Association of Genetic Variation at AQP4 Locus with Vascular Depression

Despite its substantial clinical importance, specific genetic variants associated with depression have not yet been identified. We sought to identify genetic variants associated with depression by (a) focusing on a more homogenous subsample (vascular depression) and (b) applying a three-stage approach. First, we contacted 730 participants with a confirmed atherosclerotic disease (coronary artery disease) from a population-based study population (German Myocardial Infarction Family Study IV) for psychiatric assessment with the Mini International Neuropsychiatric Interview. Second, we genotyped these patients using genome-wide single nucleotide polymorphism (SNP) arrays. Third, we characterized the SNP via in-silico analysis. The final sample consisted of 342 patients (78.3% male, age = 63.2 ± 9.9 years), 22.8% with a severe depressive disorder. Variant rs528732638 on chromosome 18q11.2 was a genome-wide significant variant and was associated with 3.6-fold increase in the odds of lifetime depression. The locus belongs to a linkage disequilibrium block showing expression quantitative trait loci effects on three putative cis-regulated genes, including the aquaporin 4 (AQP4) locus. AQP4 is already known to mediate the formation of ischemic edema in the brain and heart, increasing the size and extent of resulting lesions. Our findings indicate that AQP4 may also play a role in the etiopathology of vascular depression.


Introduction
Unipolar depression is a common disorder, with 150 million men and women affected worldwide. With 65.5 million disability-adjusted life years (DALYs) worldwide, it is also one of the leading causes of loss of healthy life years [1]. Regarding the etiology, a bio-psycho-social model is assumed, with an estimated heritability of 35% [2]. However, no major gene locus has been consistently shown to be associated with unipolar depression [3][4][5][6], which may be attributable to a β error due to sample heterogeneity. Thus, research focusing on a more homogenous subsample holds promise, such as analyzing subjects with vascular depression. The vascular depression concept evolved at the end of the last century based on consistent differences between unipolar depression manifesting before or after the age of 50 (early onset depression (EOD) or late onset depression (LOD)) [7,8]. Regarding the clinical presentation, patients with LOD tend to exhibit more substantial psychomotor retardation and executive dysfunction but a less depressed affect than patients with EOD [9,10] and tend to respond poorly to treatment [11,12].
The term "vascular depression" places cerebrovascular lesions at the heart of the etiopathology of LOD. This assumption is based on morphological, functional, and epidemiological data: patients with LOD show greater carotid pathology [13] and more cerebrovascular lesions (such as white matter hyperintensities, silent brain infarctions, and microbleeds) [14,15] than patients with EOD. Additionally, cerebrovascular lesions are associated with a higher prevalence of depressive episodes [16]. The relevance of these cerebrovascular lesions is supported by functional imaging studies showing altered connectivity in patients with LOD, which led to the disconnectivity hypothesis of vascular depression [10,[17][18][19]. Furthermore, epidemiological studies link unipolar depression with vascular risk factors [20] and with vascular diseases such as dementia, stroke, myocardial infarction and coronary artery disease (CAD) [21][22][23]. The prevalence of vascular depression in the elderly is estimated to range from 2.4 to 3.4% [24,25]. Regarding the genetics of vascular depression, the only study published to date failed to identify an association with the SLC6A4 locus of the serotonin transporter gene [26]. Therefore, we conducted a genome-wide association study (GWAS) of severe depressive disorders in patients with a confirmed atherosclerotic disorder, namely coronary artery disease.

Study Protocol and Study Subjects
The study protocol was approved by the Ethics Committee of the University of Lübeck (ID 04/041) and is depicted in Figure 1. Regarding our study population, we chose to study patients with CAD which correlates highly with cerebrovascular disease [27,28] and thus constitutes an acceptable surrogate marker. When admitted to the University Hospital of Lübeck, Germany, for cardiac catheterization between February 2004 and December 2012, 730 patients with coronary heart disease (154 females, 576 males; mean age = 59.5 years; standard deviation (SD) = 9.8 years; previously described as the German Myocardial Infarction Family Study IV [29]), provided fully informed consent to the use of their data and genetic material for research purposes and to be contacted for further research purposes, in accordance with the Declaration of Helsinki. Of these individuals, 342 patients agreed to participate in our telephone survey when contacted by phone several years later (median = 1.8 years), between March 2013 and January 2015. Exclusion criteria were a lack of informed consent, an inability to participate in a telephone survey (due to cognitive deficits, hearing impairment, aphasia, or insufficient language skills) and the exclusion of CAD during cardiac catheterization.

Assessment of Depression
To further reduce the risk of a β error, we focused on severe depressive disorders (Major Depressive Episode and Dysthymia) which were diagnosed during the telephone survey using the Mini International Neuropsychiatric Interview (M.I.N.I. 5.0.0) [30]. The M.I.N.I. is a short, structured diagnostic interview for mental disorders according to Diagnostic and Statistical Manual of Mental Disorders (DSM) IV with good psychometric qualities. As the timing of the interview was arbitrary, current severe depressive disorders were not differentiated from previous, currently remitted disorders. A detailed report on the psychiatric morbidity found in our sample has been published elsewhere [31].

DNA Isolation and Genotyping
DNA was isolated at the Institute for Cardiogenetics at the University of Lübeck according to standard protocols. Genotyping was performed on a Affymetrix Genome-Wide Human SNP Array 6.0 (Thermo Fisher Scientific, Santa Clara, CA 95051, USA) in cooperation with the Helmholtz-Zentrum in Munich, Germany.

Pre/Post-Processing and Test of Genome-Wide Associations
In our analysis, we used variants that were previously imputed into the 1000 Genomes Phase 3 reference panel. Variants with poor imputation quality (i.e. info < 0.8), a minor allele frequency (MAF) < 0.05 and failing the Hardy-Weinberg-Equilibrium (P < 10 −4 ) were filtered out. We defined cases as patients with CAD who were diagnosed with severe depressive disorder in the interview (regardless

Assessment of Depression
To further reduce the risk of a β error, we focused on severe depressive disorders (Major Depressive Episode and Dysthymia) which were diagnosed during the telephone survey using the Mini International Neuropsychiatric Interview (M.I.N.I. 5.0.0) [30]. The M.I.N.I. is a short, structured diagnostic interview for mental disorders according to Diagnostic and Statistical Manual of Mental Disorders (DSM) IV with good psychometric qualities. As the timing of the interview was arbitrary, current severe depressive disorders were not differentiated from previous, currently remitted disorders. A detailed report on the psychiatric morbidity found in our sample has been published elsewhere [31].

DNA Isolation and Genotyping
DNA was isolated at the Institute for Cardiogenetics at the University of Lübeck according to standard protocols. Genotyping was performed on a Affymetrix Genome-Wide Human SNP Array 6.0 (Thermo Fisher Scientific, Santa Clara, CA 95051, USA) in cooperation with the Helmholtz-Zentrum in Munich, Germany.

Pre/Post-Processing and Test of Genome-Wide Associations
In our analysis, we used variants that were previously imputed into the 1000 Genomes Phase 3 reference panel. Variants with poor imputation quality (i.e., info < 0.8), a minor allele frequency (MAF) <0.05 and failing the Hardy-Weinberg-Equilibrium (p < 10 −4 ) were filtered out. We defined cases as patients with CAD who were diagnosed with severe depressive disorder in the interview (regardless of remission status) and controls as patients with CAD for whom a lifetime diagnosis of severe depressive disorder had been excluded in the interview. Associations were tested with SNPTEST (Department of Statistics, University of Oxford, Oxford, UK) using the additive model [32]. Subsequently, variants with an association level of p < 10 −5 were clustered into independent loci. Loci with at least two variants displaying p < 10 −5 were retained.

Functional Annotation and Software
Statistical analyses of demographic and clinical characteristics were performed using the Software Package for the Social Sciences (SPSS) (IBM, Armonk, NY, USA). We annotated variants that passed our pre-selected filtering criteria using the Genehopper database (DB) [33]. The Genehopper DB is a relational database that integrates genetic and biomedical data from many public resources. Variants were annotated based on linkage disequilibrium (LD) information [34], expression quantitative trait loci (eQTL) mappings [35][36][37], and topologically associated domain boundaries (TADs) [38].

Demographic and Clinical Characteristics
Of the initial 730 patients, 112 patients were known to be deceased, 75 patients refused to participate, 10 patients were not able to participate due to cognitive limitations, 17 patients were not able to participate because of an inability to communicate (e.g., insufficient language skills, hearing impairment, or aphasia), 50 patients were not able to be reached within four attempts, 118 patients were not able to be reached due to change in contact details, and 6 patients could not be interviewed for other reasons. The final sample consisted of 342 participants (78.3% males; mean age = 63.2 years; SD = 9.9 years) with a lifetime prevalence of severe depressive disorders of 23.0%. The prevalence of severe depressive disorders was higher after manifestation of CAD than before (17.4% vs. 8.8%; McNemar's χ2 (1) = 11.05; p = 0.001), although the corresponding time interval was shorter (median = 7 years vs. median = 54 years; Wilcoxon's Z = −15.36; two-tailed p < 0.001). The initial manifestation of 81.0% of severe depressive disorders lay after the initial manifestation of CAD-which was on average at 53.6 years (SD = 9.9 years)-and thus the majority of severe depressive disorders in our sample fit the definition of LOD.

In-Silico Characterization of Variant rs528732638
We next assessed the putative effects of variant rs528732638 and its highly correlated variants (R 2 > 0.8) on gene expression, according to LD information from 1000 Genomes Phase 3 individuals with a European background. Variant rs528732638 shares an LD block with 50 highly correlated variants (Table S2). This LD block showed eQTL effects on five genes in multiple tissues ( Table 2 and  Table S3) with the eQTL on AQP4 having the overall lowest p-value (p = 4.7 × 10 −10 in blood). We used information about TAD boundaries to group genes into cisand trans-located genes with respect to the LD block of rs528732638 (Table S4). Three of the five genes (AQP4, AQP4-AS1 and CHST9) are putative cis-regulated genes, i.e., these genes are located within a single spatial compartment in which increased physical interaction occurs, that is between regulatory DNA elements and gene promoters. Table 2. Variant rs528732638 and its high LD variants showed expression quantitative trait loci (eQTL) effects on multiple tissues.

Variant rs528732638 on Chromosome 18q11.2 Represents the First Vascular Depression Locus
Variant rs528732638 on chromosome 18q11.2 was associated with a 3.6-fold increase in the odds of lifetime depression in patients with CAD (95% CI = 2. 38-5.50). It is the first locus meeting the formal threshold for genome-wide significance [39] of 7.2 × 10 −8 for an association with depression in patients with coronary artery disease.
Variant rs528732638 is located between the genes AQP4-AS1 (Aquaporin 4 antisense RNA 1) and CHST9 (Carbohydrate sulfotransferase 9) and shares an LD block with 50 highly correlated variants. Prior to our study, the GWAS catalog for rs528732638 or one of its LD variants contained no entries [40]. In our study, however, this LD block showed eQTL effects on five genes, three of which (AQP4, AQP4-AS1 and CHST9) are putative cis-regulated genes.
Outside of the central nervous system, AQP4 is expressed in myocytes in both cardiac and skeletal muscles [66]. Regarding coronary artery disease, AQP4 is involved in cell swelling and cardiac dysfunction in murine myocardial infarction models [67,68]. AQP4 knock-out in mice reduces the infarct size [69], and tigacelor-mediated cardioprotection relies on AQP4 downregulation [70].
In summary, AQP4 mediates the formation of ischaemic edema in the brain and heart, enhancing cytotoxicity and thus the number and extent of lesions. Based on our finding that variant rs528732638 which belongs to a LD block with eQTL effects on AQP4 is associated with depression in CAD patients, we speculate that AQP4 may play a role in the etiopathology of vascular depression via promotion of ischemic cytotoxicity.
Further studies that aim to replicate our findings in independent, possibly non-European samples, explore expression of AQP4 in atherosclerosis-relevant tissues as a function of genotype at variant rs528732638, and explore the expression of AQP4 in depressed patients with and without atherosclerosis, and functional studies (e.g., animal models) are the next potential steps to elucidating the role of AQP4 in vascular depression.

Limitations of the Study
Nevertheless, limitations of this study must be considered. First, our findings are solely based on a case-control study, not including unbiased prospective population-based cohorts. Second, our sample size is too small to draw definite conclusions. Third, due to the high morbidity in the sample and the length of the follow-up period, our study suffered from a high drop-out rate. As the prognosis of patients with CAD and depression is poorer than that of patients with CAD that are not depressed [71][72][73], selection bias may have occurred, leading to an underestimation of the prevalence of depression and reducing the power of our study. Thus, we may have missed other depression loci in CAD patients. Fourth, we did not diagnose cerebrovascular disease directly using cerebral imaging. Fifth, our finding has not yet been replicated in an independent sample. Sixth, the LD block of the locus we identified also exerts eQTL effects on the CHST9 gene. Although no previous data have implicated CHST9 in the etiopathology of vascular depression, we cannot yet exclude this possibility.

Conclusions
We conducted the first GWAS in a psychiatrically characterized cohort of patients with a confirmed atherosclerotic disorder, namely coronary artery disease. We thus identified a potential locus for vascular depression on chromosome 18q11.2. The lead variant rs528732638 belongs to a LD block with eQTL effects on putative cis-regulated genes (AQP4, AQP4-AS1, and CHST9). As AQP4 is already known to increase the size and extent of ischemic lesions in the heart and brain, we speculate that AQP4 may play a role in the etiopathology of vascular depression. Replication steps including non-European cohorts are needed to confirm this finding and establish the functional role of the AQP4 locus in the etiopathology of depression in CAD patients.

Acknowledgments:
The authors would like to thank Marie Luise Bussmann and Sandra Köhne for assistance with data acquisition.

Conflicts of Interest:
The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.