Major depressive disorder (MDD) is a chronic psychiatric illness and remains one of the most important contributors to morbidity and mortality.Reference Ayuso-Mateos, Vazquez-Barquero, Dowrick, Lehtinen, Dalgard and Casey1 The development of novel interventions has been hampered by a deficient understanding of the underlying neurobiology. Despite convincing evidence for a genetic contribution to disease susceptibility (estimated heritability of 37%), there has been a dearth of substantive genetic findings of MDD, with the lack of success attributed to the phenotypic and aetiologic heterogeneity.Reference Flint and Kendler2, Reference Hyde, Nagle, Tian, Chen, Paciga and Wendland3 In our previous study, we performed whole-genome sequencing of 5303 Chinese women with recurrent MDD selected to reduce phenotypic heterogeneity and 5337 controls (CONVERGE sample), and revealed the SIRT1 gene, located on chromosome 10q21.3, as one of the first two genes successfully linked to MDD.4
SIRT1 is characterised as a class-III histone deacetylase, which can deacetylate numerous substrates of histones and non-histone proteins and thereby influence gene expression and cellular physiology.Reference Vaquero, Sternglanz and Reinberg5 Supportive of the genetic observation in the CONVERGE sample, results from several independent studies have suggested the involvement of SIRT1 in MDD. First, SIRT1 expression is markedly downregulated in the peripheral of people with MDD when compared with healthy individuals and those with remitted MDD cases.Reference Kishi, Yoshimura, Kitajima, Okochi, Okumura and Tsunoka6, Reference Luo and Zhang7 Second, inhibition of hippocampal SIRT1 function by either pharmacologic or genetic treatment led to an elevation in depression-like behaviours, whereas SIRT1 activation could block both the development of depression-related phenotypes and abnormal dendritic structures.Reference Abe-Higuchi, Uchida, Yamagata, Higuchi, Hobara and Hara8 Similarly, increased expression of SIRT1 in the nucleus accumbens was observed in depressed mice. Pharmacologic or genetic activation of SIRT1 in the nucleus accumbens increased both depression- and anxiety-like behaviours and, conversely, inhibition of SIRT1 reduced these behaviours.Reference Kim, Hesterman, Call, Magazu, Keeley and Armenta9 Third, it has been shown that resveratrol-triggered SIRT1 activation increased neurogenesis in the hippocampus of aged rats and prevented ageing-associated memory loss and mood dysfunction.Reference Kodali, Parihar, Hattiangady, Mishra, Shuai and Shetty10 These above lines of evidence partially revealed the mechanisms of SIRT1 in the pathogenesis of MDD, especially the involved brain regions. However, direct evidence as to whether genetic variants across the SIRT1 gene locus can affect brain structure is lacking. In this study, we performed imaging genetics to assess whether the SIRT1 variants affect brain structure using high-resolution magnetic resonance imaging (MRI) data of 92 healthy individuals from Eastern China.
Method
This study was approved by the Medical Research Ethics Committee of Wuxi Mental Health Center, China (no. WXMHCIRB2013LLKY001). Written informed consent was obtained from each person before participation in this study.
Participants
A total of 92 healthy individuals (41 males and 51 females, aged 40.3 ± 14.9 years) were enrolled from the Wuxi Mental Health Center, Nanjing Medical University, China. All controls were assessed using the Chinese version of the Modified Structured Clinical Interview for DSM-IV-TR Axis-I Disorders Non-patient Edition (SCID-I/NP). Individuals with a history of major psychiatric disorders or suicidal behaviour were excluded. Individuals who had a first-degree relative with a history of severe mental disorder or suicidal behaviour were also excluded. All participants were of Chinese Han origin, and none of belonged to extended family.
MRI
T1-weighted structural MRIs (sMRIs) were conducted on a 3.0-T scanner (Siemens MAGNETOM Trio with Tim, syngo MR B17 software, Erlangen, Germany) using a three-dimensional volumetric sequence (repetition time/echo time/inversion time 2530/3.44/900 ms, flip angle 7°, field of view 256 mm2, voxel size 1 × 1 × 1 mm3, pixel bandwidth 190 Hz, total scan time 6.6 min).
Voxel-wise morphometric MRI data processing
The T1-weighted sMRI data were preprocessed with the Statistical Parametric Mapping 12 (SPM12, www.fil.ion.ucl.ac.uk/spm/software) by using unified segmentation in which image registration, bias correction and tissue classification are performed using a single integrated algorithm.Reference Ashburner and Friston11 In this way, brains were segmented into grey matter, white matter and cerebrospinal fluid and nonlinearly transformed into the standard Montreal Neurological Institute space.Reference Friston, Ashburner, Frith, Poline, Heather and Frackowiak12 Unmodulated normalised parameters were used to segment the brain into probabilistic maps of grey matter, white matter and cerebrospinal fluid. The resulting grey matter images consisting of voxel-wise grey matter density were resliced to 3 × 3 × 3 mm3, resulting in 53 × 63 × 46 voxels, which was then smoothed with an 8-mm full width at half-maximum Gaussian kernel. A mask was then generated to include only the segmented grey matter voxels inside the brain, as described in our previous study.Reference Sui, Pearlson, Du, Yu, Jones and Chen13 Finally, voxels in the grey matter maps of each participant were collapsed into a one-dimensional vector and stacked, forming a population matrix (number of participants [N participant] × number of voxels [N voxel]).
Genotyping
Genomic DNA was extracted from peripheral blood cells according to the standard phenol–chloroform method. Genotyping of the SIRT1 gene locus was performed on the human Illumina PsychArray-24 (Illumina, USA), strictly according to the manufacturer's instructions. The single nucleotide polymorphism (SNP) data were subjected to a series of standard quality control procedures by using PLINK (Massachusetts General Hospital, Boston, USA; http://pngu.mgh.harvard.edu/puurcell/plink), which included checking for missing data, gender mismatch relatedness, Hard–Weinberg equilibrium and minor allele frequency as described elsewhere.Reference Anderson, Pettersson, Clarke, Cardon, Morris and Zondervan14, Reference Purcell, Neale, Todd-Brown, Thomas, Ferreira and Bender15
Statistical analysis
For each SNP, we divided the participants into three categories: AA (assigned as ‘0’), Aa (assigned as ‘1’) and aa (assigned as ‘2’). ‘A’ and ‘a’ indicated the major and minor allele, respectively. The additive dosage value was regressed against the voxel-wise grey matter density by using a multiple linear regression framework controlling for age, gender and diagnosis (when applicable). To correct for multiple comparisons, a Monte Carlo cluster simulation was performed to identify any cluster that shows significant correlation between SNPs and grey matter density.Reference Genovese, Lazar and Nichols16 The algorithm of cluster-wise correction for multiple comparisons by a Monte Carlo cluster simulation was implemented in the framework of AFNI software,Reference Cox17 and 1000 iterations were performed in the simulation procedure. Significance was thresholded at the uncorrected voxel-wise P-value of 0.001, followed by the family-wise error-corrected cluster-wise P-value of 0.05. To compare the grey matter density between any two groups, a post hoc test was performed by using the two-tailed t-test, with a statistical significance level set at P < 0.05.
Results
After genotyping and quality control, three SNPs – rs4746720, rs10823112 and rs3758391 – across the SIRT1 locus were finally analysed (Table 1). As shown in Fig. 1a, significant association between rs4746720 and grey matter density was observed in two brain regions: the orbital part of the right inferior frontal gyrus (cluster size 76 voxels) and the orbital part of the left inferior frontal gyrus (cluster size 104 voxels) (family-wise, error-corrected P < 0.05; voxel-wise P < 0.001). The sMRI images for all the slices can be found in the Supplementary materials available at https://doi.org/10.1192/bjp.2018.270. Post hoc comparisons between different genotypes of rs4746720 revealed that the carriers of the risk allele (G) predicted higher mean grey matter density than the carriers of the non-risk allele (Fig. 1b). No significant sMRI–SNP association was observed for either rs10823112 or rs3758391 after correcting for multiple comparisons (both family-wise, error-corrected P > 0.05).
Fig. 1 Significant association of the SIRT1 rs4746720 with cortical grey matter density in two clusters. (a) Spatial map of the two clusters where the SIRT1 rs4746720 genotypes had an effect on grey matter density. A multiple regression model and voxel-wise t-test was applied to test the correlation between rs4746720 and grey matter density. The colour bar specified the statistics of the t-test. Significance was thresholded at the uncorrected voxel-wise P-value of 0.001 and the family-wise, error-corrected cluster-wise P-value of 0.05. (b) Detailed summary of the two clusters. Coordinates referred to the Montreal Neurological Institute (MNI) space in mm. (c) The rs4746720 risk-allele (G) carriers exhibited higher mean grey matter density in the two clusters compared with non-risk allele carriers (two-tailed t-test). Frontal_Inf_Orb_L, orbital part of the left inferior frontal gyrus; Frontal_Inf_Orb_R, orbital part of the right inferior frontal gyrus; GM, grey matter.
Table 1 The SIRT1 single nucleotide polymorphisms from this study and their association with major depressive disorder from the CONVERGE samples
The first six columns give the chromosome (Chr.), genomic position (Pos.), SNP identifier (RSID), location in the gene locus (Location), reference allele (Ref.) on National Center for Biotechnology Information Build GRCh37 and alternative allele (Alt.) called in 10 640 CONVERGE samples (5303 cases, 5337 controls). The next three columns show the alternative allele frequency (Freq.) in three superpopulations according to the 1000 Genome Project.Reference Abecasis, Auton, Brooks, DePristo and Durbin18 The final four columns present the results of association testing with major depressive disorder in CONVERGE samples: odds ratio of association with major depressive disorder with respect to the alternative allele and standard error in the odds ratio were obtained from a logistic regression model, P-values of association were obtained from a linear-mixed model with a genetic relatedness matrix containing all samples and linkage disequilibrium of each SNP with the index SNP rs12415800 from the original manuscript in EAS populations. SNP, single nucleotide polymorphism; EAS, East Asian; EUR, European; AFR, African.
We next investigated whether these SNPs are associated with MDD in the CONVERGE sample. Supporting the imaging genetic results, genome-wide significant association with MDD was only observed for rs4746720 (P = 3.32 × 10−08, odds ratio 1.161), but not for either rs10823112 (P = 0.042) or rs3758391 (P = 0.001) (Table 1). Given that rs12415800, the top-risk SNP close to the SIRT1 locus from the CONVERGE sample, was not successfully genotyped in the current study,4 we next investigated whether rs4746720 was in linkage disequilibrium with rs12415800. Interestingly, strong linkage disequilibrium between rs4746720 and rs12415800 was observed in East Asian populations from the 1000 Genomes Project pilot study (r 2 = 0.839; Fig. 2), strongly suggesting that the sMRI–SNP association might not be generated by chance.
Fig. 2 Linkage disequilibrium plots between rs12415800 and adjacent single nucleotide polymorphisms (SNPs) across the SIRT1 gene locus in Asian populations and their association with major depressive disorder in the CONVERGE sample. Statistical results of association with major depressive disorder were extracted from the CONVERGE samples.4 A physical map of the region is given and depicts known genes within the region. The recombination rates expressed in centimorgans (cM) per megabase (Mb) (National Center for Biotechnology Information Build GRCh37; light blue lines) are shown on the right y-axis. Position in Mb is on the x-axis. Linkage disequilibrium of each SNP, with the top SNP rs12415800 displayed as a purple diamond, is indicated by its colour (data from the 1000 Genomes Project, pilot 1, Asian panel). The plots were drawn using LocusZoom.Reference Pruim, Welch, Sanna, Teslovich, Chines and Gliedt19 Chr10, chromosome 10.
Discussion
Neuroimaging has been widely applied to identify the key brain regions implicated in the pathophysiology of MDD, and imaging genetic methods have also been applied to investigate the functions of specific MDD-risk SNPs in previous studies.Reference Chang, Hoshina, Zhang, Ma, Cao and Wang20, Reference Li, Luo, Rietschel, Lewis, Mattheisen and Muller-Myhsok21 In contrast to white matter, which shows linear increase in volume, previous studies have suggested a pattern of rapid neurogenesis and related volume increase of grey matter from early childhood to the adolescent period, followed by a process of selective elimination and myelination which leads to volume loss and thinning.Reference Giedd and Rapoport22 Associations of cortical and subcortical volume with people with MDD have been extensively reported. For instance, based on data from 8927 participants using an individual participant data-based meta-analysis approach, the MDD working group within the Enhancing Neuro Imaging Genetics through Meta-Analysis (ENIGMA) reported subcortical volume differences, which were greatest in the hippocampus, between people with MDD and healthy controls.Reference Schmaal, Veltman, van Erp, Samann, Frodl and Jahanshad23 Again, the same group collected brain MRI data from >10 000 people and found significant differences in cortical thickness in 13 out of 68 regions examined, encompassing the medial prefrontal cortex, rostral anterior and posterior cingulated cortex, insula and fusiform gyrus.Reference Schmaal, Hibar, Samann, Hall, Baune and Jahanshad24 More importantly, the development of cortical and subcortical grey matter regions is largely influenced by genetic factors, suggesting a neurobiological basis underlying the risk variants identified in MDD genetic studies.Reference Hibar, Stein, Renteria, Arias-Vasquez, Desrivieres and Jahanshad25, Reference Lippard, Jensen, Wang, Johnston, Spencer and Pittman26
In the current study, we found significant association between rs4746720, a genome-wide significant SNP located within the SIRT1 locus, and grey matter density in the inferior frontal cortex, with the risk allele-G carriers predicting higher mean grey matter than the non-risk allele carriers. In support of the result observed in our study, accumulating evidence from sMRI studies have suggested the involvement of neuroanatomical changes in the frontal lobe in mood disorders, including both MDD and bipolar disorder.Reference Bora, Fornito, Pantelis and Yucel27, Reference Konarski, McIntyre, Kennedy, Rafi-Tari, Soczynska and Ketter28 For example, abnormalities of both grey matter and cortical thickness have been reported in the inferior frontal gyrus in patients with MDD.Reference Bora, Fornito, Pantelis and Yucel27, Reference Truong, Minuzzi, Soares, Frey, Evans and MacQueen29 Furthermore, through systematic meta-analysis, Kempton et al Reference Kempton, Salvador, Munafo, Geddes, Simmons and Frangou30 found that, compared with the structure of a healthy brain, MDD was associated with smaller volumes of the frontal lobe and several other brain regions. More importantly, people with MDD showed increased rates of subcortical grey matter hyperintensities compared with healthy controls.Reference Kempton, Salvador, Munafo, Geddes, Simmons and Frangou30
We next moved from the imaging genetic finding to the putative mechanisms for the inferior frontal gyrus. It has been reported that the inferior frontal gyrus is involved in a number of cognitive processes of potential relevance to MDD, including response inhibition, set switching, socioemotional learning and sustained attention.Reference Liakakis, Nickel and Seitz31 Moreover, the patients with MDD showed dysfunction of language processing and cognitive performance, possibly due to impairment of the network of the ‘theory of mind’, which includes inferior frontal gyrus.Reference Mason and Just32 These studies strongly suggest the possibility that loss of the functional integrity of the inferior frontal gyrus may underlie trait dysfunction in MDD.
Notwithstanding its significant strengths, our study has some inherent shortcomings due to its design. First, we employed healthy participants but not people with MDD. Although the use of healthy controls for imaging genetics at the level of brain function avoids potential confounders related to chronic illness and medical treatment, it is unclear whether grey matter density in the inferior frontal cortex was correlated with the symptoms of depression. Second, the sample size (N = 92) recruited in our study was relatively small when compared with other neuroimaging studies,Reference Schmaal, Veltman, van Erp, Samann, Frodl and Jahanshad23, Reference Schmaal, Hibar, Samann, Hall, Baune and Jahanshad24 which may obscure other true correlations. Third, we performed the whole brain-wide correction for each association test, however, whole brain-wide correction might preclude some positive associations of moderate association. We acknowledge that brain regions other than the inferior frontal cortex might also be involved in the pathogenesis of depression. Third, it is worth noting that only Han Chinese women with severe MDD were recruited in our CONVERGE MDD genome-wide association study.4 On one hand, this could definitely have minimised the phenotypic and aetiologic heterogeneity, allowing for the discovery of the risk variants; but on the other hand, this could have precluded some promising risk variants.
In sum, our results provide evidence for a potential role of the SIRT1 gene in the brain, implying a possible pathophysiological mechanism underlying susceptibility to major depression. However, additional investigation of the underlying molecular pathways within affected brain regions is warranted.
Supplementary material
Supplementary material is available online at https://doi.org/10.1192/bjp.2018.270.
Funding
This work was supported by the Fundamental Research Funds for the Central Universities (no. 2682018CX38 to S.R.), the National Natural Science Foundation of China (no. 81471364 to F.Z., no. 81471367 and 61773380 to J.S.), the Strategic Priority Research Program of the Chinese Academy of Sciences (no. XDBS01040100 to J.S.) and Brain Science and Brain-inspired Technology Plan of Beijing City (no. Z181100001518005 to J.S.).
Acknowledgements
We thank all the voluntary DNA donors in this study. We thank staff at the blood laboratory who helped collect the blood samples.
eLetters
No eLetters have been published for this article.