Is Peripheral BDNF Promoter Methylation a Preclinical Biomarker of Dementia?

Abstract

Brain-derived neurotrophic factor (BDNF) has been implicated in dementia. Preliminary evidence suggests that BDNF DNA methylation may be a diagnostic biomarker of dementia, but the potential pre-clinical utility remains unclear. Participants in the ESPRIT study were assessed for cognitive function and dementia (DSM-IV criteria) over 14 years. BDNF exon 1 promoter methylation was measured in blood at baseline (n = 769) and buccal samples during follow-up (n = 1062). Genotyping was carried out for several common BDNF SNPs, including Val66Met (rs6265) and APOE ɛ4. Multivariable logistic regression analyses determined the association between BDNF methylation and both prevalent and incident dementia. Adjustment for gender, age, education, APOE ɛ4 genotype, body mass index, depression, and type 2 diabetes, as well as possible effect modification by gender and genetic variation were also investigated. Weak evidence of an association between lower blood methylation and dementia was observed at one of 11 sites investigated (Δ–0.5%, 95% CI:–0.9,–0.04, p = 0.03, p = 0.22 adjusted for multiple comparisons). Buccal methylation at two other sites was associated with 14-year incident dementia cases prior to adjustment for multiple comparisons only, and the effect sizes were small (Δ+0.3%, OR:1.57, SE:0.30, p = 0.02, p = 0.14 adjusted and Δ–1.5%, OR:0.85, SE:0.06, p = 0.03, p = 0.14 adjusted). Genetic variation in the BDNF gene did not modify these associations, and no gender-specific effects were observed. There was only a weak correlation between blood and buccal BDNF log-methylation at two sites (both r=–0.11). There was no strong evidence that blood or buccal BDNF exon 1 promoter DNA methylation is associated with prevalent or incident dementia, and reported associations would not remain after adjustment for multiple testing.

Keywords

biomarkers blood dementia DNA methylation epigenetics

INTRODUCTION

Dementia is the umbrella term used to describe a group of symptoms that results in long-term decline of cognitive functioning, which is significant enough to affect daily life [1]. It is a prevalent, debilitating condition and one of the leading contributors to the overall burden of disease, with significant impacts on the individual, family, and society [2]. Over 50 million people worldwide are thought to have dementia, and dementia prevalence is projected to rise to > 150 million by 2050 [3]. Further, it is a multi-factorial disease with complex etiology that is yet to be completely defined [4].

There is strong interest in identifying robust biomarkers for dementia which would assist in accurate and timely diagnosis. A biomarker which could be readily analyzed from a blood sample or buccal swab would enable a relatively easy and non-invasive, cost effective screening [5, 6]. As dementia pathophysiology manifests many years prior to symptom onset, a preclinical biomarker would hold greatest utility, with the future possibility for early intervention [7].

Epigenetic biomarkers are a promising new class of measures with a growing body of evidence indicating their potential clinical utility [8]. Epigenetics, refers to dynamic modifications to the genome, with the potential to impact its activity, in response to environmental stimuli or genetic factors [9]. The most well studied epigenetic mechanism is DNA methylation, wherein a methyl group is covalently attached to cytosine of cytosine-phosphate-guanine (CpG) base pairs within DNA [10]. DNA methylation is largely conserved across tissues, although there are key sites of differential methylation associated with tissue specificity [11].

Recent research suggests the disruption of epigenetic processes in the brain in individuals with dementia [12]. Correlations have been made between blood and brain methylation at certain CpG sites, suggesting that methylation measures in blood have the potential to represent methylation levels in the brain [13]. Buccal epithelial cells are also a possible surrogate for neural tissues in methylation studies, as both tissues are derived from embryonic ectoderm during development [14]. Although epigenetic biomarker research in dementia is still in its infancy, there is some preliminary data to suggest that differently methylated genes can be found in the blood of demented patients compared to controls [15].

Brain-derived neurotrophic factor (BDNF) is a neurotrophin protein, important in the regulation of neuronal activity and neurogenesis [16]. BDNF is thought to be neuroprotective against amyloid-β toxicity, one of the primary hallmarks of Alzheimer’s disease (AD), the most common form of dementia [17, 18]. As well, lower serum BDNF levels have been reported in individuals with dementia [19].

Our recent systematic review of the utility of blood DNA methylation as a biomarker for dementia, suggested that methylation of BDNF was one of the more promising candidates [15]. Two small case-control studies (n = 106 and 18) reported higher BDNF DNA methylation in dementia cases compared to controls [20, 21]. A third study (n = 39) however, found no association [22]. Importantly, all three studies investigated DNA methylation in individuals with dementia, so they could not conclude whether these marks were present prior to diagnosis, and thus their potential as pre-symptomatic biomarkers is unknown.

An important consideration when looking at DNA methylation is underlying genetic variation [23, 24], but none of the previous case-control studies have considered this. BDNF single nucleotide polymorphism (SNP) rs6265, also known as Val66Met, has been associated with dementia neuropathophysiology, particularly in AD [25] and cognitive impairment [26]. In addition to variation in the BDNF gene, APOE is the single biggest genetic risk factor for late-onset dementia [27, 28]. It is possible that genetic variation in the APOE gene can also interact to modify the association between BDNF DNA methylation and dementia.

The primary aim of this study was to determine whether blood DNA methylation levels at BDNF were associated with the prevalence and incidence of dementia in a community-based population. Secondary aims were to ascertain whether underlying genetic variation influences these associations, and to investigate the correlation between blood and buccal BDNF methylation.

METHODS

ESPRIT cohort

This study included participants from the Enquête de Sant $\overset{´}{e}$ Psychologique–Risques, Incidence et Traitement (ESPRIT) cohort, a longitudinal study of psychiatric disorders in an aging French population [29]. To be included in ESPRIT, participants needed to be over 65 years old, not institutionalized, and residing in the Montpellier region. They were predominantly (>99%) of white European ancestry [30]. Participants answered standardized questionnaires, and underwent extensive clinical assessments at their inclusion, and at up to seven follow-up waves (with approximately 2–3 years between each). Written and informed consent was given by each participant, and ethical approval was granted by the regional ethics committee (Ethical Committee of Sud M $\overset{´}{e}$ diterran $\overset{´}{e}$ e III and University Hospital of Kremlin-Bicêtre, France, project number DGS2007-0477).

Assessment of cognitive function

Cognitive tests were administered by trained staff at baseline and each follow-up interview. These included global cognition (Mini-Mental State Examination (MMSE) [31]), visual memory (Benton’s Visual Retention Test [32]), verbal fluency (Isaacs Set Test [33]), psychomotor speed (Trail Making Test, Part A [34]), and executive function (Trail Making Test, Part B [34]).

Diagnosis of dementia

Dementia status was assessed over a 14-year period. A preliminary diagnosis and classification of dementia at each follow-up examination was made by clinical study investigators according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) revised criteria [35, 36]. These were further validated by a national panel of neurologists, independent of the ESPRIT investigators. The date of dementia onset was the date of the follow-up interview when dementia was diagnosed. All-cause dementia was the primary outcome in this study, but probable and possible AD were also considered.

Other characteristics

Education was classed as having completed at least secondary school education or not. MMSE, as a measure of global cognitive function, was considered as normal (>26) or low (≤26). Smokers were classed as ever smoking (>1 pack year), versus not smoking (<1 pack year). Alcohol use was categorized as either high (≥24 grams per day) or low (≤24 grams per day). Participants were categorized as obese if body mass index (BMI) was ≥30, and diabetes was recorded as high glucose levels or treatment. Depression was defined as detailed previously [37]. Past or current depressive disorders were diagnosed by trained professionals according to DSM-IV criteria. The Center for Epidemiologic Studies Depression (CES-D) scale was used to assess depressive symptoms [38]. Participants were categorized as having depression if they had a current diagnosis of major depression or a CES-D score≥16, which is representative of moderate to severe depression.

Methylation analysis for BDNF exon I promoter

Methylation data were generated using blood samples from 769 participants collected at baseline, and buccal samples for 1062 collected at approximately the fourth wave (7 year) follow-up. Genomic DNA was extracted from white blood cells in 15 ml EDTA using Amersham-Pharmacia Biotechusing DNA extraction kits [39]. Four-hundred nanograms of extracted DNA was used for bisulphite conversion (EZ-96 DNA Methylation-Lightning MagPrep kit, Irvine, USA). Genomic DNA was extracted from buccal swabs by salting out, and bi-sulphite converted using methods described above. DNA was subsequently amplified in triplicate from both tissues.

An assay targeting a CpG island in BDNF exon I promoter region was designed using Agena Bioscience Epidesigner (epidesigner.com), covering 254 base pairs (chr11 : 27,744,025 –27,744,278, hg19). The amplification primers were: forward: 5’-aggaagagagGGTAGAGGTAGGGAGATTTTATGTT AG-3’ reverse: 5’-cagtaatacgactcactatagggagaaggct CCTACCCCCACTCTAATTAAAACAA-3’.

The Sequenom MassARRAY EpiTYPER platform was used to obtain DNA methylation data for 11 analytical CpG units (Supplementary Table 1) [40]. These units comprised 16 individual CpG sites, with some units having multiple CpG sites. In the case of a single methylation measure (unit) comprising multiple individual sites, the average methylation was given. The average methylation level for samples with technical replicate values within 10% of the median were included in analysis. Three-hundred and eighty-three participants with analyzable blood methylation data had matching buccal methylation data. To normalize both blood and buccal methylation data, values were log-transformed for analysis. Where raw methylation values were ‘0’, data were replaced with ‘0.0001’ to avoid undefined log-transformed results. Where evidence of association was seen between log-methylation and any variable, difference in mean untransformed data is reported.

BDNF and APOE genotyping

Genotyping was performed for seven polymorphisms across the BDNF gene region, including rs6265, rs11030101, rs28722151, rs7103411, rs962369, rs908867, and rs1491850. These were chosen as they have previously been implicated in neurological disorders [37]. BDNF genotyping was carried out using buccal DNA, using the KBioscience Competitive Allele-Specific PCR SNP genotyping system (KASPar) and performed by KBiosciences [41]. APOE genotyping was carried out at Lille Genopole in Northern France (http://www.genopole-lille.fr/spip/) using DNA extracted from fasting blood samples (Puregene Kit, Qiagen, France) [42].

BDNF genotype frequencies can be seen in Supplementary Table 2. Binary variables were generated, grouped as either homozygous for major allele, or heterozygous/homozygous for minor allele (Supplementary Table 3). For example, rs6265 genotypes were grouped as homozygous major (Val/Val (GG)) or heterozygous (Val/Met (GA)) plus homozygous minor (Met/Met (AA)). In the case of APOE, it was grouped as at least one ɛ4 allele (cytosine at either rs429358 or rs7412), versus none (Table 1 and Supplementary Table 4).

Table 1

Baseline characteristics of participants according to prevalent and 14-year incident dementia (prevalent analysis: n = 769, incident analysis n = 749)

Baseline characteristic	Prevalent dementia		p	Incident dementia		p
	No n=749 (97.4%)	Yes n=20 (2.6%)		No n=679 (90.7%)	Yes n=70 (9.3%)
	Mean (SD)			Mean (SD)
Age	72.95 (5.40)	80.4 (6.75)	<0.0001	72.69 (5.35)	75.43, (5.25)	<0.0001
	n (%)			n (%)
Female	422 (56.3%)	11 (55%)	0.90	383 (56.4%)	39 (55.7%)	0.91
Male	327 (43.7%)	9 (45%)		296 (43.6%)	31 (44.3%)
At least 1 APOEɛ4 allele	157 (20.9%)	9 (45%)	0.01	130 (19.1%)	27 (38.5%)	<0.0001
Lower education	512 (68.4%)	12 (60%)	0.43	465 (68.5%)	47 (67.1%)	0.82
Low MMSE, ≤26¹	235 (31.5%)	20 (100%)	<0.0001	204 (30.1%)	31 (44.9%)	0.01
Ever smoker²	311 (42.1%)	5 (27.2%)	0.22	278 (41.5%)	33 (48.5%)	0.26
Heavy alcohol use, ≥24 g/day³	158 (21.6%)	1 (5%)	0.07	141 (21.2%)	17 (25.4%)	0.43
Obese, BMI≥30⁴	67 (9%)	3 (15%)	0.35	62 (9.2%)	5 (7.1%)	0.55
Depression, CESD16 +⁵	222 (30%)	4 (21.1%)	0.40	192 (28.6%)	30 (42.9%)	0.01
Diabetes⁶	74 (9.9%)	4 (20%)	0.14	61 (8.9%)	13 (18.8%)	0.009

Data not available for ¹3, ²13, ³17, ⁴1, ⁵8, and ⁶1 participants. APOE, Apolipoprotein E; BMI, body mass index; CESD16+, Center for Epidemiologic Studies Depression scale with a score of 16 or over; MMSE, Mini-Mental State Examination.

Statistical analysis

T-tests were used to determine the unadjusted association between blood log-methylation and prevalent dementia at baseline, incident dementia (over 14-year period) and all dementia cases (at baseline or over follow-up). Cohen’s d was used to calculate the effect size. Logistic regression analysis was used to determine the association between blood log-methylation at baseline and dementia, accounting for a number of possible confounding factors. This included factors which are known to be associated with dementia risk and could possibly be associated with variations in DNA methylation levels, for example age, gender (women have increased risk of dementia [43], and there are clear DNA methylation differences between men and women [44]), APOE ɛ4 genotype, education level [45], BMI [46], depression [47], and diabetes [48]. Interactions were also assessed between BDNF methylation and APOE ɛ4 status, gender, and rs6265; however, no evidence of effect modification was found.

To determine whether the findings were consistent across tissues, logistic regression analysis was also used to investigate the association between buccal DNA methylation and incident dementia. Correlation coefficients were used to determine the extent to which there was a linear correlation between blood and buccal BDNF methylation levels. Equivalence testing was carried out using two one-sided tests (TOST) with the R Package “TOSTER” [49] to provide further support for null findings. Adjustment for multiple comparisons was carried out using the Benjamini-Hochberg (BH) procedure, and adjusted BH p-values reported for each test [50, 51].

RESULTS

Study population

Characteristics of the participants included in the study are shown in Table 1. Individuals with prevalent dementia were older than individuals without dementia. Older age was also predictive of 14-year incident dementia. Lower MMSE was associated with both prevalent and incidence dementia, as was the APOE ɛ4 allele. Depression and diabetes also differed between individuals with incident (but not prevalent) dementia and individuals without. No other characteristics differed significantly between groups.

The initial analysis of prevalent dementia includes all 769 participants with blood samples (n = 20 prevalent dementia, n = 749 without dementia). The subsequent analysis of incident dementia involves only 749 participants, as the 20 individuals with prevalent dementia were excluded.

Unadjusted associations between blood BDNF DNA methylation and dementia

There was no strong evidence for an association between BDNF methylation and dementia prevalence at any CpG unit (Supplementary Figure 1). The only small differences that were observed was slightly lower BDNF methylation in individuals with dementia, compared to those without at CpG_6 (T-test: Δ –0.9%, 95% CI: –1.8,0.9, p = 0.08, TOST: 90% CI:–0.016,–0.001, p < 0.001), and CpG_7_8_9 (T-test: Δ –0.5%, 95% CI:–1.1,0.04, p = 0.07, TOST: 90% CI: –0.01,0, p < 0.001); however, these did not reach significance. Over the 14-year follow-up period, 70 participants were diagnosed with dementia, and 679 remained without dementia. No evidence was observed of an association between BDNF DNA methylation and incident dementia (Fig. 1). When combining both prevalent and incidence cases (Supplementary Figure 2), slightly lower baseline BDNF methylation at CpG_6 (Δ –0.5%, 95% CI: –0.9,–0.04, p = 0.03, p = 0.22 BH adjusted, Cohen’s d: 0.24) was observed in individuals with dementia, compared to those without dementia. When considering individuals diagnosed with AD specifically (n = 54) prevalent and incidence cases of total dementia), results remained very similar (e.g., CpG_6 (Δ –0.6%, 95% CI:–1.2,–0.05, p = 0.03, p = 0.37 BH adjusted, Cohen’s d: 0.30). Neither observation passed adjustment for multiple comparisons.

Fig.1

Unadjusted blood log-methylation levels of incident dementia (n = 70) versus control participants (n = 679) in BDNF exon I promoter region. Ave indicates average methylation where all methylation data was present across all CpGs (n = 522). Error bars indicate 95% confidence intervals.

BDNF methylation and genetic variation

There was no association between genetic variation in APOE ɛ4, rs6265, rs1491850, rs28722151, and rs7103411 and BDNF methylation (data not shown). Before BH adjustment, an association was seen between rs692369 and lower methylation at CpG_10 (Δ –1.3%, 95% CI:–0.06,–2.5, p = 0.04, p = 0.38 BH adjusted), rs11030101 and higher methylation at CpG_25 (Δ+0.6%, 95% CI:0.07,1.1, p = 0.03, p = 0.27 BH adjusted), and rs908867 and higher methylation at CpG_6 (Δ+0.5%, 95% CI:0.09,0.9, p = 0.02, p = 0.20 BH adjusted). None of the other BDNF genetic variants were associated with either prevalent or incidence dementia (Supplementary Table 3).

Regression analysis

As there was some evidence that BDNF methylation at CpG_6 and CpG_7_8_9 may be associated with dementia, logistic regression models were generated with adjustment for potential confounding factors. The addition of gender, age, education, and APOE ɛ4 genotype had little influence on the results (Table 2), nor did the addition of obesity, depression and diabetes on CpG_6 (OR: 0.74, SE: 0.10, p = 0.02, p = 0.23 BH adjusted), but it did for CpG_7_8_9 (OR: 0.85, SE: 0.11, p = 0.21, p = 0.61 BH adjusted).

Table 2

Multivariate adjusted¹ logistic regression of model for select units of blood-based methylation

	Prevalent dementia				Incident dementia				All dementia cases²
CpG Unit	OR	SE	p	Adjusted p	OR	SE	p	Adjusted p	OR	SE	p	Adjusted p
6	0.62	0.15	0.04	0.22	0.75	0.10	0.04	0.42	0.72	0.09	0.008	0.09
7_8_9	0.60	0.15	0.04	0.22	0.86	0.12	0.30	0.83	0.81	0.10	0.09	0.47

¹Adjusted for sex, age, education, and APOE ɛ4 genotype. ²Diagnosis of dementia at baseline or at any time over follow-up.

Buccal BDNF DNA methylation and incident dementia

Buccal methylation data was available from 1062 participants (n = 97 incident dementia cases, 965 without dementia). Characteristics of participants that provided buccal samples are shown in Supplementary Table 4. Age, APOE ɛ4 genotype, MMSE, BMI, and diabetes differed significantly between controls and the incident dementia group. To determine whether findings in blood could be replicated in buccal tissue, logistic regression was performed (Supplementary Table 5). We found no evidence of an association between CpG_6 methylation and dementia status (Fig. 2). However, higher methylation at CpG_7_8_9 was associated with incident dementia (Δ+0.3%, 95% CI:0.02,0.5, p = 0.04, Cohen’s d: –0.22), even after adjusting for confounding factors (Model 2), including sex, age, education, APOE ɛ4 genotype, obesity, and diabetes (OR: 1.57, SE: 0.30, p = 0.02). DNA methylation at CpG_14 was also associated with incident dementia using log transformed data (Δ –0.43%, 95% CI:–0.10,–0.75, p = 0.0097, Cohen’s d 0.30) and remained after adjustments for confounding factors (Model 2, OR: 0.85, SE: 0.06, p = 0.03). After adjustment for multiple comparisons however, none of the associations remained significant.

Fig.2

Unadjusted buccal log-methylation levels of incident dementia (n = 97) versus control participants (n = 965) in BDNF exon I promoter region. Ave indicates average methylation where all methylation data was present across all CpGs (n = 835). Error bars indicate 95% confidence intervals.

Correlation between blood and buccal BDNF DNA methylation

Of the participants included in this study, 383 participants had methylation data for both blood and buccal tissues. This included 35 individuals with incident dementia and 348 who were without dementia (Supplementary Table 6). Using log-methylation values, CpG_6 (n = 378, r=–0.11) and CpG_15 (n = 377, r=–0.11) showed a weak negative correlation between blood and buccal methylation (Supplementary Figure 3).

DISCUSSION

We found there is no strong evidence that blood methylation of BDNF has the potential to be a biomarker for preclinical or diagnosed dementia. Differential methylation at two CpGs were associated with prevalent dementia, one of which was also associated with incident, and prevalent plus incident dementia. A higher buccal methylation at one CpG was associated with incident dementia, even after adjusting for multiple covariates. However, the effects were small, and equivalence testing indicated that null observations were statistically equivalent to zero. Furthermore, these associations would not remain after adjustment for multiple comparisons. Genetic variation across the BDNF region was only weakly associated with three specific BDNF methylation sites, but in these instances, did not modify the association between methylation and dementia status. Likewise, consideration of APOE ɛ4 genotype or potential gender-specific associations did not modify the findings.

There have been only three previous case-control studies which have measured BDNF methylation and investigated its association with prevalent dementia; they were limited in size and did not consider confounding factors. One of these studies included 44 individuals with AD (mean age 80, SD 8.92) and 62 control individuals (mean age 79.63, SD 7.85) all of whom were recruited from hospitals in China [20]. AD was diagnosed using ICD-10 criteria [52], as well as a medical history, brain imaging, and cognitive screening tests including MMSE. Controls were free of any neurological disorders, although exact assessments undertaken are not clear. The methylation of four CpG units of BDNF exon I promoter region, 120 bases downstream from the region within our assay, was measured in peripheral blood using pyrosequencing. All four CpGs were found to be higher in the AD group compared to controls, as was average methylation over these four sites (AD 9.50% ±4.43, CT 7.45% ±2.77, p = 0.004). When stratified by sex, only one of the four sites remained significant in males, while another was significant in females. However, the number of males or females were not stated. It appears that no adjustment was made for potential confounding factors.

Another study included 6 controls and 12 AD cases (mean age 66.5, SD 5.07, 60% women) consisting of outpatients from a memory clinic and a University hospital in Japan [21]. Probable AD was diagnosed using National Institute of Neurology and Communicative Disorder and Stroke/Alzheimer Disease and Related Disorder Association (NINCDS/ADRDA) criteria, as well as a behavioral test (behave-AD) and cognitive tests including MMSE and the Frontal Assessment Battery [53 –55]. No cognitive assessment of controls was reported; however, those with a self-reported history of any psychiatric disorder were excluded. In this study they also measured blood BDNF methylation at exon I promoter region (covering 20 CpG sites in a different CpG island), using a slightly different method (ABI 3730 DNA analyzer) approximately 21,000 base pairs downstream from the region targeted in our study. In a similar manner to the previous study, they found total methylation ratio (including all 20 CpGs) to be significantly higher in the AD group compared to controls (5.08±5.52% versus 2.09±0.81%; p = 0.04). One particular site was much higher in the AD group compared to controls (4.85% ±6.91, 0.00% ±0.00, p = 0.01), although only univariate analysis was presented.

While these two studies conflict with the findings from our analysis, involving a larger cohort of population-based individuals, a third case-control reported findings that support those of our work. This study involving Caucasian Italian men (20 AD, 19 controls; mean age 75 (SD 7), found no association between blood BDNF methylation and AD status [22]. AD was diagnosed using the DSM-III-R and NINCDS/ADRDA criteria, including cognitive assessments using MMSE. It was unclear if control participants were cognitively assessed. Methylation was measured in circulating leukocytes (using methylation-specific RT-PCR). It is unclear which region of the BDNF gene was investigated, and they did not state the number of CpG sites analyzed.

In addition to these case-control studies which specifically focused on BDNF, there have now been a number of epigenome-wide associated DNA methylation studies of dementia, predominantly using blood [15, 56]. None of these studies identified the BDNF gene or nearby region, as one of the significant findings. This aligns with our current study, where there was no strong evidence in adjusted analysis, for an association between blood BDNF methylation and both prevalent and incident dementia.

The contrasting findings of previous studies could be due to a number of reasons. As part of our previously mentioned systematic review, we assessed the quality of evidence in all included studies, using Joanna Briggs Institute criteria [57]. The three aforementioned studies scored quite low (44.4%–66.7%) in critical appraisal. The primary reason for this was that these studies failed to identify, or adjust for possible confounding factors. The two studies to find associations between BDNF blood methylation and dementia were small, and potentially underpowered to find such large methylation differences, suggesting the possibility of false positive results. Given this, the strength of evidence within these studies is weak, and could thus be the cause of contradictory results. Based on our sample size and the number of prevalent and incident dementia cases, our study had over > 80% power (at 5% significance level) to detect relatively small effect sizes in blood of≥1.3% and 0.64% (SD 2) for prevalent and incident dementia respectively, and in buccal of≥0.61% (SD 2) for incident dementia. Thus, our study was adequately powered to observe previously reported effect sizes (Δ 2.99% and Δ 5.07%, respectively) [20, 21]. However, as all studies examined slightly different genomic regions, it is hard to make direct comparisons. Finally, previous studies included different ethnic groups (Chinese, Japanese, Italian, French). As it is well recognized that there are ethnic-specific genetic and epigenetic patterns, as well as associations with health outcomes, this must be considered as one possible reason for the discrepancy in findings between studies (e.g., in Asian versus Caucasian populations) [58, 59].

An advantage of our study over previous studies was the opportunity to assess blood-based methylation as a preclinical biomarker of incident dementia. All previous studies have focused on prevalent AD only. Differences between our study and previous studies are unlikely to be due to the analysis of all-cause dementia versus AD. Indeed, when we looked specifically at AD, the findings remained similar. Another strength of our study was the consideration of genetic variation, in particular APOE ɛ4 and rs6265 genotypes, which are of particular interest for dementia. There was no evidence that genetic variation influenced or modified the association between BDNF DNA methylation and dementia.

While it has been common to use both blood and buccal tissues in some methylation studies, to our knowledge this is the first study to investigate correlations between inter-individual blood and buccal tissue BDNF methylation. Only very weak correlations were found between blood and buccal tissues at two CpGs, and only CpG_15 remained associated when looking at raw methylation values. Neither passed adjustment for multiple comparisons. It should be noted, however, that the biological samples were not taken at the same time, and given that DNA methylation can change over time, this could partly account for the weak correlations across tissue [60, 61]. The inverse correlations were somewhat surprising; however, DNA methylation is known to vary across tissues [11]. A previous study has explored the inter-individual correlation between methylation of brain and peripheral tissues such as blood, buccal, and saliva [62]. They reported that the proportion of correlated CpGs within BDNF between brain and peripheral tissues is low (6 to 18%). Similar to our study, they also found relatively weak correlations (r = 0.15 to 0.19), however this was between brain and peripheral tissues, rather than between blood and buccal tissue.

Future directions

As dementia is a progressive disease, and DNA methylation changes over time, future studies should consider looking at measures of mild cognitive impairment (MCI), which can be considered to precede dementia [63]. Two studies by the same group measured BDNF methylation using pyrosequencing (3 CpGs in promoter I and 4 CpGs promoter IV) in peripheral blood of controls (n = 728, mean age 66.9, SD:6.32) and participants with amnestic MCI (n = 506, mean age 67.2, SD:6.95) [64, 65]. They also followed the 506 individuals with MCI over five years and compared methylation between those that remained cognitively stable (n = 330, mean age 69.9, SD:4.01), and those that converted to AD (n = 128, mean age 76.0, SD:4.82). These studies found that methylation was associated with MCI versus controls, higher at two CpGs in each promoter, as well as average methylation (MCI: 16.11% ±4.48, Controls: 14.85% ±4.07, p < 0.001). The methylation of these same four CpGs was found to be significantly higher in AD converted from MCI, compared to stable MCI, as well as average methylation (AD: 19.92% ±5.89 versus MCI: 17.43% ±5.10, p < 0.001). Results were adjusted for age, education, and gender. Cognitive scores were also added in multivariate Cox regression, although it is not clear which measure, or if multiple were included. In these models they found one CpG in promoter IV to be associated with AD conversion from MCI (HR = 3.51, p = 0.013). They also found that the AA genotype of rs6265 was associated with methylation at this same CpG, which interacted with the risk of MCI compared to controls, (OR: 1.23, 95% CI: 1.12–1.30, p = 0.019), as well as progression from MCI to AD (OR:1.40, 95% CI: 1.20–1.48, p = 0.003). These findings are important, as distinguishing between individuals that have stable MCI and those that progress to dementia, brings further understanding of the relationship between DNA methylation and dementia progression.

Within the BDNF gene region there are thought to be 11 functional exons, which include 9 different promoter regions [66]. In our study, we have only looked at one area of promoter I. Other promoter regions could be of interest, for example, promoter IV, where differences in methylation have been found to be associated with other neurological disorders [67, 68]. Another option beyond BDNF methylation could be in epigenome wide association studies (EWAS), where instead of looking at one defined genomic region of 20 CpG units, they offer data on hundreds of thousands of sites across the genome. This is an important consideration as it allows the entire epigenome, hence the epigenetic dysregulation of potentially multiple genes, to be analyzed at once.

As mentioned, previous studies were case-control in design, and included individuals already diagnosed with dementia. Case-control studies are instrumental in bringing understanding to the relationship between blood-based DNA methylation and the disease state, and perhaps useful in diagnosing dementia. However, whether these and other DNA methylation differences can also be detected prior to the appearance of dementia symptoms, remains to be determined. Thus, longitudinal studies like ESPRIT with stored biospecimens, that include ongoing cognitive examination, will prove very useful for future studies in this field.

Another future consideration, particularly for studies that examine CpG sites which share probes with EWAS arrays, is to look at the correlation between blood-based DNA methylation and brain DNA methylation. This will bring a deeper understanding to the relationship between blood based DNA methylation biomarkers of disease, and DNA methylation related neuropathophysiology in the brain [13]. This can be done using a publicly available online tool which compares DNA methylation at specific sites, between blood and brain tissues. The tool was created using 450K Illumina methylation data from both blood and multiple brain regions, and included participants who were neuropathologically unaffected, as well those that had a range of neuropathology [69]. In our study, CpG_26 is the same genomic site as a probe on the 450K Illumina Array (cg15462887). Using the tool, it can be seen that blood-based methylation at this site is positively correlated to methylation of the superior temporal gyrus in the brain (r = 0.29, p = 0.01) [69]. A high correlation between blood and brain tissue methylation at CpGs associated with dementia, could suggest that blood is an ideal surrogate tissue for investigating dementia related methylation changes in the brain at those specific CpG sites.

Conclusion

The results of this study provide little support for methylation of the exon I promoter region of BDNF in blood and buccal as a preclinical or diagnostic biomarker of dementia. Further, equivalence testing of null observations indicated that these findings were statistically equivalent to zero, further supporting the overall conclusion of no association. Recommendations for further studies in this area include large longitudinal cohort studies which investigate DNA methylation patterns across the entire epigenome. Such studies should collect biological samples across multiple time points to investigate dynamic methylation in regard to disease progression. Ongoing cognitive assessment will enable clear distinctions to be made between cognitively healthy participants, individuals with MCI, and those with dementia. Importantly, covariate factors, including genetic variation, need to be considered and included in statistical analysis. DNA methylation in its very nature is known to be influenced by not only environmental, but also genetic factors. If an epigenetic biomarker can be identified with sufficient sensitivity and specificity, it could be used for more accurate and less invasive dementia diagnosis. Such a biomarker could also permit preclinical dementia screening and may be useful when monitoring the effectiveness of intervention treatments in individuals with early clinical symptoms, or for those participating in studies of new therapeutic agents.

Footnotes

ACKNOWLEDGMENTS

The authors acknowledge the Laboratoire d’Analyse Génomique – Centre de Ressources Biologiques (LAG-CRB) – BB-0033-00071 – Institut Pasteur de Lille, F-59000 Lille, France for managing human samples. The authors would also like to acknowledge Dr. Audrey Gabelle, Memory Research and Resources Center, Department of Neurology, CHU Montpellier, Montpellier, France for her role if validating the dementia diagnosis.

The ESPRIT project is financed by the regional government of Languedoc-Roussillon, the Agence Nationale de la Recherche Project 07 LVIE 004, and an unconditional grant from Novartis. This work was also supported by France Alzheimer. The funders had no role in the design and conduct of the study; in data collection, management, analysis or interpretation of the data and were not involved with the writing, preparation, review or approval of the manuscript. PF is funded by RTP stipend PhD scholarship, awarded by Monash University and the Australian Government. JR is funded by a Dementia Research Leader fellowship (APP1135727 from the National Health & Medical Research Council (NHMRC), Australia.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Robinson

, Tang

, Taylor

(2015) Dementia: Timely diagnosis and early intervention. BMJ 350, h3029.

Livingston

, Sommerlad

, Orgeta

, Costafreda

, Huntley

, Ames

, Ballard

, Banerjee

, Burns

, Cohen-Mansfield

, Cooper

, Fox

, Gitlin

, Howard

, Kales

, Larson

, Ritchie

, Rockwood

, Sampson

, Samus

, Schneider

, Selbaek

, Teri

, Mukadam

(2017) Dementia prevention, intervention, and care. Lancet 390, 2673–2734.

Patterson

(2018) World Alzheimer Report 2018. The state of the art of dementia research: New frontiers. Alzheimer’s Disease International, London.

Ryan

, Fransquet

, Wrigglesworth

, Lacaze

(2018) Phenotypic heterogeneity in dementia: A challenge for epidemiology and biomarker studies. Front Public Health 6, 181.

O’Bryant

, Mielke

, Rissman

, Lista

, Vanderstichele

, Zetterberg

, Lewczuk

, Posner

, Hall

, Johnson

, Fong

, Luthman

, Jeromin

, Batrla-Utermann

, Villarreal

, Britton

, Snyder

, Henriksen

, Grammas

, Gupta

, Martins

, Hampel

(2017) Blood-based biomarkers in Alzheimer disease: Current state of the science and a novel collaborative paradigm for advancing from discovery to clinic. Alzheimers Dement 13, 45–58.

François

, Leifert

, Martins

, Thomas

, Fenech

(2014) Biomarkers of Alzheimer’s disease risk in peripheral tissues; focus on buccal cells. Curr Alzheimer Res 11, 519–531.

Jack

Jr. , Knopman

, Jagust

, Petersen

, Weiner

, Aisen

, Shaw

, Vemuri

, Wiste

, Weigand

, Lesnick

, Pankratz

, Donohue

, Trojanowski

(2013) Tracking pathophysiological processes in Alzheimer’s disease: An updated hypothetical model of dynamic biomarkers. Lancet Neurol 12, 207–216.

Andersen

, Tost

(2018) A summary of the biological processes, disease-associated changes, and clinical applications of DNA methylation. Methods Mol Biol 1708, 3–30.

Fenoglio

, Scarpini

, Serpente

, Galimberti

(2018) Role of genetics and epigenetics in the pathogenesis of Alzheimer’s disease and frontotemporal dementia. J Alzheimers Dis 62, 913–932.

10.

, Zhang

(2014) DNA methylation in mammals. Cold Spring Harb Perspect Biol 6, a019133.

11.

Slieker

, Relton

, Gaunt

, Slagboom

, Heijmans

(2018) Age-related DNA methylation changes are tissue-specific with ELOVL2 promoter methylation as exception. Epigenetics Chromatin 11, 25.

12.

Delgado-Morales

, Esteller

(2017) Opening up the DNA methylome of dementia. Mol Psychiatry 22, 485–496.

13.

Hannon

, Lunnon

, Schalkwyk

, Mill

(2015) Interindividual methylomic variation across blood, cortex, and cerebellum: Implications for epigenetic studies of neurological and neuropsychiatric phenotypes. Epigenetics 10, 1024–1032.

14.

Braun

, Han

, Hing

, Nagahama

, Gaul

, Heinzman

, Grossbach

, Close

, Dlouhy

, Howard

, Kawasaki

, Potash

, Shinozaki

(2019) Genome-wide DNA methylation comparison between live human brain and peripheral tissues within individuals. Transl Psychiatry 9, 47.

15.

Fransquet

, Lacaze

, Saffery

, McNeil

, Woods

, Ryan

(2018) Blood DNA methylation as a potential biomarker of dementia: A systematic review. Alzheimers Dement 14, 81–103.

16.

Lima Giacobbo

, Doorduin

, Klein

, Dierckx

, Bromberg

, de Vries

EFJ

(2019) Brain-derived neurotrophic factor in brain disorders: Focus on neuroinflammation. Mol Neurobiol 56, 3295–3312.

17.

Song

, Yu

, Tan

(2015) Brain-derived neurotrophic factor in Alzheimer’s disease: Risk, mechanisms, and therapy. Mol Neurobiol 52, 1477–1493.

18.

Lane

, Hardy

, Schott

(2018) Alzheimer’s disease. Eur J Neurol 25, 59–70.

19.

TKS

, Ho

CSH

, Tam

WWS

, Kua

, Ho

(2019) Decreased serum brain-derived neurotrophic factor (BDNF) levels in patients with Alzheimer’s disease (AD): A systematic review and meta-analysis. Int J Mol Sci 20, 257.

20.

Chang

, Wang

, Ji

, Dai

, Xu

, Jiang

, Hong

, Ye

, Zhang

, Zhou

, Liu

, Li

, Chen

, Li

, Zhou

, Zhuo

, Zhang

, Yin

, Mao

, Duan

, Wang

(2014) Elevation of peripheral BDNF promoter methylation links to the risk of Alzheimer’s disease. PLoS One 9, e110773.

21.

Nagata

, Kobayashi

, Ishii

, Shinagawa

, Nakayama

, Shibata

, Kuerban

, Ohnuma

, Kondo

, Arai

, Yamada

, Nakayama

(2015) Association between DNA methylation of the BDNF promoter region and clinical presentation in Alzheimer’s disease. Dement Geriatr Cogn Dis Extra 5, 64–73.

22.

Carboni

, Lattanzio

, Candeletti

, Porcellini

, Raschi

, Licastro

, Romualdi

(2015) Peripheral leukocyte expression of the potential biomarker proteins Bdnf, Sirt1, and Psen1 is not regulated by promoter methylation in Alzheimer’s disease patients. Neurosci Lett 605, 44–48.

23.

Fisher

, Wang

, Deng

, Sarnowski

, Cupples

, Liu

C-T

(2018) Do changes in DNA methylation mediate or interact with SNP variation? A pharmacoepigenetic analysis. BMC Genet 19, 70.

24.

Lim

, Rainey-Smith

, Lim

, Laws

, Gupta

, Porter

, Bourgeat

, Ames

, Fowler

, Salvado

, Villemagne

, Rowe

, Masters

, Zhou

, Martins

, Maruff

(2017) BDNF Val66Met in preclinical Alzheimer’s disease is associated with short-term changes in episodic memory and hippocampal volume but not serum mBDNF. Int Psychogeriatr 29, 1825–1834.

25.

Shen

, You

, Joseph

, Mirzaei

, Klistorner

, Graham

, Gupta

(2018) BDNF polymorphism: A review of its diagnostic and clinical relevance in neurodegenerative disorders. Aging Dis 9, 523–536.

26.

Franzmeier

, Ren

, Damm

, Monte-Rubio

, Boada

, Ruiz

, Ramirez

, Jessen

, Duzel

, Rodriguez Gomez

, Benzinger

, Goate

, Karch

, Fagan

, McDade

, Buerger

, Levin

, Duering

, Dichgans

, Suarez-Calvet

, Haass

, Gordon

, Lim

, Masters

, Janowitz

, Catak

, Wolfsgruber

, Wagner

, Milz

, Moreno-Grau

, Teipel

, Grothe

, Kilimann

, Rossor

, Fox

, Laske

, Chhatwal

, Falkai

, Perneczky

, Lee

, Spottke

, Boecker

, Brosseron

, Fliessbach

, Heneka

, Nestor

, Peters

, Fuentes

, Menne

, Priller

, Spruth

, Franke

, Schneider

, Westerteicher

, Speck

, Wiltfang

, Bartels

, Araque Caballero

, Metzger

, Bittner

, Salloway

, Danek

, Hassenstab

, Yakushev

, Schofield

, Morris

, Bateman

, Ewers

(2019) The BDNF Val66Met SNP modulates the association between beta-amyloid and hippocampal disconnection in Alzheimer’s disease. Mol Psychiatry. doi: 10.1038/s41380-019-0404-6.

27.

Riedel

, Thompson

, Brinton

(2016) Age, APOE and sex: Triad of risk of Alzheimer’s disease. J Steroid Biochem Mol Biol 160, 134–147.

28.

Tan

, Hyman

, Tan

JJX

, Hess

, Dillon

, Schellenberg

, Besser

, Kukull

, Kauppi

, McEvoy

, Andreassen

, Dale

, Fan

, Desikan

(2017) Polygenic hazard scores in preclinical Alzheimer disease. Ann Neurol 82, 484–488.

29.

Ritchie

, Artero

, Beluche

, Ancelin

, Mann

, Dupuy

, Malafosse

, Boulenger

(2004) Prevalence of DSM-IV psychiatric disorder in the French elderly population. Br J Psychiatry 184, 147–152.

30.

Ancelin

, Carriere

, Scali

, Ritchie

, Chaudieu

, Ryan

(2013) Angiotensin-converting enzyme gene variants are associated with both cortisol secretion and late-life depression. Transl Psychiatry 3, e322.

31.

Folstein

, Folstein

, McHugh

(1975) “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12, 189–198.

32.

Benton

, Silvan

(1992) Benton Visual Retention Test – Fifth Edition. The Psychological Corporation, San Antonio, TX.

33.

Isaacs

, Kennie

(1973) The Set test as an aid to the detection of dementia in old people. Br J Psychiatry 123, 467–470.

34.

Corrigan

, Hinkeldey

(1987) Relationships between parts A and B of the Trail Making Test. J Clin Psychol 43, 402–409.

35.

American Psychiatric Publishing Inc (1994) Diagnostic and statistical manual of mental disorders, 4th ed, American Psychiatric Publishing, Inc., Arlington, VA, US.

36.

First

, Frances

, Pincus

(2002) DSM-IV-TR handbook of differential diagnosis, American Psychiatric Publishing, Inc., Arlington, VA, US.

37.

Januar

, Ancelin

, Ritchie

, Saffery

, Ryan

(2015) BDNF promoter methylation and genetic variation in late-life depression. Transl Psychiatry 5, e619.

38.

Beekman

, Deeg

, Van Limbeek

, Braam

, De Vries

, Van Tilburg

(1997) Criterion validity of the Center for Epidemiologic Studies Depression scale (CES-D): Results from a community-based sample of older subjects in The Netherlands. Psychol Med 27, 231–235.

39.

Ritchie

, Jaussent

, Stewart

, Dupuy

A-M

, Courtet

, Ancelin

M-L

, Malafosse

(2009) Association of adverse childhood environment and 5-HTTLPR Genotype with late-life depression. J Clin Psychiatry 70, 1281–1288.

40.

Saffery

, Gordon

(2015) Time for a standardized system of reporting sites of genomic methylation. Genome Biol 16, 85.

41.

Raitio

, Harinen

, Lampinen

(2012) Validation of KASPar™ Dual FRET Based SNP Genotyping Assay, Thermo Fisher Scientific, pp. 1–5.

42.

Ryan

, Carriere

, Carcaillon

, Dartigues

, Auriacombe

, Rouaud

, Berr

, Ritchie

, Scarabin

, Ancelin

(2014) Estrogen receptor polymorphisms and incident dementia: The prospective 3C study. Alzheimers Dement 10, 27–35.

43.

Andrew

, Tierney

(2018) The puzzle of sex, gender and Alzheimer’s disease: Why are women more often affected than men? Women’s Health 14, 1745506518817995.

44.

Masser

, Hadad

, Porter

, Mangold

, Unnikrishnan

, Ford

, Giles

, Georgescu

, Dozmorov

, Wren

, Richardson

, Stanford

, Freeman

(2017) Sexually divergent DNA methylation patterns with hippocampal aging. Aging Cell 16, 1342–1352.

45.

Ngandu

, von Strauss

, Helkala

, Winblad

, Nissinen

, Tuomilehto

, Soininen

, Kivipelto

(2007) Education and dementia: What lies behind the association? Neurology 69, 1442–1450.

46.

Geurts

, Dugue

, Joo

, Makalic

, Jung

, Guan

, Nguyen

, Grove

, Wong

, Hodge

, Bassett

, FitzGerald

, Tsimiklis

, Baglietto

, Severi

, Schmidt

, Buchanan

, MacInnis

, Hopper

, Pankow

, Demerath

, Southey

, Giles

, English

, Milne

(2018) Novel associations between blood DNA methylation and body mass index in middle-aged and older adults. Int J Obes (Lond) 42, 887–896.

47.

Chen

, Meng

, Pei

, Zheng

, Leng

(2017) A review of DNA methylation in depression. J Clin Neurosci 43, 39–46.

48.

Bansal

, Pinney

(2017) DNA methylation and its role in the pathogenesis of diabetes. Pediatr Diabetes 18, 167–177.

49.

Lakens

(2017) Equivalence tests: A practical primer for t tests, correlations, and meta-analyses. Soc Psychol Personal Sci 8, 355–362.

50.

Benjamini

, Hochberg

(1995) Controlling the false discovery rate: A practical and powerful approach to multiple testing. J Royal Stat Soc Series B (Methodological) 57, 289–300.

51.

Benjamini

, Heller

, Yekutieli

(2009) Selective inference in complex research. Philos Trans A Math Phys Eng Sci 367, 4255–4271.

52.

World Health Organization (1993) The ICD-10 classification of mental and behavioural disorders, World Health Organization, Geneva.

53.

McKhann

, Drachman

, Folstein

, Katzman

, Price

, Stadlan

(1984) Clinical diagnosis of Alzheimer’s disease: Report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology 34, 939–944.

54.

Reisberg

, Borenstein

, Salob

, Ferris

, Franssen

, Georgotas

(1987) Behavioral symptoms in Alzheimer’s disease: Phenomenology and treatment. J Clin Psychiatry 48, 9–15.

55.

Dubois

, Slachevsky

, Litvan

, Pillon

(2000) The FAB: A Frontal Assessment Battery at bedside. Neurology 55, 1621–1626.

56.

Fransquet

, Ryan

(2019) The current status of blood epigenetic biomarkers for dementia. Crit Rev Clin Lab Sci 56, 435–457.

57.

The Joanna Briggs Institute (2016) Joanna Briggs Institute Reviewers’ Manual: 2016 edition. The Joanna Briggs Institute, Australia.

58.

Chu

, Yang

(2017) Interethnic DNA methylation difference and its implications in pharmacoepigenetics. Epigenomics 9, 1437–1454.

59.

Galanter

, Gignoux

, Oh

, Torgerson

, Pino-Yanes

, Thakur

, Eng

, Hu

, Huntsman

, Farber

, Avila

, Brigino-Buenaventura

, LeNoir

, Meade

, Serebrisky

, Rodríguez-Cintrón

, Kumar

, Rodríguez-Santana

, Seibold

, Borrell

, Burchard

, Zaitlen

(2017) Differential methylation between ethnic sub-groups reflects the effect of genetic ancestry and environmental exposures. eLife 6, e20532.

60.

Horvath

(2013) DNA methylation age of human tissues and cell types. Genome Biol 14, 3156.

61.

Luo

, Hajkova

, Ecker

(2018) Dynamic DNA methylation: In the right place at the right time. Science 361, 1336–1340.

62.

Braun

, Han

, Hing

, Nagahama

, Gaul

, Heinzman

, Grossbach

, Close

, Dlouhy

, Howard

3rd , Kawasaki

, Potash

, Shinozaki

(2019) Genome-wide DNA methylation comparison between live human brain and peripheral tissues within individuals. Transl Psychiatry 9, 47–47.

63.

Ravaglia

, Forti

, Maioli

, Martelli

, Servadei

, Brunetti

, Pantieri

, Mariani

(2006) Conversion of mild cognitive impairment to dementia: Predictive role of mild cognitive impairment subtypes and vascular risk factors. Dement Geriatr Cogn Disord 21, 51–58.

64.

Xie

, Liu

, Jiang

, Zhang

, Cui

, Zhang

, Xu

(2017) DNA methylation and Tag SNPs of the BDNF gene in conversion of amnestic mild cognitive impairment into Alzheimer’s disease: A cross-sectional cohort study. J Alzheimers Dis 58, 263–274.

65.

Xie

, Xu

, Liu

, Jiang

, Zhang

, Cui

, Zhang

, Xu

(2017) Elevation of peripheral BDNF promoter methylation predicts conversion from amnestic mild cognitive impairment to Alzheimer’s disease: A 5-year longitudinal study. J Alzheimers Dis 56, 391–401.

66.

Zheleznyakova

, Cao

, Schiöth

(2016) BDNF DNA methylation changes as a biomarker of psychiatric disorders: Literature review and open access database analysis. Behav Brain Funct 12, 17.

67.

Lieb

, Dreimüller

, Wagner

, Schlicht

, Falter

, Neyazi

, Müller-Engling

, Bleich

, Tadić

, Frieling

(2018) BDNF plasma levels and BDNF exon IV promoter methylation as predictors for antidepressant treatment response. Front Psychiatry 9, 511.

68.

K-S

, Won

, Kang

, Chang

, Yoon

H-K

, Tae

, Kim

Y-K

, Lee

M-S

, Joe

S-H

, Kim

, Ham

B-J

(2016) Brain-derived neurotrophic factor promoter methylation and cortical thickness in recurrent major depressive disorder. Sci Rep 6, 21089.

69.

Hannon

, Lunnon

, Schalkwyk

, Mill

. Blood brain DNA methylation comparison tool, https://epigenetics.essex.ac.uk/bloodbrain/, Accessed 10 May 2019.