Original Article

Molecular Psychiatry (2011) 16, 516–532; doi:10.1038/mp.2010.38; published online 30 March 2010

Poor replication of candidate genes for major depressive disorder using genome-wide association data

F J Bosker1,8, C A Hartman1,8, I M Nolte2,8, B P Prins2, P Terpstra2, D Posthuma3,4, T van Veen5, G Willemsen3, R H DeRijk5, E J de Geus3, W J Hoogendijk6, P F Sullivan7, B W Penninx1,5,6, D I Boomsma3, H Snieder2,8 and W A Nolen1,8

  1. 1Department of Psychiatry, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands
  2. 2Unit of Genetic Epidemiology and Bioinformatics, Department of Epidemiology, University Medical Center Groningen, University of Groningen, Groningen, The Netherlands
  3. 3Department of Biological Psychology, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands
  4. 4Department of Medical and Functional Genomics, VU University Medical Center, Amsterdam, The Netherlands
  5. 5Department of Psychiatry, Leiden University Medical Center, Leiden, The Netherlands
  6. 6Department of Psychiatry, VU University Medical Center, Amsterdam, The Netherlands
  7. 7Department of Genetics, University of North Carolina, Chapel Hill, NC, USA

Correspondence: Dr FJ Bosker, Department of Psychiatry, University Medical Center Groningen, University of Groningen, PO Box 30.001, 9700 RB Groningen, The Netherlands. E-mail: f.j.bosker@psy.umcg.nl

8These authors contributed equally to this work.

Received 3 August 2009; Revised 18 February 2010; Accepted 22 February 2010; Published online 30 March 2010.



Data from the Genetic Association Information Network (GAIN) genome-wide association study (GWAS) in major depressive disorder (MDD) were used to explore previously reported candidate gene and single-nucleotide polymorphism (SNP) associations in MDD. A systematic literature search of candidate genes associated with MDD in case–control studies was performed before the results of the GAIN MDD study became available. Measured and imputed candidate SNPs and genes were tested in the GAIN MDD study encompassing 1738 cases and 1802 controls. Imputation was used to increase the number of SNPs from the GWAS and to improve coverage of SNPs in the candidate genes selected. Tests were carried out for individual SNPs and the entire gene using different statistical approaches, with permutation analysis as the final arbiter. In all, 78 papers reporting on 57 genes were identified, from which 92 SNPs could be mapped. In the GAIN MDD study, two SNPs were associated with MDD: C5orf20 (rs12520799; P=0.038; odds ratio (OR) AT=1.10, 95% CI 0.95–1.29; OR TT=1.21, 95% confidence interval (CI) 1.01–1.47) and NPY (rs16139; P=0.034; OR C allele=0.73, 95% CI 0.55–0.97), constituting a direct replication of previously identified SNPs. At the gene level, TNF (rs76917; OR T=1.35, 95% CI 1.13–1.63; P=0.0034) was identified as the only gene for which the association with MDD remained significant after correction for multiple testing. For SLC6A2 (norepinephrine transporter (NET)) significantly more SNPs (19 out of 100; P=0.039) than expected were associated while accounting for the linkage disequilibrium (LD) structure. Thus, we found support for involvement in MDD for only four genes. However, given the number of candidate SNPs and genes that were tested, even these significant may well be false positives. The poor replication may point to publication bias and false-positive findings in previous candidate gene studies, and may also be related to heterogeneity of the MDD phenotype as well as contextual genetic or environmental factors.


candidate genes; genome-wide association study; major depressive disorder; replication; single-nucleotide polymorphisms



Major depressive disorder (MDD) is a multi-factorial disease, with both genetic and environmental factors likely to have a role in its etiology. The Netherlands Study of Depression and Anxiety (NESDA; www.nesda.nl) and the Netherlands Twin Registry (NTR; www.tweelingenregister.org) took part in the Genetic Association Information Network (GAIN; http://www.fnih.org/GAIN) project to enable a genome-wide association study (GWAS) using a 600K Perlegen chip (Perlegen Sciences, Mountain View, CA, USA).1 Within the GAIN MDD study,2 1862 participants with a diagnosis of MDD and 1860 controls at low liability for MDD were selected for genome-wide genotyping.

A GWAS approach allows a hypothesis-free search for potential new susceptibility genes. The downside of a GWAS is that a strict statistical adjustment for the large number of single-nucleotide polymorphisms (SNPs; in the GAIN MDD study 435291 SNPs) is required before associations can be considered significant on a genome-wide level,3 and replication of such findings in independent cohorts is mandatory to exclude false-positive findings.4

Another potential use of a GWAS is to use the results for a large-scale replication study of previous candidate gene studies. Application of such previous knowledge within the context of a GWAS allows less stringent significance thresholds than those for the hypothesis-free GWAS analysis.5, 6 So far, the role of candidate genes in MDD has been the subject of many association studies. Unfortunately, there is little consistency between candidate gene studies for multifactorial diseases such as MDD (see, for example, Hirschhorn et al.7 and Munafo8).

In the current study we attempted to replicate significant findings from previous candidate gene studies in MDD. To this end we conducted a systematic review of the literature and selected those genes that were reported to be significantly associated with MDD at least once. The GWAS data from the GAIN MDD study were used to screen all the identified candidate genes in two ways: (1) for association with the specific SNPs reported in the literature; and (2) for association with any of the SNPs (genotyped or imputed) from the Perlegen chip within the identified genes.


Materials and methods

Selection of studies reporting on candidate genes

To prevent any bias from the results of the GAIN MDD study, we identified candidate genes for MDD before the results from the GAIN MDD study became available. Therefore, a so-called ‘enhanced search’ was performed in Medline through PubMed on 1 September 2007 using the following search terms: ((‘genes’ (TIAB) NOT Medline(SB])) OR ‘genes’ (MeSH terms) OR gene (text word)) OR (‘genes’(MeSH terms) OR genes (text word)) OR snp (all fields) OR (‘single nucleotide polymorphism’ (text word OR ‘polymorphism, single nucleotide’ (MeSH terms) OR snps (text word)) OR (‘genetic polymorphism’ (text word) OR ‘polymorphism, genetic’ (MeSH terms) OR polymorphism (text word)) OR polymorphisms (all fields) AND (‘major depressive disorder’ (text word) OR ‘depressive disorder, major’ (MeSH Terms) OR major depression (text word)) AND ‘humans’ (MeSH terms). This resulted in 641 hits. We additionally scrutinized the reference list of the meta-analysis of genetic studies on MDD by Lopez-Leon et al.9 that appeared online on 16 October 2007, shortly after the end date of our search, resulting in an additional 39 hits of possibly relevant papers. These researchers used somewhat broader search terms than we did, and their search ran until June 2007; therefore, as a final check, we searched the literature using their search terms for the period June 2007 to September 2007 not covered in their paper. This yielded an additional 110 hits. Of all these papers we retrieved the abstracts, and if considered relevant, the full paper.

In the next step, we selected all papers fulfilling the following five inclusion criteria: (1) The study had to be a candidate gene case–control association study. Linkage and fine-mapping studies were excluded. (2) The primary diagnosis of the patients in the candidate gene study had to be MDD to enable comparison with the GAIN MDD study. Therefore, we excluded studies involving patients: (i) with a depressive episode in the course of bipolar disorder, (ii) with a primary psychotic disorder such as schizophrenia and a secondary depression, (iii) with a seasonal affective disorder not being MDD with a seasonal pattern, (iv) with a primary anxiety disorder (that is, panic disorder, agoraphobia or social phobia) or obsessive compulsive disorder and a secondary depression and (v) with MDD plus an additional specific comorbid disorder or condition, for example, MDD plus alcoholism. However, we allowed subgroups within MDD, for example, MDD in women or in men, recurrent MDD or early-onset MDD. (3) The sample of a specific study consisted of at least 30 patients with MDD and 30 healthy controls. (4) The findings on the association with MDD of any variant within the candidate gene (either a SNP, a microsatellite marker or a haplotype) had reached a statistical significance at the level of P<0.05. (5) Finally, the genetic association had to be with the diagnosis of MDD and not with other aspects such as associated personality features (for example, neuroticism) or factors related to treatment response. This resulted in 78 papers.

In order not to miss potential true-positive findings, we did not exclude candidate genes with associated markers that had low allele frequencies or that deviated from Hardy–Weinberg equilibrium. Two investigators (FB and CH) independently made a selection from the initial list of papers using the above-mentioned criteria. When both reached consensus, the paper was included or excluded; in case of disagreement, consensus was obtained with assistance of two other authors (WN and HS). As a final step, one author (IN) double checked whether all selected papers fulfilled the aforementioned selection criteria 1–5. Figure 1 shows a flowchart of how we retrieved the 78 papers for the present study.

Figure 1.
Figure 1 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Selection procedure of studies reporting on candidate genes. Abbreviations: LL, Lopez-Leon; PM, PubMed.

Full figure and legend (134K)

Bioinformatic tools

For many SNPs no reference SNP identification number (rs-id) was given in the original papers, but codes based on position (for example, 677C/T in MTHFR or Tyr129Ser in HTR3B) or even own codes (for example, s1-s5 in AVPR1B) were given. To retrieve rs-ids for these SNPs, we used searches in PubMed or in the SNP database of the National Center for Biotechnology Information (NCBI) (http://www.ncbi.nlm.nih.gov/SNP/) using the ‘Geneview’ option in conjunction with NCBI's MapViewer with the human genome assembly build 37.1 (http://www.ncbi.nlm.nih.gov/projects/mapview/). Six SNPs remained that could not be easily found in this way. We contacted the corresponding authors of the papers and used NCBI's Primer-Blast and SNP-Blast (http://blast.ncbi.nlm.nih.gov/Blast.cgi) to map these SNPs using the provided primer sequences.


The 1862 MDD cases included in the GAIN MDD study were mainly from NESDA, a longitudinal cohort study designed to be representative of individuals with depressive and/or anxiety disorders10 and were recruited from mental health-care organizations (N=785), primary care (N=603) and community samples (N=314). Additional cases came from the NTR (N=160). Regardless of recruitment setting, similar inclusion and exclusion criteria were used to select MDD cases. Inclusion criteria were a lifetime diagnosis of MDD according to DSM-IV (Diagnostic and Statistical Manual, Fourth Edition)11 as diagnosed through the Composite International Diagnostic Interview (CIDI Version 2.1.12), age 18–65 years and self-reported western European ancestry. People who were not fluent in Dutch and those with a primary diagnosis of a psychotic disorder, obsessive compulsive disorder, bipolar disorder or severe alcohol or substance use disorder were excluded.

Most of the 1860 control subjects were from the NTR (N=1,703) and additional controls from NESDA (N=157). Longitudinal phenotyping in NTR included assessment of depressive symptoms (through multiple instruments), anxiety and neuroticism. Inclusion required no report of MDD at any measurement occasion and never scoring high (>0.5 s.d.) on a general factor score based on combined measures of neuroticism, anxiety and depressive symptoms. Finally, controls and their parents were required to have been born in the Netherlands or Western Europe. Only one control per family was selected. NESDA controls had no lifetime diagnosis of MDD or an anxiety disorder as assessed by the CIDI and reported low depressive symptoms at baseline. For more details, see Boomsma et al.2


Perlegen Sciences performed all genotyping according to strict standard operating procedures. DNA samples from cases and controls were randomly assigned to plates, shipped to Perlegen and identified only by barcode. High-density oligonucleotide arrays were used yielding 599164 SNPs. Eight SNPs with duplicate numbers were deleted and 73 mitochondrial SNPs were removed for later analysis. From the remaining 599083 SNPs on the Perlegen chip, 435291 passed quality control tests. A total of 280 subjects were excluded because of various quality control issues, leaving 1738 cases (93.4%) and 1802 controls (96.9%) in the final analysis data set. For more details see Boomsma et al.2 and Sullivan et al.4


Not all SNPs selected from the literature were present on the genotyping array. On the basis of the linkage disequilibrium (LD) structure between SNPs we followed an imputation procedure to predict non-genotyped SNPs using the HapMap CEU data (release 22, build 36) as the reference database. The imputation was performed by IMPUTE version 0.3.2 using the default settings and the recommended number 11418 for the effective population size of Caucasians.13 In this way we extended the genome-wide autosomal SNP data set from 427049 to 2467430 SNPs. For our candidate genes this meant an extension from 851 to 4955 SNPs. However, the quality of the imputation was low for 85 SNPs (SNPTEST proper_info <0.5). These SNPs were excluded leaving 4870 SNPs for analysis. No SNP had a minor allele frequency of <1%.

Association test

The association between MDD and the autosomal SNP data was tested using a frequentist case–control test provided in the software package SNPTEST version 1.1.4 using allele dosages with sex as a covariate to adjust for the slight imbalance in the percentage of females between cases (69.6%) and controls (62%),2, 4 and the ‘proper’ option to account for the uncertainty of the genotypes that were imputed.13 In addition, 7988 genotyped SNPs on the X chromosome were analyzed in PLINK version 1.0314 using logistic regression with sex as a covariate. SNPs on the Y chromosome (n=15) and SNPs mapped to ambiguous locations (n=239) were not analyzed.

Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for the significant candidate SNPs according to the disease model from the original article using counted or estimated numbers for genotyped or imputed SNPs, respectively. To establish true replication, we checked whether the effect was for the same allele and in the same direction.

Permutation procedure

To facilitate interpretation of the significance of our findings for each SNP, we calculated three P-values by permutation: (1) a crude uncorrected significance, (2) a gene-wide significance (corrected for all SNPs in the gene) and (3) an overall significance (corrected for all SNPs in all selected genes). The crude uncorrected P-value was determined as the fraction of permutations that yielded a P-value that was smaller than the observed one. This P-value was determined to validate the permutation procedure as it should be similar to the P-value calculated by SNPTEST or PLINK. The gene-wide significance of a particular SNP was defined as the fraction of permutations that the P-value of any SNP in the candidate gene concerned was smaller than the one observed for that SNP. This procedure corrects the P-value for testing multiple SNPs within a gene. On the basis of the rationale that each selected gene is a candidate for MDD (hypothesis driven), no correction for all SNPs in the entire gene set is necessary. Nonetheless, we also calculated an overall significance (that is, corrected for all SNPs in all candidate genes) by determining an overall P-value for each SNP as the fraction of permutations for which any of the SNPs in any of the candidate genes had a P-value smaller than the observed one. For each of the three P-values we conducted 10000 permutations. Case–control status was randomly permuted within males and females separately, hence leaving the number of affected males and females intact.

In addition to the three SNP-specific P-values, we also determined whether the number of SNPs within a candidate gene with an original P-value of <0.05 based on SNPTEST or PLINK (see above) was higher than expected. For each permutation the number of SNPs in each candidate gene that was significant at P<0.05 was recorded. The fraction of permutations with a higher number of significant SNPs than originally observed determined the significance of the number of significant SNPs of that candidate gene. By permuting case–control status the LD structure of the genes is preserved; that is, the resulting significance is corrected for possible high correlation between SNPs. The number of permutations for this test was 10000 as well.



Literature search

The 78 papers that resulted from our systematic literature search reported 115 statistically significant differences between MDD cases and healthy controls in 57 genes: for 96 SNPs, 7 microsatellite markers (that is, length polymorphisms), 11 haplotypes and one protein polymorphism in the haptoglobin gene15 (Table 1). Twenty-nine SNPs were identified by an rs-id, whereas 67 SNPs were only specified by a location code, restriction enzyme or author-designed code. For 64 of these 67 SNPs, rs-ids could be obtained, whereas two rs-ids were not found (246G/A in CCKAR16 and –7054C/A in DRD216). In addition, for SNP 1463G/A in TPH2, others have tried to replicate the association, but the SNP seemed to be nonexistent.17, 18 Thus, we could map 93 SNPs.

For the seven microsatellite markers, no LD data with SNPs are known and hence we cannot determine whether these are covered by any of the available SNPs in the corresponding genes. Among them is the 5-HTTLPR 44-bp deletion in the promoter region of the serotonin transporter gene SLC6A4 (SERT), which has attracted considerable attention in various previous candidate gene studies for MDD.

Association in the GAIN MDD data with specific SNPs from the literature

Of the 93 selected SNPs in the 57 candidate genes, 61 were either present (n=18) on the Perlegen array or could be imputed (n=43). Four additional SNPs could be tagged by other available SNPs (Table 2). Two of these were not genotyped in the Centre d'Etude du Polymorphisme Humain, Utah (CEU) population of the HapMap Phase2 (release 22, build 36) project, but were available in the Japanese individuals from Tokyo (JPT) and Chinese Han individuals from Beijing (CHB) populations, in which they showed high LD (r2=0.97 and 1.00, respectively) with at least one other available SNP. The two other SNPs were genotyped in the CEU population and could be tagged by available SNPs with r2=1, but were for unreported reasons not included in the HapMap reference file used for the imputation procedure as provided on the website of the imputation software package IMPUTE (https://mathgen.stats.ox.ac.uk/impute/impute_v0.5.html). A total of 28 SNPs were neither genotyped directly nor imputed nor could be tagged, and hence were not available for replication.

We investigated which of the 65 available or tagged SNPs could be confirmed in our data (Table 2). SNPs rs12520799 in C5orf20 (DCNP1) (P=0.038; OR AT=1.10, 95% CI 0.95–1.29; OR TT=1.21, 95% CI 1.01–1.47) and rs16139 in NPY (P=0.034; OR C allele=0.73, 95% CI 0.55–0.97) replicated the reported effect in the literature. In addition, three SNPs in the ACE gene selected from the literature, which were in strong LD with each other (r2=0.78–1.00), were also significant in the GAIN MDD sample (rs4333: P=0.029; rs4329: P=0.030; rs4461142: P=0.036). However these three SNPs showed effects in the opposite direction compared with previous results19 (TC Baghai, personal communication). Hence, these SNPs do not represent a true replication.

Candidate genes from the literature in the GAIN MDD study

As in different populations different SNPs might have a role, we also studied all SNPs present on the genotyping array or available through imputation in an area from 5-kb upstream to 5-kb downstream of each selected candidate gene to cover the promoter and 3′ untranslated region, respectively. Of the 57 genes, 49 were covered by one or more SNPs that were present on the Perlegen chip. For another six genes no genotyped SNPs were available but imputed ones were. Neither genotyped nor imputed SNPs were available for two genes (AVPR1B and CHRFAM7A). In total, 4870 SNPs, of which 820 were genotyped and 4019 were imputed, covered the 55 candidate genes ranging from 1 SNP for AR to 642 SNPs for PDE11A (Figure 2). We noted a significant correlation of 0.75 between gene size and the number of genotyped SNPs (including genes on the X chromosome) and a correlation of 0.96 between gene size and total number of SNPs (excluding genes on the X chromosome, as SNPs on the X chromosome were not imputed).

Figure 2.
Figure 2 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Coverage by genotyped (black bar) and imputed (grey bar) SNPs and size (white circles) of the 57 selected candidate genes (above: genes with less than or equal to50 SNPs; below: genes with >50 SNPs).

Full figure and legend (140K)

For 28 of the 55 genes, one or more SNPs were found to be different between MDD cases and healthy controls in the GAIN MDD GWA scan at a significance level of P<0.05 (Table 3). The remaining 27 genes were not associated with MDD as none of the SNPs reached P<0.05. The smallest P-value was observed for SNP rs769178 in the TNF gene region (P=0.00029; OR T allele =1.35, 95% CI 1.13–1.63). The minor allele T was observed in 8.0% of the MDD cases and 6.1% of the controls.

With genes covered by a large number of SNP markers, the expected minimal P-value will decrease purely as a result of chance alone as a function of the number of SNPs. Thus, we used gene-wide P-values from the permutation procedure that corrects for the number of SNPs within a candidate gene and only SNP rs769178 in TNF remained significant (P=0.0034; Figure 3 and Table 3). The second strongest associated SNP, which was observed in DISC1 (rs7533169: P-value=0.0025), became nonsignificant after this correction (gene-wide P-value=0.28), because in this gene there were 491 SNPs and apparently the small P-value was observed purely based on chance. In addition to the gene-wide significance, we also determined overall significance per SNP accounting for testing 4870 SNPs in 55 candidate genes. In that case none of the SNPs remained significant, not even the TNF SNP rs769178 (overall P=0.33).

Figure 3.
Figure 3 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Observed minimal P-value versus the expected minimal P-value based on the number of SNPs in the gene.

Full figure and legend (61K)

The significance levels per SNP are one way of testing the true value of the selected candidate genes. If many SNPs in a candidate gene are associated, this could also indicate potential involvement of a gene in the disorder under study.20 We noticed, for instance, that in the norepinephrine transporter SLC6A2 (NET), 19 of the 100 SNPs were significant at a P level of <0.05 (Table 3). The permutation procedure that tested whether the number of SNPs that were significant at 0.05 was larger than expected, revealed that this finding for SLC6A2 was indeed significant (P=0.039). Figure 4 shows that the 19 SNPs are scattered in the right half of the gene; 17 of them lie in three independent haplotype blocks and the other 2 SNPs are not in LD with the haplotype blocks or with each other. For none of the other candidate genes this permutation test revealed a significant result.

Figure 4.
Figure 4 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Association of SLC6A2 (NET) with MDD. Along the x axis the location of all 100 SNPs within 5kb of SLC6A2 is shown and the –log10(P-value) is on the y axis. A total of 19 SNPs have a P-value of <0.05, which is significantly more than expected (P=0.039). Of these, 17 lie in three haplotype blocks (squares–upward triangles (light grey) and downward triangles (dark grey)) (r2 values HapMap CEU: squares–upward triangles: 0.020; squares–downward triangles: 0.661; and upward triangles–downward triangles: 0.013). The other two (black diamonds) SNPs are not in LD with the haplotype blocks or with each other. The light grey line shows the recombination rate in this area (axis on the right).

Full figure and legend (59K)



Several genome-wide linkage studies of MDD have been published (reviewed in Boomsma et al.2), but the GAIN MDD study is among the first GWAS in MDD.4, 21 We used these GWAS data as a large-scale replication of previously reported significant findings from candidate gene studies in MDD. To this end, we first conducted a systematic review of the literature and identified a total of 57 genes for which a significant association with MDD has been reported at least once. Fifty-five of these genes could be included in our replication study, with either genotyped or imputed SNPs available from the GWAS data. With a sample size (1738 cases and 1802 controls) by far exceeding all previous candidate gene studies (with a mean sample size of 164 MDD cases and 252 controls, and only 1 study with >1000 MDD cases and controls),22 we found minimal support for involvement in MDD for only 4 out of 55 candidate genes: C5orf20 (P=0.038), NPY (P=0.034), TNF (P=0.0034) and SLC6A2 (P=0.039). Replication of these genes was based on three different statistical approaches. First, the involvement of C5orf20 (rs12520799) and NPY (rs16139) constituted a direct replication at SNP level of previously identified SNPs associated with MDD. Second, studying the selected candidate genes for all SNPs present on the genotyping array, TNF (rs76917) was identified as the only gene for which the association with MDD remained gene-wide significant. Third, the potential involvement of SLC6A2 (NET) in MDD was derived on the basis of a statistically significant number of SNPs (that is, 19 out of 100) associated with MDD within this gene, which could not be explained solely by high LD between the SNPs. Note that the previous evidence in the selected studies for the involvement of C5orf20, NPY, TNF and SLC6A2 in MDD did not stand out as particularly strong; that is, ORs were mostly in the moderate range (minimum OR=0.33 and maximum OR=2.41), and none of the P-values in these studies was smaller than P<0.001, not even for the largest study on C5orf20 with a total sample size of 864. NPY was the one exception in which the allele was present in 6.3% of the controls but not in the patients, thus suggesting a strong effect (OR=0). However, this was based on only 51 patients.23, 24, 25, 26

An important question is why so few candidate genes were replicated by our GWAS, whereas the sample size of our study was so much larger than any of the 78 selected candidate gene studies. One possible—and in our opinion most likely—explanation is publication bias; positive results have a better chance of being published than negative results. This would also imply that many previously reported positive findings were actually false-positive findings (type 1 errors), probably amplified by insufficient correction for multiple testing. Testing this, for example, with funnel plots, is not appropriate for the approach followed here as our literature search strategy did not include negative candidate gene studies. Second, given that MDD is a rather heterogeneous disorder, and that it was diagnosed with different instruments across studies, previous and current study samples may have differed phenotypically (see Hek et al.27 for a discussion of this point). A third explanation is that associations between genes and etiologically complex diseases may depend on genetic (gene × gene interactions or epistasis) and environmental (gene × environment interactions) contexts, which may differ in samples from different populations.28 Thus, previous and current samples may have been different genetically or in their contextual factors.

Given the modest support for the replicated candidate genes (one P-value <0.01 and three P-values <0.05) it is possible that even the four replications of our analysis are false-positive results. With the number of SNPs (n=65) been tested and under the null hypothesis of no true associations of any of the candidate SNPs in previous studies, one would expect three significant findings. Thus, it is possible that our two significant findings in C5orf20 and NPY are false-positive results. In addition, under a similar null hypothesis for the 55 candidate genes, two significant findings were expected, implying that the single gene-wide significant result of TNF might also be a false-positive result. This is supported by the lack of overall significance for the TNF SNP (P=0.33). And finally, the significant finding for SLC6A2, the only gene showing a larger number of significant SNPs than expected, might also be a chance finding.

On the other hand, the above calculations may be too conservative when assuming that at least some of the previously found associations of candidate genes with MDD were true findings. In that case, our approach of replicating candidate genes is more or less hypothesis driven, thus not requiring the same multiple testing penalty as the genome-wide approach.4 Nonetheless, the few replications in our study are rather sobering and to uncover whether the replicated SNPs and genes are truly associated to MDD, confirmation in independent samples is crucial. As such, meta-analytical results from the Psychiatric GWAS Consortium are also eagerly awaited.29

From the most recent meta-analysis of genetic studies on MDD, Lopez-Leon et al.9 concluded that statistically significant evidence exists for six MDD susceptibility genes, that is, APOE, DRD4, GNB3, MTHFR, SLC6A3 and SLC6A4. Our study offers little support for these genes. Given our sample size, we had 80% power to detect ORs of >1.15 for allele frequencies >5% under an additive disease model. All of the above genes meet these criteria. The association of GNB3 and MTHFR with MDD was directly tested but could not be replicated in the GAIN MDD sample, although the effect of MTHFR showed a trend in the expected direction (P=0.074; OR TT versus CC=1.14). In addition, APOE, DRD4 and SLC6A4 were not associated with MDD in our study, but the previously identified genetic variants were length polymorphisms instead of SNPs. Hence, these could not directly be tested and as it is unsure whether these variants are tagged by the SNPs in the corresponding genes, we cannot refute the associations. SLC6A3 was not identified in our literature search as a candidate gene for MDD, because individual studies did not report significant effects for this gene30, 31 and it only reached significance in the pooled meta-analysis. We also tested for significance in our sample post hoc and did not find any association (59 SNPs, most significant SNP: rs27072, P=0.096).

In the context of the non-replication of many of the selected candidate genes as discussed earlier, two limitations of our study need to be noted. First, we did not have direct or indirect information on one-third of the candidate SNPs, as these were not present on the genotyping chip and could not be imputed using the HapMap CEU data. Second, we were unable to test candidate length polymorphisms previously associated with MDD. Among these was the 44-bp insertion/deletion polymorphism (or 5-HTTLPR) in the promoter region of the serotonin transporter gene SLC6A4 (SERT). Length polymorphisms are often difficult to tag with single SNPs because LD information between SNPs and length polymorphisms is either unavailable or LD with SNPs is insufficiently strong.32 However, a recent study by Wray et al.33 identified a two-SNP haplotype proxy for 5-HTTLPR.

In conclusion, the GWAS data of the GAIN MDD study failed to replicate all but four of the previously reported candidate gene associations with MDD. However, given the number of candidate SNPs and genes that were tested, even these significant may well be false-positives, implying that we found no replication at all. This might point to publication bias and false-positive findings in previous studies, and also to heterogeneity of the MDD phenotype as well as variations in contextual genetic or environmental factors.


Conflict of interest

The authors declare no conflict of interest.



  1. Manolio TA, Rodriguez LL, Brooks L, Abecasis G, Ballinger D, Daly M et al. New models of collaboration in genome-wide association studies: the Genetic Association Information Network. Nat Genet 2007; 39: 1045–1051. | Article | PubMed | ISI | ChemPort |
  2. Boomsma DI, Willemsen G, Sullivan PF, Heutink P, Meijer P, Sondervan D et al. Genome-wide association of major depression: description of samples for the GAIN Major Depressive Disorder Study: NTR and NESDA biobank projects. Eur J Hum Genet 2008; 16: 335–342. | Article | PubMed | ISI | ChemPort |
  3. The Wellcome Trust Case-Control Consortium. Genome-wide association study of 14,000 cases of seven common diseases and 3,000 shared controls. Nature 2007; 447: 661–678. | Article | PubMed | ISI | ChemPort |
  4. Sullivan PF, de Geus EJ, Willemsen G, James MR, Smit JH, Zandbelt T et al. Genome-wide association for major depressive disorder: a possible role for the presynaptic protein piccolo. Mol Psychiatry 2009; 14: 359–375. | Article | PubMed | ISI | ChemPort |
  5. Arking DE, Pfeufer A, Post W, Kao WH, Newton-Cheh C, Ikeda M et al. A common genetic variant in the NOS1 regulator NOS1AP modulates cardiac repolarization. Nat Genet 2006; 38: 644–651. | Article | PubMed | ISI | ChemPort |
  6. Roeder K, Bacanu SA, Wasserman L, Devlin B. Using linkage genome scans to improve power of association in genome scans. Am J Hum Genet 2006; 78: 243–252. | Article | PubMed | ISI | ChemPort |
  7. Hirschhorn JN, Lohmueller K, Byrne E, Hirschhorn K. A comprehensive review of genetic association studies. Genet Med 2002; 4: 45–61. | Article | PubMed | ISI | ChemPort |
  8. Munafo MR. Candidate gene studies in the 21st century: meta-analysis, mediation, moderation. Genes Brain Behav 2006; 5(Suppl 1): 3–8. | Article | PubMed | ISI | ChemPort |
  9. Lopez-Leon S, Janssens AC, Gonzalez-Zuloeta Ladd AM, Del-Favero J, Claes SJ, Oostra BA et al. Meta-analyses of genetic studies on major depressive disorder. Mol Psychiatry 2008; 13: 772–785. | Article | PubMed | ISI | ChemPort |
  10. Penninx BW, Beekman AT, Smit JH, Zitman FG, Nolen W, Spinhoven P et al. The Netherlands Study of Depression and Anxiety (NESDA): rationale, objectives and methods. Int J Methods Psychiatr Res 2008; 17: 121–140. | Article | PubMed | ISI |
  11. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. American Psychiatric Association: Washington, DC, 1994.
  12. World Health Organization. Composite International Diagnostic Interview (CIDI), Version 2.1.. World Health Organization: Geneva, Switzerland, 1997.
  13. Marchini J, Howie B, Myers S, McVean G, Donnelly P. A new multipoint method for genome-wide association studies by imputation of genotypes. Nat Genet 2007; 39: 906–913. | Article | PubMed | ISI | ChemPort |
  14. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MA, Bender D et al. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am J Hum Genet 2007; 81: 559–575. | Article | PubMed | ISI | ChemPort |
  15. Maes M, Delanghe J, Scharpe S, Meltzer HY, Cosyns P, Suy E et al. Haptoglobin phenotypes and gene frequencies in unipolar major depression. Am J Psychiatry 1994; 151: 112–116. | PubMed | ISI | ChemPort |
  16. Koks S, Nikopensius T, Koido K, Maron E, Altmae S, Heinaste E et al. Analysis of SNP profiles in patients with major depressive disorder. Int J Neuropsychopharmacol 2006; 9: 167–174. | Article | PubMed | ISI | ChemPort |
  17. Zhang X, Gainetdinov RR, Beaulieu JM, Sotnikova TD, Burch LH, Williams RB et al. Loss-of-function mutation in tryptophan hydroxylase-2 identified in unipolar major depression. Neuron 2005; 45: 11–16. | Article | PubMed | ISI | ChemPort |
  18. Blakely RD. Overview: a rare opportunity or just one less reason to be depressed. Neuron 2005; 48: 701–702; author reply 705–706. | Article | PubMed | ISI | ChemPort |
  19. Baghai TC, Binder EB, Schule C, Salyakina D, Eser D, Lucae S et al. Polymorphisms in the angiotensin-converting enzyme gene are associated with unipolar depression, ACE activity and hypercortisolism. Mol Psychiatry 2006; 11: 1003–1015. | Article | PubMed | ISI | ChemPort |
  20. Arking DE, Khera A, Xing C, Kao WH, Post W, Boerwinkle E et al. Multiple independent genetic factors at NOS1AP modulate the QT interval in a multi-ethnic population. PLoS ONE 2009; 4: e4333. | Article | PubMed | ChemPort |
  21. Muglia P, Tozzi F, Galwey NW, Francks C, Upmanyu R, Kong XQ et al. Genome-wide association study of recurrent major depressive disorder in two European case-control cohorts. Mol Psychiatry; advance online publication, 23 December 2008; doi:10.1038/mp.2008.131. | Article | PubMed | ChemPort |
  22. Lucae S, Salyakina D, Barden N, Harvey M, Gagne B, Labbe M et al. P2RX7, a gene coding for a purinergic ligand-gated ion channel, is associated with major depressive disorder. Hum Mol Genet 2006; 15: 2438–2445. | Article | PubMed | ISI | ChemPort |
  23. Willis-Owen SA, Shifman S, Copley RR, Flint J. DCNP1: a novel candidate gene for major depression. Mol Psychiatry 2006; 11: 121–122. | Article | PubMed | ISI | ChemPort |
  24. Heilig M, Zachrisson O, Thorsell A, Ehnvall A, Mottagui-Tabar S, Sjogren M et al. Decreased cerebrospinal fluid neuropeptide Y (NPY) in patients with treatment refractory unipolar major depression: preliminary evidence for association with preproNPY gene polymorphism. J Psychiatr Res 2004; 38: 113–121. | Article | PubMed | ISI |
  25. Inoue K, Itoh K, Yoshida K, Shimizu T, Suzuki T. Positive association between T-182C polymorphism in the norepinephrine transporter gene and susceptibility to major depressive disorder in a Japanese population. Neuropsychobiology 2004; 50: 301–304. | Article | PubMed | ISI | ChemPort |
  26. Jun TY, Pae CU, Hoon H, Chae JH, Bahk WM, Kim KS et al. Possible association between -G308A tumour necrosis factor-alpha gene polymorphism and major depressive disorder in the Korean population. Psychiatr Genet 2003; 13: 179–181. | Article | PubMed | ISI |
  27. Hek K, Mulder CL, Luijendijk HJ, van Duijn CM, Hofman A, Uitterlinden AG et al. The PCLO gene and depressive disorders: replication in a population-based study. Hum Mol Genet 2009; 19: 731–734. | Article | PubMed | ISI | ChemPort |
  28. Martinez FD. Gene-environment interactions in asthma: with apologies to William of Ockham. Proc Am Thorac Soc 2007; 4: 26–31. | Article | PubMed |
  29. Psychiatric GWAS Consortium Steering Committee. A framework for interpreting genome-wide association studies of psychiatric disorders. Mol Psychiatry 2009; 14: 10–17. | Article | ISI |
  30. Frisch A, Postilnick D, Rockah R, Michaelovsky E, Postilnick S, Birman E et al. Association of unipolar major depressive disorder with genes of the serotonergic and dopaminergic pathways. Mol Psychiatry 1999; 4: 389–392. | Article | PubMed | ISI | ChemPort |
  31. Manki H, Kanba S, Muramatsu T, Higuchi S, Suzuki E, Matsushita S et al. Dopamine D2, D3 and D4 receptor and transporter gene polymorphisms and mood disorders. J Affect Disord 1996; 40: 7–13. | Article | PubMed | ISI | ChemPort |
  32. Su S, Zhao J, Bremner JD, Miller AH, Tang W, Bouzyk M et al. Serotonin transporter gene, depressive symptoms and interleukin-6. Circ Cardiovasc Genet 2009; 2: 614–620. | Article | PubMed | ISI | ChemPort |
  33. Wray NR, James MR, Gordon SD, Dumenil T, Ryan L, Coventry WL et al. Accurate, large-scale genotyping of 5HTTLPR and flanking single nucleotide polymorphisms in an association study of depression, anxiety, and personality measures. Biol Psychiatry 2009; 66: 468–476. | Article | PubMed | ISI | ChemPort |
  34. Arinami T, Yamada N, Yamakawa-Kobayashi K, Hamaguchi H, Toru M. Methylenetetrahydrofolate reductase variant and schizophrenia/depression. Am J Med Genet 1997; 74: 526–528. | Article | PubMed | ISI | ChemPort |
  35. Kelly CB, McDonnell AP, Johnston TG, Mulholland C, Cooper SJ, McMaster D et al. The MTHFR C677 T polymorphism is associated with depressive episodes in patients from Northern Ireland. J Psychopharmacol 2004; 18: 567–571. | Article | PubMed | ISI | ChemPort |
  36. Lemonde S, Turecki G, Bakish D, Du L, Hrdina PD, Bown CD et al. Impaired repression at a 5-hydroxytryptamine 1A receptor gene polymorphism associated with major depression and suicide. J Neurosci 2003; 23: 8788–8799. | PubMed | ISI | ChemPort |
  37. Choi MJ, Kang RH, Ham BJ, Jeong HY, Lee MS. Serotonin receptor 2A gene polymorphism (-1438A/G) and short-term treatment response to citalopram. Neuropsychobiology 2005; 52: 155–162. | Article | PubMed | ISI | ChemPort |
  38. Choi MJ, Lee HJ, Lee HJ, Ham BJ, Cha JH, Ryu SH et al. Association between major depressive disorder and the -1438A/G polymorphism of the serotonin 2A receptor gene. Neuropsychobiology 2004; 49: 38–41. | Article | PubMed | ISI | ChemPort |
  39. Du L, Bakish D, Lapierre YD, Ravindran AV, Hrdina PD. Association of polymorphism of serotonin 2A receptor gene with suicidal ideation in major depressive disorder. Am J Med Genet 2000; 96: 56–60. | Article | PubMed | ISI | ChemPort |
  40. Lerer B, Macciardi F, Segman RH, Adolfsson R, Blackwood D, Blairy S et al. Variability of 5-HT2C receptor cys23ser polymorphism among European populations and vulnerability to affective disorder. Mol Psychiatry 2001; 6: 579–585. | Article | PubMed | ISI | ChemPort |
  41. Yamada K, Hattori E, Iwayama Y, Ohnishi T, Ohba H, Toyota T et al. Distinguishable haplotype blocks in the HTR3A and HTR3B region in the Japanese reveal evidence of association of HTR3B with female major depression. Biol Psychiatry 2006; 60: 192–201. | Article | PubMed | ISI | ChemPort |
  42. Mendlewicz J, Oswald P, Claes S, Massat I, Souery D, Van Broeckhoven C et al. Patient-control association study of substance P-related genes in unipolar and bipolar affective disorders. Int J Neuropsychopharmacol 2005; 8: 505–513. | Article | PubMed | ISI | ChemPort |
  43. Covault J, Pettinati H, Moak D, Mueller T, Kranzler HR. Association of a long-chain fatty acid-CoA ligase 4 gene polymorphism with depression and with enhanced niacin-induced dermal erythema. Am J Med Genet B Neuropsychiatr Genet 2004; 127B: 42–47.
  44. Su QR, Su LY, Su HR, Chen Q, Ren GY, Yin Y et al. Polymorphisms of androgen receptor gene in childhood and adolescent males with first-onset major depressive disorder and association with related symptomatology. Int J Neurosci 2007; 117: 903–917. | Article | PubMed | ISI | ChemPort |
  45. Saab YB, Gard PR, Yeoman MS, Mfarrej B, El-Moalem H, Ingram MJ. Renin-angiotensin-system gene polymorphisms and depression. Prog Neuropsychopharmacol Biol Psychiatry 2007; 31: 1113–1118. | Article | PubMed | ChemPort |
  46. Holmes C, Russ C, Kirov G, Aitchison KJ, Powell JF, Collier DA et al. Apolipoprotein E: depressive illness, depressive symptoms, and Alzheimer's disease. Biol Psychiatry 1998; 43: 159–164. | Article | PubMed | ISI | ChemPort |
  47. Fan PL, Chen CD, Kao WT, Shu BC, Lung FW. Protective effect of the apo epsilon2 allele in major depressive disorder in Taiwanese. Acta Psychiatr Scand 2006; 113: 48–53. | Article | PubMed | ISI | ChemPort |
  48. van West D, Del-Favero J, Aulchenko Y, Oswald P, Souery D, Forsgren T et al. A major SNP haplotype of the arginine vasopressin 1B receptor protects against recurrent major depression. Mol Psychiatry 2004; 9: 287–292. | Article | PubMed | ISI | ChemPort |
  49. Hwang JP, Tsai SJ, Hong CJ, Yang CH, Lirng JF, Yang YM. The Val66Met polymorphism of the brain-derived neurotrophic-factor gene is associated with geriatric depression. Neurobiol Aging 2006; 27: 1834–1837. | Article | PubMed | ISI | ChemPort |
  50. Ribeiro L, Busnello JV, Cantor RM, Whelan F, Whittaker P, Deloukas P et al. The brain-derived neurotrophic factor rs6265 (Val66Met) polymorphism and depression in Mexican-Americans. NeuroReport 2007; 18: 1291–1293. | Article | PubMed | ISI | ChemPort |
  51. Schumacher J, Jamra RA, Becker T, Ohlraun S, Klopp N, Binder EB et al. Evidence for a relationship between genetic variants at the brain-derived neurotrophic factor (BDNF) locus and major depression. Biol Psychiatry 2005; 58: 307–314. | Article | PubMed | ISI | ChemPort |
  52. Hashimoto R, Okada T, Kato T, Kosuga A, Tatsumi M, Kamijima K et al. The breakpoint cluster region gene on chromosome 22q11 is associated with bipolar disorder. Biol Psychiatry 2005; 57: 1097–1102. | Article | PubMed | ISI | ChemPort |
  53. Massat I, Souery D, Del-Favero J, Nothen M, Blackwood D, Muir W et al. Association between COMT (Val158Met) functional polymorphism and early onset in patients with major depressive disorder in a European multicenter genetic association study. Mol Psychiatry 2005; 10: 598–605. | Article | PubMed | ISI | ChemPort |
  54. Funke B, Malhotra AK, Finn CT, Plocik AM, Lake SL, Lencz T et al. COMT genetic variation confers risk for psychotic and affective disorders: a case control study. Behav Brain Funct 2005; 1: 19. | Article | PubMed |
  55. Pae CU, Yu HS, Kim TS, Lee CU, Lee SJ, Jun TY et al. Monocyte chemoattractant protein-1 (MCP1) promoter -2518 polymorphism may confer a susceptibility to major depressive disorder in the Korean population. Psychiatry Res 2004; 127: 279–281. | Article | PubMed | ISI | ChemPort |
  56. Comings DE, Wu S, Rostamkhani M, McGue M, Iacono WG, MacMurray JP. Association of the muscarinic cholinergic 2 receptor (CHRM2) gene with major depression in women. Am J Med Genet 2002; 114: 527–529. | Article | PubMed | ISI |
  57. Lai IC, Hong CJ, Tsai SJ. Association study of nicotinic-receptor variants and major depressive disorder. J Affect Disord 2001; 66: 79–82. | Article | PubMed | ISI | ChemPort |
  58. Claes S, Villafuerte S, Forsgren T, Sluijs S, Del-Favero J, Adolfsson R et al. The corticotropin-releasing hormone binding protein is associated with major depression in a population from Northern Sweden. Biol Psychiatry 2003; 54: 867–872. | Article | PubMed | ISI | ChemPort |
  59. Liu Z, Zhu F, Wang G, Xiao Z, Wang H, Tang J et al. Association of corticotropin-releasing hormone receptor1 gene SNP and haplotype with major depression. Neurosci Lett 2006; 404: 358–362. | Article | PubMed | ISI | ChemPort |
  60. Villafuerte SM, Del-Favero J, Adolfsson R, Souery D, Massat I, Mendlewicz J et al. Gene-based SNP genetic association study of the corticotropin-releasing hormone receptor-2 (CRHR2) in major depression. Am J Med Genet 2002; 114: 222–226. | Article | PubMed | ISI |
  61. Dorado P, Penas-Lledo EM, Gonzalez AP, Caceres MC, Cobaleda J, Llerena A. Increased risk for major depression associated with the short allele of the serotonin transporter promoter region (5-HTTLPR-S) and the CYP2C9*3 allele. Fundam Clin Pharmacol 2007; 21: 451–453. | Article | PubMed | ISI | ChemPort |
  62. LLerena A, Berecz R, Dorado P, Gonzalez AP, Penas LEM, De La Rubia A. CYP2C9 gene and susceptibility to major depressive disorder. Pharmacogenomics J 2003; 3: 300–302. | Article | PubMed | ISI | ChemPort |
  63. Hashimoto R, Numakawa T, Ohnishi T, Kumamaru E, Yagasaki Y, Ishimoto T et al. Impact of the DISC1 Ser704Cys polymorphism on risk for major depression, brain morphology and ERK signaling. Hum Mol Genet 2006; 15: 3024–3033. | Article | PubMed | ISI | ChemPort |
  64. Dikeos DG, Papadimitriou GN, Avramopoulos D, Karadima G, Daskalopoulou EG, Souery D et al. Association between the dopamine D3 receptor gene locus (DRD3) and unipolar affective disorder. Psychiatr Genet 1999; 9: 189–195. | Article | PubMed | ISI | ChemPort |
  65. Serretti A, Lilli R, Di Bella D, Bertelli S, Nobile M, Novelli E et al. Dopamine receptor D4 gene is not associated with major psychoses. Am J Med Genet 1999; 88: 486–491. | Article | PubMed | ISI | ChemPort |
  66. Tsai SJ, Wang YC, Hong CJ, Chiu HJ. Association study of oestrogen receptor alpha gene polymorphism and suicidal behaviours in major depressive disorder. Psychiatr Genet 2003; 13: 19–22. | Article | PubMed | ISI |
  67. Geng YG, Su QR, Su LY, Chen Q, Ren GY, Shen SQ et al. Comparison of the polymorphisms of androgen receptor gene and estrogen alpha and beta gene between adolescent females with first-onset major depressive disorder and controls. Int J Neurosci 2007; 117: 539–547. | Article | PubMed | ISI | ChemPort |
  68. Thomson PA, Wray NR, Thomson AM, Dunbar DR, Grassie MA, Condie A et al. Sex-specific association between bipolar affective disorder in women and GPR50, an X-linked orphan G protein-coupled receptor. Mol Psychiatry 2005; 10: 470–478. | Article | PubMed | ISI | ChemPort |
  69. Henkel V, Baghai TC, Eser D, Zill P, Mergl R, Zwanzger P et al. The gamma amino butyric acid (GABA) receptor alpha-3 subunit gene polymorphism in unipolar depressive disorder: a genetic association study. Am J Med Genet B Neuropsychiatr Genet 2004; 126B: 82–87.
  70. Tadokoro K, Hashimoto R, Tatsumi M, Kosuga A, Kamijima K, Kunugi H. The Gem interacting protein (GMIP) gene is associated with major depressive disorder. Neurogenetics 2005; 6: 127–133. | Article | PubMed | ISI | ChemPort |
  71. Cao MQ, Hu SY, Zhang CH, Xia DS. Study on the interrelationship between 5-HTTLPR/G-protein beta3 subunit (C825 T) polymorphisms and depressive disorder. Psychiatr Genet 2007; 17: 233–238. | Article | PubMed | ISI |
  72. Lee HJ, Cha JH, Ham BJ, Han CS, Kim YK, Lee SH et al. Association between a G-protein beta 3 subunit gene polymorphism and the symptomatology and treatment responses of major depressive disorders. Pharmacogenomics J 2004; 4: 29–33. | Article | PubMed | ISI | ChemPort |
  73. Zill P, Baghai TC, Zwanzger P, Schule C, Minov C, Riedel M et al. Evidence for an association between a G-protein beta3-gene variant with depression and response to antidepressant treatment. NeuroReport 2000; 11: 1893–1897. | Article | PubMed | ISI | ChemPort |
  74. Pae CU, Yu HS, Kim JJ, Lee CU, Lee SJ, Lee KU et al. BanI polymorphism of the cytosolic phospholipase A2 gene and mood disorders in the Korean population. Neuropsychobiology 2004; 49: 185–188. | Article | PubMed | ISI | ChemPort |
  75. Kolsch H, Ptok U, Majores M, Schmitz S, Rao ML, Maier W et al. Putative association of polymorphism in the mannose 6-phosphate receptor gene with major depression and Alzheimer's disease. Psychiatr Genet 2004; 14: 97–100. | Article | PubMed | ISI |
  76. Schulze TG, Muller DJ, Krauss H, Scherk H, Ohlraun S, Syagailo YV et al. Association between a functional polymorphism in the monoamine oxidase A gene promoter and major depressive disorder. Am J Med Genet 2000; 96: 801–803. | Article | PubMed | ISI | ChemPort |
  77. Yu YW, Tsai SJ, Hong CJ, Chen TJ, Chen MC, Yang CW. Association study of a monoamine oxidase a gene promoter polymorphism with major depressive disorder and antidepressant response. Neuropsychopharmacology 2005; 30: 1719–1723. | Article | PubMed | ISI | ChemPort |
  78. Du L, Bakish D, Ravindran A, Hrdina PD. MAO-A gene polymorphisms are associated with major depression and sleep disturbance in males. NeuroReport 2004; 15: 2097–2101. | Article | PubMed | ISI | ChemPort |
  79. Kunugi H, Hashimoto R, Yoshida M, Tatsumi M, Kamijima K. A missense polymorphism (S205 L) of the low-affinity neurotrophin receptor p75NTR gene is associated with depressive disorder and attempted suicide. Am J Med Genet B Neuropsychiatr Genet 2004; 129B: 44–46.
  80. van Rossum EF, Binder EB, Majer M, Koper JW, Ising M, Modell S et al. Polymorphisms of the glucocorticoid receptor gene and major depression. Biol Psychiatry 2006; 59: 681–688. | Article | PubMed | ISI | ChemPort |
  81. van West D, Van Den Eede F, Del-Favero J, Souery D, Norrback KF, Van Duijn C et al. Glucocorticoid receptor gene-based SNP analysis in patients with recurrent major depression. Neuropsychopharmacology 2006; 31: 620–627. | Article | PubMed | ISI | ChemPort |
  82. Wong ML, Whelan F, Deloukas P, Whittaker P, Delgado M, Cantor RM et al. Phosphodiesterase genes are associated with susceptibility to major depression and antidepressant treatment response. Proc Natl Acad Sci USA 2006; 103: 15124–15129. | Article | PubMed | ChemPort |
  83. Ryu SH, Lee SH, Lee HJ, Cha JH, Ham BJ, Han CS et al. Association between norepinephrine transporter gene polymorphism and major depression. Neuropsychobiology 2004; 49: 174–177. | Article | PubMed | ISI | ChemPort |
  84. Nobile M, Cataldo MG, Giorda R, Battaglia M, Baschirotto C, Bellina M et al. A case-control and family-based association study of the 5-HTTLPR in pediatric-onset depressive disorders. Biol Psychiatry 2004; 56: 292–295. | Article | PubMed | ISI | ChemPort |
  85. Hoefgen B, Schulze TG, Ohlraun S, von Widdern O, Hofels S, Gross M et al. The power of sample size and homogenous sampling: association between the 5-HTTLPR serotonin transporter polymorphism and major depressive disorder. Biol Psychiatry 2005; 57: 247–251. | Article | PubMed | ISI | ChemPort |
  86. Cervilla JA, Rivera M, Molina E, Torres-Gonzalez F, Bellon JA, Moreno B et al. The 5-HTTLPR s/s genotype at the serotonin transporter gene (SLC6A4) increases the risk for depression in a large cohort of primary care attendees: the PREDICT-gene study. Am J Med Genet B Neuropsychiatr Genet 2006; 141B: 912–917.
  87. Grunblatt E, Loffler C, Zehetmayer S, Jungwirth S, Tragl KH, Riederer P et al. Association study of the 5-HTTLPR polymorphism and depression in 75-year-old nondemented subjects from the Vienna Transdanube Aging (VITA) study. J Clin Psychiatry 2006; 67: 1373–1378. | Article | PubMed | ISI |
  88. Munafo MR, Clark TG, Roberts KH, Johnstone EC. Neuroticism mediates the association of the serotonin transporter gene with lifetime major depression. Neuropsychobiology 2006; 53: 1–8. | Article | PubMed | ISI | ChemPort |
  89. Mann JJ, Huang YY, Underwood MD, Kassir SA, Oppenheim S, Kelly TM et al. A serotonin transporter gene promoter polymorphism (5-HTTLPR) and prefrontal cortical binding in major depression and suicide. Arch Gen Psychiatry 2000; 57: 729–738. | Article | PubMed | ISI | ChemPort |
  90. Ogilvie AD, Battersby S, Bubb VJ, Fink G, Harmar AJ, Goodwim GM et al. Polymorphism in serotonin transporter gene associated with susceptibility to major depression. Lancet 1996; 347: 731–733. | Article | PubMed | ISI | ChemPort |
  91. Liu W, Gu N, Feng G, Li S, Bai S, Zhang J et al. Tentative association of the serotonin transporter with schizophrenia and unipolar depression but not with bipolar disorder in Han Chinese. Pharmacogenetics 1999; 9: 491–495. | PubMed | ISI | ChemPort |
  92. Domotor E, Sarosi A, Balogh G, Szekely A, Hejjas K, Sasvari-Szekely M et al. Association of neurocognitive endophenotype and STin2 polymorphism in major depressive disorder]. Neuropsychopharmacol Hung 2007; 9: 53–62. | PubMed |
  93. Bozina N, Mihaljevic-Peles A, Sagud M, Jakovljevic M, Sertic J. Serotonin transporter polymorphism in Croatian patients with major depressive disorder. Psychiatr Danub 2006; 18: 83–89. | PubMed | ChemPort |
  94. Battersby S, Ogilvie AD, Smith CA, Blackwood DH, Muir WJ, Quinn JP et al. Structure of a variable number tandem repeat of the serotonin transporter gene and association with affective disorder. Psychiatr Genet 1996; 6: 177–181. | Article | PubMed | ISI | ChemPort |
  95. Gutierrez B, Pintor L, Gasto C, Rosa A, Bertranpetit J, Vieta E et al. Variability in the serotonin transporter gene and increased risk for major depression with melancholia. Hum Genet 1998; 103: 319–322. | Article | PubMed | ISI | ChemPort |
  96. Schahab S, Heun R, Schmitz S, Maier W, Kolsch H. Association of polymorphism in the transcription factor LBP-1c/CP2/LSF gene with Alzheimer's disease and major depression. Dement Geriatr Cogn Disord 2006; 22: 95–98. | Article | PubMed | ISI | ChemPort |
  97. Tan EC, Chan AO, Tan CH, Mahendran R, Wang A, Chua HC. Case-control and linkage disequilibrium studies of the tryptophan hydroxylase gene polymorphisms and major depressive disorder. Psychiatr Genet 2003; 13: 151–154. | Article | PubMed | ISI |
  98. Gizatullin R, Zaboli G, Jonsson EG, Asberg M, Leopardi R. Haplotype analysis reveals tryptophan hydroxylase (TPH) 1 gene variants associated with major depression. Biol Psychiatry 2006; 59: 295–300. | Article | PubMed | ISI | ChemPort |
  99. Zill P, Baghai TC, Zwanzger P, Schule C, Eser D, Rupprecht R et al. SNP and haplotype analysis of a novel tryptophan hydroxylase isoform (TPH2) gene provide evidence for association with major depression. Mol Psychiatry 2004; 9: 1030–1036. | Article | PubMed | ISI | ChemPort |
  100. Zhou Z, Roy A, Lipsky R, Kuchipudi K, Zhu G, Taubman J et al. Haplotype-based linkage of tryptophan hydroxylase 2 to suicide attempt, major depression, and cerebrospinal fluid 5-hydroxyindoleacetic acid in 4 populations. Arch Gen Psychiatry 2005; 62: 1109–1118. | Article | PubMed | ISI | ChemPort |
  101. Koido K, Koks S, Nikopensius T, Maron E, Altmae S, Heinaste E et al. Polymorphisms in wolframin (WFS1) gene are possibly related to increased risk for mood disorders. Int J Neuropsychopharmacol 2005; 8: 235–244. | Article | PubMed | ISI | ChemPort |


We thank Ting Wu for making the flow diagram in Figure 1.