Lower DNA methylation levels in CpG island shores of CR1, CLU, and PICALM genes in the blood of Alzheimer's disease patients

The aim of the present study was to (1) investigate the relationship between late onset AD and DNA methylation levels in the top six Alzheimer's disease (AD)-related genes in blood and (2) examine its applicability to the diagnosis of AD. We examined methylation differences at CpG island shores in the top six genes using Sanger sequencing, and one of two groups of 48 AD patients and 48 elderly controls was used for a test or replication analysis. We found that methylation levels in three out of the six genes, CR1, CLU, and PICALM, were lower in AD subjects. The combination of CLU methylation levels and the APOE genotype classified AD patients with AUC=0.84 and 0.80 in the test and replication analyses, respectively. Our results implicate methylation differences at the CpG island shores of AD-related genes in the onset of AD and suggest their diagnostic value.


Introduction
Moderate concordance rates for Alzheimer's disease (AD) among genetically identical twins suggest the possible involvement of epigenetics in the etiology of AD [1,2]. Among epigenetic components, DNA methylation and chromatin modifications are of great interest in AD because they have been shown to change with aging [3,4], which is the main cause of the disease, and their unintended changes affect gene expression [5]. Since DNA methylation differences in genomes between AD patients and cognitive normal elderly individuals are expected to not only reveal AD-susceptible genes, but also provide biomarkers for clinical purposes, they have extensively examined in blood and brains using gene-specific and genome-wide approaches [6][7][8][9][10].
Sanger or pyrosequencing and DNA methylation arrays have generally been employed for region-specific and genome-wide analyses, respectively. Although large numbers of differentially methylated genes have been reported, only a few genes in the brain and blood have been independently confirmed [7,11]. These findings suggest that methylation anomalies in AD samples are so marginal that depending on the sample size or sensitivity of the procedures employed, they cannot reproducibly be detected [12,13].
Another possibility is that genomic regions with evident differences were not simply targeted by these studies. The whole-genome bisulfite sequencing of large samples of . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https: //doi.org/10.1101//doi.org/10. /2020 cognitively normal and AD subjects with sufficient coverage may be the only approach to circumvent these limitations; however, this is prohibitive for large-scale studies due to the associated costs.
We are interested in identifying differentially methylated regions (DMRs) in blood DNA that may be used for the diagnosis of AD and its prognosis. We assumed that DMRs in AD-related genes, such as those identified by genome-wide association studies (GWASs) [14,15], in the brain or any other tissues confer susceptibility to the onset of AD through subtle, but long-lasting expression changes in these genes, similar to single nucleotide polymorphisms (SNPs). Based on this assumption, we herein examined whether DMRs exist in AD-associated genes in the blood of late-onset AD (LOAD) patients. We focused on a genic region designated as CpG island shores in AD-associated genes because methylation levels at CpG island shores are vulnerable under various conditions, such as tissue differentiation, reprogramming, aging, and disease including AD [16][17][18][19][20]. CpG island shores are defined as 2-kb-long regions that lie on both sides of a CpG island [16]. CpG islands are, on average, 1000 base pairs (bp) in length, characterized by dense clusters of CpG dinucleotides, and located around the promoter regions of 70% of human genes [21]. CpG islands are exempt from DNA methylation irrespective of gene expression, and the further away CpG dinucleotides are . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https: //doi.org/10.1101//doi.org/10. /2020 from CpG islands, the higher the chance of methylation. Therefore, methylated cytosines appear to be symmetrically distributed with respect to promoter CpG islands, and the cytosines of CpGs in CpG island shores are often slightly and moderately methylated [16]. A previous study demonstrated that gene expression levels were negatively associated with methylation levels at CpG island shores [16].
We herein demonstrated that, in blood, AD-related DMRs existed in CpG island shores in three out of six AD-related genes examined, CR1 (complement receptor 1), CLU (clusterin), and PICALM (phosphatidylinositol-binding clathrin assembly protein), and also that with the combination of APOE (apolipoprotein E) genotypes, the methylation levels of CLU in blood may be effectively used to differentiate AD patients from cognitive elderly with high sensitivity and specificity.
. CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Ethics statements
The present study was conducted with the informed consent of all individuals and with the approval of the Ethical Committee of the National Center for Geriatrics and Gerontology (NCGG). The design and performance of the present study involving human subjects were clearly described in a research proposal. All participants were voluntary and completed informed consent in writing before registering at the NCGG biobank, which collects human biomaterials and data for geriatrics research.

Patient sample collection
All 293 serum subjects and the associated clinical data were provided by the NCGG Biobank. Ninety-six subjects were patients with AD, 40 with VaD, 34 with DLB, and 27 with FTD, and 96 subjects were cognitively normal elderly controls (hereafter, controls). One of two groups of 48 AD patients and 48 age-and gender-matched controls was used as a test group and the other as a replication group.
The cognitive status and severity of dementia were assessed by the Mini-Mental State Examination (MMSE). The status of the APOE e4 allele genotype (the main genetic risk factor for AD) and the MMSE score were obtained from the NCGG biobank. All . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https: //doi.org/10.1101//doi.org/10. /2020 7 subjects were >60 years in age. All control subjects had a MMSE score of >25.

Methylation analyses
The methylation levels of target regions were assessed by Sanger sequencing, pyrosequencing, or by both methods. In either method, 200 ng of genomic DNA was initially treated with sodium bisulfite to convert non-methylated cytosines to uracil using the EZ DNA Methylation-Gold Kit (Zymo research) following the manufacturer's instructions. Target regions were PCR amplified by primer sets designed by the web tool, MethPrimer [22], for Sanger sequencing or by Pyrosequencing Assay Design Software ver.2.0 provided by Qiagen for pyrosequencing. The conditions of PCR, lengths of amplicons, and sequences of primers were shown in S1 Table. In Sanger sequencing, amplicons were cloned into the pGEM-T vector (Promega), which was then used to transform the E. coli strain, DH5α (Takara Bio). Plasmid DNA was amplified from each of 32 colonies by Templiphi (GE Healthcare) and used as a template for a dideoxy sequencing reaction (BigDye ver.3.1, Applied Biosystems). Sequences were elucidated on a Genetic Analyzer 3500 or 3130xL (Applied Biosystems), and data containing a minimum of twenty sequences were analyzed by QUMA [23] to quantify the mean methylation levels of PCR products. In pyrosequencing, the mean . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Identification of CpG island shores
To identify CpG island shores in AD-associated genes, we plotted CpG dinucleotide sites along the genes using a program designated CyGnusPlotter, which was written for the present study and is freely available via the internet (https://github.com/kzyskgch/CyGnusPlotter). These diagrams showed that, as reported in many human genes, there was a cluster of CpGs at their promoter regions, named the CpG island. Since CpG island shores are defined as regions that immediately flank CpG islands, the exact boundary was only identified after the edges of the hypomethylated region were confirmed.

Genotypes of SNPs
The genotypes of AD-related SNPs in CR1, CLU, and PICALM loci in our samples were provided by the NCGG Biobank, which supported a genome-wide association study (GWAS) in Japan to identify SNPs linked to AD. Since most samples used in the present study were part of larger samples utilized for GWAS, we were able to obtain the genotypes of SNPs in CR1, CLU, and PICALM loci in . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https: //doi.org/10.1101//doi.org/10. /2020 9 substantial numbers of our samples after due formalities.

Statistical analyses
Comparisons between two and more groups were made using the Student's t-test and ANOVA, respectively, by PRISM ver.5 (GraphPad Software, Inc., San Diego, CA, USA). The sensitivity and specificity of the measured variables for the diagnosis of AD were evaluated using a receiver operating characteristic (ROC) curve. ROC plots were drawn by Minitab 17 (Minitab Inc.). Regarding multiple biomarkers, logistic regression analyses were conducted to derive analytical expression for the risk of AD using methylation levels as continuous variables and the APOE genotype as a nominal variable. The classification performance of CLU methylation, the APOE genotype, and the combination of both, were assessed using the area under the ROC curve (AUC).
ROC is a plot of the probability of correctly classifying positive samples against the rate of incorrectly classifying true negative samples. Therefore, the AUC measure of an ROC plot is a measure of predictive accuracy. A DeLong test was used to compare AUC between groups [24]. All tests were two-tailed, and a p value <0.05 was considered to be significant.
. CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Selection of target regions
We selected target regions in the CpG island shores of six known AD-related genes in the AlzGene database ([25], http://www.alzgene.org/) for methylation analyses: APOE, CLU, CR1, PICALM, BIN1, and ABCA7 as follows: we initially referred to the distribution of CpG dinucleotides along the genes (Fig 1 and S1-S5 Figs). Promoter CpG islands were easily recognized as a cluster of CpG dinucleotides around the 5' ends of the genes, and regions juxtaposed to the promoter CpG islands were putative CpG island shores. We PCR-amplified portions of the putative CpG island shores at which at least several CpGs existed and quantified their methylation levels. Slightly or moderately methylated portions that met the criteria for CpG island shores were used in subsequent analyses. These portions were only found in a limited space in the case of the APOE gene (region II in Fig 1). The schematic shows the distribution of CpG dinucleotides along the APOE gene. The position of the transcription start site is defined as +1; hence, "-2,000" in the parenthesis . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14. Moderate DNA methylation was observed at a limited region (II) juxtaposed to the CpG island, a region called the "CpG island shore". Abbreviations: APOE, apolipoprotein E.
. CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

DNA methylation differences
We quantified the methylation levels of target regions in the six genes in 48 control and AD blood samples by cloning and sequencing PCR products from bisulfite-treated genomic DNAs. Demographic data for the samples included in the present study are shown in Table 1. When stratified by the disease status, we found that the mean methylation levels in three out of the six genes, PICALM, CR1, and CLU, were significantly lower in AD than in control samples (Fig 2). These results suggested that the development of AD was related to methylation changes in the CpG island shores of AD-related genes. AD-related hypomethylation in blood samples was also reported for TREM2 (Triggering receptor expressed on myeloid cells 2), another AD-related gene [27]. We judged that the region surveyed in their study corresponded to the CpG island shore of TREM2 based on its location (S6 Fig) and methylation level [27]. We compared the methylation levels of the same region in our control and AD samples by following the procedure reported previously. As shown in Fig 2, the mean methylation level was significantly lower in AD than in control samples, which was consistent with previous findings [27].
. CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14.20035683 doi: medRxiv preprint 1 3 . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Fig 2. Methylation rates at CpG island shores in seven AD-associated genes in the peripheral blood of control and AD subjects.
Each dot represents the mean methylation rate at the CpG island shores of the indicated genes in a subject. Subjects were composed of a group of 48 cognitive normal elderly (Ctrl) and 48 AD patients, named the test group, except for TREM2, for which another set of 96 subjects, named the replication group, was also examined. Methylation rates were quantified by Sanger sequencing, except for TREM2, for which pyrosequencing was employed [27]. The relative position of the CpG island shore investigated in APOE was region II in Fig 1, while those in the other six genes are shown in S1-S6 Figs.
The horizontal lines and error bars show the mean ± standard deviation. p-values were obtained using the Student's t-test. Abbreviations: Ctrl, control; AD, Alzheimer's disease

Replication and verification of methylation differences
To assess the reproducibility of the results obtained above, we compared the methylation levels of APOE and CLU in the replication set of 48 blood samples of control and AD using the same method as for the test set, and confirmed the absence and presence of methylation differences in APOE and CLU, respectively (Fig 3A, B).
. CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14.20035683 doi: medRxiv preprint 1 5 We then verified the methylation levels of CLU in two sets of blood samples using pyrosequencing, which directly measures the methylation levels of PCR products ( Fig   3C, D). Lower methylation levels in AD samples were confirmed in either set of blood samples. These results indicate that methylation differences between controls and AD patients were reproducible.

No effect of SNPs, MMSE, or age on methylation
SNPs that correlate with the methylation degree of nearby CpG sites in cis are called methylation quantitative trait loci (mQTLs). Previous studies suggested that . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14.20035683 doi: medRxiv preprint 1 6 DNA methylation acts as an intermediary of genetic risk [28][29][30] and mQTLs associated with AD have been reported [31,32]. Since the six AD-associated genes examined above were all discovered through GWAS, AD-related SNPs exist that are linked to CLU, PICALM, and CR1, in which lower methylated regions in AD blood were identified. Therefore, these SNPs may cause methylation differences in the three genes. To examine this possibility, we assessed the relationship between AD-related SNPs and the methylation degree, but found no relationship in any of the genes ( Fig   4A-C). These results demonstrated that these SNPs and methylation are independently associated with the onset of AD. The independence of AD pathology-related methylation changes from AD risk variants has also been reported [10,30].
We investigated whether the degree of methylation in the three genes was related to the scores of the Mini-Mental State Examination (MMSE) and age (Fig 4D-I) based on Pearson's correlation coefficient. However, no relationships were observed in any of the three genes. These results indicate that methylation levels in the genes do not reflect the severity of disease and they are unrelated to aging in the elderly. . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Methylation in non-AD dementia
To clarify how specific the methylation changes detected may be in dementia, we examined the methylation levels of APOE and CLU in blood samples of control, AD, dementia with Lewy bodies (DLB), vascular dementia (VaD), and frontotemporal dementia (FTD). Although no significant differences were observed in the methylation levels of APOE between the control and any dementia (Fig 5A), CLU methylation levels were lower in DLB than in the control, similar to AD samples ( Fig   5B). This result indicates that CLU hypomethylation is related to limited types of dementia.

dementia.
Methylation rates at the CpG island shores of APOE and CLU in DLB, VaD, and FTD were quantified by Sanger sequencing and differences relative to control subjects in the test group were assessed using one-way ANOVA with Dunnett's multiple comparison . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14.20035683 doi: medRxiv preprint 1 9 post-test. The positions of the CpG island shores examined were the same as those in

Blood-brain methylation discordance
In view of etiology, it is important to establish whether methylation changes in blood also occur in the brains of the same individuals. Since we had no brain samples available from the same individuals that provided the blood samples used in this study, we needed to clarify whether the degree of DNA methylation in the brain reflects that in blood in the same individuals at the three AD-related hypomethylated regions. To achieve this, we utilized two online searchable databases, designated five probes against the regions we surveyed: three probes for the CLU region and one each for CR1 and PICALM. We found that the CpG sites targeted by the probes in the . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14.20035683 doi: medRxiv preprint 2 0 brain showed either low or moderate methylation levels, suggesting that they were in the CpG island shores. However, interindividual variations in the methylation levels of these CpGs were less in the brain than in blood, and no correlation was noted between blood and any brain regions, except for a CpG site in PICALM in the entorhinal cortex (S8C Fig). These results indicated that overall methylation levels in the CpG island shores of the three genes in blood do not reflect those in brain.

Classification utility
The identification of subjects at high risk of developing AD is important for early interventions and clinical trials. Multiple logistic regression (MLR) analyses were used to derive linear classifier models that differentiate control and AD subjects using the data of test and replication groups (S3 Table). Methylation levels of CLU in the test and replication groups were employed for the analyses with the number of APOE ε4 alleles as covariates since the APOE epsilon4 (ε4) allele is the strongest genetic risk factor for AD. The performance of each model was assessed by ROC, which establishes classification values, and AUC, which assesses multimarker classification performance. Stepwise feature selection selected the top-performing linear combination of CLU methylation and the APOE genotype as MLR variables in . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14.20035683 doi: medRxiv preprint 1 both groups, which attain AUCs of 0.83 and 0.85 by the data of the test and replication groups, respectively (Fig 6A). Since slightly better classification performance was obtained with the data of the replication group, we applied the model to the data of the test group, which still yielded high AUC of 0.80 (Fig 6B). In both cases, the differences in AUCs between APOE ε4 data alone and the combination of methylation and APOE ε4 data were significant (DeLong's test; p=0.0083 for A, p= 0.0064 for B). . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
. CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Discussion
In the present study, Sanger sequencing of more than twenty thousand clones from 96 blood samples quantified the methylation levels of the CpG island shores of the top six AD-associated genes in the AlzGene database [25], which provides a comprehensive field synopsis of genetic association research conducted on AD. Although it was a very low throughput, we detected decreases in DNA methylation in three out of the six genes.
Furthermore, we replicated AD-related DNA hypomethylation in the CpG island shore of TREM2 in our blood samples using pyrosequencing. While significant DNA methylation changes were not observed in APOE, BIN1, or ABCA7, the regions surveyed were limited fractions of the CpG island shores in these genes, and, thus, we cannot deny the possibility of AD-related methylation changes in other regions in these genes. Alternatively, AD-associated DNA methylation changes might occur in a tissue-specific manner since CpG dinucleotides with AD pathology-related methylation were found in the ABCA7 and BIN1 loci in the brain's DNA [10,20,35].
Genome-wide surveys that aimed to detect AD-related methylation changes in blood have been conducted using Illumina HM450K and 850K (MethylationEPIC) arrays; however, AD-associated hypomethylation was not detected in CR1, CLU,or PICALM [8,36,37]. This discrepancy may be explained by the modest mean differences detected in CR1, CLU, and PICALM, which were all less than 10%, because depending on the probes, these . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
It is reasonable to assume that methylation changes relevant to AD, if any, occur in the brain, a prominently affected tissue, and that blood methylation changes most likely reflect peripheral responses to the disorder rather than causally related variations. Thus, methylation changes in blood have been searched for as surrogate markers for brain methylation changes in most cases. Contrary to expectations, co-methylation changes that occur in both the blood and brains of AD patients were shown to be limited to a subset of DNA methylation sites [8,36]. Methylation changes in the brains with neuropathology were not replicated in CD4+ lymphocytes [38] or in the whole blood in AD [9,39] from the same individuals. Similarly, in silico analyses in the present study showed no relationships between methylation in blood and the brain at CpG sites in AD-associated DMRs, and the DMRs found in blood were not detected in previous studies that searched for DMRs in AD brains [9,10]. These results were consistent with the finding that for the majority of DNA methylation sites, interindividual variations in whole blood are not a strong predictor of those in the brain [34]. Therefore, the methylation changes detected in blood in the present study appear not to occur in AD brains. On the other hand, the chance of AD-related methylation changes in three out of the six AD-linked genes in blood appeared to be more than coincidence. A previous study reported that age-related cognitive impairment was improved . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14.20035683 doi: medRxiv preprint 2 5 by the introduction of young blood in old mice (heterochronic parabiosis) [40,41]; therefore, epigenetic changes that confer susceptibility to AD do not necessarily appear to be confined to neuronal cells and may also occur in blood.
A possible change that concomitantly occurs with the hypomethylation of CpG island shores is an increase in the expression of genes because CpG island shore methylation is strongly and negatively related to gene expression [16]. Although it currently remains unclear whether CpG island shore hypomethylation in CLU, CR1, and PICALM is related to their up-regulation, the inverse correlation is true for TREM2 in blood [27], and it may also be the case for CLU because the transcript levels of CLU were shown to be elevated in the blood of AD patients [42,43]. Based on the long duration of this disease, even small changes in DNA methylation in AD-related genes, such as those observed in the present study, that accompany small changes in their expression may confer susceptibility to the progression of AD.
Although further studies are needed to clarify whether hypomethylation in AD-related genes in blood is causal or consequential, these DNA methylation signatures have potential as clinical biomarkers for AD. We herein demonstrated that in combination with the APOE genotype, CLU methylation in CpG island shores offered good prediction performance for the diagnosis of AD in our subjects. Since the methylation levels of CR1, CLU, and PICALM are not related to MMSE scores, methylation levels in the CpG island . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14.20035683 doi: medRxiv preprint 2 6 shores of these genes in the blood of AD patients may have already been lower than those in controls before the onset of the disease. This is worth testing with samples in longitudinal studies. Furthermore, we discovered that the hypomethylation of CLU occurred in the blood of DLB, but not in VaD or FTD. This result suggests that hypomethylation is a disease-specific phenomenon and may reflect some shared underlying pathophysiological mechanisms between the two diseases, as suggested in a previous study [44], while these results may also be explained, in part, by possible diagnostic misclassification or the comorbidity of DLB with AD. Neurodegenerative disease-specific differential DNA methylation has also been reported for the ANK1 gene [45].
Ideal biomarkers for AD were proposed to have sensitivity and specificity of >80% [46]; however, the diagnostic ability obtained by the combinatorial use of CLU methylation and the APOE genotype did not meet these criteria. Therefore, better DMRs for the clinical use of blood DNA in AD need to be identified. Further efforts to identify DMRs in other AD-associated genes will be a rational approach of choice. The success of combined uses of methylation rates and another variable regarding AD, such as the neuritic plaque burden [9, 10] and cognitive measure, [8,47], may be referred to for unbiased screening of AD-related methylation markers in blood.
. CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Acknowledgments
We thank the NCGG Biobank for providing the study materials, clinical information, and technical support.
. CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14.20035683 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

5.
Pal S, Tyler JK. Epigenetics and aging. Science advances. 2016;2(7):e1600584. . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14. . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

15.
Lambert JC, Heath S, Even G, Campion D, Sleegers K, Hiltunen M, et al. . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14. . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14.  . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14.20035683 doi: medRxiv preprint 4 0 The figure is drawn similar to that in S1 Fig, except for the dashed line, which indicates the region for which methylation levels were quantified by pyrosequencing. . CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14.20035683 doi: medRxiv preprint 1 The figure is drawn similar to that in S1 Fig, except that the underlined region was the region for which methylation levels were analyzed by pyrosequencing.

S7 Fig.
In silico examination of blood-brain methylation concordance by the online tool,

BECon.
Each plot shows the inter-individual variability of the methylation level at a CpG site across 16 subjects. The five CpGs examined were derived from three AD-related DMRs identified in the present study: one CpG in CR1 (A), three CpG sites in CLU (B), and one CpG site in PICALM (C). The numbers in "Correlation" columns in the tables indicate Spearman's rank correlation coefficients (r S or ρ), which were obtained by comparisons of methylation levels between blood and either one of three different cortical regions (Broadmann area 10 (BA10), prefrontal cortex; Broadmann area 7 (BA7), parietal cortex; and Broadmann area 20 (BA20), temporal cortex). CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.03.14.20035683 doi: medRxiv preprint 4 2 (B), and a CpG site in PICALM (C), in blood and four brain regions (PFC, prefrontal cortex; EC, entorhinal cortex; STG, superior temporal gyrus; CER, cerebellum) from the same individual donors were plotted in the rectangles, and the correlation of DNA methylation in blood with the four brain regions are plotted in square boxes. Methylation data were generated by Hannon et al [34]. S1 Table. Primers and PCR conditions for bisulfite genomic sequencing.

S2 Table. Association of known AD-related SNPs with AD in NCGG samples (test and
replication groups).

S3 Table. Results of binominal regression analyses
. CC-BY 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.
is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10. 1101/2020