OGDHL Variant rs2293239: A Potential Genetic Driver of Chinese Familial Depressive Disorder

Depressive disorders are a severe psychiatric and social problem that affect more than 4% of the global population. Depressive disorders have explicit hereditary characteristics; however, the precise driving genetic force behind these disorders has not yet been clearly illustrated. In the present study, we recruited a three-generation Chinese pedigree in which 5 of 17 members had long-term depression. We conducted whole-exome sequencing to identify the genetic mutation profiles of the family, and a list of susceptible genetic variations that were highly associated with depression onset was revealed via multiple omics analysis. In particular, a non-synonymous single nucleotide variation in the oxoglutarate dehydrogenase-like (OGDHL) gene, rs2293239 (p.Asn725Ser), was identified as one of the major driving genetic forces for depression onset in the family. This variant causes an important conformational change in the transketolase domain of OGDHL, thus reducing its binding affinity with the cofactor thiamine pyrophosphate and eventually resulting in the abnormal accumulation of glutamate in the brain. Brain imaging analysis further linked the rs2293239 variant with an enlarged amygdala and cerebellum in depressive family members. In summary, the present study enhances the current genetic understanding of depressive disorders. It also provides new options for prioritizing better clinical therapeutic regimens, as well as identifying a new protein target for the design of highly specific drugs to treat depressive disorders.


INTRODUCTION
Depressive disorders, or depression, refers to common but severe mental disorders that affect more than 264 million people worldwide (1). They are also a leading cause of disability, and place heavy economic burdens on both patients' families and society (2). Unfortunately, clinical therapy for depressive disorders, and in particular major depressive disorder (MDD), usually results in poor outcomes because of the heterogeneous pathophysiology of these disorders. Any one or a combination of biological, physical, genetic, and social factors may account for depression onset.
Although great efforts have been made over many years, the driving force underlying depressive disorders remains unclear in most cases, which has little benefit for improving therapeutic regimens (3,4).
Depressive disorders are heritable. Early twin studies estimated the heritability of depression to be about 36% (5,6). For MDD, heritability was determined to be approximately 32% in a study that measured genomic similarity among unrelated individuals (7). The inheritance of depression from generation to generation makes it feasible to reveal the potential driving genetic force of this disorder. For example, Hu conducted comparative genotyping of a case-control cohort and reported that a single nucleotide polymorphism, rs25531 in SLC6A4, likely impacts depression onset by interfering with the serotonin pathway (8). In addition, Zubenko et al. performed a linkage analysis on 81 families and identified 19 loci that were suspected to relate to depressive disorders (9). Camp et al. reported three loci related to depression or anxiety based on 87 Utah pedigrees (10). Furthermore, Kendler and Flint summarized a number of genes (5HTTP/SLC6A4, APOE, DRD4, GNB3, HTR1A, MTHFR, and SLC6A3) linked to susceptibility to heterogeneous depression by interpreting a large amount of candidate gene literature in 2014 (11). However, despite convincing evidence for their genetic contribution to disease susceptibility, only a few of these genes or loci have substantial molecular evidence to support the diagnoses of the members of this family and clinical therapy (11,12). In recent years, genome-wide association studies (GWAS) have been applied to the search for genetic associations in various psychiatric diseases. Several biomarker gene variants have been proposed for depression in different ethnic populations, including BICC1 rs9416742 in a UK population (13), SIRT1 rs12415800 in a Chinese Han population (14), and rs12462886 of a non-coding region in a US population (15). However, the results of GWAS for MDD have been questioned because of a Abbreviations: OGDHL, Oxoglutarate Dehydrogenase L; SLC6A4, Sodiumand-chloride-dependent solute carrier family 6 gene number 4; 5-HTTP, 5 lack of generalizability and interpretability (11). It has also been proposed that previously identified candidate genes are likely to be false positives (16). In particular, the variants identified using GWAS approaches usually have low penetrance, and few have had their associations with depression onset confirmed molecularly. To date, strong depression-associated variants have not yet been identified; a large gap remains for the translation of our current knowledge of depressive disorders to efficient clinical therapy.
To narrow this gap, in the present study, we conducted a pedigree analysis in a three-generation Chinese family with depression. We first portrayed the genetic mutation profiles for every family member, from which we identified rare variants with high penetrance that were associated with familial depression. We then performed a protein structure-activity relationship analysis and multimodal brain image analyses to interpret the potential roles of the selected variants in depression onset.

This study was approved by the Clinical Research Ethics
Committee of the Xiamen Xianyue Hospital, Fujian, China. All experimental protocols were performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from all participants after a complete description of the study.

Family Pedigree
The pedigree in this study was a three-generation family of 17 members, five of whom were diagnosed with different levels of depressive disorders by multiple doctors from the Xiamen Xianyue Hospital. Diagnoses were made strictly according to the guidelines of the International Classification of Diseases, Tenth Revision (ICD-10), and pathophysiology was carefully assessed by reviewing a thorough history, examination, and workup of each patient. The pedigree at the time of this study is illustrated in Figure 1 and demographic information is shown in Table 1.
For each family member, 1 mL of peripheral blood was collected in the hospital. Subsequently, plasma separation was conducted by centrifugation at 800 ×g for 10 min and 2,500 ×g for 10 min consecutively. The separated leukocytes were stored at −80 • C before the DNA was extracted.

Mutation Profiling by Whole-Exome Sequencing
The genomic DNA of leukocyte samples was extracted using the EZ-10 Spin Column Blood Genomic DNA Purification Kit (Sangon Biotech Co, Ltd., Shanghai, China). The DNA concentration was measured using the Qubit R DNA Assay Kit and Qubit R 2.0 Fluorometer (Life Technologies, CA, USA). For each sample, 0.6 µg of DNA was used as the input material for the DNA sample preparation. The whole exome was captured using the Agilent SureSelect Human All Exon Kit (Agilent Technologies, CA, USA) and the library for sequencing was prepared according to the manufacturer's instructions. WES was performed by Novogene (Beijing, China) using the Illumina Hiseq platform in a 150-base pair (bp) paired-end mode.

Differential Mutation Density
The DMD was determined by summarizing the mutation difference between the depressive and healthy members of the family for every 1,000,000 bp (1 Mbp) segment of chromosome: other modules (-normalization and -matrix) were applied to generate a z-score of read depth and normalization, and the hidden Markov model (-discover) was applied to identify CNVs. The GATK method used the Determine Germline Contig Ploidy module to determine autosomal and allosomal contig ploidy, and subsequently used the Germline CNV Caller module to call CNVs. In the current study, only consensus CNVs that were called by both tools were retained for differential analysis. The CNV differential analysis was conducted between the depressive group and the healthy group using a self-written shell script. The differential CNVs, if available, were further functionally annotated to evaluate their connection with depression using Classify CNV (24) software (v.1.1.0) and referring to the human genome (GRCh38).

Identification of Depression-Associated Variants
We performed a series of bioinformatic analyses one by one to identify depression-associated variants (Figure 2), as follows.
(1) A comparison of mutation profiles was performed to extract the differential, non-synonymous, and exonic variants. At the time of this investigation, the third generation of the family were mostly teenagers, when depression is likely under progression but symptoms have not yet appeared. Thus, members of the third generation were excluded from the case-control comparative analysis. The third-generation member G3-F-02, who was diagnosed with depression just before the initiation of this study, was the only exception; she was involved in the variant detection. (2) Genetic segregation analysis was conducted to confirm whether the variants satisfied a Mendelian inheritance model using the segreg program in the Statistical Analysis for Genetic Epidemiology package (25) (v.6.4). Sporadic variants that did not fit the Mendelian inheritance model were excluded because they did not contribute much to familial trait inheritance statistically according to the program. (3) Common variants, with an allele frequency of more than 0.05 in the Exome Aggregation Consortium (ExAC) (26) and the 1000 genome project (27,28), were excluded. Odds ratios (ORs) were calculated for every retained rare variant following conventional practices, based on the allele frequency difference between the psychosis FIGURE 2 | Schema of the discovery of driver variants for the familial depressive disorder.
(ExAC_Psych) and non-psychosis (ExAC_nonPsych) subsets of ExAC release 1.0 (GRCh38  (38), and PhyloP (39). The variants that were predicted deleterious by at least one algorithm were considered to have high penetrance for depression. Variants with more hits were considered to be more deleterious. The most deleterious variants were considered to be driving variants for depression and were adopted for later functional analyses.

Protein Structure-Activity Relationship Analysis
Because the human OGDHL protein structure was not yet available in the Protein Data Bank (PDB) database, we used homology modeling to simulate the structure for the structure-activity relationship analysis. The simulation was conducted using the commercial software Molecular Operating Environment (MOE), adopting the multifunctional 2oxoglutarate metabolism enzyme (PDBID: 2yic) as the template coupled with the cofactor thiamine pyrophosphate (TPP), Ca 2+ , and Mg 2+ . The simulated protein structure was stabilized and optimized in a water environment (310 K temperature) using GROMACS software (40) (v.2018.4) for 20 ns molecular dynamics with both force fields: amber general force field (41) and amber ff99SB-ILDN (42). Furthermore, we simulated the mutant protein (OGDHL N725S) by changing the amino acid at position 725 (NP_001334748.1) from asparagine (N) to serine (S), followed by 50 ns molecular dynamics under the same conditions as for the wild-type protein simulation. The structureactivity relationship analysis was performed by evaluating the conformational change and binding affinity between the OGDHL wild type and N725S mutant using MOE software.

Image Data Acquisition and Processing
Magnetic resonance imaging (MRI) of the brain, which included structural MRI (sMRI), functional MRI (fMRI), and diffusion imaging (dMRI), were scanned for every family member when possible. For various reasons, all three imaging modalities were not used for all family members. The image data acquisition statuses are shown in Figure 1A. All images were taken using a MAGNETOM Skyra 3T MRI (Siemens AG, Erlangen, Germany The raw brain images were converted into NIfTI format using dcm2niix (43) (v.25) and were then output for external analyses. For the sMRI images, we used the FIRST module (44) (run_first_all) in FMRIB Software Library (FSL) software (45) (v.6.0.1) to extract the subcortical structure segmentation. Overall, 15 subcortical segments (left-thalamus, left-caudate, left-putamen, left-pallidum, left-hippocampus, left-amygdala, left-accumbens-area, brain-stem, right-thalamus-proper, rightcaudate, right-putamen, right-pallidum, right-hippocampus, right-amygdala, and right-accumbens-area) and a cerebellum segment were extracted in this study. The volumes for the subcortical structures and cerebellum were determined using the "fslmath" program of FSL. The volume comparison between the healthy control and depression groups was performed using a randomization test in a self-written R script; its robustness was evaluated using the randomization test. For the fMRI images, raw images were processed for motion correction, field unwarping, normalization, and bias field correction using fMRIPrep (46) software. In the normalized images, the blood oxygen levels of 39 distinct brain regions (nodes) were determined using the Python package nilearn (47), referring to the multi-subject dictionary learning (MSDL) atlas (48). For each family member, the covariance values of the blood oxygen levels between the nodes were calculated to form a 39 × 39 connection matrix. The connectome map was built upon the connection matrix by setting an edge_threshold = 0.99, in which the edge stood for the connectivity [calculated by the tangent (49)] between the nodes. The connection map difference between the healthy control and depression groups was determined using the FSL module randomize (50). The dMRI images were preprocessed using the tract-based spatial statistics (TBSS) module (51) of FSL software to create the fractional anisotropy (FA) image, register the FA image, skeletonize the mean FA image, and project the FA information onto the mean FA skeleton image. The difference in the fiber tract skeleton between the healthy control and depressive groups was then evaluated using the FSL module randomize (50).

Identification of Potential Driver Variants in Familial Depression
WES of the peripheral blood of all 17 family members yielded a mean sequencing depth of 258.94× (ranging from 234.26× to 294.52×) after quality control (Supplementary Table S1). The sequencing coverage was deep enough for robust variant calling. Overall, 141,022 single nucleotide variants (SNVs) and 19,148 insertion/deletion variants (InDel) were obtained. The distribution of SNVs and InDels in each family member is illustrated in Figure 1A. The family members had almost the same distributions of the different variant types, indicating that the variant calling from WES was well processed.
Before searching for depression-associated variants, we first examined chromosomal integrity by measuring the CNVs caused by abnormal chromosomal events, such as duplication, deletion, rearrangement, or recombination. There were no significant chromosomal differences between the depressive and healthy family members. In particular, no significant CNV differences were observed in the chromosomal region 22q11.2; this region has been linked with schizophrenia and other neuropsychiatric/behavioral disorders. Furthermore, we analyzed the DMD landscapes per 1 Mbp segment of chromosomes ( Figure 1B). Three hotspots of DMD (with DMD > 20) were detected, at 12q12 (DMD = 32.9833), 14q32.33 (DMD = 24.5667), and 15q26.3 (DMD = 22.3000). These three DMD hotspots may be chromosomal fragments that are mutual to the depressive members, and are inherited from generation to generation ( Figure 1B). These chromosomal fragments covered 19 genes and 976 variants: four genes in 12q12 (SLC2A13, LRRK2, MUC19, and CNTN1), five genes in 14q32.33 (TDRD9, RASPG, KIF26A, C14orf180, and TMEM179), and 10 genes in 15q26.3 (LRRK, CHSY1, SELENOS, SNRPA1, PCSK6, TM2D3, TARS3, OR4F6, OR4F15, and OR4F4). Although a literature search of these genes and variants revealed some clues of their association with depressive or other affective disorders, a series of bioinformatic analyses (including OR association analyses and deleteriousness analyses) on these variants did not confirm their significant association with or penetrance for depression in our pedigree. Hence, we no longer considered the variants at these depression-specific hotspots as the candidate driving force of familial depressive disorders.
To identify potential depression-associated variants, we conducted a series of bioinformatic analyses. The case-control analysis extracted 159 differential, non-synonymous, and exonic variants between the depressive and other family members, including 144 SNVs and 15 InDels. According to the segregation analysis based on the whole mutation profiles, the familial depression likely fit either an autosomal dominant or a recessive Mendelian inheritance gene model ( Table 2). Hence, variants that did not fit Mendelian inheritance were excluded; 93 variants were consequently retained, including 84 SNVs and nine InDels. Furthermore, because the estimated incidence rate of depression worldwide is about 2-6% (52), driver genetic variants are more likely to be rare variants than common variants. Accordingly, common variants (with an allele frequency > 0.05 in the population) were also removed from the candidate list by referring to ExAC and the 1000 Genome project. Moreover, a deleteriousness analysis was conducted to evaluate the penetrance of the variants on depression. Only four SNV variants were considered to have high penetrance for depression: OGDHL rs2293239, TBX1 rs41298838, IL16 rs201457933, and FBXO15 rs79499419 ( Table 3).
Of these four candidate variants, we speculated that OGDHL rs2293239 was most likely the driving genetic force behind depression in this family. This speculation was based on several pieces of evidence: (1) OGDHL rs2293239 was predicted as highly deleterious by 9/13 algorithms; (2) this variant had a slightly higher frequency in the CONVERGE Chinese depressive women cohort (14) than in the non-psychotic controls (OR = 1.2602); (3) a literature search revealed that OGDHL is related to psychological diseases such as Alzheimer's disease (53), childhood disintegrative disorder (54), and cerebral atrophy (55); and (4) OGDHL is highly expressed in all regions of the human brain according to the Tissue Atlas database (56,57). Together, these findings suggest that OGDHL rs2293239 is connected with depression. The genetic status of OGDHL rs2293239 in the present family is illustrated in Figure 3A, and the genotypes of this allele were further confirmed by additional Sanger sequencing (Figure 3B). Of the 17 family members, all depressive members carried this variant in a heterozygous genotype. However, we must note that four healthy thirdgenerations also carried this heterozygous mutation but had not yet exhibited depressive or other psychotic disorders (at the time of writing).

Mutations in OGDHL May Cause Impaired Glutamate Metabolism
OGDHL is known compete with oxoglutarate dehydrogenase (OGDH) in forming the complex E1 subunit, which is essential for catalyzing the conversion of 2-oxoglutarate (α-ketoglutarate) to succinyl-coenzyme A and CO 2 in the citric acid cycle. According to Sen et al. (58) and Bunik et al. (53,59), OGDH and OGDHL are isozymes that are located in different organs. OGDHL is mainly located at the mitochondrial matrix in the brain, and uses TPP as a cofactor (Figure 3C).
OGDHL has two functional domains: the E1 dehydrogenase component (PfamID: PF00676) and the transketolase pyrimidine binding domain (PfamID: PF02779) ( Figure 3B). The OGDHL rs2293239 variant that was identified in the family was located in the transketolase pyrimidine binding domain, and resulted in a non-synonymous translation from arginine to serine at position 725 (N725S). Because there are currently no OGDHL structures available in the PDB database, we constructed one using homology modeling. The OGDHL was found to work as a dimer, and cofactor TPP binding occurred at the site between two monomers. We simulated the homodimer of OGDHL coupled with TPP as well as with Ca 2+ and Mg 2+ (Figure 3D). The binding pattern was analyzed and the critical residues within are illustrated in Figure 3D.
To evaluate the impact of the rs2293239 mutation on protein activity, we also simulated the structure of the OGDHL N725S variant. Compared with the wild-type protein, the N725S mutation induced a 5.39 Å conformational change (Figure 3D), which caused the binding affinity of the TPP-OGDHL dimer to decrease from −12.94 kcal/mol (wild-type) to −8.75 kcal/mol (N725S variant). It was speculated that this change in OGDHL activity may eventually impair glutamate metabolism and raise glutamate concentrations in the brain. Therefore, we measured glutamate levels in the peripheral blood (because of the difficulty in obtaining cerebrospinal fluid) of the family members. The depressive family members had higher blood glutamate levels than the healthy family members (analysis of variance). This

Endophenotype of the OGDHL Variant
Brain imaging can provide helpful information for identifying endophenotypes in the central nervous system that are caused by genetic mutations. Previously, Yoon et al. reported that the homozygous mutation S778L in OGDHL, which is located at a domain near rs2293239, caused severe neurodegeneration in a 13-year-old patient (55). In the current study, we performed brain imaging analyses for the family members in as much detail as possible (Figure 1). The images were evaluated in three separate analyses. First, we compared the volumes of 15 distinct subcortical structures between five depressive family members and eight healthy family members. There was a significant (ttest, p = 0.03908) volume increase in the left amygdala in the depressive family members (Figure 4A). This result was consolidated by the permutation test, which randomized the depressive and healthy members 10,000 times (p < 0.05) for the differential volume analysis. This finding agreed with the results of previous studies, that MDD patients have comparatively larger amygdalae (60)(61)(62). However, there were no significant volume differences in the left amygdala between nine rs2293239 carriers and four family members with wild-type OGDHL. Interestingly, a slight but non-significant volume increase (t-test, p = 0.07) in the left cerebellum was also noted in the nine rs2293239 carriers (Figure 4B). In both comparative analyses, no changes in the shape of subcortical or cerebellar structures were observed.
Second, we examined the integrity of neural tract skeletons by analyzing dMRI images using a TBSS method. MRI data did not reveal any significant differences among the five patients diagnosed with depression, including structural, functional, and DTI alterations. No significant skeleton defects or structural differences were identified between either the rs2293239 carriers or the depressive family members and the controls (Figure 4C). Finally, we attempted to evaluate changes in brain activity. However, in our fMRI analysis of a time-series of blood oxygen level changes, we did not detect any significant differences in activity in the connections between 39 distinct brain regions ( Figure 4D). Therefore, it was considered that the OGDHL rs2293239 variant was unlikely to induce depression by interfering with cerebral cortex activity in this family.

OGDHL Variant May Be One of the Major Genetic Factors of Depression Onset
Depressive disorders are heterogeneous in their pathophysiology. To date, identifying the major pathological factors for depression onset in each individual has been something of a "mission impossible, " and the development of precise therapeutic regimens for individuals has also been challenging. Genetic mutations are estimated to account for approximately one-third of all depressive disorders (5,6,63,64). Because genetic factors are inherited from generation to generation, they are the most likely common pathological factors to be identified for depressive disorders. Many studies of twin, pedigree, and case-control cohorts have been performed worldwide, and have identified a variety of depression susceptibility genes, such as SLC6A4, APOE, DRD4, GNB3, HTR1A, MTHFR, and SLC6A3 (11). Unfortunately, however, many of these susceptibility genes have not been be repeated in larger populations or in other cohorts (16). GWAS studies have complemented the population gap, but the common somatic variants identified by these studies mostly have low penetrance for depressive disorders. The poor interpretation and repetition of GWAS results have also meant that such findings have been questioned (11,16). Notably, using low-coverage whole-genome sequencing, a 2015 casecontrol study of 5,303 Chinese women with recurrent MDD and 5,337 controls identified two risk loci, at noncoding regions in SIRT1 and LHPP (14). A similar result was obtained in a large Japanese cohort study, which identified that the SIRT1 rs10997875 variant was associated with MDD (65). Furthermore, a mouse study suggested that hippocampal SIRT1 signaling can mediate depression-like behaviors (66), and a clinical trial also observed significantly reduced peripheral blood SIRT1 mRNA in depressive patients compared with healthy controls (67). Together, these findings indicate that SIRT1 is highly susceptible to depression and may be a good therapy target. However, opposing findings have also been reported; in one study, mice with global SIRT1 overexpression had elevated anxiety and increased susceptibility to depression (68). In the current study, we also evaluated SIRT1 variants and their penetrance in the family. Of three distinct susceptible variants (an exonic variant, rs2273773, and two novel intronic variants) that were identified in and near SIRT1, none showed a significant association with depression. In fact, none of the common variants identified in this family were associated with depression.
Instead of seeking common somatic variants, we discovered rare germline variants with high penetrance for depression in the present study. Via a series of bioinformatic analyses, we mined four potential driver variants: OGDHL rs2293239, TBX1 rs41298838, IL16 rs201457933, and FBXO15 rs79499419 ( Table 3). These variants satisfied multiple criteria: a strong association with depression, low frequency in the population (less than or nearly equal to the depression incidence rate in Chinese), high deleteriousness, and fit a Mendelian inheritance model.
T-box transcription factor (TBX1) is a probable transcriptional regulator that is involved in development. TBX1 is responsible for most of the physical malformations present in 22q11.2 deletion syndrome, or DiGeorge syndrome, which is a congenital disease that has been implicated in various behavioral abnormalities including schizophrenia and other neuropsychiatric/behavioral disorders (69)(70)(71). A recent study also reported that TBX1 in the mouse hippocampus might be linked to monosaccharides-D-ribose-induced depressive-like behavior (72). In the present study, the rare variant rs41298838 was predicted to be highly deleterious to TBX1 function, which made it a potential driving variant for familial depression. However, a close review of this variant in the general population revealed that it has a comparatively high allele frequency (more than 10%) in non-psychotic Asians, according to the ExAC database. In addition, WES analysis and karyotyping did not identify 22q11.2 deletions or other chromosomal  arrangements in the family members. These data challenge the idea that TBX1 rs41298838 might be a good candidate for familial depression. Interleukin (IL)-16 is the ligand of cluster of differentiation (CD)4, and can stimulate a migratory response in CD4 + lymphocytes, monocytes, and eosinophils. Several studies have used IL-6, as well as other cytokines, as indicators for monitoring the immune response in the treatment and pathogenesis of affective disorders (73,74). However, no substantial evidence indicates that IL-6 is directly involved in the development of depressive or other affective disorders. F-box only protein 15 (FBXO15) is the substrate-recognition component of the E3 ubiquitin ligase complex, which is involved in ubiquitin-proteasome-mediated protein degradation. To date, knowledge of the function of FBXO15 remains very limited, and its association with affective disorders or nerve diseases has not yet been well studied. Nonetheless, considering the uncertain relevance of IL-16 and FBXO15 for depressive disorders, we did not prioritize them in the functional investigation. Compared with TBX1, IL-16, and FBXO15, we found that OGDHL was strongly associated with depressive disorders in a range of different aspects. First, a series of bioinformatic analyses indicated that the OGDHL rs2293239 variant had a strong association with depression onset in the present family. The rs2293239 variant is a very rare mutation in global population (about 0.1%); it is comparatively more frequent in Asian populations (about 2%), and its allele frequency is nearly equal to the incidence rate of depression in Asian populations (75). A previous case-control cohort study (14) reported that the rs2293239 variant occurs more frequently in Chinese depressive women than in nonpsychotic controls (OR = 1.2602). These data suggest that rs2293239 may not be a common driving genetic force for all cases of depressive disorders. However, the rs2293239 mutation in the transketolase pyrimidine binding domain is likely very deleterious, as predicted by nine algorithms in the present study. Via a structure-activity relationship analysis, we revealed that the mutation may cause an important conformational change (about 5.39 Å) in the protein, which would decrease the binding affinity of OGDHL dimer with its cofactor TPP. OGDHL, which competes with the dominant player OGDH, is the rate-limiting component of the oxoglutarate dehydrogenase complex in glutamate metabolism. Inhibition of OGDH(L) results in a two-to three-fold increase in glutamate in neuronal cells (76). Glutamate is a major neurotransmitter in more than 80% of neurons (77), and abnormal glutamate production in neurons and glial cells can cause severe depressive symptoms (78). Thus, we measured serum glutamate levels in the depressive family members. An additional literature search also supported the relationship between OGDHL and psychological diseases. For example, an early study proposed that OGDHL might be linked to neurotransmitter synthesis and Alzheimer's disease (53). Furthermore, aberrant OGDHL copy number may be associated with childhood disintegrative disorder (54). In particular, the homozygous mutation rs773888308 (S778L) of OGDHL, which is a mutation at the transketolase pyrimidine binding domain near rs2293239 (N725S), was found to cause a severely hypoplastic corpus callosum and abnormal cerebellum in a 13-year-old patient (55). This previous work strongly supports our speculation that the rs2293239 mutation likely impairs OGDHL catalytic activity, thus resulting in increased intracellular glutamate and eventually inducing depression.

OGDHL Variant May Cause Brain Structural Changes
In 2017, Yoon et al. attributed the homozygous OGDHL mutation rs773888308 (S778L) to severe neurodegeneration and encephalatrophy in a girl (55). In the present study, we also measured brain structural changes using MRI techniques, and revealed that depressive family members had a significantly enlarged left amygdala compared with healthy family members. This finding agreed with the results of previous cohort studies, that MDD patients have a larger amygdala and smaller anterior cingulate compared with normal controls, with medium to large effect sizes (60)(61)(62). Notably, the endophenotype change in the amygdala was not observed in all carriers of OGDHL rs2293239 in the family; three third-generation young variant carriers did not exhibit changes in either the volume or the shape of the amygdala. MDD patients have been reported to have comparatively enlarged amygdalae (79), however this phenomenon may be related to long-term antidepressant use given that the enlargement is less reliably observed in nonmedicated patients with MDD (80). It has been suggested that anti-depressive pharmacotherapy associated induction of brainderived neurotrophic factor and other neuroprotective factors may have a volume-augmenting effect on the amygdala (81). However, these three variant carriers had a comparatively, although non-significantly (p = 0.07), enlarged left cerebellum compared with the non-carriers. In addition, no neural tract defects were identified in the depressive family members via dMRI examination. Moreover, a further case-control comparative analysis of the connectome of different cerebral cortex nodes (regions) did not identify any significant changes in cerebral activity, as indicated by a time-series of blood oxygen levels. It is therefore reasonable to suspect that the changes in cerebral substructures caused by OGDHL rs2293239 may account for the depressive disorders of this family. However, because this genetic alteration was heterozygotic, it may be that OGDHL protein function is only partially defective, and that brain structural changes progress so slowly that depressive symptoms are not apparent in teenagers. It will be necessary to perform animal experiments in the future to validate this hypothesis. Taking all of our results and the evidence from previous literature together, we propose that OGDHL rs2293239 is likely one of the major genetic driving forces of depression onset in the present family, via the following mechanism ( Figure 5). The rs2293239 variant decreases the binding affinity of the OGDHL dimer to its crucial catalytic cofactor TPP, thus weakening its transketolase activity. This alters the glutamate metabolism balance in cells, resulting in increased glutamate concentrations in the brain. As a result of longterm glutamate imbalance, brain structures may change, such as enlargement of the amygdala or cerebellum; furthermore, as a result of glutamate unbalance and brain structural changes, depressive symptoms appear. This hypothesis is partially supported by a recent study that demonstrated that glutamate concentrations in the amygdala may be related to depressive moods (82).
Nevertheless, there are several uncertainties in the aforementioned hypothesis. The impact of the rs2293239 variant on glutamate metabolism has not yet been experimentally validated, although both computational simulations in the present study and previous work (76) suggest that an inhibition or defect of OGDHL can cause aberrant glutamate metabolism. In the family in the present study, the glutamate concentrations in the amygdala, cerebellum, and other brain structures are unknown because of the difficulty in obtaining cerebrospinal fluid. However, the relatively high peripheral serum glutamate concentration in this family partially supports our hypothesis because rs2293239 is a germline mutation, and should therefore take effect all over the body. Furthermore, causality has not yet been confirmed between brain structural changes and this variant. Although a study by Yoon et al. (55) suggested a strong association between an OGDHL homozygotic mutation and neurodegeneration, the penetrance of heterozygotic mutations for brain structural changes requires further verification by animal experiments. Notably, the rs2293239 variant was the most likely driving mutation mined from hundreds of thousands of germlines SNVs, Indels, and CNVs via a series of conditions. It is not the only genetic factor in familial depression, but it might be the most dominant one. Indeed, the relatively weak association (OR = 1.2602) of this variant with Chinese depressive women (14) suggests that it may not be a common driving genetic force for all cases of depressive disorders.

CONCLUSION
We identified a novel rare variant of OGDHL, rs2293239, that may serve as the driving force for depression onset in a Chinese pedigree. Combining multi-omics integration and multimodal imaging analysis, or imaging genetics, we revealed the possible mechanism underlying familial depression in this pedigree. This work enhances the current molecular understanding of the complex pathogenesis of depressive disorders. Furthermore, it provides new options for prioritizing better clinical therapeutic regimens and suggests a new protein target for the design of highly specific drugs to treat depressive disorders.

DATA AVAILABILITY STATEMENT
The datasets generated and analyzed during the present study are available from the corresponding author upon reasonable request.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Clinical Research Ethics Committee of the Xiamen Xianyue Hospital, Fujian, China. The patients/participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
CZ, ZZ, and HT conceived and designed research. ZP, ZZ, and WW collected data and conducted research. TF, ZL, XL, GD, and GH analyzed and interpreted data. WW and CZ wrote the initial paper. HT and GC revised the paper. WW and GC had primary responsibility for final content. All authors read and approved the final manuscript.

FUNDING
This study was supported by the National Natural Science Foundation of China (81871018 to WW).