DNA methylation patterns of LINE-1 and Alu for pre-symptomatic dementia in type 2 diabetes

The identification of early markers of dementia is important for higher-risk populations such as those with type 2 diabetes (T2D). Retrotransposons, including long interspersed nuclear element 1 (LINE-1) and Alu, comprise ~40% of the human genome. Although dysregulation of these retrotransposons can induce aberrant gene regulation and genomic instability, their role in the development of pre-symptomatic dementia (PSD) among T2D patients is unknown. Here, we examined locus-specific changes in LINE-1 and Alu methylation in PSD and the potential to offset these changes via supplementation with folate and vitamin B12. We interrogated DNA methylation patterns corresponding to 22,352 probes for LINE-1 and Alu elements using publicly-available Illumina Infinium 450K methylation datasets from i) an 18-month prospective study in 28 T2D patients (GSE62003) and ii) an intervention study in which 44 individuals were supplemented with folic acid (400 μg/day) and vitamin B12 (500 μg/day) over two years (GSE74548). We identified 714 differentially methylated positions (DMP) mapping to retrotransposons in T2D patients who developed PSD in comparison to those who did not (PFDR < 0.05), comprised of 2.4% (228 probes) of all LINE-1 probes and 3.8% (486 probes) of all Alu probes. These loci were enriched in genes with functions related to Alzheimer's disease and cognitive decline, including GNB5, GNG7 and PKN3 (p < 0.05). In older individuals supplemented with folate/vitamin B12, 85 (11.9%) PSD retrotransposon loci showed significant changes in methylation (p < 0.05): participants with the MTHFR CC genotype predominantly showed hypermethylation at these loci, while hypomethylation was observed more frequently in those with the TT genotype. In T2D patients, LINE-1 and Alu elements are differentially methylated in PSD in a locus-specific manner and may offer clinical utility in monitoring risk of dementia. Further work is required to examine the potential for dietary supplementation in lowering the risk of PSD.

Introduction Type 2 diabetes mellitus (T2D) is a metabolic disease affecting more than 200 million people worldwide in 2010, and global prevalence is predicted to increase to more than 400 million people by 2030 [1]. T2D is a recognised risk factor for dementia and for cognitive dysfunction in older people, affecting domains of processing speed, cognitive flexibility, learning and memory [2][3][4][5][6][7]. Alzheimer's disease (AD) is the most common cause of dementia and its development is associated with insulin resistance in the brain [8]. Up to 80% of AD patients have either impaired fasting blood glucose or T2D, and these conditions are associated with cell death in the brain and loss of β-cells in the pancreas [9].
To date, the characteristics of T2D patients who go on to develop dementia are unclear. Epigenetic changes occur during cognitive dysfunction and the development of dementia in T2D patients and, therefore, have potential as disease biomarkers. Retrotransposons, including long interspersed nuclear element 1 (LINE-1) and Alu, comprise approximately 40% of the human genome [10]. There are one-million copies of Alu and over 500,000 copies of LINE-1 in the human genome, which can replicate and integrate themselves into new locations throughout the genome [11][12][13]. The evolutionary age of these elements, that is the timing of their insertion into the human genome, can be estimated through sequence variants that facilitate their categorisation into subfamilies. Younger LINE-1 and Alu subfamilies have retained the ability to retrotranspose within the human genome [12,14], and incorporation of these elements can interfere with the expression of adjacent genes, disrupting regulatory sequences of exon-intron interactions [15][16][17]. Aberrant activity of transposable elements has been implicated in neurological disorders, such as multiple sclerosis and AD, and can promote neuronal loss [18][19][20]. Furthermore, LINE-1 and Alu methylation has been widely measured as surrogate marker for global DNA methylation [15,[21][22][23][24][25]. Hypomethylation of these elements is associated with genomic instability [26] and occurs in multiple diseases including cancer [15,22,27] and AD [28]. Hypomethylation of retrotransposons is an essential factor promoting retrotransposition and becomes more common during ageing [15,29,30]. Blood methylation signatures of pre-symptomatic dementia (PSD) in older individuals with T2D have been identified, including differential methylation of loci mapping to genes linked to dementia [31]. Additionally, retrotransposons, including LINE-1 and Alu, are active in the human central nervous system throughout life [32,33]. Therefore, we hypothesise that DNA methylation changes within LINE-1 or Alu elements are potential novel markers of early-stage dementia in T2D.
Folate and vitamin B 12 are the source of coenzymes in one-carbon metabolism [34], and are involved in the maintenance of normal epigenetic patterns [35] by generating S-adenosylmethionine (SAM), the main methyl group donor for DNA methylation. Low serum concentrations of vitamin B 12 and folate are observed frequently in AD patients [36] and can disrupt one-carbon metabolism and, therefore, SAM production [37]. Folate is essential for the development of the central nervous system, and folate insufficiency leads to high circulating concentrations of homocysteine [38], which are associated with cardiovascular disease [39,40], venous thrombosis, cancer, psoriasis, T2D [39], renal and thyroid dysfunction [41], and with the development of dementia and AD [42,43]. The well-characterised polymorphism within the methylenetetrahydrofolate reductase (MTHFR) gene at position 677 (C677T) results in a dose-dependent decrease in enzyme activity [44], leading to reduced SAM production, increased serum homocysteine and decreased serum folate concentration [45,46]. However, the potential for intervention with nutrients involved in one-carbon metabolism to offset the epigenetic changes associated with PSD in T2D patients is unknown. DNA methylation patterns are well known to display tissue-specific differences [47], but most loci show similar DNA methylation levels across a wide variety of tissue types. Equally, epigenetic changes in disease and in response to external stimuli may be unique to target tissues or may be more widespread throughout the body, and this is an important consideration when seeking to identify blood-based epigenetic biomarkers of disease. Changes in DNA methylation measured in the blood have been reported in response to a variety of environmental exposures [48,49] and lifestyles [50], and also in the blood of patients with variety of pathologies [51,52], but their role in disease aetiology can be complex. For example, differences in DNA methylation patterns observed in the blood of cancer patients may relate to changes in leukocyte composition of the blood as part of the body's immune response to the tumour [53]. In the case of dementia, there is evidence that epigenetic marks in blood can serve as biomarkers of early disease, which may be related to brain-blood barrier leakage in the early stages of Alzheimer disease [54]. Most blood-based studies have identified differences in DNA methylation associated with pre-symptomatic dementia (PSD) but with little correlation to changes in brain tissue [55], which may imply that they are associated with risk factors for the disease (e.g. chronic inflammation) rather than informing upon epigenetic changes in the brain. Indeed, some of the genes whose differential methylation with PSD are unique to blood have roles in inflammatory processes [56]. However, some genes show disruption in both blood and brain tissue, such as differential methylation of the OXT gene that encodes the neuropeptide oxytocin [56]. Furthermore, in an analysis of blood samples taken from older individuals with T2D, genes with known functional relevance to dementia showed DNA methylation changes that were associated with PSD [31].
In this study, publicly-available Illumina Infinitum 450K methylation datasets were utilised to examine DNA methylation of retrotransposons in blood of older T2D patients with and without PSD, and to examine the potential of supplementation with folic acid and vitamin B 12 to offset the epigenetic changes observed in PSD.

Study design
Publicly available Illumina Infinium 450K datasets from two studies were used to analyse retrotransposon methylation (Fig 1). We used DNA methylation data from the Israel Diabetes and Cognitive Decline (IDCD) study (GSE62003) [31] to identify epigenetic changes in PSD. This study included 36 older T2D patients (mean age = 73.9 years) with a clinical dementia rating (CDR) of 0.0 at baseline, of whom 18 showed a decline in cognition to CDR = 0.5 at 18-months follow-up, while 18 controls remained without dementia (CDR = 0). T2D with PSD patients was age, gender, education, and HbA1C matched to the T2D without PSD (S1 Table). Secondly, data from an intervention study of folic acid and vitamin B 12 supplementation (GSE74548) [57] were used to determine the effect of these nutrients on methylation of retrotransposons. This study was a part of the B-PROOF study and recruitment criteria have been described in previous study [58]. Briefly, 44 older participants (mean age = 70.8 years) were supplemented with 400 μg/day folic acid and 500 μg/day vitamin B 12 over two years and 43 participants constituted the placebo group. Both groups were non-smokers (25 and 31 former smokers in supplemented and placebo groups, respectively). No excessive alcohol users were recruited. Levels of the C-reactive protein (CRP) at baseline was � 10 mg/L. In both studies, DNA was extracted from peripheral blood samples and DNA methylation was analysed using the Illumina Infinium 450K methylation microarray platform at two time points (baseline and 18-month follow-up in the IDCD study; and baseline and two-year follow-up in the folic acid and vitamin B 12 intervention study). MTHFR genotype data were available for the dietary intervention study (22 participants with 677CC and 22 participants with 677TT). Ethical approval for the longitudinal IDCD study was granted by the Institutional Review Board of Mount Sinai and the Helsinki committees of Sheba and MHS [59], and for the dietary intervention study by the Academic Hospital Maastricht [57].

Data analysis
Datasets from the intervention study and IDCD study were normalised using the wateRmelon method, respectively [31,57]. Data from the two studies were analysed separately and therefore normalisation of the two datasets together was not required. After data filtering and quality control, 15 subjects who developed dementia and 13 who had normal cognitive function were retained from within the IDCD dataset. Probes on the Illumina Infinium 450K methylation microarray mapping to LINE-1 and Alu elements were extracted using RepeatMasker and categorised according to evolutionary age of subfamilies into oldest (L1M; AluJ), intermediate (L1P and L1PB; AluS), and youngest (L1HS and L1PA; AluY) [20]. Top differentially methylated positions (DMPs) were identified in PSD patients by paired-sample t-tests and ANOVA, with correction for multiple hypothesis testing using the Benjamini-Hochberg method and significance defined as P FDR < 0.05. The genomic locations (TSS200, TSS1500, 5'UTR, 3' UTR, 1 st Exon, and gene body) of probes were identified through the Illumina annotation. The pathway analysis with significantly differentially methylated loci was performed using Web-Gestalt (WEB-based Gene SeT AnaLysis Toolkit) [60].

Statistical analysis
Differentially methylated loci were identified by paired-sample t-tests and ANOVA. Paired ttest was used to identify significant changes from baseline in DNA methylation of retrotransposons in both PSD and supplementation datasets. Fisher's exact test was used to identify enrichment by genomic location and evolutionary age of retrotransposons. ANOVA was conducted to assess changes in DNA methylation by genomic location. Correlations between DNA methylation and plasma homocysteine and serum vitamin B 12 concentrations were assessed by Pearson's correlation coefficient. The impact of MTHFR genotype upon the effects of folate supplementation was assessed by paired t-test. All analyses were performed using IBM SPSS statistical software program (version 24) and R Studio (version 1.1.442). Data are presented as means ± SD unless otherwise stated. Probabilities (P) were adjusted for multiple hypothesis testing by the Benjamini-Hochberg method, with P FDR < 0.05 considered significant.

Distribution of retrotransposon (LINE-1 and Alu) probes by subfamily and genomic location
A total of 9,139 probes mapping to LINE-1 elements and 12,339 probes mapping to Alu sequences on the 450K array were analysed in relation to PSD. Of the LINE-1 probes, 7,389 map to L1M (old) elements, 1,460 to L1P and L1PB (intermediate) elements, and 290 to L1HS and L1PA (youngest) elements (Fig 2A). Of the Alu probes, 3,456 map to AluJ (oldest) elements, 7,440 map to AluS (intermediate) elements, and 1,443 to AluY (youngest) elements. The majority of the interrogated repetitive element loci were located within gene bodies (50%), while 30% were in TSS1500 and 3% in TSS200 ( Fig 2B); the rest of the elements were located in the 5'UTR (13%) and 3'UTR (4%). DNA methylation across all retrotransposons by genomic location did not significantly differ between PSD and non-PSD cases (Fig 2C).

DNA methylation changes within LINE-1 and Alu elements in PSD
Changes in DNA methylation of retrotransposons at the 18-month follow-up compared with baseline were examined for all participants. Among the 21,478 of retrotransposon loci, there were 714 significant differentially methylated positions (DMPs; P FDR < 0.05) in PSD at followup (Fig 3A and S2 Table), corresponding to 2.4% of all LINE-1 and 3.8% of all Alu probes analysed, and enriched significantly within the AluJ and AluS families (p < 0.0001). Of these, only three probes were also significantly different in non-PSD subjects at follow-up. The majority of the loci were hypermethylated (424 Alu loci and 155 LINE-1 loci; P FDR < 0.05), while 62 Alu loci and 73 LINE-1 loci were significantly hypomethylated (P FDR < 0.05) (Fig 3B and 3C). The DNA methylation changes (follow-up vs baseline) by retrotransposon subfamily evolutionary age are shown in Fig 3D. The oldest LINE-1 subfamilies (L1M) demonstrated the largest variation in magnitude of DNA methylation changes (Δβ) within LINE-1 elements [standard deviation (SD) = 0.041] and tended to be hypomethylated at the 18-month followup, while the youngest Alu subfamilies (AluY) had the greatest variation in methylation (SD = 0.016) of all Alu elements (Fig 3D). We did not observe changes in DNA methylation at non-retrotransposon probes mapping in close proximity to the DMPs, suggesting that these changes are highly loci specific.

The effects of folic acid and vitamin B 12 supplementation on DNA methylation at DMPs of PSD
To assess the effects of supplementation with folic acid and vitamin B 12 on DNA methylation of the 714 PSD-associated DMPs, an independent intervention study dataset was used (supplementation group: n = 44; placebo group: n = 43). DNA methylation at 85 of the 714 loci were significantly altered (P FDR < 0.05) following the two-year intervention. An additional 12 loci which overlapped with the placebo group were excluded from further analysis. Methylation at 15 DMPs, nine of which mapped to L1M elements that were hypomethylated in the PSD group, significantly increased (P FDR < 0.05) after folic acid and vitamin B 12 supplementation (Fig 6A, top-left quarter and Table 1). Four of these 15 loci were also related to AD: cg14051544 of HAL1-2a-MD in the 5'UTR of Solute Carrier Family 7 Member 14 (SLC7A14); cg22023664 of L1M in the 5'UTR of GNG7; cg23342367 of L1M in the gene body of Tumor Protein D52 (TPD52); and cg23304647 of L1M in the gene body of G Protein Subunit Alpha 12 (GNA12). An additional 29 probes within AluJ (35%), AluS (24%), AluY (7%), of L1M (24%), L1PA (7%) and L1PB (1%) elements were positively correlated with the development of PSD and displayed significantly decreased DNA methylation (paired t-test, P FDR < 0.05) following dietary intervention (Fig 6A, bottom-right quarter and S4 Table). However, another 29 loci were hypermethylated both in the vitamin-supplemented and PSD groups (Fig 6A, topright quarter and S5 Table). When examined by MTHFR genotypes (677CC and 677TT), DNA methylation levels of 71 of the 714 PSD-associated DMPs were significantly altered in participants with the MTHFR CC genotype (8 hypomethylated, 19 hypermethylated) (Fig 6B), whereas 89 loci were significantly different in participants carrying the MTHFR TT genotype (61 hypomethylated, 6 hypermethylated) after intervention (Fig 6C).
A threshold of Δβ value = ±0.05 was used to identify loci displaying substantial changes in methylation levels following dietary supplementation. For all participants, methylation at 10 repetitive element loci increased significantly while one decreased significantly following supplementation (S6 Table), which corresponded to DMPs that were hypomethylated (10 loci) and hypermethylated (1 locus) in PSD. With both the CC and TT genotypes of MTHFR, methylation of cg05310486 (AluS) and cg23911433 (L1M) increased by Δβ > 0.05. However, in participants with the CC genotype, methylation of 10 other loci (cg27324781, cg23304647, cg14567424, cg15980656, cg26381210, cg12785183, cg10959668, cg24534583, cg06305891, and cg23342367), increased by Δβ > 0.05 after vitamin supplementation and had lower methylation within the PSD group, while methylation of cg01168028 was increased by supplementation and within PSD patients (Fig 6B). In those with the TT genotype only, methylation of three CpG loci (cg23489236, cg10675515, and cg27133230) decreased by Δβ > 0.05 following dietary supplementation and methylation at these loci correlated inversely with DNA methylation changes in PSD. In addition, methylation of cg07291836 decreased both in the PSD group and after intervention (Fig 6C).

Discussion
In this study, we identified 714 loci mapping to retrotransposons (LINE-1 and Alu) that were differentially methylated in T2D patients with PSD compared to those with normal cognitive function, with such changes enriched within older families (L1M, AluJ, and AluS). LINE-1 elements are enriched within intergenic regions of the genome while Alu elements are preferentially integrated upstream of genes and within introns [61], but for both there are extensive reports on their impact upon gene expression via cis-and trans-regulation. In concordance with their genomic distribution, the probes on the Illumina Infinium 450K mapping to retrotransposons predominantly correspond to TSS1500, gene bodies and intergenic regions, with relative depletion close to transcriptional start sites and within exons [20]. As changes at those probes mapping to TSS1500 are more likely to have a functional impact upon gene expression, such as through promoter hypomethylation, we observed enrichment of DMPs at TSS1500. This finding supports the hypothesis that changes in retrotransposon DNA methylation may have influence through mediation of neighbour gene expression. Further, we observed that a number of PSD-associated DMPs showed inverse changes in methylation with folic acid and vitamin B 12 supplementation. To our knowledge, this is the first study to analyse methylation of retrotransposons by their evolutionary age in people with T2D-associated PSD and in response to dietary intervention. Changes in methylation of LINE-1 or Alu elements during development and ageing may cause DNA damage and double-strand DNA breaks that contribute to the process of neurodegeneration [62,63]. Moreover, our locus-specific analysis revealed changes in elements mapping to genes associated with dementia including GNB5, GNG7, and PKN3. Specifically, we have identified hypermethylation of an AluSx element at the 5' region of the GNB5 promoter, and an AluSg element within the PKN3 promoter. We observed enrichment of DMPs within AluJ and AluS sequences. While the reasons for this are not currently clear, it may suggest that such elements are implicated in PSD not through retrotransposition, but rather through modifying the expression of proximal genes. Younger repetitive element subfamilies such as L1HS and AluY are known to be more active in retrotransposition [12,64], and therefore the differential methylation of AluJ and AluS elements are unlikely to result in somatic retrotransposition events. Loss of GNG7 is associated with dysregulation of the adenylyl cyclase signalling pathway that is associated with a number of neurodegenerative disorders [65]. GNB5 is similarly implicated in regulation of adenylyl cyclase signalling, and genetic variants within the gene are associated with impaired cognition [66], while Gnb5 knockout mice display severe impairment of development and motor co-ordination [67]. PKN3 activity increases in response to insulin signalling [68] and may have a neuroprotective effect in response to hypoxia [69]. Furthermore, PKN family members have previously been implicated in the biology of AD [70]. Our study has therefore identified epigenetic changes within retrotransposons mapping to genes with critical functions in neuronal biology and the development of neurodegenerative disorders. However, further functional studies will be required to examine the subsequent impact upon the expression of these genes and fully elucidate the role of retrotransposons in dementia biology. Genome-wide hypomethylation of LINE-1 and Alu elements causes loss of genomic integrity [71,72]. Folic acid and vitamin B 12 are key players in one-carbon metabolism which generates SAM [73,74], the universal methyl group donor used for DNA methylation. We observed that these nutrients modified DNA methylation of retrotransposons, including PSDassociated DMPs. Although these nutrients increased methylation of some CpG sites within retrotransposons that are also hypermethylated in PSD (Fig 6), this effect was small for most of the analysed loci (e.g. the change in methylation of cg02276269 and cg19764326 was below 0.05). Interestingly, nine of the 15 loci displaying hypermethylation after supplementation with B-vitamins were negatively correlated with plasma homocysteine concentrations, while DNA methylation at six loci were positively correlated with serum vitamin B 12 levels. Repetitive elements display differential sensitivity to environmental exposures [23], and it may be that epigenetic changes within repetitive elements in response to internal stimuli or environmental factors can serve to modify the expression of genes implicated in PSD development. DNA methylation of old LINE-1 subfamily (L1M) seems to be vulnerable to dietary exposure than younger subfamilies.
Additionally, we observed differences in retrotransposon DNA methylation by MTHFR genotype following dietary supplementation with folic acid and vitamin B 12 . Specifically, higher methylation levels were observed at some loci in in participants with the MTHFR CC genotype (the most common variant of MTHFR), whereas lower levels were observed at the same loci in participants with the MTHFR TT genotype. In those carrying the MTHFR TT genotype, MTHFR enzyme activity is lower leading to decreased SAM production and, potentially, resulting in hypomethylation events. Therefore, and in concurrence with our previous study [35], individuals with the MTHFR CC genotype may show greater benefit from supplementation with folic acid and vitamin B 12 than those with the MTHFR TT genotype.
Our study has limitations. The number of participants in each of the two studies was relatively small (38 participants for analysis of changes in PSD and 44 participants receiving dietary supplementation). Despite the measures in the study design and our analysis to increase statistical power and limit the risk of yielding false positives, we cannot exclude the possibility that some of the DMPs identified in this study may be artefacts. Our results will need to be confirmed in independent cohorts, and the role of retrotransposon methylation in the initiation of dementia will need to be examined further through in vitro studies. Secondly, the data on effects of folic acid and vitamin B 12 supplementation were obtained from healthy older people and not in T2D patients. Due to the absence of cognition data, we cannot exclude the possibility that some of the participants within the dietary supplementation study may have PSD; therefore, care must be taken in extrapolating the results to the development of PSD in those with T2D. However, both studies were undertaken in older people and the intervention and follow-up periods were of similar duration (i.e. one-and-a-half and two years respectively), the resulting data may be compatible.
To the best of our knowledge, there are no publicly-available RNA-seq datasets available to examine the link between altered DNA methylation and gene expression to understand the underlying mechanisms of PSD in T2D. Future studies may require such analysis to reveal the functionality of these DMRs in the aetiology of PSD in T2D patients.

Conclusions
Our study is the first to report that locus-specific changes in DNA methylation within retrotransposons is associated with PSD incidence in T2D patients, and that the methylation of these sites might be modifiable by dietary supplementation with folic acid and vitamin B 12 . Further, we observed that the effects of these nutrients on DNA methylation differ by MTHFR genotype. If confirmed in future larger studies, these findings may contribute to the identification of epigenetic biomarkers of dementia risk in those with T2D and will strengthen the evidence for dietary intervention to reduce the risk of dementia among high-risk T2D patients.
Supporting information S1