Genetic Landscape of Major Depressive Disorder: Assessment of Potential Diagnostic and Antidepressant Response Markers

Abstract Background The clinical heterogeneity in major depressive disorder (MDD), variable treatment response, and conflicting findings limit the ability of genomics toward the discovery of evidence-based diagnosis and treatment regimen. This study attempts to curate all genetic association findings to evaluate potential variants for clinical translation. Methods We systematically reviewed all candidates and genome-wide association studies for both MDD susceptibility and antidepressant response, independently, using MEDLINE, particularly to identify replicated findings. These variants were evaluated for functional consequences using different in silico tools and further estimated their diagnostic predictability by calculating positive predictive values. Results A total of 217 significantly associated studies comprising 1200 variants across 545 genes and 128 studies including 921 variants across 412 genes were included with MDD susceptibility and antidepressant response, respectively. Although the majority of associations were confirmed by a single study, we identified 31 and 18 replicated variants (in at least 2 studies) for MDD and antidepressant response. Functional annotation of these 31 variants predicted 20% coding variants as deleterious/damaging and 80.6% variants with regulatory effect. Similarly, the response-related 18 variants revealed 25% coding variant as damaging and 88.2% with substantial regulatory potential. Finally, we could calculate the diagnostic predictability of 19 and 5 variants whose positive predictive values ranges from 0.49 to 0.66 for MDD and 0.36 to 0.66 for response. Conclusions The replicated variants presented in our data are promising for disease diagnosis and improved response outcomes. Although these quantitative assessment measures are solely directive of available observational evidence, robust homogenous validation studies are required to strengthen these variants for molecular diagnostic application.


INTRODUCTION
Major depressive disorder (MDD) is a common, complex, and debilitating neuropsychiatric disorder that contributes a significant burden on both the individual and society (Norkeviciene et al., 2022).It is primarily characterized by symptoms that include low mood, anhedonia, fatigue, alteration in sleep and appetite, lack of concentration, feelings of guilt, helplessness, and in the worst cases, suicidal ideations (Malhi and Mann, 2018).According to the WHO, nearly 4.4% of the world population are affected by depression, with a significant rate of suicidal deaths of approximately 1 million per year (World Health Organization, 2017).
To date, the diagnosis of MDD revolves around the symptom-based approach, which often becomes difficult due to the lack of defined natural boundaries and overlapping symptoms between affective disorders (Feczko et al., 2019;García-Gutiérrez et al., 2020).Furthermore, the treatment of depression involves a trial-and-error approach, wherein monotherapy initially is favored followed by multi-drug therapy in case patients fail to respond to initial therapy.Nevertheless, one-third of the patients do not adequately respond to the existing pharmacotherapy regimen.This variability could be a consequence of intrinsic biological (inter-individual genetic variations) and environmental heterogeneity among MDD patients (Fabbri et al., 2019;Srivastava et al., 2019).Therefore, it is imperative to identify certain bio-signatures (such as single nucleotide polymorphisms (SNPs)) associated with the MDD susceptibility and antidepressant response, enabling the appropriate diagnosis and promoting personalized medication/therapy.
The etiology of MDD is multifactorial, involving a complex interaction of both genetic and environmental factors.The genetic contribution to pathophysiology has been demonstrated by a plethora of genetic association studies in addition to twin studies that reported the heritability of MDD to be approximately 40% (Sullivan et al., 2000;Shadrina et al., 2018;Schwabe et al., 2019).The existing literature regarding the genetics of MDD and treatment response is enormous and complex.The 2 complementary approaches primarily exploited in genetic studies are genome-wide association studies (GWAS) and candidate gene studies.Until recently, the majority of candidate studies primarily focused on genes related to a widely accepted hypothesis based on a serotonergic, hypothalamic-pituitary-adrenal (HPA) axis, neural plasticity, glutamatergic system considering their role in MDD pathophysiology and the fact that they might be the targets of antidepressant drugs (Basu et al., 2015(Basu et al., , 2019)).Most of them have reported a positive association with the disease susceptibility and treatment response despite that low power of the study and contradictory findings limit their clinical applicability.Moreover, a recent study by Border et al. (2019) on a large sample pool examined the association of 18 highly studied candidate gene polymorphisms with depression and found no substantial contribution of any of the polymorphisms to depression liability.On the other hand, earlier GWAS studies failed to detect SNPs of a genome-wide significance level (Lewis et al., 2010;Muglia et al., 2010;Rietschel et al., 2010;Shi et al., 2011), but recent GWAS and meta-analyses have revealed 15 (Hyde et al., 2016), 16 (Howard et al., 2021), 44 (Wray et al., 2018), and 102 (Howard et al., 2019) genetic loci to be significantly associated with MDD susceptibility.However, these results across studies were not consistently reproducible.These inconsistencies likely may be explained by methodological differences such as the definition of depression and sampling strategies employed.We speculate positive association in a single study can be argued as false positive; thus, reproducible associations hold promise and would have tremendous importance in prognosis, diagnosis, and therapeutics.
Therefore, in the present study, we systematically collated the available genetic association data and identified the replicated genetic variants and performed their functional annotation and diagnostic predictability analysis in MDD susceptibility and antidepressant response, independently, via a PubMed-based search strategy.Our work is an attempt to find if any genetic variant holds the potential for clinical translation based on the available observational evidence.

METHODS
A systematic search was performed using MEDLINE database wherein we independently retrieved the literature for all genetic association studies related to MDD susceptibility and antidepressant response.

Acquisition of MDD Risk-Associated Studies-
A comprehensive PubMed search (https://pubmed.ncbi.nlm.nih.gov/) was conducted using a combination of the following medical subject heading terms "major depressive disorder", "major depression", "depressive disorder", "genetics", "gene", "SNP", "polymorphism", and "variants" with AND/OR Boolean operators to retrieve relevant publications until June 2021.Articles were manually screened for titles, abstracts, and full texts by 2 authors (P.S. and A.S.) independently.Any discrepancies were resolved on consensus agreement with a third author (R.K.).Two authors (D.G. and S.T.) cross-checked the data to ensure consistency.Additionally, the bibliography of included articles was subsequently screened for additional references.The inclusion criteria were as follows: (1) patients primarily diagnosed as MDD; (2) a case-control study-design; (3) the study must examine the association between a variant/s and MDD; and (4) study published in English.Exclusion criteria were as follows: (1) non-genetic studies; (2) other than case-control study design; (3) studies focusing on the antidepressant response; (4) in vitro and animal studies; (5) enrolled patients who had depressive episodes due to other psychiatric disorder or studies evaluating other phenotypes in depressive patients (suicide etc.); (6) articles with only an abstract and no full text available; and (7) review articles, meta-analysis, case studies, commentary, and editorials.

Acquisition of Antidepressant Response-Associated Studies-
To retrieve genetic association studies related to antidepressant response, the search string using "antidepressants", "SSRIs", "MAOIs", "SNRIs", "TCAs", "Atypical antidepressants" AND "pharmacogenetics", "pharmacogenomics", genetics", "gene", "SNP", "polymorphism", "variants" with AND/OR was conducted.Articles were extracted until June 2021.Manual screening of the titles and abstracts and further full-text evaluation was performed.Furthermore, bibliographic cross-reference was also included for additional studies.Articles were included if (1) the patients had a primary diagnosis of major depressive disorder; (2) the study examined the association between the candidate or genome-wide variants and antidepressant response; (3) the study should have dichotomous groups (e.g., between responder and non-responder or remitter and non-remitter); (4) response/ remission assessed by using relevant severity scales such as Hamilton Rating Scale for Depression (HAMD), Montgomery and Åsberg Depression Rating Scale (MADRS) etc.; and

Functional Annotation
The SNPs were mapped to genes using National Center for Biotechnology Information (NCBI) dbSNP following the GRCh38.p13genome assembly.To predict the functional consequences of the included genetic variants, we used different in silico prediction tools to assess the deleterious effect of the amino acid change in case of a missense variant and the regulatory potential of intronic SNPs (Kaur et al., 2014).The different tools used for function prediction were SIFT (Ng and Henikoff, 2001), PolyPhen2 (Adzhubei et al., 2010), RegulomeDB 2.0.3 (Boyle et al., 2012), and SNP info (Xu and Taylor, 2009).Additionally, we checked the genes of these replicated variants: (1) for which behavior/neurological phenotype was observed in knockout mice and the phenotype data were extracted from the Mouse Genome Informatics database (http://www.informatics.jax.org/); (2) assessed whether the gene is a known target of an antidepressant drug as detailed in the drug-gene interaction database (www.DGidb.com);and (3) whether the gene is preferentially expressed in the brain means the average expression in all brain tissues was higher than the average expression in non-brain tissues.This was assessed by using gene expression data from all 53 tissues of the Gene-Tissue expression Consortium.Data obtained from each tool have been described in supplementary Tables 5 and 6.

Assessment of Diagnostic Predictability of Genetic Variants
The diagnostic predictability of genetic variants indicates its ability to precisely predict the occurrence of a disease phenotype in patients carrying the risk allele.We attempted to evaluate the diagnostic predictability for genetic variants with more than 1 publication of disease-variant association as well as response-variant association.True positives (TP) and false negatives (FN) are values representing the risk allele and wild-type allele carrier, respectively, in cases.Conversely, true negatives (TN) and false positives (FP) are defined as wild-type and risk allele carriers, respectively, in controls.The TP rate or sensitivity is calculated as ΣTP/(ΣTP + ΣFN) and specificity is ΣTN/(ΣTN + ΣFP).Furthermore, the positive predictive values are calculated as ΣTP/(ΣTP + ΣFP).

Acquisition of MDD Risk-Associated Studies-
A systematic search strategy for identifying MDD associated genes extracted 10 631 genetic association studies, which were further reduced to 280 articles of relevance after title and abstract screening.Among these excluded articles, 2382 non-human/other language, 6081 other disorder/phenotype/comorbidity, 201 association study of MDD with other phenotype, 120 no association, and 1567 other articles including non-genetic studies, review articles, editorials, meta, comments, and letters were also removed.The remaining 280 articles were then searched for their full text, and 63 articles were again excluded from the study because they did not meet the inclusion criteria, leading to a final of 217 articles for data extraction and processing.Twelve of these were GWAS studies and 205 were candidate association studies (Figure 1).

Acquisition of Antidepressant Response-Associated Studies-
Similarly, a systematic search strategy for identifying antidepressant response associated genes extracted 7892 genetic association studies, which was further reduced to 345 articles of relevance after title and abstract screening.Among these excluded articles, 3109 were non-human, 160 were other language, 1768 were not an antidepressant response study, 533 were other disorders, 39 were methylation/miRNA, protein-related studies, 82 no association studies and 1856 other articles including non-genetic studies, review articles, genome-wide studies, editorials, and letters were also removed.The remaining articles were then searched for their full text and 222 articles were again excluded from the study, as they did not meet the inclusion criteria, leading to a final of 128 articles (including 5 articles that are additionally added through cross-references) for data extraction and processing.Of these, 11 were GWAS response studies and 117 were candidate response association studies (Figure 2).

Candidate Genetic Association Studies Related to MDD Risk-
We have summarized the main characteristics of the identified studies in supplementary Table 1.Only significant P values and ORs are presented.The 205 articles reported nominal significant associations (P < .05).These 205 articles reported significant results for 378 polymorphisms in 156 unique genes.A total of 196 339 individuals (n = 80 891 cases and n = 115 448 controls) were included in all the studies.The 27 genetic variants were confirmed by at least 2 studies.The number of patients ranged from 11 to 1738, and their age varied from 14 to 102 years.The ratio of male to female was 1.30 and 1.35 in cases and controls, respectively.Of all the candidate association studies, 90 studies were performed in East Asia, 87 in Europe, 11 in America, 7 in the Caucasian population, 4 in Middle Eastern countries, 3 in South Asia, 2 included mixed population, 1 from African population, 1 involving white participants and 1 with white non-Hispanic individuals.The most used MDD diagnostic criterion was based on the DSM (n = 164), followed by The International Classification of Diseases (ICD) (n = 21), studies that used both DSM and ICD criteria (n = 7), and others (n = 11).

GWAS Studies Related to MDD Risk-
A total of 12 GWAS determine the association of genetic variants in MDD.These 12 articles reported significant results for 819 polymorphisms in 387 genes.In total, 268 481 patients with MDD and 841 656 controls were included.The mean age of the participants was 45.55 years for MDD cases and 49.71 years for controls.The overall ratio of male to female was 1.12.Most of the studies were performed in the European population (n = 9), Han Chinese (n = 1), and mixed population (n = 2).The MDD patients were diagnosed based on Structured Clinical Interview fulfilling DSM/ICD criteria (n = 11), although in 1 study self-reported MDD patients were also recruited.The majority of studies included patients with MDD/unipolar depression/recurrent depression (n = 10), except in 2 studies, where 1 included recurrent early-onset MDD patients and in another broad depression and probable MDD patients were recruited in the study.All studies employed an array-based genotyping method.The characteristics of the included studies and genotypic/allelic distributions of the polymorphisms are shown in supplementary Table 2.

Candidate Genetic Association Studies Related to Antidepressant Response-
A total 116 articles reported nominal significant associations (P < .05).The characteristics of the included studies and genotypic/allelic distributions of the polymorphisms are shown in supplementary Table 3.The table includes only those studies that were either in the form of responder vs non-responder (n = 64) or remitter vs non-remitter (n = 34).Thirteen studies examined both response as well as remission status.Additionally, 6 studies checked the association in treatment-resistant patients.A total of 21 753 (12 577 responders and 9029 non-responders) and 11 299 (4546 remitters and 4906 non-remitters) patients were included in the studies.The recruited cohort were of East Asian (55.17%),European (27.58%),American (7.75%), and other (9.48 %) origin.All the studies recruited their patients based on either DSM (n = 112) or ICD-10 (n = 4) diagnostic criteria.Remission was defined as a final Hamilton Rating Scale for Depression (HAM-D) or Montgomery and Åsberg Depression Rating Scale (MADRS) total score of 7 or less, and response was defined as at least a 50% decrease in HAM-D or MADRS total score.Maximum studies have employed HAM-D scale (n = 69), followed by Quick Inventory of Depressive Symptomatology (QIDS) (n = 5), MADRS (n = 2), and Beck's Depression Inventory (n = 1) for accessing the response in patients.Similarly, for remission assessment HAMD/ MADRS scale (n = 39) and QIDS (n = 5) were employed on the patients.Furthermore, we observed the majority of studies were based on selective serotonin reuptake inhibitors (SSRIs) (n = 69) followed by mixed therapy (n = 33), serotonin and norepinephrine reuptake inhibitors (n = 9), atypical (n = 3), and tricyclic antidepressants (n = 2).Of these, 60 studies were on monotherapy.There were some studies (n = 6) that determined the response association with multi-drug therapy and further also examined the association based on monotherapy.The follow-up period in all the studies ranged from 2 to 18 weeks.In total, 97 unique genes and 217 variants were found to be significantly associated with antidepressant response.

GWAS Studies Related to Antidepressant Response-
A total of 11 studies have examined association between SNPs and antidepressant response (supplementary Table 4).There were 3 response studies, 3 remission studies, and 5 associated with both response and remission status.A total 39 533 patients (23 141 responder and 11 322 non-responders; 2392 remitter and 1413 non-remitter) were included in either a discovery or replication cohort of all the studies.Included studies were performed in different ethnic populations, from Korean (n = 3), European (n = 2), Japanese (n = 1), Mexican American (n = 1), White non-Hispanic (n = 1), Caucasian (n = 1), and Mixed (n = 3).The male to female ratio in response studies was 1.32 and in remission studies was 0.79.We could not determine the mean age inof all the studies due to the heterogeneity in data reporting as well as the unavailability of raw data in some of the studies.All studies employed the DSM/ICD-10 diagnostic criteria to determine the MDD phenotype except 1 study, which utilized a questionnaire designed by 23andMe to recruit self-reported patients.These 11 studies reported significant results for 704 polymorphisms in 315 genes.Only rs1908557 SNPs have shown genome-wide significance with a P value of 2.6 × 10 −8 .

Reproducibility of Findings
To narrow the focus of the results presented above, we chose to concentrate our analysis on the replicated genetic variants and genes that were reported 2 or more times across studies.Of all the positive associations, we found a total of 34 replicated variants with MDD susceptibility.However, 3 variants (rs120074175/ TPH2, rs6195/NR3C1, and rs6189+rs6190/NR3C1) had MAF <0.05 and hence were removed from further analysis.The remaining 31 replicated variants from candidate studies (n = 24), GWAS (n = 3), and both candidate and GWAS studies (n = 4) are shown in Table 1 and Figure 3A.Additionally, information related to the number of studies reporting no association between these variants and MDD susceptibility is provided in supplementary Table 7A.Further, 5-HTTLPR/SLC6A4 was the genetic factor most frequently investigated (n = 10), followed by rs6265/BDNF (n = 8), rs4680/COMT (n = 6), rs1801133/MTHFR (n = 4), and others.Interestingly, of all 31 replicated variants, we observed 13 variants (3 from GWAS, 7 from candidate, and 3 from both), that is, C allele of rs2273289/PLOD1 in 2 studies, C allele of rs2715148/PCLO in 2 studies, C allele of rs2423618/LINC00687 in 2 studies, T allele of rs1801133/MTHFR in 4 studies, T allele of rs5443/GNB3 in 4 studies, G allele of rs242939/CRHR1 in 3 studies, G allele of rs6295/HTR1A in 3 studies, A allele of rs1006737/CACNA1C in 2 studies, T allele of rs4880/SOD2 in 2 studies, C allele of rs1801131/MTHFR in 2 studies, A allele of rs2715147/PCLO, G allele of rs9416742/BICC1, and A allele of rs999845/BICC1 showed consistency in reporting the risk allele, whereas 18 variants had inconsistent risk alleles across studies.Moreover, we also checked the reproducibility of genes where 27 genes in candidate studies, 8 genes in GWAS, and 7 genes in both candidate and GWAS studies were replicated, which is represented in Table 2 and Figure 3B.
On the other hand, we found 15 genetic variants in candidate, 2 in GWAS, and 1 common in candidate and GWAS studies for drug response, as shown in Table 3 and Figure 4A, whereas information related to the number of studies reported no association with the antidepressant response is provided in supplementary Table 7B.SNPs located in SLC6A4, BDNF, HTR2A, HTR1A, GNB3, CYP1A2, NR3C1, TPH2, COMT, GRIK4, FKBP5, CFAP61, and UBE3C were positively replicated in independent populations.Moreover, the only variant that showed a positive association in both GWAS and candidate studies was rs6127921 of BMP7 gene.In addition, 18 genes in candidate studies, 11 in GWAS, and 4 genes in both candidate and GWAS studies were found to be replicated in response studies, as represented in Table 4 and Figure 4B.

Functional Annotation of Reproducible Findings
This study further sought to explore the molecular consequences of these reproducible genetic variants (missense and non-coding) from MDD susceptibility and antidepressant response studies by using different computational methods.The different functionality prediction servers such as SIFT, PolyPhen-2, RegulomeDB, and SNP info were utilized.SIFT and PolyPhen-2 sieved the deleterious non-synonymous SNPs, followed by RegulomeDB, which provides information about regulatory SNPs.The prediction analysis of the effect of SNPs on specific functions, including splicing regulations, miRNA binding site, regulatory potential, and conserveness, was performed using the FuncPred tool of SNPinfo.Consequently, we checked a total of 31 replicated variants (7 missense, 14 intronic, 6 UTR/upstream, 1 downstream and 3 synonymous) resulting from MDD susceptible studies shown in supplementary Table 5.Here, out of 7 missense SNPs, SIFT predicted 1 SNP (14.2%)

S.
No. Table 2. Continued as deleterious and the rest 7 (85.7%)as tolerated.PolyPhen-2 predicted 2 SNPs (28.5%) as damaging and 5 SNPs (71.4 %) as neutral or benign.Importantly, only nonsynonymous/missense variant rs1801133/MTHFR was predicted to be deleterious and probably damaging by SIFT and PolyPhen-2 respectively.In addition, rs6265/BDNF was also predicted as possibly damaging by PolyPhen-2 though tolerated by SIFT.These findings were further strengthened by the ClinVar database as rs1801133 was reported to be pathogenic and rs6265 as a risk factor for other psychiatric disorders such as schizophrenia and bipolar, respectively.Out of all 31 variants, SNPinfo predicted 1 variant (i.e., rs2715148/PCLO) affects the miRNA binding site activity and 7 variants (rs4680, rs4880, rs1801131, rs2522833, rs1045642, rs4343, rs2522833) were found to affect the splicing activity.Moreover, 26.6% of SNPs had >80% conservation and 16.6% of SNPs had >40% regulatory potential.Further, among the 31 variants evaluated with RegulomeDB, 25 variants were predicted to have a regulatory effect ranging from 1d to 5. Out of these 25 variants, 2 variants (i.e., rs1800629/TNF, score = 1d; and rs1801131/MTHFR, score = 1f) got the lowest scores and thus are most likely to be involved in eQTL functions.Similarly, 4 variants (rs4343/ACE, rs2242446/SLC6A2, rs41423247/NR3C1, rs2273289/PLOD1) were assigned a rank of 2b, which indicates TF binding + any motif + DNase Footprint + DNase peak.We also observed, of 31 replicated variants, 26 (83.8%) and 16 (51.6%)genes were found to have behavioral/neurological knockout mouse model and are known targets of antidepressant drugs, respectively.Further, 12 (38.7%)genes have shown to be preferentially expressed in the brain (supplementary Table 5).

Diagnostic Predictability
The allele frequency of reproducible variants was utilized for sensitivity and specificity analysis (Table 1).The majority of replicated findings were inconsistent in their reported risk allele such as the 5-HTTLPR, which appeared in 10 studies, the short (S) form was reported in 7 studies, and long (L) form in 3 studies.Therefore, studies reported the same type of risk allele were pooled together for sensitivity and specificity analysis.G allele of rs6265/BDNF had the highest sensitivity (0.78), whereas G allele of rs242939/CRHR1 had the highest specificity (0.93).Moreover, the positive predictive values for 19 genetic variants ranged from 0.49 to 0.66.Most of them had moderate PPV.However, we could not perform the diagnostic predictability for a few variants (n = 6) due to unavailable raw data.Additionally, if a replication of variant is due to 2 studies that reported the 2 different risk alleles (n = 6), then assessment of sensitivity and specificity could not be calculated.
In response studies, we could only assess the diagnostic predictability of 5 out of 18 variants (PPV between 0.36 and 0.66) because the majority of replicated findings were inconsistent in their response status, drug class, associated phenotype, population, and reported risk allele and therefore could not be cumulated/pooled (Table 3).

DISCUSSION
This study comprehensively evaluated the original articles on genetic association studies of MDD susceptibility and antidepressant treatment response, independently.A total of 217 studies of MDD susceptibility and 128 studies of drug response met our a priori inclusion criteria and were described in the present study.The results of this article have shown that the majority of the positive associations were confirmed by only 1 study and therefore cannot exclude the possibility of having been obtained by chance, and thus are not sufficient to establish a link with MDD susceptibility or antidepressant treatment response.Therefore, the rest of the discussion is limited to reproducible findings at the SNP or gene level generated in more than 1 independent study.
In total, we found 31 and 18 replicated variants in MDD susceptibility and antidepressant drug response, respectively.Further, taking all these findings into account, it appears most studies used the candidate gene approach.As a result, a large number of SNPs and genes were found to be replicated in candidate studies compared with GWAS studies, which demonstrate the surprisingly small overlap of genetic variants across studies.Also, none of the variants from candidate studies appeared among the top hits in any of the identified GWAS.Nevertheless, we identified an important overlap of positive association of 4 (rs2522833/PCLO, rs2715147/PCLO, rs9416742/BICC1, and rs999845/BICC1) and 1 (rs6127921/BMP7) variants between candidate and GWAS studies in MDD susceptibility and antidepressant response, respectively.Replication of variants between candidate and unbiased GWAS studies greatly increases the acceptability of genotype and phenotype association.Besides replication, additional evidence regarding the biological relevance of PCLO, BICC1, and BMP7 has also been reported in the literature.A meta-analysis by Hek et al. (2010) provided additional evidence of significant association (P = 2.16 × 10 −3 ) between SNP rs2522833/PCLO and depressive disorders in a population-based study.In addition, the rs2522833 variant causes change in amino acid from serine to alanine in a piccolo C2A calcium binding domain (Verbeek et al., 2012), whose overexpression causes depression-like behavior in mice (Furukawa-Hibi et al., 2010).Similarly, expression of BICC1 was reported to be upregulated in the post-mortem brains of individuals with MDD and expression levels of the same observed to be reduced after antidepressant treatment in rat models of chronic stress (Ota et al., 2015).Furthermore, a significant reduction in the expression of the BMP7 gene in the locus coeruleus region of post-mortem brains of individuals with MDD has also been shown in the literature (Ordway et al., 2012).Thus, these variants would most likely be of considerable interest in future studies.
Further, the functional annotation of reproducible variants of MDD susceptibility revealed 2 non-synonymous/missense variants (rs1801133/MTHFR and rs6265/BDNF) as deleterious/damaging and 6 variants (rs1800629/TNF, rs1801131/MTHFR, rs4343/ACE, rs2242446/SLC6A2, rs2273289/PLOD1, and rs41423247/NR3C1) with a score of <3 had high regulatory effect in disease condition.On the other hand, functional annotation of variants from antidepressant treatment response predicted 1 non-synonymous/missense variant (rs6265/BDNF) as damaging and 3 non-coding variants (rs2470890/CYP1A2, rs6313/HTR2A, and rs41423247/NR3C1) with a high regulatory effect.Moreover, diagnostic predictability of reproducible variants revealed moderate PPV ranges from 0.49 to 0.66 in MDD susceptibility and 0.36 to 0.66 in antidepressant drug response.This quantitative measure offers a substantial hint toward a direction of predictability and also corroborates increasing the significance of replicated findings.
Of all the positive association studies, 5-HTT gene-linked polymorphic region (5-HTTLPR) polymorphism has gained particular attention for being the most replicated variant in both disease (n = 10) and drug response studies (n = 22).The 2 allelic variants, S and L, are due to deletion and insertion of 44 bp in the promoter region, respectively.The S allele has been demonstrated to have lower transcriptional activity than the L allele, resulting in the reduction of serotonin transporter in the cell membrane.Consequently, the reduction of 5-HTT causes the imbalance of serotonin concentration and function (Fratelli et al., 2020).However, we observed a significant inconsistency in the reported risk allele across studies.For example, in disease, 5HTTLPR       appeared in 10 studies, wherein the S form was reported as risk allele in 7 studies and the L form in 3 studies.The heterogeneity in reported risk allele even within the same geographical population weakens our confidence in proposing them as a probable genetic diagnostic marker.In a similar line, there were 17 (rs6265, rs4680, rs2242446, rs1045642, rs9340799, rs2234693, rs4291, rs6311, rs1360780, rs4713916, rs1800532, rs1800629, rs4295, rs6195, rs2522833, rs4343, and rs41423247) and 9 (5HTTLPR , rs6265, rs7997012, rs6295, rs5443, rs2470890, rs6313, rs2171363, rs1954787) replicated variants associated with MDD susceptibility and drug response, respectively, where authors have found discrepancy in reported risk allele, suggesting a more robust future study design for such candidate gene studies.
Therefore, reproducible variants reporting a single risk allele across studies are of prime importance and could be taken forward for diagnostic and therapeutic application.In total 14 variants in MDD susceptibility and 6 in antidepressant response reported the consistent findings in terms of risk allele.However, due to the unavailability of raw data, we could assess the diagnostic predictability of only 7 and 2 variants in disease and drug response, respectively.Interestingly, we observed variants of FKBP5 and HTR2A genes are being commercially utilized for diagnostic application in MDD gene panels (Rubinstein et al., 2013), and the replication of variants in these genes, rs1360780/FKBP5 (PPV = 0.53), rs4713916/FKBP5 (PPV = 0.53), and rs6311/HTR2A (PPV = 0.54), are demonstrated in our analysis.However, the diagnostic predictability of these variants were found to be moderate.This observation inclined our interest to all the other reproducible variants with moderate diagnostic predictability and suggested the importance of other potential variants for diagnostic purposes.And because there are no specific guidelines to date regarding selection of diagnostic or predictive markers, we are proposing a few probable candidates (shown in Table 5) for diagnostic/predictive genetic panel for MDD susceptibility and to predict antidepressant response.
In addition to reproducibility with a single risk allele, these variants have clinical relevance as well.For example, out of 7, there were missense (rs1801133/MTHFR, rs1801131/MTHFR, rs4880/SOD2) variants including 1 SNP (rs1801133/MTHFR) annotated as having a deleterious/damaging effect.Two of the SNPs (rs4880/SOD2 and rs1801131/MTHFR) were identified to be present at the splice site.In addition, of all the variants predicted to have a regulatory role in transcription factor binding site activity, importantly, rs1801133/MTHFR (score = 1F) involved eQTL function as well.The majority of genes have shown to be preferentially expressed in the brain and are known targets of antidepressant drugs.Moreover, all 7 variants from disease and 2 variants from drug response were consistently found to be associated with moderate positive predictive value.Hence, this work holds promise by highlighting biologically important markers for MDD susceptibility and antidepressant response assessment.
Even though this study is an attempt to provide a wide landscape of genetic literature, we must acknowledge the limitations as well.First and foremost, our search is limited to only MEDLINE and studies in English, so there is the possibility of non-inclusion of relevant data outside these criteria.Second, in antidepressant response search data were excluded where response is assessed based on percentage improvement and where data was not compared between responder and non-responder or remitter and non-remitter groups.However, the exclusion was according to the protocol, as we have aimed to specifically focus on studies considered dichotomous groups.Moreover, heterogeneity in

Gene
Table 4. Continued antidepressant class, response status, inconsistent risk allele, and unavailability of raw data limit us to calculate the predictability of most of the replicated polymorphisms.Nevertheless, we maintained the consensus across studies by calculating the unadjusted P value and OR for risk allele wherever raw data (allele frequency) was provided.We found most findings were inconsistent in reporting the risk allele even in the same population across studies, which precluded us from drawing any robust conclusion by nullifying the assumption of population-specific risk allele.The reason for these inconsistencies could be addressed by small sample size and large publication bias in addition to other technical or methodological differences among studies.Further, GWAS published to date have largely not replicated the candidate gene polymorphisms conducted over the past many years, which may be attributed to the polygenic nature of the disease.
Lastly, the use of sensitivity and specificity as quantitative assessment to estimate diagnostic predictability was solely based on available evidence.With the publication of more new studies, such measures may vary.Hence, our study only suggests a direction of predictability and should be used judiciously for clinical interpretations.

CONCLUSIONS AND FUTURE DIRECTIONS
This study attempted to provide a holistic landscape of all genetic association studies related to MDD risk and antidepressant response.Despite the substantial heterogeneity in genetic association across studies, study design, and drugs administered, we identified 31 and 18 genomic regions that showed independent replication in MDD susceptibility and antidepressant response studies, respectively.Importantly, 4 variants (rs2522833/PCLO, rs2715147/PCLO, rs9416742/ BICC1, and rs999845/ BICC1) in MDD and 1 (rs6127921/BMP7) variant with antidepressant response are of high potential as they are significantly associated in both candidate and GWAS studies.Further, the variants significantly replicated in the candidate as well as GWAS studies, with consistent risk allele and moderate positive predictive values, that is, 7 (rs1801133, rs5443, rs242939, rs1006737, rs4880, rs6295, rs1801131) with MDD, and 2 (STin2, rs41423247) with antidepressant drug response, are of high interest for diagnostic and therapeutic applications after appropriate validation.Besides, this article would serve as the basis for selecting the candidate variants, and researchers can directly take up these replicated variants (31 in MDD disease and 18 in drug response) for performing future meta-analysis and functional validation.Moreover, it is also important to emphasize on remaining significant associations found in a single study.Hence, replication of these genetic associations are crucial to ascertain the significance, whether random or causal.

Figure 1 .
Figure 1.Identification and selection process of relevant genetic association studies of major depressive disorder (MDD).

Figure 2 .
Figure 2. Identification and selection process of relevant genetic association studies of antidepressant response.

Figure 3 .
Figure 3. (A) Significantly associated reproducible* genetic variants: 24 in candidate studies, 3 in genomewide studies, and 4 variants that are common in both candidate and genomewide association studies (GWAS).Color represents the studies; gradient represents the no. of studies/ evidence.Label order: gene, SNPs, No. of studies/evidence.*Minimum in 2 studies.(B) Reproducible genes from genome-wide and candidate association studies: 27 in candidate studies, 8 in genome-wide studies, and 7 variants that are common in both candidate and GWAS studies.

Figure 4 .
Figure 4. (A) Significantly associated reproducible* genetic variants: 15 in candidate studies, 2 in genomewide studies, and 1 variant that is common in both candidate and genomewide association studies (GWAS).Color represents the studies; gradient represents the no. of studies/ evidence.Label order: gene, SNPs, No. of studies/evidence.*Minimum in 2 studies.(B) Reproducible genes from genome-wide and candidate drug response association studies: 18 in candidate studies, 11 in genome-wide studies, and 4 variants that are common in both candidate and GWAS studies.

Table 1 .
List of all Reproducible Genetic Variants With Their Sensitivity, Specificity, and PPV in Candidate Studies, GWAS Studies, and in Both Related to MDD

Table 3 .
List of All Reproducible Genetic Variants With Their Sensitivity, Specificity, and PPV in Candidate Studies, GWAS Studies and in Both Related to Antidepressant Response

Table 4 .
List of All Reproducible Genes in GWAS Studies, Candidate Studies, and in Both Related to Antidepressant Response