Arranz et al. Translational Psychiatry (2021)11:5 https://doi.org/10.1038/s41398-020-01139-z Translational Psychiatry ARTICLE Open Access A genome-wide methylation study reveals X chromosome and childhood trauma methylation alterations associated with borderline personality disorder María J. Arranz1,2, Cristina Gallego-Fabrega 1,3, Ana Martín-Blanco2,4,5, Joaquim Soler2,4,5, Matilde Elices2,4,5, Elisabet Dominguez-Clavé4, Juliana Salazar6, Daniel Vega5,7, Laia Briones-Buixassa7 and Juan Carlos Pascual 2,4,5 Abstract Borderline personality disorder (BPD) is a severe and highly prevalent psychiatric disorder, more common in females than in males and with notable differences in presentation between genders. Recent studies have shown that epigenetic modifications such as DNA methylation may modulate gene × environment interactions and impact on neurodevelopment. We conducted an epigenome wide study (Illumina Infinium HumanMethylation450k beadchip) in a group of BPD patients with (N = 49) and without (N = 47) childhood traumas and in a control group (N = 44). Results were confirmed in a replication cohort (N = 293 BPD patients and N = 114 controls) using EpiTYPER assays. Differentially methylated CpG sites were observed in several genes and intragenic regions in the X chromosome 1234567890():,; 1234567890():,; 1234567890():,; 1234567890():,; (PQBP1, ZNF41, RPL10, cg07810091 and cg24395855) and in chromosome 6 (TAP2). BPD patients showed significantly lower methylation levels in these CpG sites than healthy controls. These differences seemed to be increased by the existence of childhood trauma. Comparisons between BPD patients with childhood trauma and patients and controls without revealed significant differences in four genes (POU5F1, GGT6, TNFRSF13C and FAM113B), none of them in the X chromosome. Gene set enrichment analyses revealed that epigenetic alterations were more frequently found in genes controlling oestrogen regulation, neurogenesis and cell differentiation. These results suggest that epigenetic alterations in the X chromosome and oestrogen-regulation genes may contribute to the development of BPD and explain the differences in presentation between genders. Furthermore, childhood trauma events may modulate the magnitude of the epigenetic alterations contributing to BPD. Introduction associated with severe psychosocial deterioration and high Borderline personality disorder (BPD) is a severe psy- use of mental health resources3. BPD may result from the chiatric disorder with a prevalence ranging from 0.5 to interaction between genetic and stressful environmental 5.9%1,2. BPD is characterized by distorted sense of self, factors such as early childhood traumatic experiences4–6. unstable interpersonal relationships, strong emotional BPD is relatively more common in females than in males: reactions, and self-harming behaviours. It is also about 75% of people with a clinical diagnosis of BPD are females7,8. In addition, the disorder is manifested differ- entially in men and in women, with notable differences in Correspondence: Juan Carlos Pascual (jpascual@santpau.cat) personality traits9. However, the reasons behind these 1 2 Fundació Docència i Recerca Mutua Terrassa, Terrassa, Spain gender differences in prevalence and presentation and the Centro de Investigación Biomédica en Red de Salud Mental (CIBERSAM), biological factors contributing to this disorder are not Institut de Recerca Biomèdica Sant Pau (IIB-Sant Pau), Barcelona, Spain Full list of author information is available at the end of the article yet clear. These authors contributed equally: María J. Arranz, Cristina Gallego-Fabrega © The Author(s) 2021 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/. Arranz et al. Translational Psychiatry (2021)11:5 Page 2 of 10 Twin and family studies have proven the genetic con- completed self-reported questionnaires designed to dis- tribution to BPD, with heritability values of 35–67% for card BPD symptoms or the presence of childhood trauma. BPD8,10. Several studies have associated genetic variants BPD sample: patients were recruited from the outpatient in serotonin (TPH1 and 5-HTT) and COMT genes with BPD unit from Hospital de la Santa Creu i Sant Pau. BPD10. Variants in FKBP5 and CRHR1 genes involved in Inclusion criteria: (1) BPD diagnosis according to DSM-IV the regulation of the hypothalamus–pituitary–adrenal criteria and assessed through two semi-structured diag- (HPA) axis, which is deregulated in psychiatric disorders, nostic interviews, (2) age 18–50 years, (3) no current were also found associated with BPD risk11. Genome-wide episode of any Axis I disorder according to DSM-IV cri- association studies have revealed several risk genes teria, and (4) no severe physical conditions such as (DPYD, PKP4) shared with other psychiatric conditions organic brain syndrome, neurological disease, or mental such as bipolar disorder, major depression and schizo- deficiency. The following clinical scales were used for the phrenia12. Finally, several studies have observed assessment: Structured clinical interview for DSM IV axis gene–environmental interactions between early life II personality disorders (SCID II)30, Revised Diagnostic experiences and monoaminergic or HPA axis genes Interview for Borderlines (DIB-R)31, McLean Screening including SCL6A4, COMT, BDNF, TPH1, FKBP5 and Instrument for Borderline Personality Disorder (MSI- CRHR2 associated with BPD2,11,13,14. However, these BPD), and Childhood Trauma Questionnaire—Short findings have not been consistently replicated. Form (CTQ-SF)32. This questionnaire retrospectively Recent studies have shown that epigenetic regulation assesses childhood abuse and neglect. It evaluates five factors such as DNA methylation may modulate gene × types of childhood trauma: sexual, physical, and emotional environment interactions and impact on neurodevelop- abuse, and physical and emotional neglect. The ques- ment. Environmental factors such as early life stress and tionnaire provides an overall rating and a specific score childhood maltreatment alter DNA methylation pat- for every subscale (from 5 to 25), as well as cut-off points terns15–20 and potentiate maladaptive behaviours and to classify each trauma according to the severity of the psychiatric disorders21–25. However, relatively few studies exposure. According to previous studies which have have investigated the contribution of epigenetic factors to described a positive correlation between the severity and BPD. In previous studies, we found alterations in the number of type of maltreatments, and the methylation methylation patterns of NR3C1 in BPD patients with and of several genes17,27, we considered a positive history of without childhood traumas26, in agreement with Perroud trauma if severity was at least moderate in two or more and collaborators’ results27. Altered methylation patterns subscales (i.e. cut-off scores of 8 or higher for sexual of HTR2A, MAOA, MAOB and COMT have also been abuse, 10 or higher for physical abuse, 13 or higher for observed in BPD subjects28. Finally, a genome-wide emotional abuse, 10 or higher for physical neglect, and 15 methylation study found that BPD and life adverse or higher for emotional neglect). According to the CTQ- events were associated with the methylation status of SF questionnaire, 49 BPD patients fulfilled the criteria for several genes (IL17RA, KCNQ2 and EFNB1, among childhood traumatic experiences (BPD+T), whereas 47 others)29. did not fulfil the criteria (BPD-T). In summary, there is strong evidence of the genetic and environmental contribution to the aetiology of BPD and Replication sample epigenetic regulation may act as a modulator of this The replication cohort included 407 subjects (114 CTL interaction. The main aim of this study is to discern the and 293 BPD). Healthy controls had no history of mental epigenetic events contributing to BPD and to investigate illness, drug use, or treatment with psychotropic medi- the contribution of childhood trauma to those events. cation. Patients were recruited from the BPD outpatient units of two hospitals (Hospital de la Santa Creu I Sant Materials and methods Pau and the Consorci Sanitari de l’Anoia, Barcelona). The Participants same inclusion criteria were used for the discovery and Discovery sample replication samples. This sample included N = 147 A total of N = 140 subjects including 44 healthy con- patients with available information for childhood trau- trols and 96 individuals with a BPD diagnosis were matic experiences according to the CTQ-SF32. A total of recruited for the discovery genome-wide DNA methyla- N = 81 patients (BPD+T) fulfilled the criteria of trau- tion study. Given the higher prevalence of BPD in women matic experiences in childhood, whereas N = 66 patients and to decrease genetic heterogeneity, all recruited indi- (BPD-T) did not fulfil the criteria. All patients and con- viduals were female of Caucasian origin. Control sample: trols were assessed by clinical staff experienced in the Healthy controls (CTL) with no history of mental illness, evaluation of BPD. All participants were also females of drug use or treatment with psychotropic medication were Caucasian origin. Almost 90% of BPD+T patients were recruited from the local population. Control participants exposed to moderate to severe emotional abuse and 60% Arranz et al. Translational Psychiatry (2021)11:5 Page 3 of 10 were exposed to sexual abuse. No significant differences suitable to be analysed by EpiTYPER®. Two measures in demographics or clinical severity were observed were captured for each CpG site and then averaged to between BPD patients with and without trauma (Sup- avoid off-measurements. Pairs with a standard deviation plementary Table 1). bigger than 10% were excluded as part of the quality The study was carried out in accordance with the control. Declaration of Helsinki and was approved by our research ethics committee. Written informed consent was obtained Statistical analyses from all participants who did not receive any economic Genome-wide DNA methylation study compensation. Methylation levels of each CpG site were expressed as β-values, ranging from 0 to 1 (unmethylated and com- Methylation studies pletely methylated, respectively). Differentially methy- Genome-wide DNA methylation assay, quality control and lated CpG sites (DMCs) were calculated using linear pre-processing models as described in the limma package39. Three dif- DNA was extracted from whole blood (Autopure Qia- ferent analyses were performed. First, DNAm levels were gen) following the manufacturer’s instructions. DNA compared between BPD patients and controls (BPD vs. bisulphite treatment and PCR amplification were per- CTL). Second, BPD+T patients were compared to BPD- formed by means of EpiTech Bisulfite and the PyroMark T patients and controls pooled together (BPD+T vs. PCR kits (Qiagen), respectively. Samples were randomly CTL & BPD-T). Third, DNAm levels from BPD patients distributed for all phenotypes of interest (www. who had experienced childhood trauma were compared randomyzer.org). Genome-wide DNAm was assessed with levels from patients without childhood trauma using the Infinium HumanMethylation450K BeadChip (BPD+T vs. BPD-T). Analyses were adjusted by age, PC, (Illumina Inc, San Diego, CA). Quality control (QC) and presence or absence of trauma in the first analysis, metrics (hybridization rate, bisulfate conversion, number and age and PC for the second and third analyses. The of beads per array and background correction) were two first PCs that accounted for 91% of variability were examined to determine the success of the bisulfite con- included in the analyses. Statistical significant values version and subsequent array hybridization. Probe filter- were set at p value <10−07 as described by Rakyan et al.40. ing was performed by removing those with detection Differentially methylated regions (DMRs) were also cal- p values >0.05, and were not represented by a minimum culated using the Bumphunter method; statistical sig- of three beads on the array, as described elsewhere33,34. nificant values were set at p value <0.05. Samples with poor bisulfite conversion were also removed. Single-nucleotide polymorphism-related probes35 and Replication study all multi-hit probes were excluded, as described in the Methylation levels of CpG sites an CpG units were ChAMP Bioconductor package36. A set of quantile nor- expressed as discreet proportions ranging from 0 to 1 by malizations was performed using a background adjust- intervals of 0,2 (0 = unmethylated and 1 = completely ment between-array normalization and a dye bias methylated). DMCs were tested using general linear correction, following previous recommendations34. Mul- models. As in the genome-wide DNA methylation study, tidimensional scaling, principal components (PC) analyses three different analyses were performed: BPD vs. CTL, and singular value decomposition were used to check for BPD+T vs. BPD-T & CTL, BPD+T vs. BPD-T, using the unknown population structures, possible batch effects, as same models. The R statistical computing environment well as other technical artefacts. Next, the Reference- (3.6.3 version) was used to perform all statistical analyses Based Method37 was used to adjust the data by cell-type and plots. proportion, using the champ.refbase function. Finally, the Combat function38, implemented via the ChAMP Bio- Gene set enrichment analyses (GSEA) conductor package, was performed to correct technical The Molecular Signatures Database (MSigDB)41 was used batch effects. to identify enriched gene sets. CpG sites with unadjusted p values <10−03 were selected from each discovery analysis, EpiTyper assay resulting in 723 unique genes from BPD vs. CTL analysis, High-Resolution Quantitative Methylation Profiling 1603 from BPD+T vs. BPD-T & CTL and 196 genes from with EpiTYPER® and the MassARRAY® System (Agena BPD+T vs. BPD-T. We evaluated the overlap between our Bioscience, San Diego, California), was used to validate three gene sets with two MSigDB collections, Gene the results from the discovery phase. EpiDesigner soft- Ontology (GO) biological processes ontology and Hallmark ware for genomic target selection and PCR primer design gene sets. Similar analyses were performed with genes was used to select top CpG sites from the discovery study mapping statistically significant DMRs. Arranz et al. Translational Psychiatry (2021)11:5 Page 4 of 10 Results The second analysis (N = 49 BPD+T patients vs. N = Genome-wide DNA methylation study 91 BPD-T & CTL) revealed five CpGs with lower After applying quality controls, 140 individuals and methylation levels in the BPD+T group (POU5F1 424,616 probes (87.4% of the total probes contained in the cg15948871, q value = 3.86 × 10−04; GPR55 cg24915915, HumanMethylation450 BeadChip array) were included in q value = 1.15 × 10−02; GGT6 cg04511534, q value = the analyses. 1.15 × 10−02; TNFRSF13C cg00253346, q value= 1.15 × The first analysis comparing DNAm levels between BPDs 10−02; FAM113B cg05478172, q value = 2.61 × 10−02) (N = 96) and CTL (N = 44) showed six CpG sites (PQBP1 after adjusting by age and PC (Table 2 and Supplemen- cg10030436, q value (FDR adjusted p value) = 3.04 × 10−02; tary Fig. 2). intergenic cg07810091 q value = 3.04 × 10−02; ZNF41 Comparisons between BPD+T (N = 49) and BPD-T cg22713892, q value = 3.15 × 10−02; RPL10 cg02871887, (n = 47) patients did not show statistically significant q value = 3.15 × 10−02; intergenic cg24395855, q value = results at the 1 × 10−7 level. However, 4 of the top 20 CpG 3.15 × 10−02; TAP2 cg20156774, q value = 4.86 × 10−02) findings coincided with top findings in the second analysis with differential methylation in our multivariate model (intergenic cg24786705, unadjusted p value = 4.11 × 10−6; (considering age, PC, and presence or absence of childhood POU5F1 cg15948871, unadjusted p = 1.04 × 10−5; GGT6 trauma as covariates (Table 1). All six CpGs had lower cg04511534, unadjusted p = 2.02 × 10−5; SLC22A7 methylation levels in BPD patients than in controls (Fig. 1). cg16465769, unadjusted p = 2.75 × 10−5 (Table 3 and In general, the differences in methylation levels with con- Supplementary Fig. 3). All four CpG sites showed lower trols were clearer in BPD+T patients than in BPD-T methylation levels in BPD+T compared to BPD-T individuals (Supplementary Fig. 1). subjects. Table 1 Top 20 differentially methylated CpG sites resulting from the comparison between BPD (N = 96) and CTL (N = 44) methylation levels. CpG site P value Q value B Chr. Chr. position Gene Gene region cg10030436 1.16E−07 3.04E−02 5.94 X 48755633 PQBP1 5′UTRa cg07810091 1.43E−07 3.04E−02 5.73 X 153626455 IGR cg22713892 3.21E−07 3.15E−02 4.95 X 153575581 ZNF41 5′UTR cg02871887 3.31E−07 3.15E−02 4.92 X 47518016 RPL10 TSS200 cg24395855 3.71E−07 3.15E−02 4.81 X 32782796 IGR cg20156774 6.82E−07 4.82E−02 4.21 6 3982903 TAP2 Body cg06753086 1.12E−06 6.29E−02 3.73 1 67430099 IGR cg11780549 1.19E−06 6.29E−02 3.68 17 48755716 ASPSCR1 Body cg24047905 1.40E−06 6.62E−02 3.51 17 3977385 IGR cg23390865 1.70E−06 6.70E−02 3.33 X 30174190 LOC100133957 TSS1500 cg11435369 1.74E−06 6.70E−02 3.30 7 55670466 ADAP1 Body cg17322683 2.04E−06 7.22E−02 3.15 6 67032472 RNF39 3′UTR cg12054566 2.54E−06 8.31E−02 2.93 X 1510000 OGT 1stExon cg04433201 3.02E−06 8.48E−02 2.77 17 48534493 FAM101B TSS1500 cg25375329 3.10E−06 8.48E−02 2.74 X 942874 RNF113A TSS1500 cg19880947 3.19E−06 8.48E−02 2.71 17 145011888 SLC43A2 Body cg11147309 3.52E−06 8.49E−02 2.62 8 45737011 PLEC1 Body cg11020638 3.63E−06 8.49E−02 2.59 17 1818886 STAT5A Body cg07478007 3.80E−06 8.49E−02 2.55 1 27188505 ATAD3B Body cg16061947 4.19E−06 8.89E−02 2.45 11 65550444 ADRBK1 TSS1500 Q value indicates FDR corrected p values. B indicates log-odds that the gene is differentially expressed. Bold letters represent statistically significant results. a CpGs in two studies: BPD vs. CTL and BPD+T vs. BPD-T & CTL. Arranz et al. Translational Psychiatry (2021)11:5 Page 5 of 10 Fig. 1 Significantly differentially methylated CpG sites resulting from the comparison between BPD (N = 96) and CTL (N = 44). Y-axis represents methylation levels (B values after combat and champ.refbase adjustment, 0 = not methylated, 1 = fully methylated). Interestingly, five of the six most significant CpGs when analysing the results of comparing BPD+T vs. BPD- findings in the BPD vs. CTL comparison (first analysis) T data. were in the X chromosome whereas no statistically sig- Comparisons between BPD and CTL subjects revealed nificant associations were detected with X chromosome 33 DMRs, most of them in the chromosome 6 (major CpG sites in the other two analyses. Supplementary Fig- histocompatibility complex locus) and in the X chromo- ure 4 shows all Manhattan and Quantile–Quantile (Q–Q) some. GSEA of the 33 genes included within these plots of DMCs resulting from different comparisons. regions, using GO biological processes gene sets, revealed GSEA of CpG results, using GO biological processes a significant presence of genes involved in antigen pro- gene sets, revealed a significant presence of genes involved cessing and presentation, and in immune response (Sup- in neurogenesis, neuron and cell differentiation and cell plementary Table 7). Analysis of BPD+T patients vs. CTL transport when comparing BPD patients and control and BPD-T individuals uncovered 16 DMR regions that individuals (Supplementary Table 2). Similar findings were enriched for genes involved in embryonic develop- were observed when investigating genes with CpG sites ment and regulation of gliogenesis (Supplementary Table differentially methylated between BPD+T patients vs. 8). Analyses of DMRs in BPD+T vs. BPD-T revealed 13 CTL and BPD-T individuals, with regulation of cell dif- regions although no significant gene set enrichment ferentiation and neurogenesis being the top findings within these regions was observed. (Supplementary Table 3). GSEA analyses of the results from BPD+T vs. BPD-T comparisons revealed a sig- Replication study nificant presence of DMCs in genes involved in cellular Four hundred and seven individuals (N = 293 BPDs, macromolecule location and in extracellular structure including 147 with childhood trauma information −81 organization (Supplementary Table 4). GSEA using the BPD+T and 66 BPD−T−, and N = 114 CTL) and 15 CpG Hallmark database showed a significant number of DMCs sites (top results from analyses 1, 2 and 3 plus 2 CpGs not in genes involved in early and late oestrogen response and statistically significant that were amongst the top findings myogenesis when comparing BPD vs. CTL and BPD+T in two comparisons) were included in the replication vs. CTL & BPD-T (Supplementary Tables 5 and 6). study. Five of the selected CpGs (ZNF41 cg22713892; However, no statistically significant results were observed TAP2 cg20156774; BNIP3 cg05762616; GGT6 Arranz et al. Translational Psychiatry (2021)11:5 Page 6 of 10 Table 2 Top 20 differentially methylated CpG sites resulting from the comparison between BPD+T patients (N = 49) and BPD-T patients and CTL (N = 91). CpG site P value Q value B Chr. Chr. position Gene Gene region cg15948871 9.09E−10 3.86E−04 10.57 6 31139620 POU5F1 TSS1500a cg24915915 6.01E−08 1.15E−02 6.45 2 231799064 GPR55 IGRa cg04511534 8.10E−08 1.15E−02 6.16 17 4463371 GGT6 Bodya cg00253346 1.09E−07 1.16E−02 5.87 22 42324281 TNFRSF13C TSS1500 cg05478172 3.07E−07 2.61E−02 4.86 12 47628981 FAM113B Body cg16465769 7.10E−07 5.03E−02 4.04 6 43266151 SLC22A7 1stExona cg26529771 1.07E−06 6.20E−02 3.64 16 89260650 CDH15 Body cg10030436 1.17E−06 6.20E−02 3.56 X 48755633 PQBP1 5′UTRb cg07823755 1.36E−06 6.44E−02 3.40 2 234668355 UGT1A10 Body cg07583088 1.73E−06 7.25E−02 3.17 16 1181617 IGR cg20587394 1.88E−06 7.25E−02 3.09 12 54332209 HOXC13 TSS1500 cg13231322 2.21E−06 7.35E−02 2.94 1 32234411 IGR cg24786705 2.25E−06 7.35E−02 2.92 5 34500318 IGRa cg14153321 2.49E−06 7.49E−02 2.82 22 38246051 EIF3L Body cg24073755 2.66E−06 7.49E−02 2.76 1 231955433 DISC2 TSS1500 cg02224755 2.82E−06 7.49E−02 2.70 16 1608969 IFT140 Body cg14788655 3.41E−06 8.33E−02 2.52 11 40224146 LRRC4C 5′UTR cg25929552 3.53E−06 8.33E−02 2.48 2 67551782 IGR cg01619107 3.94E−06 8.57E−02 2.38 12 52515354 IGR cg10188253 4.22E−06 8.57E−02 2.31 19 43908315 TEX101 5′UTR Q value indicates FDR corrected p values. B indicates log-odds that the gene is differentially expressed. Bold letters represent statistically significant results. a CpGs in two studies: BPD+T vs. BPD-T & CTL and BPD+T vs. BPD-T. b CpGs in two studies: BPD vs. CTL and BPD+T vs. BPD-T & CTL. cg04511534; SLC22A7 cg16465769) were not suitable for significant in the EWAS. No other significant association EpiTYPER® analyses and only ten CpGs were studied was found in the replication study, although the direction (Supplementary Table 9). All samples and amplicons of the epigenomic alterations coincided with that of the passed Quality Control analyses. Only two of the four discovery study. CpGs investigated were found statistically significant (PQBP1 cg10030436, p = 1.84 × 10−02 and RPL10 Discussion cg02871887, p = 2.24 × 10−02) when comparing BPD and The present study investigated methylation patterns in CTL subjects. Nevertheless, the methylation levels in all patients with BPDs in comparison with controls and CpGs investigated were lower in BPD patients than in investigated the possible contribution of childhood trau- controls, coinciding with the findings in the mas to the differences observed. discovery study. As expected, different methylation patterns were The PXDN CpG (cg10888111) was found differentially observed in patients when compared to controls, as well methylated in the BPD+T vs. BPD−T & CTL comparison as in individuals with childhood trauma and without. The (p = 5.33 × 10−3), and in the BPD+T vs. BPD−T com- methylation differences observed between BPD patients parison (p = 3.26 × 10−2). BPD+T subjects showed lower and controls did not coincide, in general, with the dif- methylation levels in both analyses, mirroring the findings ferences observed when analysing childhood trauma. For of the discovery study. Interestingly, the PXDN instance, comparisons between BPD patients and controls cg10888111 was also found differentially methylated when revealed several X chromosome regions differentially comparing BPD and CTL subjects in the replication study methylated, an observation that was not replicated when (p = 2.25 × 10−4), although it was not found statistically comparing subjects with or without childhood trauma. Arranz et al. Translational Psychiatry (2021)11:5 Page 7 of 10 Table 3 Top 20 differentially methylated CpG sites resulting from the comparison between BPD+T patients (N = 49) and BPD-T patients (N = 47). CpG site P value Q value B Chr. Chr. Position Gene Gene region cg01652665 1.25E−06 5.30E−01 3.71 14 36987722 NKX2-1 Body cg24786705 4.11E−06 8.73E−01 2.55 5 34500318 IGRa cg10888111 6.44E−06 9.11E−01 2.12 2 1637068 PXDN 3′UTRa cg15948871 1.01E−05 9.46E−01 1.68 6 31139620 POU5F1 TSS1500a cg05762616 1.41E−05 9.46E−01 1.35 10 133787150 BNIP3 Body cg13911052 1.47E−05 9.46E−01 1.32 2 176978909 IGR cg24846397 1.56E−05 9.46E−01 1.26 10 133787150 CASP7 TSS1500 cg24915915 1.99E−05 9.54E−01 1.02 2 231799064 IGR cg04511534 2.02E−05 9.54E−01 1.01 17 4463371 GGT6 Bodya cg16465769 2.75E−05 0.999985 0.71 6 43266151 SLC22A7 1stExona cg01862363 3.21E−05 0.999985 0.56 17 8928028 NTN1 Body cg12575928 3.37E−05 0.999985 0.51 11 22687660 GAS2 TSS1500 cg02481697 3.68E−05 0.999985 0.43 17 8928012 NTN1 Body cg20989454 3.82E−05 0.999985 0.39 11 613478 IRF7 Body cg24319547 3.90E−05 0.999985 0.37 6 33168857 SLC39A7 1stExon cg17207064 4.08E−05 0.999985 0.33 2 17982866 IGR cg00914033 4.58E−05 0.999985 0.22 10 81664698 IGR cg27190024 4.82E−05 0.999985 0.17 7 94034528 COL1A2 Body cg10969178 5.55E−05 0.999985 0.03 6 31838402 SLC44A4 Body cg06759559 6.37E−05 0.999985 −0.10 1 7028331 CAMTA1 Body Q value indicates FDR corrected p values. B indicates log-odds that the gene is differentially expressed. a CpGs in two studies: BPD+T vs. BPD-T & CTL and BPD+T vs. BPD-T. These X chromosome epigenomic alterations may par- disorders42 and with Renpenning syndrome43. The zinc- tially explain the different gender prevalence and pre- finger protein 41 (ZNF41) participates in nucleic acid sentation of BPD. In support of this hypothesis, gene binding and transcription regulation. ZNF41 mutations enrichment analyses also revealed epigenomic alterations have been associated with cognitive deficits44. However, in genes controlling oestrogen regulation, although fur- this finding was not confirmed in later studies45. The ther confirmation of these findings is required. Methyla- ribosomal protein L10 (RPL10) is a component of the 60S tion alterations were also observed in genes involved in ribosomal subunit and participates in RNA binding and neurogenesis, neuron differentiation, development, reg- transcription and translation regulation. RPL10 dysfunc- ulation and morphogenesis, as revealed by the GSEA, tion disrupts neurodevelopment and causes X-linked whereas DMRs analyses revealed enrichment of genes ribosomopathy characterized by syndromic intellectual involved in the regulation of immune response processes. disability and epilepsy46. RPL10 mutations have been The discovery study revealed that methylation levels of linked to autism spectrum disorder (ASD) mechanisms47. 5 X chromosome CpG sites within the PQBP1, ZNF41, Cognitive and personality traits share common genetic RPL10 genes and two intragenic regions were lower in factors and symptomatology48. For instance, impulsivity BPD patients than in controls. Additionally, a CpG site in and emotional dysregulation are symptoms common to the transporter 2 member of the ATP-binding cassette individuals with personality disorders or significant subfamily B (TAP2, chromosome 6) was also found to be intellectual disability49. These epigenetic alterations may less methylated in BPD patients than in controls. The partially explain the similarities in symptomatology Polyglutamine binding protein 1 (PQBP1) is involved in between BPD, ASD and intellectual disability. Another X transcription activation. PQBP1 mutations have been chromosome intragenic CpG island (cg24395855) found linked with intellectual disability and neurodegenerative to be associated with BPD risk is within a region not Arranz et al. Translational Psychiatry (2021)11:5 Page 8 of 10 previously associated with mental disorders or its symp- patients with childhood trauma presenting lower levels tomatology. Less clear is the link observed between BPD than those without trauma, thus replicating the previous risk and low methylation of the TAP2 gene, involved in finding. NKX2-1 regulates expression of thyroid and antigen presentation. Mutations in TAP2 have been morphogenesis genes. Several studies report a clear associated with several diseases including cancer, tuber- association between alterations in NKX2-1 and thyroid culosis, diabetes and arthritis although no link to mental and lung cancer, although it has also been hypothesized to disorders has been reported. play a role in the development of schizophrenia through Interestingly, the differences in methylation levels were the regulation of implicated pathways. PXDN is a perox- bigger when comparing BPD patients with trauma vs. idase involved in external matrix formation associated controls than when comparing BPD without trauma vs. with obesity risk, ovarian56 and prostate cancer57, and controls. These observations suggest that childhood more interestingly, with ASD58. A deletion of a chromo- trauma confers epigenomic alterations that modulate or some 2 region including the PXDN gene has also been increase BPD presentation and symptomatology. associated with intellectual disability and obesity59. Epigenome wide comparisons between BPD+T and However, these two later reports have not been replicated. BPD-T & CTLS revealed significant differences in the Although no significant gene enrichment results were level of methylation of CpG islands in five genes, obtained when analysing BPD+T and BPD−T, several top including POU5F1, GPR55, GGT6, TNFRSF13C, and findings replicated those found when analysing trauma in FAM113B, none of them located in the X chromosome. the whole cohort (i.e. neurogenesis and cell replication), The methylation levels of these islands were lower in BPD suggesting that they are true results. patients with childhood trauma than in BPD patients Previous reports have linked methylation levels in the without childhood trauma and controls. The POU5F1 gene NR3C1, involved in stress response, to trauma, major (OCT4) is a transcription factor that plays a key role in depression, post-traumatic stress disorder and personality embryonic development and stem cell pluripotency, and it disorders60. In a previous study we reported an associa- has been linked to several forms of cancer. The GPR55 tion between higher methylation levels in NR3C1 and gene codes for a putative G-protein-coupled cannabinoid BPD risk and childhood trauma26, a finding that coincided receptor that activates a variety of transduction signal with those of Perroud and colleagues27. Although we did pathways. GPR55 polymorphisms have been associated not find statistically significant differences in the NRC31 with anorexia nervosa50, cancer and Crohn’s disease and it methylation sites, we did find a nominal association also has been hypothesized to play a potential role in obesity51. showing higher methylation of a NRC31 CpG site in BPD No clear links between this receptor and mental disorders patients. A previous study had also found increased has been reported, although in a previous study increased methylation levels of 5-HT2A, MAOA, MAOB and S- levels of GPR55 were found in mouse models of RETT COMT genes in 26 BPD patients compared with 11 syndrome52. GGT6 is a gamma-glutamyltransferase that controls28. We were not able to replicate these findings, as plays a key role in glutathione homoeostasis by providing we detected hypo and hypermethylation sites in all those substrates for its synthesis53. It has been associated with genes when comparing BPD patients and controls, and cancer and rheumatoid arthritis with no links with mental none of these differences reached statistical significance. disorders. TNFRSF13C enhances B cell survival and reg- Differences in sample size and sites analysed may explain ulates B cell population. Recently, a microdeletion in 22q this discrepancy. Interestingly, an EWAS study performed including this gene was associated with intellectual dis- in BPD patients also found epigenomic alterations in ability ASD54. FAM113B main function is the modifica- genes involved in cell and neurogenesis regulation asso- tion of biopolymers on the cell surface and has been ciated with the severity of the disorder, with their top associated with lung cancer55. GSA of these results con- finding in the X chromosome29. However, their results are firmed that childhood trauma altered the methylation not comparable to ours as male and female BPD patients levels of genes involved in cell regulation and neurogen- were compared to major depression patients in their esis. Additionally, DMRs analyses revealed enrichment of study, with no general population controls considered. genes involved in the regulation of gliogenesis and Our sample has several limitations. First, the moderate embryonic development. sample size used in the discovery sample hinders the Although no statistically significant findings were reliability of the results. To minimize false-positive find- obtained when comparing patients with and without ings, we investigated the top findings in a second larger childhood trauma, probably due to the reduced sample replication sample. Most of the investigated findings were size, several interesting associations were observed, confirmed in the replication sample, or showed similar including regions in the genes NKX2-1, PXDN and directions as in the discovery study, confirming their POU5F1. Interestingly, POU5F1 was also found differen- reliability. Secondly, the controls used in the discovery tially methylated in this reduced sample, with BPD sample did not experience childhood trauma according to Arranz et al. Translational Psychiatry (2021)11:5 Page 9 of 10 the questionnaires and childhood trauma data were not References available in all replication samples. This information 1. Widiger, T. A. & Weissman, M. M. Epidemiology of borderline personality disorder. Hosp. Community Psychiatry 42, 1015–1021 (1991). might help to infer the contribution of childhood trauma 2. Bassir Nia, A. et al. Past, present, and future of genetic research in borderline to epigenomic changes. Thirdly, smoking habits were not personality disorder. Curr. Opin. Psychol. 21, 60–68 (2018). considered in the analyses, although they are similar in 3. Alvarez-Tomás, I. et al. Long-term course of borderline personality disorder: a prospective 10-year follow-up study. J. Pers. Disord. 31, 590–605 (2017). BPD patients and unaffected individuals. Fourthly, we did 4. Beauchaine, T. P., Klein, D. N., Crowell, S. E., Derbidge, C. & Gatzke-Kopp, L. not have detailed information on comorbidities (i.e. Multifinality in the development of personality disorders: a biology x sex x ADHD or affective disorders) nor on pharmacological environment interaction model of antisocial and borderline traits. Dev. Psy- chopathol. 21, 735–770 (2009). treatment received by the study subjects. Both issues may 5. Crowell, S. E., Beauchaine, T. P. & Linehan, M. M. A biosocial developmental hinder the analyses and difficult the obtention of clear model of borderline personality: elaborating and extending Linehan’s theory. results. Additionally, we did not have data on body mass Psychol. Bull. 135, 495–510 (2009). 6. Porter, C. et al. Childhood adversity and borderline personality disorder: a index, socioeconomic status or education and did not meta-analysis. Acta Psychiatr. Scand. 141, 6–20 (2020). include these variables in the analyses. Fourthly, cell-type 7. Ten Have, M. et al. Prevalence rates of borderline personality disorder proportions were not known in the replication sample, so symptoms: a study based on the Netherlands Mental Health Survey and Incidence Study-2. BMC Psychiatry 16, 249 (2016). the results of our discovery study may not be entirely 8. Skoglund, C. et al. Familial risk and heritability of diagnosed borderline per- reproducible in the replication study. Finally, although we sonality disorder: a register study of the Swedish population. Mol. Psychiatry conducted our study on DNA extracted from blood cells (2019). 9. Sansone, R. A. & Sansone, L. A. Gender patterns in borderline personality instead of brain tissue, there is a high correlation (r = disorder. Innov. Clin. Neurosci. 8, 16–20 (2011). 0.86) between blood and brain methylation status61. 10. Amad, A., Ramoz, N., Thomas, P., Jardri, R. & Gorwood, P. Genetics of borderline In summary, we found epigenomic alterations in X personality disorder: systematic review and proposal of an integrative model. Neurosci. Biobehav. Rev. 40, 6–19 (2014). chromosome and oestrogen regulation genes that may 11. Martín-Blanco, A. et al. The role of hypothalamus-pituitary-adrenal genes and contribute to the development of BPD and explain, at childhood trauma in borderline personality disorder. Eur. Arch. Psychiatry Clin. least partially, the differences in presentation between Neurosci. 266, 307–316 (2016). 12. Witt, S. H. et al. Genome-wide association study of borderline personality genders. Furthermore, childhood trauma events modu- disorder reveals genetic overlap with bipolar disorder, major depression and lated the magnitude of the epigenomic alterations con- schizophrenia. Transl. Psychiatry 7, e1155 (2017). tributing to BPD, confirming previous findings. 13. Martín-Blanco, A. et al. An exploratory association study of the influence of noradrenergic genes and childhood trauma in borderline personality disorder. Psychiatry Res. 229, 589–592 (2015). Acknowledgements 14. Amad, A., Ramoz, N., Peyre, H., Thomas, P. & Gorwood, P. FKBP5 gene variants We would like to thank all the study participants for their kind contribution. and borderline personality disorder. J. Affect. Disord. 248, 26–28 (2019). This study was supported by Centro de Investigación Biomédica en Red de 15. Thomas, M. et al. Increased BDNF methylation in saliva, but not blood, of Salud Mental (CIBERSAM) and by a grant from Instituto de Salud Carlos III (PI14/ patients with borderline personality disorder. Clin. Epigenet. 10, 109 (2018). 00214) and co-financed by the European Regional Development Fund (ERDF). 16. Bockmühl, Y. et al. Methylation at the CpG island shore region upregulates Nr3c1 promoter activity after early-life stress. Epigenetics 10, 247–257 (2015). Author details 17. Perroud, N. et al. Childhood maltreatment and methylation of the gluco- 1 Fundació Docència i Recerca Mutua Terrassa, Terrassa, Spain. 2Centro de corticoid receptor gene NR3C1 in bipolar disorder. Br. J. Psychiatry 204, 30–35 Investigación Biomédica en Red de Salud Mental (CIBERSAM), Institut de (2014). Recerca Biomèdica Sant Pau (IIB-Sant Pau), Barcelona, Spain. 3Stroke 18. Thaler, L. et al. Methylation of BDNF in women with bulimic eating syndromes: Pharmacogenomics and Genetics Group, Institut de Recerca Biomèdica Sant associations with childhood abuse and borderline personality disorder. Prog. Pau (IIB-Sant Pau), Barcelona, Spain. 4Department of Psychiatry, Hospital de la Neuropsychopharmacol. Biol. Psychiatry 54, 43–49 (2014). Santa Creu i Sant Pau, Barcelona, Spain. 5Department of Psychiatry and 19. Tyrka, A. R. et al. Childhood maltreatment and methylation of FK506 binding Forensic Medicine & Institute of Neurosciences, Universitat Autònoma de protein 5 gene (FKBP5). Dev. Psychopathol. 27, 1637–1645 (2015). Barcelona, Bellaterra, Spain. 6Translational Medical Oncology Laboratory, 20. Marinova, Z., Maercker, A., Grünblatt, E., Wojdacz, T. K. & Walitza, S. A pilot Institut de Recerca Biomèdica Sant Pau (IIB-Sant Pau), Bellaterra, Spain. investigation on DNA methylation modifications associated with complex 7 Psychiatry and Mental Health Department, Hospital of Igualada, Consorci posttraumatic symptoms in elderly traumatized in childhood. BMC Res. Notes Sanitari de l’Anoia & Fundació Sanitària d’Igualada, Igualada, Spain 10, 752 (2017). 21. Nestler, E. J., Peña, C. J., Kundakovic, M., Mitchell, A. & Akbarian, S. Epigenetic Conflict of interest basis of mental illness. Neuroscientist 22, 447–463 (2016). The authors declare that they have no conflict of interest. 22. Wolf, E. J. et al. Traumatic stress and accelerated DNA methylation age: a meta- analysis. Psychoneuroendocrinology 92, 123–134 (2018). Publisher’s note 23. Kuehner, J. N., Bruggeman, E. C., Wen, Z. & Yao, B. Epigenetic regulations in Springer Nature remains neutral with regard to jurisdictional claims in neuropsychiatric disorders. Front. Genet. 10, 268 (2019). published maps and institutional affiliations. 24. Sumner, J. A., Colich, N. L., Uddin, M., Armstrong, D. & McLaughlin, K. A. Early experiences of threat, but not deprivation, are associated with accelerated Supplementary Information accompanies this paper at (https://doi.org/ biological aging in children and adolescents. Biol. Psychiatry 85, 268–278 10.1038/s41398-020-01139-z). (2019). 25. Bustamante, A. C., Armstrong, D. L. & Uddin, M. Epigenetic profiles associated with major depression in the human brain. Psychiatry Res. 260, 439–442 (2018). Received: 18 March 2020 Revised: 2 December 2020 Accepted: 7 December 26. Martín-Blanco, A. et al. Association between methylation of the glucocorticoid 2020 receptor gene, childhood maltreatment, and clinical severity in borderline personality disorder. J. Psychiatr. Res. 57, 34–40 (2014). Arranz et al. Translational Psychiatry (2021)11:5 Page 10 of 10 27. Perroud, N. et al. Increased methylation of glucocorticoid receptor gene 46. Bourque, D. K. et al. A de novo mutation in RPL10 causes a rare X-linked (NR3C1) in adults with a history of childhood maltreatment: a link with the ribosomopathy characterized by syndromic intellectual disability and severity and type of trauma. Transl. Psychiatry 1, e59 (2011). epilepsy: A new case and review of the literature. Eur. J. Med. Genet. 61, 28. Dammann, G. et al. Increased DNA methylation of neuropsychiatric genes 89–93 (2018). occurs in borderline personality disorder. Epigenetics 6, 1454–1462 (2011). 47. Klauck, S. M. et al. Mutations in the ribosomal protein gene RPL10 suggest a 29. Prados, J. et al. Borderline personality disorder and childhood maltreatment: a novel modulating disease mechanism for autism. Mol. Psychiatry 11, genome-wide methylation analysis. Genes Brain Behav. 14, 177–188 (2015). 1073–1084 (2006). 30. First, M., Gibbon, M., Spitzer, R., Benjamin, L. User’ Guide for the Structured 48. Briley, D. A. & Tucker-Drob, E. M. Comparing the developmental genetics of Clinical Interview for DSM-IV Axis II Personality Disorders: SCID-II (American Psy- cognition and personality over the life span. J. Pers. 85, 51–64 (2017). chiatric Publishing, 1997). 49. Cowan, A. Too many feelings: a case series of individuals with borderline 31. Barrachina, J. et al. Axis II comorbidity in borderline personality disorder is personality disorders and intellectual disability. J. Child Dev. Disord. 4, https:// influenced by sex, age, and clinical severity. Compr. Psychiatry 52, 725–730 (2011). doi.org/10.4172/2472-1786.100072 (2018). 32. Bernstein, D. P. et al. Development and validation of a brief screening version 50. Ishiguro, H. et al. Functional polymorphism in the GPR55 gene is associated of the Childhood Trauma Questionnaire. Child Abuse Negl. 27, 169–190 (2003). with anorexia nervosa. Synapse 65, 103–108 (2011). 33. Touleimat, N. & Tost, J. Complete pipeline for Infinium(®) Human Methylation 51. Moreno-Navarrete, J. M. et al. The L-α-lysophosphatidylinositol/GPR55 system 450K BeadChip data processing using subset quantile normalization for and its potential role in human obesity. Diabetes 61, 281–291 (2012). accurate DNA methylation estimation. Epigenomics 4, 325–341 (2012). 52. Vigli, D. et al. Chronic treatment with the phytocannabinoid Cannabidivarin 34. Pidsley, R. et al. A data-driven approach to preprocessing Illumina 450K (CBDV) rescues behavioural alterations and brain atrophy in a mouse model of methylation array data. BMC Genomics 14, 293 (2013). Rett syndrome. Neuropharmacology 140, 121–129 (2018). 35. Zhou, W., Laird, P. W. & Shen, H. Comprehensive characterization, annotation 53. Heisterkamp, N., Groffen, J., Warburton, D. & Sneddon, T. P. The human and innovative use of Infinium DNA methylation BeadChip probes. Nucleic gamma-glutamyltransferase gene family. Hum. Genet. 123, 321–332 (2008). Acids Res. 45, e22 (2017). 54. Upadia, J. et al. A previously unrecognized 22q13.2 microdeletion syndrome 36. Tian, Y. et al. ChAMP: updated methylation analysis pipeline for Illumina that encompasses TCF20 and TNFRSF13C. Am. J. Med. Genet. A 176, BeadChips. Bioinformatics 33, 3982–3984 (2017). 2791–2797 (2018). 37. Houseman, E. A. et al. DNA methylation arrays as surrogate measures of cell 55. Hong Y. et al. Epigenome-wide association analysis of differentially methylated mixture distribution. BMC Bioinformatics 13, 86 (2012). signals in blood samples of patients with non-small-cell lung cancer. J. Clin. 38. Johnson, W. E., Li, C. & Rabinovic, A. Adjusting batch effects in microarray Med. 8, 1307 (2019). expression data using empirical Bayes methods. Biostatistics 8, 118–127 (2007). 56. Zheng, Y. Z. & Liang, L. High expression of PXDN is associated with poor 39. Ritchie, M. E. et al. limma powers differential expression analyses for RNA- prognosis and promotes proliferation, invasion as well as migration in ovarian sequencing and microarray studies. Nucleic Acids Res. 43, e47 (2015). cancer. Ann. Diagn. Pathol. 34, 161–165 (2018). 40. Rakyan, V. K., Down, T. A., Balding, D. J. & Beck, S. Epigenome-wide association 57. Dougan, J. et al. Proteomics-metabolomics combined approach identifies studies for common human diseases. Nat. Rev. Genet. 12, 529–541 (2011). peroxidasin as a protector against metabolic and oxidative stress in prostate 41. Liberzon, A. et al. Molecular signatures database (MSigDB) 3.0. Bioinformatics cancer. Int. J. Mol. Sci. 20, 3046 (2019). https://doi.org/10.3390/ijms20123046. 27, 1739–1740 (2011). 58. Meyer, K. J., Axelsen, M. S., Sheffield, V. C., Patil, S. R. & Wassink, T. H. Germline 42. Okazawa, H. PQBP1, an intrinsically disordered/denatured protein at the mosaic transmission of a novel duplication of PXDN and MYT1L to two male crossroad of intellectual disability and neurodegenerative diseases. Neurochem half-siblings with autism. Psychiatr. Genet. 22, 137–140 (2012). Int. 119, 17–25 (2018). 59. Bonaglia, M. C., Giorda, R. & Zanini, S. A new patient with a terminal de novo 43. Jeong, H. I. et al. First Korean case of renpenning syndrome with novel 2p25.3 deletion of 1.9 Mb associated with early-onset of obesity, intellectual mutation in PQBP1 diagnosed by targeted exome sequencing, and literature disabilities and hyperkinetic disorder. Mol. Cytogenet. 7, 53 (2014). review. Ann. Clin. Lab. Sci. 48, 522–527 (2018). 60. Watkeys, O. J., Kremerskothen, K., Quidé, Y., Fullerton, J. M. & Green, M. J. 44. Shoichet, S. A. et al. Mutations in the ZNF41 gene are associated with cog- Glucocorticoid receptor gene (NR3C1) DNA methylation in association with nitive deficits: identification of a new candidate for X-linked mental retarda- trauma, psychopathology, transcript expression, or genotypic variation: a tion. Am. J. Hum. Genet. 73, 1341–1354 (2003). systematic review. Neurosci. Biobehav. Rev. 95, 85–122 (2018). 45. Piton, A., Redin, C. & Mandel, J. L. XLID-causing mutations and associated 61. Braun, P. R. et al. Genome-wide DNA methylation comparison between live genes challenged in light of data from large-scale human exome sequencing. human brain and peripheral tissues within individuals. Transl. Psychiatry 9, 47 Am. J. Hum. Genet. 93, 368–383 (2013). (2019).