Transcriptional and Cellular Signatures of Morphometric Similarity Remodeling in Major Depressive Disorder

Wei Liao (  weiliao.wl@gmail.com ) University of Electronic Science and Technology of China https://orcid.org/0000-0001-7406-7193 Jiao Li MOE Key Laboratory for Neuroinformation, University of Electronic Science and Technology of China Jakob Seidlitz National Institute of Mental Health https://orcid.org/0000-0002-8164-7476 John Suckling University of Cambridge Feiyang Fan University of Electronic Science and Technology of China Gongjun Ji The First A liated Hospital of Anhui Medical University Yao Meng University of Electronic Science and Technology of China Siqi Yang University of Electronic Science and Technology of China Kai Wang Department of Neurology, the First A liated Hospital of Anhui Medical University Huafu Chen University of Electronic Science and Technology of China


INTRODUCTION
Major depressive disorder (MDD) is a prevalent worldwide psychiatric disease that 54 often first occurs in adolescence 1 . Despite significant efforts, our current 55 understanding of its pathophysiology is unclear with inconsistent brain architectural 56 changes 2 and the variable effects of treatment 3 . Although neuroimaging studies show 57 some focal structural alterations 4 , functionally MDD is increasingly recognized as a 58 disorder involving brain "disconnectivity" 5 .   Genetic factors play important roles in brain connectomes 17,18 , and brain-wide gene  In this study, we investigated MDD-related morphometric disconnections and their 95 relationships with transcriptomic profiles in discovery and replication independent 96 samples. We tested four key hypotheses: i) that MS remodeling in MDD is associated 97 with anatomically patterned gene expressions, using the AHBA; ii) that the resultant 98 gene enrichments were specifically associated with genes that were differentially   (Table S1). There was 111 no difference in image quality, age and sex between patients and HC for both 112 discovery and replication samples (Supplemental Result 1 and Fig. S1).

114
Morphometric similarity remodeling in MDD 115 We first calculated the MS connection weights (a 308×308 matrix) from inter-regional In the discovery sample, summing regional MS weights across all regions for each  Table S2). The identically derived 132 cross-sectional MDD-HC t-map from replication sample was significantly spatially 133 related to the discovery sample (r(306) = 0.43, pspin = 0.0002; Fig. S4A&B).

144
Next, to identify locations of case-control differences, we divided the cortex using two  Table S4 and Fig. S5D.

153
The case-control t-map was significantly spatially correlated with the mean control   Table S5). An exploratory correlation analysis was also performed across 168 all D-K regions. We found that dorsal lateral prefrontal cortex exhibited a significant  Cortical gene expression related to regional MS differences 176 We used the Allen Human Brain Atlas (AHBA) (http://human.brain-map.org), a whole-177 brain transcriptomic dataset, to obtain brain gene expressions (Supplemental Result  Figure 3C).

192
In total, 2,984 genes constituted the regional MS difference gene list in MDD patients.

256
For visualization, uncorrected overlapping ontology terms are shown in Fig. S9. 257 Significantly overlapping ontology terms support the generalized relationship 258 between gene expression and the MS differences in MDD.

260
Transcriptional signatures for canonical cell types 261 To further refine our analysis, and considering cellular diversity in the brain, we took   Cellular characterization of the MDD-related genes 385 We showed that cellular organization of the human brain provides a biological 386 mechanism that can translate genes of MDD-related brain alterations into MDD- Case-control analysis of the MS network 473 The regional MS was calculated by using the sum of weighted correlation coefficients 474 between a given region and its correlations to all other regions. To estimate the spatial 475 pattern, regional MS was averaged across all HC participants. To examine the case-476 control differences, a generalized linear model was used with regional MS values as 477 the dependent variables. Age, sex, and education level were added as covariates.

478
Significance was set at p < 0.05 with false-discovery rate (FDR) correction for multiple 479 comparisons across regions.

481
The above case-control analyses were also used for the replication samples. To test 482 replicability of regional MS differences, a spatial similarity analysis was conducted on  Table S6). The AHBA dataset was processed  The degree of overlapped genes was measured by the odds ratio (OR). We are grateful to all the participants in this study. We thank International Science 546 Editing (http://www.internationalscienceediting.com) for editing this manuscript. This The authors report no biomedical financial interests or potential conflicts of interest.  across multimodal magnetic resonance imaging features (e.g., myelination, gray matter, and curvature) to produce a 308 × 308 matrix (depicted by a subdivision of the Desikan-Killiany atlas, D-K 308). Then, casecontrol differences across regions were computed. (B) Gene expression pro les. Gene expression pro les from the Allen Human Brain Atlas in 152 regions (left hemisphere only) were averaged across six postmortem brains. Partial least squares regression was then used to identify imaging-transcriptomic associations. Finally, enrichment analysis was performed on the gene list associated with PLS1.

Figure 2
Case-control differences of regional morphometric similarities. (A) Mean regional morphometric similarity (MS) pattern of healthy controls (HCs). The frontal and temporal lobes exhibited high MS values, whereas the occipital and somatosensory cortices showed low MS values. (B) Case-control comparison (t-map) of regional MS in discovery samples ( rst row, unthresholded). Seven cortical regions showed statistically signi cant differences (bottom row, pFDR < 0.05). (C) A scatterplot of the mean regional MS (x-axis) and case-control t-map (y-axis) ( rst row). Most cortical regions exhibited de-differentiation (34%) and decoupling (41%) in MDD patients (bottom row).

Figure 3
Gene expression pro les related to morphometric similarity differences. (A) Regional morphometric similarity (MS) differences in the left hemisphere (unthresholded). (B) A weighted gene expression map of regional PLS1 scores in the left hemisphere (unthresholded). (C) A scatterplot of regional PLS1 scores (a weighted sum of 10,027 gene expression scores) and regional MS differences. (D) Ranked PLS1 loadings. (E) Major depressive disorder (MDD)-related genes from in-situ hybridization in the adult human brain positively (e.g., CUX2; Pearson's r(150) = 0.50, p < 0.0001) and negatively (e.g., TAC1; Pearson's r(150) = -0.53, p < 0.0001) correlated with regional MS differences.  is represented by a circle node, where its size is proportional to the number of input genes included in that term, and its color represents its cluster identity (i.e., nodes of the same color belong to the same cluster).  Cell type-speci c expression to MS differences-related genes. (A) Regional gene expression maps of each cell type from overlapping genes between PLS1− gene list and each cell type-speci c genes. (B) The number of overlapping genes for each cell type. (C) Gene ontology terms enriched for MS differencesrelated genes for the cell types

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. MDDMSSOMNCv02.pdf DataS1PLS.xlsx