Ferrer et al. Translational Psychiatry (2019)9:265 https://doi.org/10.1038/s41398-019-0601-8 Translational Psychiatry ARTICLE Open Access BDNF genetic variants and methylation: effects on cognition in major depressive disorder Alex Ferrer1,2, Javier Labad2,3,4, Neus Salvat-Pujol1, Marta Barrachina5,6, Javier Costas7, Mikel Urretavizcaya1,4,8, Aida de Arriba-Arnau1, José M. Crespo1,4,8, Carles Soriano-Mas 1,4,9, Ángel Carracedo10,11,12, José M. Menchón 1,4,8 and Virginia Soria 1,4,8 Abstract Brain-derived neurotrophic factor (BDNF) gene regulation has been linked to the pathophysiology of major depressive disorder (MDD). MDD patients show cognitive deficits, and altered BDNF regulation has a relevant role in neurocognitive functions. Our goal was to explore the association between BDNF genetic and epigenetic variations with neurocognitive performance in a group of MDD patients and healthy controls considering possible modulating factors. The sample included 134 subjects, 64 MDD patients, and 70 healthy controls. Clinical data, childhood maltreatment, and neurocognitive performance were assessed in all participants. Eleven single nucleotide polymorphisms (SNPs) and two promoter regions in the BDNF gene were selected for genotype and methylation analysis. The role of interactions between BDNF genetic and epigenetic variations with MDD diagnosis, sex, and Childhood Trauma Questionnaire (CTQ) scores was also explored. We observed significant associations between neurocognitive performance and two BDNF SNPs (rs908867 and rs925946), an effect that was significantly mediated by methylation values at specific promoter I sites. We identified significant associations between neurocognitive results 1234567890():,; 1234567890():,; 1234567890():,; 1234567890():,; and methylation status as well as its interactions with MDD diagnosis, sex, and CTQ scores. Our results support the hypothesis that BDNF gene SNPs and methylation status, as well as their interactions with modulating factors, can influence cognition. Further studies are required to confirm the effect of BDNF variations and cognitive function in larger samples. Introduction phenotypes, such as the neurotrophic signaling pathway. Major depressive disorder (MDD) is a complex and The brain-derived neurotrophic factor (BDNF) gene highly prevalent psychiatric disorder with a high impact codes for a neurotrophin that is highly expressed in the on quality of life and negative effects on mood, behavior, central nervous system (CNS), mediates a variety of and cognition1. Cognitive dysfunction in MDD patients is neuroplasticity processes, and has relevant influences on a source of disability involving deficits in visual, verbal and cognition and behavior2. BDNF is produced in the CNS working memory, attention, executive function, and pro- and released to the extracellular matrix where it can cessing speed1. interact with its receptors, tropomyosin-related kinase Several neurobiological mechanisms have been involved receptor B (TrkB) and p75 neurotrophin receptor in the pathogenesis of MDD and related cognitive (p75NTR), which mediate its effects3. Blood BDNF con- centrations, that have been observed to correlate with BDNF expression levels in the brain4, are lower in MDD Correspondence: Virginia Soria (vsoria@bellvitgehospital.cat) 1 Department of Psychiatry, Bellvitge University Hospital, Bellvitge Biomedical patients than in healthy controls5. Research Institute (IDIBELL), Neurosciences Group - Psychiatry and Mental Research on the role of BDNF single-nucleotide poly- Health, Barcelona, Spain 2 morphisms (SNPs) in risk for MDD has been centered on Department of Mental Health, ParcTaulí Hospital Universitari, Institut d’Investigació i Innovació Parc Taulí (I3PT), Sabadell, Spain rs62652. This SNP, located in exon IX, results in a valine- Full list of author information is available at the end of the article. © The Author(s) 2019 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/. Ferrer et al. Translational Psychiatry (2019)9:265 Page 2 of 10 methionine change in the pro-region of the pro-BDNF methylation levels of BDNF promoter I in patients under protein, which affects BDNF regulation. Overall, incon- treatment with antidepressants35. clusive results have been found, with positive data To our knowledge, there are no previous studies showing an association between the minor allele “A” and addressing the involvement of BDNF regulation on cog- MDD6,7 and a lack of robust association in a recent meta- nitive performance in MDD. We hypothesize that specific analysis8. The inconsistency of these results might be methylation of promoter regions of the BDNF will be explained by methodological differences, heterogeneity of associated to poorer performance in neurocognitive tasks. MDD clinical subtypes, the influence of modulating Our goal was to study the association between BDNF environmental factors and/or by epigenetic mechanisms. genetic and epigenetic variation with neurocognitive For instance, previous studies suggest that the Met allele performance and to explore the influence of each specific of rs6265 moderates the relationship between stressful life CpG site within the BDNF gene promoters on cognition, events in childhood and depression9,10. In addition, taking into account potential modulating external factors rs6265 has been also associated with brain volumes11, and confounders. presence of psychosis and suicidal behavior12, tendency to chronicity13 and treatment response in melancholic Materials and methods depression14. Study sample Since BDNF is involved in processes such as neuronal The sample consisted of 64 MDD patients (72% females, survival, neurogenesis and long-term potentiation15, ani- mean age 57.1 ± 1.3 years) diagnosed according to DSM- mal studies have reported a relevant role for rs6265 in IV-TR criteria (American Psychiatric Association, 2000) motor learning and long-term memory16, which could be and 71 healthy controls (HCs) (63% females, mean age related to changes in hippocampal synaptic plasticity via 54.2 ± 1.3 years. MDD patients were recruited from the downregulation of 5-HT3a receptors in rs6265Met car- Psychiatry Department at Bellvitge University Hospital riers17. Human clinical studies have also reported sig- (Hospitalet de Llobregat, Barcelona) while HCs were nificant associations between the BDNF genotype and recruited from the same geographic region through cognitive performance18,19, although the results are often advertisements. All participants were unrelated Cauca- inconsistent and meta-analyses have been negative20. sians of Iberian Peninsula ancestry. Genome-wide association studies are beginning to The exclusion criteria were as follows: age less than identify associations of BDNF with several behavioral and 18 years, non-Caucasian ethnicity, a diagnosis of other cognitive traits21, such as smoking initiation22, educa- psychiatric disorders including substance abuse or tional attainment23, highest math class taken23, and the dependence (except nicotine), neurological disorders “worry” genetic-defined subcluster of neuroticism24. including dementia, mental retardation, severe medical DNA methylation is an epigenetic mechanism asso- conditions, electroconvulsive therapy in the previous year, ciated with gene silencing, although some evidence sug- pregnancy or puerperium, and corticosteroid treatment in gests that it may also be associated with active gene the previous 3 months. transcription25. Research on methylation patterns in The sample partially overlaps that used in previous BDNF of MDD patients has focused on promoter regions studies36,37, which explored different hypotheses. The I and IV. Although most research has found that MDD current participants agreed to provide blood samples for patients show higher methylation levels at promoter I, DNA extraction, genotyping, and methylation studies. some studies have reported an inverse association in The Clinical Research Ethics Committee (CEIC) of the severely depressed patients26. Research on methylation of Bellvitge University Hospital approved the research pro- promoter IV has also shown increased methylation levels tocol, all participants provided written informed consent in patients with depression when compared to those in after having received a full explanation of the study. healthy controls27,28. Interestingly, a higher BDNF pro- moter methylation status has been associated with suici- Clinical assessment dal ideation and suicidal attempt history in depressive All patients met DSM-IV-TR criteria for MDD and patients29. were interviewed by experienced psychiatrists using the Regarding the influence of external factors on gene Mini-International Neuropsychiatric Interview (MINI)38. regulation through DNA methylation, there is evidence HCs were required to have no history of present or past that DNA methylation of the BDNF gene is modified by psychiatric disorders and to show a score lower than 7 in early life negative stressors30,31. Moreover, differences are the 28-item Spanish adaptation of the Goldberg General observed in specific brain areas and in a sex-specific Health Questionnaire (GHQ-28). A semistructured manner32. Other factors that have been observed to interview was administered to all participants to asses influence BDNF methylation are tobacco consumption33, sociodemographic and clinical variables, substance use age34, and pharmacological treatment, with higher and treatments. Antidepressant treatment was recorded Ferrer et al. Translational Psychiatry (2019)9:265 Page 3 of 10 using the World Health Organization Anatomical Ther- disequilibrium (LD) (r2 > 0.7) with the remaining SNPs at apeutic Chemical Classification System39, which measures minor allele frequency (MAF) > 10% from the HapMap treatment as the defined daily dose (DDD), that is, the phase III European samples. Functional SNP rs6265 and assumed average maintenance dose per day for a drug SNPs most commonly associated with mental disorders or used for its main indication in adults. cognitive dysfunction from the literature (rs2030324, The 17-item Hamilton Depression Rating Scale (HDRS) rs12273363, rs908867, and rs1491850) were prioritized as was used to assess depression severity40. Treatment tagSNPs45–48. In addition, rs11030094, rs11602246, resistance in the MDD group was assessed using the and rs4923463, which were absent from the HapMap staging model as proposed by Thase and Rush41, which data, were included following Hennings, Honea, and defines 5 levels of resistance according to the number and Neves49–51. In total, 11 SNPs were selected (Fig. S1). classes of antidepressants that have failed to produce a Genotyping of the selected SNPs was performed using response. The state-trait anxiety inventory (STAI) was the MassARRAYiPLEX platform (Agena Bioscience, for- used to evaluate anxiety both in patients and HCs42. The merly Sequenom, Inc.; San Diego, California). We used Childhood Trauma Questionnaire (CTQ), which exam- mean methylation levels from the performed technical ines the exposure to five types of trauma in childhood and replicas and discarded outlying values (standard deviation has been previously validated in healthy and clinical >10%) as a quality control method. The genotyping assays populations, was administered to all participants43. Three were performed at the genotyping facilities of CeGen in types of childhood maltreatment are focused on abuse the Santiago de Compostela Node (‘Centro Nacional de (sexual, physical, and emotional) while two types of mal- Genotipado’). treatment are centered on neglect (physical and emo- tional). Each type of maltreatment is measured using five Selection of genomic regions and quantitative DNA items and each item is rated on a five-point scale. methylation analysis We selected promoters I and IV for our methylation Neuropsychological assessment study analyses since these are promoters with more The Spanish version of the Mini-mental State Exam- consistent associations with psychiatric disorders27. We ination (MMSE) was used as a screening for dementia. used two assays to cover promoter I, with Assay 1 cov- The following neuropsychological tests were administered ering the region chr11: 27744025-27744278 and Assay 2 to all participants to assess different cognitive domains: covering the region chr11: 27744414-27744653, and one (1) verbal learning and memory: Hopkins Verbal Learning assay for promoter IV, covering chr11: 27722305- Test RevisedTM (HVLT-R); (2) visual learning and mem- 27722675 (UCSC h19 assembly). The CpG sites ana- ory: Brief Visuospatial Memory Test RevisedTM (BVMT- lyzed in each region can be found in Table S1 of the R) and the Rey Complex Figure Test (RCFT), which Supplementary Material. includes copy, immediate recall, and delayed recall sub- DNA purification, bisulfite treatment, and quantitative scores; (3) working memory: Corsi Block-Tapping Test DNA methylation analysis using the MassArray platform (CBTT) and Letter-Number Span (LNS); (4) processing of SEQUENOM were performed as described52. Primers speed: Trail Making Test Part A (TMT-A), Brief Assess- were designed using MethPrimer (http://www.urogene. ment of Cognition in Schizophrenia – Symbol Coding org/ methprimer/). Primers were tagged to obtain an (BACS-SC), and category fluency (animal naming); (5) appropriate product for in vitro transcription, prevent attention/vigilance: Continuous Performance Test – abortive cycling and balance the PCR primer length Identical Pairs (CPT-IP);(6) selective attention/inter- (forward primer tag: cagtaatacgactcactatagggagaaggct, ference: Stroop test (direct subscores for words, W; colors, reverse primer tag: aggaagagag). The sequences of the C; words-colors, WC; interference); (7) reasoning and primers used for amplification are shown in Table S1 of problem solving: Neuropsychological Assessment Battery® the Supplementary Material. Mazes (NAB-Mazes); (8) executive control: Trail Making Test Part B (TMT-B). In all tests, higher scores reflected Statistical analyses better cognitive performance with the exceptions of TMT- We processed data using SPSS 19.0 (SPSS, IBM, USA) A and TMT-B, where the outcome measure is the number to perform descriptive and univariate analyses comparing of seconds needed to perform the task; thus, higher scores the MDD and HCs groups using the χ2 test for dichotomic reflected poorer cognitive performance. categorical variables and Student’s t-test for continuous variables. We previously analyzed the distribution of SNP selection and genotyping variables, and those following a skewed distribution were Blood samples were obtained close to the neu- log transformed (ln) to approximate normality, which was ropsychological assessment. The Tagger tool of Haplo- the case for the HDRS and CTQ scores and two cognitive view v.4.244 was used to select tagSNPs in linkage tests (TMT-A and TMT-B). Ferrer et al. Translational Psychiatry (2019)9:265 Page 4 of 10 Association analysis at the individual SNP level was IV of the BDNF gene) and the dependent variable (e.g. performed as implemented in PLINK 1.953 to conduct visual memory task), a positive beta coefficient would multivariate analysis based on linear regression. Statistical indicate a better cognitive performance in those indivi- significance was assessed by a permutation procedure to duals with a greater methylation whereas a negative beta estimate the significance of the best result (10,000 per- coefficient would indicate a poorer cognitive performance mutations). Three different genetic models (dominant, in those individuals with a greater methylation. recessive, and additive) were considered. These analyses The main analyses testing the hypotheses related to the represent an effective number of tests of 2.254; therefore, role of methylation of promoter I and IV of the BDNF the results were considered experiment-wise significant if gene (mean methylation values) on each cognitive test (17 p value < 0.023 after the permutation procedure. The different cognitive measures) were adjusted for multiple analyses included sex, age, years of education, MDD comparisons with the Benjamini-Hochberg procedure, diagnosis, tobacco consumption, HDRS, STAI trait sub- that allows controlling for the false discovery rate (FDR)56. score and CTQ score as covariates. The results will be adjusted for different FDR (20%, 10%, An analysis of plausible mediation through methylation and 5%) taking into account 34 comparisons (17 cognitive variables in the relationship between SNPs and neurocog- variables × 2 promoters). We further performed statistical nitive variables was performed with the PROCESS macro of analyses under an exploratory approach when studying SPSS version 21.0, developed by Hayes55. This method uses association between specific BDNF methylation CpGs and bias-corrected bootstrap confidence intervals. The results cognitive performance. For these analyses, the statistical were considered significant when the bias-corrected 95% significance level was set at p < 0.05 (bilateral), without an confidence intervals did not contain zero. adjustment for multiple comparisons, as correction for Partial correlation analyses adjusted by age, gender, and multiple testing is not strictly necessary in those analyses years of education were used to explore the relationship that are exploratory in nature57. between methylation measures and cognitive perfor- mance. We conducted a stratified analysis by diagnosis Results (HCs vs. MDD). In the MDD group, we also included Univariate analysis relevant clinical variables (HDRS, CTQ score, STAI trait Demographic and clinical variables are shown in Table 1. sub-score and antidepressant treatment). As expected, MDD patients showed higher HDRS scores Multiple linear regression analyses were carried out in than HCs. MDD patients also showed significantly higher all participants to explore the association between scores in the CTQ and STAI trait subscore. In the cog- methylation variables and neuropsychological perfor- nitive functioning analysis, MDD patients showed poorer mance, the latter being considered as dependent variables. cognitive performance than HCs on all cognitive domains Separate multiple regression analyses were performed for (Table 2). Mean methylation values at each CpG site are each methylation measure, each of which was considered shown in Table S2 of the Supplementary Material. the main independent variable. We controlled for cov- Methylation levels were higher in HCs than in MDD with ariates and potential confounders including sex, age, years significant differences in CpG 1 and 7.8.9 of promoter I of education, tobacco consumption, diagnostic group, assay 1, CpG 14 of promoter I assay 2 and CpG 11, 13 and HDRS, STAI-trait sub-score and CTQ score. Anti- 15.16.17 of promoter IV. depressant treatment was not included as a covariate since this variable presents a high collinearity with MDD Association analyses of individual SNPs diagnosis. Finally, we explored the association between All genotyped polymorphisms were in Hardy-Weinberg neuropsychological results and the interactions between equilibrium and had call rates higher than 90%. After methylation status with sex, CTQ score and MDD diag- correction for multiple testing using a 10,000 permuta- nosis. Standardized beta coefficients will be reported. tion procedure, two SNPs showed an association with Standardized beta coefficients are calculated by subtract- cognitive performance under a dominant model: ing the mean from the variable and dividing by its stan- rs908867-T (standardized β: 0.228; P value: 0.013) was dard deviation. This results in standardized variables associated with the BVMTR test outcome, and rs925946- having a mean of zero and a standard deviation of 1. The T (standardized β: 0.165; P value: 0.008) was associated standardization allows the comparison of all independent with the BACS-SC result. No other associations between variables in the equation. Standardized beta coefficients individual SNPs and cognitive performance were detec- reflect how many standard deviations a dependent vari- ted. We also tested for possible interactions between able changes, per standard deviation increase in the pre- significant SNPs with sex, CTQ and MDD diagnosis. One dictor (independent) variable. As the standardized beta significant result was found between the interaction of coefficient reflects the strength of the association between rs925946-T and MDD diagnosis with the BACS-SC an independent variable (e.g. methylation of the promoter result (MDD diagnosis [β: −0.25; p value: 0.005], Ferrer et al. Translational Psychiatry (2019)9:265 Page 5 of 10 Table 1 Demographic and clinical data of the Table 2 Neuropsychological performance in major study sample depressive disorder patients and healthy controls HCs (n = 70) MDD p value HCs (n = 70) MDD (n = 64) p value (n = 64) Cognitive domains Age (years) 54.5 (10.5) 57.1 (10.6) 0.171 Verbal learning and memory Female gender, n (%) 44 (62.9%) 46 (71.9%) 0.267 HVLT-R 23.86 (4.92) 20.37 (5.17) <0.001 Education (years) 11.6 (3.3) 9.7 (4.2) 0.004 Visual learning and memory Tobacco consumption 2 (5.3) 4.1 (8.6) 0.089 BVMT-R 22.66 (7.52) 14.72 (9.08) <0.001 (cigarettes/day) RCFT – copy 30.84 (5.34) 27.70 (7.42) 0.006 HDRS 0.7 (1.1) 12.3 (9.1) <0.001 RCFT – immediate recall 17.45 (6.36) 12.66 (6.57) <0.001 Antidepressant treatment 0 (0) 2.3 (1.3) <0.001 RCFT – delayed recall 17.76 (6.31) 12.13 (6.78) <0.001 (DDD) Working memory Treatment resistance (Thase)a – – – CBTT (non-verbal) 14.45 (3.74) 11.59 (3.76) <0.001 Stage 0 20 (31.3%) LNS (verbal) 13.72 (3.41) 10.58 (3.49) <0.001 Stage I 13 (20.3%) Processing speed Stage II 10 (15.6%) TMT-Aa (seconds) 45.71 (24.00) 65.76 (38.41) <0.001 Stage III 12 (18.8%) BACS-SC 47.73 (13.8) 33.11 (15.89) <0.001 Stage IV 9 (14.1%) Category fluency 23.74 (6.53) 20.05 (6.26) 0.001 Stage V 0 (0%) Stroop Direct W 100.90 (15.81) 92.83 (22.67) 0.021 Childhood trauma 34.3 (9.7) 38.5 (14.1) 0.043 questionnaireb Stroop Direct C 70.93 (12.17) 61.17 (14.68) <0.001 STAI trait score 15.3 (8.6) 32.3 (13.7) <0.001 Attention/vigilance CPT-IP 2.52 (0.87) 1.99 (0.88) 0.002 All variables presented in mean (SD), or n (%). Missing data: CTQ (1.5%), STAI (1%) Executive function HCs Healthy controls, MDD major depressive disorder, HDRS Hamilton Depression Rating Scale, DDD Defined Daily Doses TMT-Ba (seconds) 82.24 (48.83) 149.19 (124.23) <0.001 a Criteria for treatment resistance stages were: non-resistant or any medication trials, to date, judged to be adequate (Stage 0); failure of at least one adequate NAB-Mazes 14.80 (6.86) 8.95 (7.19) <0.001 trial of one major class of antidepressant (Stage I); failure of at least two adequate trials of at least two distinctly different classes of antidepressants Stroop Direct WC 45.06 (12.76) 36.92 (11.04) <0.001 (Stage II); Stage II resistance plus failure of an adequate trial of a tricyclic antidepressant or a first augmentation strategy (lithium or thyroid hormone) Stroop Direct Interference 3.62 (9.59) 0.38 (6.87) 0.026 (Stage III); Stage III resistance plus failure of an adequate trial of an MAOI or a second augmentation strategy (Stage IV); Stage IV resistance plus failure of an All variables presented in mean (SD), or n (%). Missing data differed for cognitive adequate course of bilateral electroconvulsive therapy (Stage V) tests: HVLT-R (0.7%), BVMT-R (2.2%), LNS (4.4%), TMT A (1.5%), BACS-SC (1.5%), Category fluency (0.7%), Stroop Direct W and C (0.7), CPT-IP (13.4%), TMT-B (4.5%), NAB-Mazes (2.2%), Stroop Direct WC and Interference (1.5%) HCs healthy controls, MDD major depressive disorder, HVLT-R Hopkins Verbal rs925946 [β: 0.081; p value: 0.226], interaction MDD x Learning Test-Revised, BVMT-R Brief Visuospatial Memory Test-Revised, RCFT Rey Complex Figure Test, CBTT Corsi Block-Tapping Test, LNS Letter Number Span, rs925946-T [β: 0.171; p value: 0.035]). We also performed TMT-A Trail Making Test part A, BACS-SC Brief Assessment of Cognition in post hoc analyses of the association between individual Schizophrenia-Symbol Coding, W words, C colors, CPT-IP Continuous Perfor- mance Test-Identical Pairs, TMT-B Trail Making Test part B, NAB-Mazes SNPs and treatment resistance in the MDD group but Neuropsychological Assessment Battery-Mazes, WC words-colors found no significant association. a TMT-A and TMT-B raw scores are shown. P values calculated upon natural log- transformed variables Mediation analysis A significant mediation was found in the association between rs908867 and the BVMTR test results through Partial correlation analyses differential methylation at the CpG 7.8.9 site of promoter We analyzed the correlation between methylation I. The bias-corrected bootstrap 95% CI indicated that the variables and neuropsychological performance. The cor- indirect effect through methylation differences was sig- relation heat map of these partial correlation analyses nificant using a dominant model (Effect: 0.84 (SE 0.49), stratified by diagnosis is included in Figs. S2 and S3 of the 95% CI, [0.15, 2.24]). No other significant mediations Supplementary Material. Antidepressant treatment was were found. not correlated with methylation measures. Ferrer et al. Translational Psychiatry (2019)9:265 Page 6 of 10 Multiple linear regression analyses Promoter I The results of multiple linear regression analyses of A higher promoter I mean methylation level was asso- mean methylation levels and neuropsychological perfor- ciated with poorer performance in the LNS test, a result mance in all participants are shown in Table 3. which remained significant with a FDR of 20% but did not reach significance at a FDR of 10%. This association was also observed when analyzing methylation levels in the region covered at Assay 2, while it did not appear in Assay Table 3 Results of multiple linear regression analyses of 1. The results of the multiple linear regression analyses of mean methylation levels and neuropsychological each CpG site in promoter I can be found in Tables S3 performance in all participants and S4 of the Supplementary Material. In Assay 1, CpG Promoter I Promoter IV site 7.8.9 was associated with poor results in most cog- nitive domains (visual learning and memory, working Mean Assay 1 Assay 2 Mean memory, processing speed, attention/vigilance, and β β β β executive functioning). An inverse relationship was observed in CpG site 10, where higher methylation pre- Verbal learning and memory dicted better results in visual learning and memory, pro- HVLT-R 0.047 0.100 –0.030 –0.013 cessing speed, attention/vigilance, and executive functioning tests. In Assay 2, poorer cognitive functioning Visual learning and memory in the LNS test was related to higher mean methylation BVMT-R 0.105 –0.042 –0.144 –0.152* levels at CpG sites 3.4.5.6 and 9. Another significant RCFT- copy 0.039 0.102 –0.047 –0.111 association was observed between higher methylation RCFT - immediate recall –0.060 <0.001 –0.108 –0.240** values at the CpG 11 site and poor performance in the BVMT-R test. RCFT - delayed recall –0.031 –0.009 –0.046 –0.261*** Working memory Promoter IV CBTT (non-verbal) 0.089 0.083 0.065 –0.014 Mean methylation levels at this promoter location were LNS (verbal) –0.193 ** –0.129 –0.207 ** –0.177 * significantly associated, at a FDR of 20%, to poorer per- formance in BVMT-R, RCFT immediate and delayed Processing speed recall, LNS and fluency tests. However, at a 5% FDR only TMT - A –0.011 –0.045 0.031 0.106 RCFT immediate and delayed recall were significant. BACS SC 0.002 –0.022 0.029 –0.048 Results of multiple linear regression analyses of each CpG Fluency –0.014 –0.029 0.009 –0.175* site in promoter IV can be found in Table S5 of the Supplementary Material. CpG sites 5, 9, and 10 presented Stroop Direct W 0.049 0.080 –0.004 0.068 an association with poorer cognition in several cognitive Stroop Direct C 0.054 0.059 0.029 –0.029 domains (verbal and visual learning and memory, working Attention/vigilance memory, processing speed and executive functioning). CPT-IP 0.032 0.069 −0.019 −0.046 Additionally, CpG site 13 showed an especially significant association with worse performance in visual learning and Executive function memory tasks (RCFT copy, immediate, and delayed recall). TMT- B –0.069 –0.053 –0.066 0.061 NAB Mazes 0.006 0.026 –0.019 –0.124 Interaction analysis Stroop Direct WC 0.105 0.087 0.091 <0.001 The results of the analyses of significant interactions between methylation variables and sex, CTQ score and Stroop Direct 0.107 0.062 0.125 –0.010 MDD diagnosis are shown in Table S6 of the Supple- Interference mentary Material. Regarding sex, the most consistent Statistically significant results are highlighted as bold (*p < 0.05; **p < 0.01; interaction was found between CpG site 22.23 of pro- *** p < 0.001). Linear regressions were adjusted by sex, age, years of education, MDD diagnosis, moter IV, in which higher methylation values were asso- tobacco consumption, HDRS, STAI trait score and CTQ score. ciated with better cognitive performance in women, β Standardized beta coefficient, HVLT-R Hopkins Verbal Learning Test-Revised, BVMT-R Brief Visuospatial Memory Test-Revised, RCFT Rey Complex Figure Test, whereas an inverse association was observed in men. CBTT Corsi Block-Tapping Test, LNS Letter Number Span, TMT-A Trail Making Test Regarding interactions between methylation variables and part A, BACS-SC Brief Assessment of Cognition in Schizophrenia-Symbol Coding, W words, C colors, CPT-IP Continuous Performance Test-Identical Pairs, TMT-B CTQ score, most significant interactions suggested that a Trail Making Test part B, NAB-Mazes Neuropsychological Assessment Battery- combination of higher methylation and a greater history Mazes, WC words-colors, MDD major depressive disorder, HDRS Hamilton Depression Rating Scale, STAI State-Trait Anxiety Inventory, CTQ Childhood of childhood trauma were associated with poorer cogni- Trauma Questionnaire tive functioning in attention, executive function and visual Ferrer et al. Translational Psychiatry (2019)9:265 Page 7 of 10 and verbal memory. Finally, when exploring the interac- Exploring the relationship between BDNF gene regula- tion with MDD diagnosis, we found that in both promoter tion through methylation and neurocognitive perfor- I and IV and in different cognitive domains, significant mance showed that, while higher methylation levels of associations between higher methylation and the presence most regions were associated with poorer cognitive per- of MDD diagnosis could be found, where higher methy- formance in the whole sample and across all cognitive lation values were associated with poorer cognition in domains, the effect of methylation in certain areas, such as MDD patients, while the inverse association was found CpG site 10 in promoter I, was associated with better in HCs. cognitive function. These results suggest that hyper- methylation at CpG site 10 could have a protective neu- Discussion rocognitive effect. Another interesting association was We have studied the contribution of BDNF genetic and found in the promoter IV region, in which CpG site epigenetic variations to cognitive functioning in MDD 13 showed a strong and specific association with impaired patients and HCs. Our main findings include the asso- visual learning and memory, while CpG sites 3, 9, and ciation between two SNPs as well as promoter methyla- 10 seemed to be associated with deficits in multiple areas. tion levels with several cognitive tasks and statistically The mechanism through which differential methylation of significant interactions with sex, CTQ scores and MDD CpG site 13 can have a specific impact on visual learning diagnosis. In addition, differential methylation at the CpG and memory, while other sites have a broader cognitive 7.8.9 site of BDNF promoter I mediated the association effect, should be the focus of future studies. Taken as a between the individual SNP rs908867 and visual memory. whole, our findings suggest an effect of BDNF gene pro- In the present study exploring the relationship between moter methylation and neurocognitive performance. Our 11 BDNF SNPs and cognitive functioning, two noncoding results are in accordance with previous data from SNPs (rs908867 and rs925946) were associated with epigenome-wide meta-analyses that reported an associa- poorer performance on cognitive tasks dealing with visual tion between DNA methylation in different genomic areas learning, memory and processing speed. Although most and cognitive performance62. research has focused on the role of the functional SNP To date, many studies have focused on the biological rs6265 in cognitive performance, there is evidence that mechanisms through which BDNF expression can influ- SNPs located in noncoding regions can also influence ence cognitive performance. Previous studies based on gene regulation58. In this line, a well-characterized non- animal models examined the influence of BDNF variations coding SNP within BDNF is rs12273363, which has been on different cognitive processes63–66. Some studies have shown to regulate BDNF transcription by altering pro- shown an association between BDNF gene methylation and moter IV activity59. Although this particular SNP was not an increased risk of developing neurodegenerative dis- associated in our sample, our findings related to associa- orders67,68, but the results were not replicated in others69. tions with two noncoding SNPs (rs908867 and rs925946) The cognitive tasks most frequently associated with lend support to the hypothesis that noncoding SNPs may methylation variables in our results are related to memory also influence BNDF activity. Several associations have processes (visual, verbal, and working memory), which are been found relating BDNF SNP rs908867 with a variety of thought to be mediated by the hippocampus and pre- clinical and neuroimaging phenotypes, including suicid- frontal cortex. The mechanisms that regulate hippo- ality60, treatment response in MDD48 and hippocampal campal function in memory have been widely studied. and cerebral atrophy49. Interestingly, in the study by Among other factors, BDNF is a key protein involved in Januar and colleagues27, only major homozygous patients hippocampal neurogenesis70,71. BDNF function is subject showed an association between BDNF promoter methy- to epigenetic control, which is sensitive to external lation and depression, and for CpG site 7.8.9, a trend modulating factors, including stress and childhood association was observed with depression, with trauma72. In this regard, animal studies have shown that rs908867 significantly modifying this relationship. Our adolescent trauma is associated with elevated corticos- results support the evidence of a relationship between terone levels and lower levels of BDNF in the dorsal rs908867 and clinical phenotypes, including cognitive hippocampus73. DNA methylation represents an epige- dysfunction, an effect that seems to be mediated by netic mechanism that contributes to the regulation of methylation at the CpG 7.8.9 site of BDNF promoter I. BDNF transcription in the CNS and is therefore involved One study including healthy individuals found an asso- in memory formation66. Research focusing on the influ- ciation between rs925946 and impairments in long-term ence of BDNF methylation and structural brain variations visual memory tasks61. Future studies focusing on the in MDD patients has described associations with cortical influence of these SNPs on BDNF transcription regulation thickness, particularly in the prefrontal and occipital are needed to describe the pathway through which they areas74, hippocampal atrophy49 and white matter integrity could be involved in cognitive function. in the anterior corona radiata75. Ferrer et al. Translational Psychiatry (2019)9:265 Page 8 of 10 Our results also show that the association between There are some limitations of our study that merit BDNF methylation and cognition can be moderated by discussion. First, we need to underscore that the asso- sex, childhood trauma or MDD diagnosis. In reference to ciation analysis between specific CpG sites and cognitive sex differences, previous human studies76 have reported performance was exploratory in nature. Thus, hypotheses higher levels of DNA methylation of promoter IV of generated by these results need to be tested in further BDNF in females. In line with these findings, animal confirmatory studies. Second, the small sample size of our studies exploring whether maltreatment or nurturing care study might have reduced the statistical power to detect is associated with differential methylation of BDNF within small effect sizes. Third, clinical variables such as child- the medial prefrontal cortex of adult rats have demon- hood trauma and anxiety questionnaires were retro- strated that methylation of DNA associated with exon IV spectively self-reported and could be influenced by recall is increased in female maltreated rats31. Interestingly, in bias and depressive state. Fourth, patients were recruited our study, we found significant interactions between in a naturalistic setting, and therefore relevant differences methylation values and female sex with several cognitive in clinical variables such as educational level existed tasks dealing with executive function and processing between groups. For this reason, we adjusted all analyses speed, which are known to involve the prefrontal cortex. by the factors that showed significant differences between No previous studies have explored the moderation effect groups and that could influence gene methylation of childhood trauma on the relationship between BDNF according to previous literature. For example, a recent methylation and cognitive functioning. In a recent twin meta-analysis reported an association between methyla- study, DNA methylation in several stress-related genes, tion at three specific genomic sites and depressive including BDNF, mediated the association between child- symptomatology; therefore, all analyses were adjusted by hood trauma and depressive symptoms at adulthood77. HDRS results79. Moreover, MDD patients were receiving Animal studies have also demonstrated that early life antidepressant treatment, which has been suggested to adversity produces persistent changes in DNA methylation influence DNA methylation at the BDNF gene35. How- of BDNF that cause altered BDNF gene expression in the ever, no significant association between antidepressant adult prefrontal cortex30. Studies in humans have shown an treatment dose (DDD) and DNA methylation at the effect of childhood abuse on the epigenetic regulation of BDNF gene was identified in our sample. Fifth, patients hippocampal glucocorticoid receptor expression, which recruited from a tertiary source may differ from could lead to cognitive dysfunction78. community-based cases, which could limit the general- Although the aim of our study was not to analyze dif- ization of results. Finally, we assessed the levels of ferences in BDNF methylation between diagnostic groups, methylation at the BDNF gene in peripheral blood cells. we observed significantly higher methylation levels at Although it might be argued that methylation studied in specific promoter regions of the BDNF gene in HCs peripheral blood cells could be different than methylation compared to MDD patients, results which contrast with in the brain, recent studies suggest that peripheral BDNF previous findings showing greater methylation in methylation closely reflects that of brain tissues80. patients26,27. MDD diagnosis moderated the association In summary, our study provides further insight into the between methylation of BDNF and cognitive functioning relationship between BDNF and cognitive performance, in memory and executive function, as MDD patients with suggesting that methylation of the BDNF gene influences increased methylation showed a poorer cognitive profile. hippocampus- and prefrontal cortex-mediated cognitive As there are no previous studies exploring this issue in tasks and that female sex, childhood trauma and MDD are relation to cognition, it is difficult to draw definitive moderators of these associations. conclusions. Our findings suggest that methylation of BDNF may contribute to the persistent cognitive deficits Acknowledgements observed in MDD patients. This study was supported in part by grants from the Carlos III Health Institute through the Ministry of Economy and Competitiveness (PI10/01753, PIE14/ To our knowledge, the present study is the first to 00034 and PI15/00662), co-funded by the European Regional Development explore the effect of BDNF regulation through DNA Fund (ERDF) “A way to build Europe”, CIBERSAM, and the Catalan Agency for methylation on cognitive performance while adjusting for the Management of University and Research Grants (AGAUR 2017 SGR 1247). We also thank CERCA Programme/ Generalitat de Catalunya for institutional modulating factors in MDD patients and HCs. Our results support. The genotyping and methylation services were carried out at CEGEN- suggest a relationship between methylation levels at spe- PRB3-ISCIII and were supported by grants PT13/0001 and PT17/0019, ISCIII- cific promoter locations and cognitive function. There- SGEFI/FEDER. Dr. Labad received an Intensification of the Research Activity Grant (SLT006/17/00012). Dr. Costas was supported by a Miguel Servet II fore, this work suggests that future research on BDNF contract from the Carlos III Health Institute (CPII16/00019). Dr. Soriano-Mas was regulation might help identify specific groups of MDD supported by a Miguel Servet contract from the Carlos III Health Institute patients who might benefit the most from intensive (CPII16/00048). The authors are very grateful to all of the study participants, as well as to the staff of the Department of Psychiatry at Bellvitge University antidepressant treatment targeting cognitive symptoms or Hospital who helped recruit the sample for this study. We would also like to neurocognitive rehabilitation. thank Sergi Sayols from the Hospital del Mar Medical Research Institute (IMIM) Ferrer et al. Translational Psychiatry (2019)9:265 Page 9 of 10 for helping in the initial statistical analysis and the technicians from the Center 10. Zhao, M. et al. BDNF Val66Met polymorphism, life stress and depression: A for Genomic Regulation (CRG) and Centro Nacional de Genotipado (CeGen). meta-analysis of gene-environment interaction. J Affect Disord. 227, 226–235 (2018). Author details 11. Cardoner, N. et al. Val66Met BDNF genotypes in melancholic depression: 1 Effects on brain structure and treatment outcome. Depress. Anxiety 30, Department of Psychiatry, Bellvitge University Hospital, Bellvitge Biomedical Research Institute (IDIBELL), Neurosciences Group - Psychiatry and Mental 225–233 (2013). Health, Barcelona, Spain. 2Department of Mental Health, ParcTaulí Hospital 12. Iga, J.-I. et al. The Val66Met polymorphism of the brain-derived neurotrophic Universitari, Institut d’Investigació i Innovació Parc Taulí (I3PT), Sabadell, Spain. factor gene is associated with psychotic feature and suicidal behavior in 3 Japanese major depressive patients. Am. J. Med. Genet. B Neuropsychiatr. Genet. Department of Psychiatry and Legal Medicine, Universitat Autònoma de Barcelona, Barcelona, Spain. 4Centro de Investigación Biomédica en Red de 144B, 1003–1006 (2007). Salud Mental (CIBERSAM), Carlos III Health Institute, Madrid, Spain. 13. Lee, Y. et al. Association between the BDNF Val66Met polymorphism and 5 chronicity of depression. Psychiatry Investig. 10, 56 (2013). Neuropathology Group, Bellvitge University Hospital, Bellvitge Biomedical Research Institute (IDIBELL), L’Hospitalet de Llobregat, Barcelona, Spain. 14. Domschke, K. et al. Brain-derived neurotrophic factor (BDNF) gene: No major 6 impact on antidepressant treatment response. Int J. Neuropsychopharmacol. Centro de Investigación Biomédica en Red de Enfermedades Neurodegenerativas (CIBERNED), Carlos III Health Institute, Madrid, Spain. 13, 93–101 (2010). 7 15. Egan, M. F. et al. The BDNF val66met polymorphism affects activity-dependent Grupo de Xenética Psiquiátrica, Instituto de Investigación Sanitaria de Santiago, Complexo Hospitalario Universitario de Santiago de Compostela, secretion of BDNF and human memory and hippocampal function. Cell 112, Servizo Galego de Saúde, Santiago de Compostela, Spain. 8Department of 257–269 (2003). Clinical Sciences, School of Medicine, Universitat de Barcelona, Barcelona, 16. Lu, B., Nagappan, G., Lu, Y. BDNF and synaptic plasticity, cognitive function, Spain. 9Department of Psychobiology and Methodology of Health Sciences, and dysfunction. Handb Exp Pharmacol. 220, 223–250 (2014). Universitat Autònoma de Barcelona, Barcelona, Spain. 10Fundación Pública 17. Hao, R. et al. BDNF val66met Polymorphism Impairs Hippocampal Long-Term Galega de Medicina Xenómica, Servicio Galego de Saúde, Instituto de Depression by Down-Regulation of 5-HT3 Receptors. Front Cell Neurosci. 11, Investigación Sanitaria de Santiago de Compostela, Santiago de Compostela, 306 (2017). Spain. 11Grupo de Medicina Xenómica, Universidade de Santiago de 18. Miyajima, F. et al. Brain-derived neurotrophic factor polymorphism Val66Met Compostela, Centro Nacional de Genotipado - Instituto Carlos III, Santiago de influences cognitive abilities in the elderly. Genes Brain Behav. 7, 411–417 Compostela, Spain. 12Centro de Investigación Biomédica en Red de (2008). Enfermedades Raras (CIBERER), Carlos III Health Institute, Madrid, Spain 19. Harris, S. E. et al. The brain-derived neurotrophic factor Val66Met poly- morphism is associated with age-related change in reasoning skills. Mol. Conflict of interest Psychiatry 11, 505–513 (2006). The authors declare that they have no conflict of interest. 20. Mandelman, S. D. & Grigorenko, E. L. BDNF Val66Met and cognition: all, none, or some? A meta-analysis of the genetic association. Genes Brain Behav. 11, 127–136 (2012). Publisher’s note 21. MacArthur, J. et al. The new NHGRI-EBI Catalog of published genome- wide Springer Nature remains neutral with regard to jurisdictional claims in association studies (GWAS Catalog). Nucleic Acids Res. 45(D1), D896–D901 published maps and institutional affiliations. (2017). 22. Furberg, H. et al. Genome-wide meta-analyses identify multiple loci associated Supplementary Information accompanies this paper at (https://doi.org/ with smoking behavior. Nat. Genet. 42, 441–447 (2010). 10.1038/s41398-019-0601-8). 23. Lee, J. J. et al. Gene discovery and polygenic prediction from a genome-wide association study of educational attainment in 1.1 million individuals. Nat. Received: 17 February 2019 Revised: 29 June 2019 Accepted: 30 July 2019 Genet. 50, 1112–1121 (2018). 24. Nagel, M. et al. Meta-analysis of genome-wide association studies for neuro- ticism in 449,484 individuals identifies novel genetic loci and pathways. Nat. Genet. 50, 920–927 (2018). 25. Uchida, S. et al. Epigenetic status of Gdnf in the ventral striatum determines References susceptibility and adaptation to daily stressful events. Neuron 69, 359–372 1. Rock, P. L., Roiser, J. P., Riedel, W. J. & Blackwell, A. D. Cognitive impairment in (2011). depression: a systematic review and meta-analysis. Psychol. Med. 44, 26. Fuchikami, M. et al. DNA methylation profiles of the brain-derived neuro- 2029–2040 (2014). trophic factor (BDNF) gene as a potent diagnostic biomarker in major 2. Hing, B., Sathyaputri, L., Potash, J. B. A comprehensive review of genetic and depression. PLoS One 6, 4–10 (2011). epigenetic mechanisms that regulate BDNF expression and function with 27. Januar, V., Ancelin, M. L., Ritchie, K., Saffery, R. & Ryan, J. BDNF promoter relevance to major depressive disorder. Am. J. Med. Genet. B Neuropsychiatr. methylation and genetic variation in late-life depression. Transl. Psychiatry 5, Genet. 177, 143–167 (2018). e619–7 (2015). 3. Thomas, K., Davies, A. Neurotrophins: a ticket to ride for BDNF. Curr. Biol. 15, 28. Kang, H. J. et al. Longitudinal associations between BDNF promoter methy- R262–R264 (2005). lation and late-life depression. Neurobiol. Aging 36, 1764.e1–1764.e7 (2015). 4. Klein, A. B. et al. Blood BDNF concentrations reflect brain-tissue BDNF levels 29. Kang, H. J. et al. BDNF promoter methylation and suicidal behavior in across species. Int J Neuropsychopharmacol. 14, 347–353 (2011). depressive patients. J. Affect Disord. 151, 679–685 (2013). 5. Molendijk, M. L. et al. Serum BDNF concentrations as peripheral manifestations 30. Roth, T. L., Lubin, F. D., Funk, A. J. & Sweatt, J. D. Lasting epigenetic influence of of depression: evidence from a systematic review and meta-analyses on 179 early-life adversity on the BDNF gene. Biol. Psychiatry 65, 760–769 (2009). associations (N=9484). Mol. Psychiatry 19, 791–800 (2014). 31. Blaze, J., Scheuing, L. & Roth, T. L. Differential methylation of genes in the 6. Hong, C. J. et al. Association study of a brain-derived neurotrophic-factor medial prefrontal cortex of developing and adult rats following exposure to genetic polymorphism and mood disorders, age of onset and suicidal maltreatment or nurturing care during infancy. Dev. Neurosci. 35, 306–316 behavior. Neuropsychobiology 48, 186–189 (2003). (2013). 7. Licinio, J., Dong, C. & Wong, M.-L. Novel sequence variations in the brain- 32. Blaze, J. & Roth, T. L. Caregiver maltreatment causes altered neuronal DNA derived neurotrophic factor gene and association with major depression and methylation in female rodents. Dev. Psychopathol. 29, 477–489 (2017). antidepressant treatment response. Arch. Gen. Psychiatry 66, 488 (2009). 33. Zeilinger S. et al. Tobacco smoking leads to extensive genome-wide changes 8. Li, M., Chang, H. & Xiao, X. BDNF Val66Met polymorphism and bipolar disorder in DNA methylation. PLoS One 8, 63812 (2013). in European populations: a risk association in case-control, family-based and 34. Boks, M. P. et al. The relationship of DNA methylation with age, gender and GWAS studies. Neurosci. Biobehav. Rev. 68, 218–233 (2016). genotype in twins and healthy controls. PLoS One 4, 21–23 (2009). 9. Hosang, G. M., Shiles, C., Tansey, K. E., McGuffin, P. & Uher, R. Interaction 35. Carlberg, L. et al. Brain-derived neurotrophic factor (BDNF) - Epigenetic reg- between stress and the BDNF Val66Met polymorphism in depression: a sys- ulation in unipolar and bipolar affective disorder. J. Affect Disord. 168, 399–406 tematic review and meta-analysis. BMC Med. 12, 1–11 (2014). (2014). Ferrer et al. Translational Psychiatry (2019)9:265 Page 10 of 10 36. Salvat-Pujol, N. et al. Hypothalamic-pituitary-adrenal axis activity and cognition 60. Ropret, S., Komel, R. & Videti, A. Neuroscience Letters Single nucleotide in major depression: the role of remission status. Psychoneuroendocrinology 76, polymorphisms in the BDNF gene and suicide in the Slovenian sample. 38–48 (2017). Neurosci. Lett. 602, 12–16 (2015). 37. Ferrer, A. et al. FKBP5 polymorphisms and hypothalamic-pituitary-adrenal axis 61. Avgan, N. et al. BDNF variants may modulate long-term visual memory per- negative feedback in major depression and obsessive-compulsive disorder. formance in a healthy cohort. Int. J. Mol. Sci. 18 (2017). https://doi.org/10.3390/ J Psychiatr Res. 104, 227–234 (2018). ijms18030655. 38. Sheehan D. V. et al. The Mini-International Neuropsychiatric Interview (M.I.N.I.): 62. Marioni, R. E. et al. Meta-analysis of epigenome-wide association studies of the development and validation of a structured diagnostic psychiatric inter- cognitive abilities. Mol Psychiatry 11, 2133–2144 (2018). view for DSM-IV and ICD-10. J. Clin. Psychiatry 59, 22–33 (1998). 63. Molteni, R., Rossetti, A. C., Savino, E., Racagni, G., Calabrese, F. Chronic mild 39. Anatomical Therapeutic Chemical (ATC) Classification System. World Health stress modulates activity-dependent transcription of BDNF in rat hippocampal Organization Collaborating Centre for Drug Statistics Methodology (2007). slices. Neural Plast. 2016, 2592319 (2016). 40. Hamilton, M. A rating scale for depression. J. Neurol. Neurosurg. Psychiatry 23, 64. Sakata, K. et al. Role of activity-dependent BDNF expression in hippocampal- 56–62 (1960). prefrontal cortical regulation of behavioral perseverance. Proc. Natl Acad. Sci. 41. Thase, M. E., Rush, A. J. When at first you don’t succeed: Sequential strategies USA 110, 15103–15108 (2013). for antidepressant nonresponders. J. Clin. Psychiatry 13, 23–29 (1997). 65. Munoz, P. C., Aspe, M. A., Contreras, L. S. & Palacios, A. G. Correlations of 42. Spielberger, C. D., Gorsuch, R. L., Lushene, P. R., Vagg, P. R. & Jacobs, A. G. recognition memory performance with expression and methylation of BDNF Manual for the State–Trait Anxiety Inventory (Form Y). https://doi.org/10.1007/ in rats. Biol. Res. 43, 251–258 (2010). 978-1-4419-9893-4 (1983). 66. Lubin, F. D., Roth, T. L. & Sweatt, J. D. Epigenetic regulation of BDNF gene 43. Bernstein, D. P. et al. Development and validation of a brief screening version transcription in the consolidation of fear memory. J. Neurosci. 28, 10576–10586 of the Childhood Trauma Questionnaire. Child Abuse Negl. 27, 169–190 (2003). (2008). 44. Barrett, J. C., Fry, B., Maller, J. & Daly, M. J. Haploview: Analysis and visualization 67. Chang, L. et al. Elevation of peripheral BDNF promoter methylation links to the of LD and haplotype maps. Bioinformatics 21, 263–265 (2005). risk of Alzheimer’s disease. PLoS One 9, e110773 (2014). 45. Lin, Y., Cheng, S., Xie, Z., Zhang, D. Association of rs6265 and rs2030324 68. Rao, J. S., Keleshian, V. L., Klein, S. & Rapoport, S. I. Epigenetic modifications in polymorphisms in brain-derived neurotrophic factor gene with Alzheimer’s frontal cortex from Alzheimer’s disease and bipolar disorder patients. Transl. disease: a meta-analysis. PLoS ONE 9, e94961 (2014). Psychiatry 2, e132–e137 (2012). 46. Liu, L. et al. Evidence of association between brain-derived neurotrophic factor 69. Wen, K. et al. The role of DNA methylation and histone modifications in gene and bipolar disorder. Psychiatr. Genet. 18, 267–274 (2008). neurodegenerative diseases: a systematic review. PLoS One 11, e0167201 47. Real, E. et al. A brain-derived neurotrophic factor haplotype is associated with (2016). therapeutic response in obsessive-compulsive disorder. Biol. Psychiatry 66, 70. Masi G., Brovedani P. The Hippocampus, neurotrophic factors and depression 674–680 (2009). possible implications for the pharmacotherapy of depression. CNS Drugs 25, 48. Gratacòs, M. et al. A brain-derived neurotrophic factor (BDNF) haplotype is 913–931 (2011). associated with antidepressant treatment outcome in mood disorders. Phar- 71. Leal, G., Afonso, P. M., Salazar, I. L., Duarte, C. B. Regulation of hippocampal macogenomics J. 8, 101–112 (2008). synaptic plasticity by BDNF. Brain Res. 1621, 82–101 (2015). https://doi.org/ 49. Honea, R. A. et al. Characterizing the role of brain derived neurotrophic factor 10.1016/j.brainres.2014.10.019. genetic variation in Alzheimer’s disease neurodegeneration. PLoS One 8, 1–9 72. Bennett, M. R. & Lagopoulos, J. Progress in Neurobiology Stress and trauma: (2013). BDNF control of dendritic-spine formation and regression. Prog. Neurobiol. 50. Hennings, J. M. et al. Possible associations of NTRK2 polymorphisms with 112, 80–99 (2014). antidepressant treatment outcome: findings from an extended tag SNP 73. Uys, J. D. et al. Developmental trauma is associated with behavioral hyper- approach. PLoS One 8, 1–12 (2013). arousal, altered HPA axis activity, and decreased hippocampal neurotrophin 51. Neves, F. S. et al. The role of BDNF genetic polymorphisms in bipolar disorder expression in the adult rat. Ann. N. Y. Acad. Sci. 1071, 542–546 (2006). with psychiatric comorbidities. J. Affect Disord. 131, 307–311 (2011). 74. Na, K.-S. et al. Brain-derived neurotrophic factor promoter methylation and 52. Villar-Menéndez, I. et al. Striatal adenosine A2A receptor expression is con- cortical thickness in recurrent major depressive disorder. Sci. Rep. 6, 21089 trolled by S-adenosyl-L-methionine-mediated methylation. Purinergic Signal. (2016). 10, 523–528 (2014). 75. Choi, S. et al. Association of brain-derived neurotrophic factor DNA methyla- 53. Purcell, S. et al. PLINK: A tool set for whole-genome association and tion and reduced white matter integrity in the anterior corona radiata in major population-based linkage analyses. Am. J. Hum. Genet. 81, 559–575 (2007). depression. J. Affect Disord. 172, 74–80 (2015). 54. González, J. R. et al. Maximizing association statistics over genetic models. 76. Ikegame, T. et al. DNA methylation analysis of BDNF gene promoters in Genet Epidemiol. 32, 246–254 (2008). peripheral blood cells of schizophrenia patients. Neurosci. Res. 77, 208–214 55. Hayes, A. F., Scharkow, M. The relative trustworthiness of inferential tests of the (2013). indirect effect in statistical mediation analysis: does method really matter? 77. Peng, H. et al. Childhood Trauma, DNA Methylation of Stress-Related Genes, Psychol Sci. 24, 1918–1927 (2013). and Depression: Findings From Two Monozygotic Twin Studies. Psychosom 56. Benjamini, Y., Hochberg, B. Controlling the false discovery rate: a practical and Med. 80, 599–608 (2018). powerful approach to multiple testing. J. R. Stat. Soc. 57, 289–300 (1995). 78. McGowan, P. O. et al. Epigenetic regulation of the glucocorticoid receptor in 57. Bender, R. & Lange, S. Adjusting for multiple testing - when and how? J. Clin. human brain associates with childhood abuse. Nat. Neurosci. 12, 342–348 Epidemiol. 54, 343–349 (2001). (2009). 58. Andersson, R. et al. An atlas of active enhancers across human cell types and 79. Jovanova, O. S. et al. DNA methylation signatures of depressive symptoms in tissues. Nature 507, 455–461 (2014). middle-aged and elderly persons: Meta-analysis of multiethnic epigenome- 59. Hing, B. et al. A polymorphism associated with depressive disorders differ- wide studies. JAMA Psychiatry 75, 949–959 (2018). entially regulates brain derived neurotrophic factor promoter IV activity. Biol. 80. Stenz, L. et al. BDNF promoter I methylation correlates between post-mortem Psychiatry 71, 618–626 (2012). human peripheral and brain tissues. Neurosci. Res. 91, 1–7 (2015).