mental Health Services, Dublin, Ireland 4 Department of Psychiatry, Beth Israel Deaconess Medical Center, 17 Department of Psychiatry, University of California, San Francisco, Harvard Medical School, Boston, MA, USA CA, USA 5 Psychiatry Neuroimaging Laboratory, Brigham and Women’s 18 Department of Biomedical Sciences, Florida State University, Hospital, Harvard Medical School, Boston, MA, USA Tallahassee, FL, USA 6 Imaging Genetics Center, Mark and Mary Stevens Neuroimaging 19 Institute for Molecular Bioscience, The University of Queensland, and Informatics Institute, Keck School of Medicine, University of Brisbane, QLD, Australia Southern California, Marina del Rey, CA, USA 20 7 Department of Psychiatry and Behavioral Sciences, The Cognitive Neuroscience Center, University Medical Center University of Minnesota, Minneapolis, MN, USA Groningen, University of Groningen, Groningen, The Netherlands 21 Department of Psychiatry and Psychotherapy, Otto von Guericke 8 FSSBI “Scientific Research Institute of Physiology & Basic University, Madgeburg, Germany Medicine”, Laboratory of Affective, Cognitive & Translational 22 Neuroscience, Novosibirsk, Russia German Center for Neurodegenerative Disease, 9 Magdeburg, Germany Department of Neuroscience, Novosibirsk State University, 23 Novosibirsk, Russia Department of Psychology, Stanford University, Stanford, CA, 10 USA University of Münster, Institute of Clinical Radiology, 24 Münster, Germany Department of Psychiatry, University of Cape Town, Cape Town, 11 South Africa Department of Psychiatry, University of Münster, 25 Münster, Germany Department of Psychiatry, Interdisciplinary Center 12 Psychopathology and Emotion regulation (ICPE), University Department of Psychiatry, The University of Melbourne, Medical Center Groningen, University of Groningen, Melbourne, VIC, Australia Groningen, The Netherlands 13 The Florey Institute of Neuroscience and Mental Heatlh, The 26 Division of Psychiatry, University of Edinburgh, Edinburgh, UK University of Melbourne, Melbourne, VIC, Australia White matter disturbances in major depressive disorder: a coordinated analysis across 20 international. . . 1525 27 Youth Mental Health Team, Brain and Mind Centre, University of Sydney, Camperdown, Australia 45 SAMRC Unit on Risk & Resilience in Mental Disorders, 28 Department of Psychiatry, Stellenbosch University, Cape Town, Department of Psychiatry & Behavioral Sciences, Stanford South Africa University, Stanford, CA, USA 46 29 Center for Depression, Anxiety, and Stress Research, McLean Department of Psychiatry, University of Marburg, Hospital, Harvard Medical School, Belmont, MA, USA Marburg, Germany 47 30 Max Planck Institute of Psychiatry, Munich, Germany Department of Psychology, The University of Minnesota, Minneapolis, MN, USA 48 Department of Psychiatry, University of Heidelberg, 31 Heidelberg, Germany Maryland Psychiatric Research Center, Department of Psychiatry, University of Maryland School of Medicine, Baltimore, MD, USA 49 West Region and Research Division, Institute of Mental Health, 32 Singapore, Singapore Sunshine Coast Mind and Neuroscience—Thompson Institute, Birtinya, QLD, Australia 50 Yong Loo Lin School of Medicine, National University of 33 Singapore, Singapore, Singapore Research Division, Institute of Mental Health, Singapore, Singapore 51 Lee Kong Chian School of Medicine, Nanyang Technological 34 University, Singapore, Singapore Division of Mind and Brain Research, Department of Psychiatry and Psychotherapy CCM, Charité—Universitätsmedizin Berlin, 52 SAMRC Unit on Risk & Resilience in Mental Disorders, corporate member of Freie Universität Berlin, Humboldt- Department of Psychiatry & Neuroscience Institute, University of Universität zu Berlin, and Berlin Institute of Health, Cape Town, South Africa Berlin, Germany 53 35 Mental Health Research Center, Moscow, Russia Department of Neurology, University of Magdeburg, 54 Magdeburg, Germany Curium-LUMC Child and Adolescent Psychiatry, Leiden 36 University Medical Center, Leiden, The Netherlands Psychiatry and Paediatrics, University of Calgary, 55 Calgary, Canada Leiden Institute for Brain and Cognition, Leiden, The Netherlands 37 56 Strategic Clinical Network for Addictions and Mental Health, Institute of Psychology, Leiden University, Leiden, The Calgary, Canada Netherlands 38 57 QIMR Berghofer Medical Research Institute, Brisbane, QLD, Instituto ITACA, Universitat Politècnica de València, Australia València, Spain 39 58 Centre for Cognitive Ageing and Cognitive Epidemiology, Department of Psychiatry and Psychotherapy, University of University of Edinburgh, Edinburgh, UK Tübingen, Tubingen, Germany 40 59 Department of Neuroscience, University of Calgary, Department of Psychiatry, Leiden University Medical Center, Calgary, Canada Leiden, The Netherlands 41 60 Alberta Children’s Hospital Research Institute, Calgary, Canada Queensland Brain Institute, The University of Queensland, 42 Brisbane, QLD, Australia Department of Psychiatry, Institute of Biomedical Research Sant 61 Pau, Barcelona, Spain Centre for Advanced Imaging, The University of Queensland, 43 Brisbane, QLD, Australia CIBERSAM, Madrid, Spain 62 44 Amsterdam Neuroscience, Amsterdam, The Netherlands Universitat Autònoma de Barcelona, Barcelona, Spain