{"paper_id":"62a8d84c-fd79-4111-8e20-a88620b7af4a","body_text":"R E S E A R C H Open Access\nThe influence of menstrual cycle and\nendometriosis on endometrial methylome\nMerli Saare 1,2,3*†, Vijayachitra Modhukur 4†, Marina Suhorutshenko 1, Balaji Rajashekar 4, Kadri Rekker 1,2,\nDeniss Sõritsa 1,2,5,6, Helle Karro 2,6, Pille Soplepmann 6, Andrei Sõritsa 5, Cecilia M. Lindgren 7, Nilufer Rahmioglu 7,\nAlexander Drong 7, Christian M. Becker 8, Krina T. Zondervan 7,8, Andres Salumets 1,2,3 and Maire Peters 1,2\nAbstract\nBackground: Alterations in endometrial DNA methylation profile have been proposed as one potential mechanism\ninitiating the development of endometriosis. However, the normal endometrial methylome is influenced by the\ncyclic hormonal changes, and the menstrual cycle phase-dependent epigenetic signature should be considered\nwhen studying endometrial disorders. So far, no studies have been performed to evaluate the menstrual cycle\ninfluences and endometriosis-specific endometrial methylation pattern at the same time.\nResults: Infinium HumanMethylation 450K BeadChip arrays were used to explore DNA methylation profiles of\nendometrial tissues from various menstrual cycle phases from 31 patients with endometriosis and 24 healthy women. The\nDNA methylation profile of patients and controls was highly similar and only 28 differentially methylated regions (DMRs)\nbetween patients and controls were found. However, the overall magnitude of the methylation differences between\npatients and controls was rather small (Δβ ranging from –0.01 to –0.16 and from 0.01 to 0.08, respectively, for hypo- and\nhypermethylated CpGs). Unsupervised hierarchical clustering of the methylation data divided endometrial samples based\non the menstrual cycle phase rather than diseased/non-diseased status. Further analysis revealed a number of menstrual\ncycle phase-specific epigenetic changes with largest changes occurring during the late-secretory and menstrual phases\nwhen substantial rearrangements of endometrial tissue take place. Comparison ofcycle phase- and endometriosis-specific\nmethylation profile changes revealed that 13 out of 28 endometriosis-specific DMRs were present in both datasets.\nConclusions:The results of our study accentuate the importance of considering normal cyclic epigenetic changes in\nstudies investigating endometrium-related disease-specific methylation patterns.\nKeywords: DNA methylation, Endometriosis, Endometrium, Epigenetics, Illumina 450K, Menstrual cycle, Microarray\nBackground\nDNA methylation, an important epigenetic mechanism\ncrucial for maintaining tissue-specific gene expression pat-\ntern [1, 2], is suggested to be one possible molecular feature\nthat contributes to the development of many human\ndiseases, including endometriosis. Deviation from normal\nDNA methylation level may lead to alterations in the cellu-\nlar microenvironment, affect gene expression and initiate\npathologic processes. During the last decade, several studies\nhave reported abnormal methylation patterns of selected\ngenes, e.g. steroidogenic factor 1 [3], progesterone receptor\nB [4], oestrogen receptor-β [5], HOXA10 [6 – 8], HOXA11\n[9], COX-2 [10] and aromatase [11], in endometriotic le-\nsions and endometria of endome triosis patients. Advance-\nments in microarray technology have now allowed to assess\nDNA methylation on a global scale; and to date, already\nfour studies, though relatively small and using different\narray platforms, have suggest ed genome-wide differences\nbetween endometriosis patients ’ endometria and lesions\n[12– 14] or between endometrial tissues of patients and\ncontrols [15]. Studies on endo metriotic lesions or stromal\ncells originating from lesions revealed clear evidence of\nepigenetic alterations that c ould be associated with the\ndisease [12– 14]. The issue whether the primary source of\nthese alterations is endometria lt i s s u eo re p i g e n e t i ca l t e r -\nations occur during the formation of lesions in abdominal\n* Correspondence: merli.saare@ut.ee\n†Equal contributors\n1Competence Centre on Health Technologies Tartu, Tartu, Estonia\n2Tartu University Women ’ s Clinic, Tartu, Estonia\nFull list of author information is available at the end of the article\n© 2016 Saare et al. Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0\nInternational License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and\nreproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to\nthe Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver\n(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.\nSaare et al. Clinical Epigenetics  (2016) 8:2 \nDOI 10.1186/s13148-015-0168-z\n\ncavity in response to change d abdominal environment has\nbeen addressed in the study by Naqvi et al. [15], who evalu-\nated endometrial DNA methylation profile of patients and\ncontrols and suggested that some epigenetic alterations\noccur already in the endometria of endometriosis patients.\nFurthermore, it has been suggested that alterations on gen-\netic and epigenetic levels during early embryogenesis may\nlead to endometriosis development because the fine-tuning\nmechanisms responsible for the correct development of the\nfemale genital system are disrupted [16, 17].\nEndometrium is a unique tissue undergoing cyclic\nbreakdown and regeneration, and similarly to other tissues\nand cell types [18, 19], has its own distinct DNA methyla-\ntion pattern that is influenced by cyclic hormonal changes\n[20]. The menstrual cycle phases of the studied women\nwere not shown in a previous study examining endomet-\nrial methylome of endometriosis patients [15]; however, in\nthe light of the recent knowledge about the significant\nimpact of menstrual cycle phases on the endometrial\nmethylome of healthy women [20], it is apparent that\nnormal cyclic epigenetic signature of endometrial tissue\nshould be considered when studying endometrial tissue-\nrelated disorders, like endometriosis.\nDuring the past 10 years, a number of studies have been\nconducted to find reliable diagnostic biomarkers for endo-\nmetriosis, unfortunately with little success. The need for\nnon-invasive or minimally invasive biomarkers is difficult\nto underestimate as the average delay between the onset of\nsymptoms and the surgical diagnosis is almost 7 years [21].\nSuch biomarkers would enable to avoid the unnecessary\nlaparoscopy while endometriosis is suspected but not\npresent and make possible to get the right diagnosis of\nendometriosis much earlier. Therefore, the aim of the\ncurrent study was to reveal potential epigenetic biomarkers\nfrom endometrial DNA of endometriosis patients’ endome-\ntria, from endometriosis centres in Tartu (Estonia) and\nOxford (UK), by considering the menstrual cycle dependent\nchanges. Furthermore, we ai med to investigate the men-\nstrual cycle-specific methylation signature to widen the\nknowledge about the epigenetic changes occurring during\nendometrial growth across the entire menstrual cycle.\nResults\nGenome-wide DNA methylation analysis of endometrial\ntissues from patients with endometriosis and healthy\nwomen\nEndometrial samples from 31 endometriosis patients and\n24 disease-free women, from menstrual (M, n = 5), prolif-\nerative (P ,n = 5), early-secretory (ES, n = 8), mid-secretory\n(MS, n = 26) and late-secretory (LS, n = 11) menstrual\ncycle phases (T able 1), were used for genome-wide DNA\nmethylation analysis.\nThe pipeline of the study is given in Fig. 1. Principal\ncomponent analysis (PCA) clustering of the normalised\ndata was used to describe the endometrial DNA methy-\nlation profiles of patients with endometriosis and healthy\nwomen (Additional file 1). Approximately 19.6 % of vari-\nation across all studied probes was accounted for in the\nfirst two principal components (12.4 % for PC1 and\n7.2 % for PC2), and no significant segregation between\npatients and controls was noticed, indicating that the\noverall DNA methylation profile between patients and\ncontrols was very similar. Still, if we compared the\nmethylation profiles of all patients with endometriosis to\nhealthy women, we found 28 differentially methylated re-\ngions (DMRs) (false discovery rate, FDR <0.05, Δβ ranging\nfrom – 0.01 to – 0.16 and from 0.01 to 0.08) from which 16\nwere associated to known genes ( PI3, SLC43A3, MGAT5B,\nMUC4, HIVEP3, FGG, CLCF1, CANT1, LTK, AHRR,\nAKR1B1, APEH, CST11, ELOVL4, HBE1 and NEGR1)\n(Additional file 2). One of the top-ranking intergenic\nDMRs was located on chromosome locus 7p15.2, about\n13 kb upstream from HOXA gene cluster.\nUnsupervised hierarchical clustering of the same data\n(Fig. 2) revealed two main branches that divided endomet-\nrial samples based on the menstrual cycle phase rather\nthan diseased/non-diseased status. The first branch in-\ncluded all LS phase samples ( n = 11), four out of five M\nphase ( n = 4) and some MS phase ( n =7 ) s a m p l e s , w h i l e\nthe other branch included the majority of samples from\nMS ( n = 19) phase, ES phase ( n = 8), P phase ( n =5 ) a n d\none remaining sample from M phase. Therefore, to con-\nsider the impact of menstrual cycle on endometriosis-\nspecific methylation signature, we determined the differ-\nences associated with menstrual cycle phases.\nMenstrual cycle-specific DNA methylation signature and\ngene ontology (GO) analysis of differentially methylated\nregions\nAs unsupervised hierarchical clustering analysis revealed\nno segregation between patients and controls, both\ngroups were combined (altogether 55 individuals) to find\nmenstrual cycle phase-specific methylation changes. The\nstudied individuals were divided into five groups accord-\ning to the menstrual cycle day at the time of biopsy\ncollection: (1) M ( n = 5), (2) P ( n = 5), (3) ES ( n = 8), (4)\nMS ( n = 26), and (5) LS phase ( n = 11) groups. To assess\nthe overall methylation pattern characteristic to each\ncycle phase, the methylation data of each phase was\ncompared to other phases. A large number of differen-\ntially hypo- and hypermethylated regions (FDR < 0.05)\nwere noticed when either or both M and LS phases were\ninvolved in comparisons, while only some DMRs were\nfound in comparisons between P , ES and MS phases\n(Table 2, Additional file 3).\nAs the endometrial tissue growth and degradation during\nthe menstrual cycle is a cont inuum where 1-cycle phase\nprogresses to another, only genes that were differentially\nSaare et al. Clinical Epigenetics  (2016) 8:2 Page 2 of 10\n\nmethylated in adjacent phases (M vs. P , MS vs. LS and LS\nvs. M) were included in the downstream analysis. The\ncomplete lists of DMRs were subjected to enrichment\nanalysis that revealed signif icant enrichment for multiple\nontology terms (the lists of Gene Ontology— GO terms and\nKyoto Encyclopaedia of Genes and Genomes— KEGG path-\nway analysis are outlined in Additional file 4 and 5).\nThe CpG island (DNA sequence at least 200 bp and\nGC content greater than 50 %) hypermethylation in gene\npromoter regions has been associated with repression of\ngene transcription and hypermethylation of regions next\nto CpG islands, island shores (2 kb regions upstream\nand downstream of the CpG islands) and shelves (4 kb\nregions upstream and downstream of the CpG islands)\nwith higher gene expression [22]. Therefore, the location\nof the differentially methylated CpG sites in relation to\ngenomic elements such as CpG islands, island shores\nand shelves, open sea (all remaining sequence) and gene\nstructure (promoter region, 5 ′ UTR, first exon, gene\nbody, 3 ′ UTR and intergenic) was analysed to investigate\ndifferential representation of functional categories between\ndifferent menstrual cycle phases (Fig. 3). The assessment of\ndistribution of hypo- and hypermethylated DMRs showed\nslight overrepresentation of CpGs located in the open sea\n(ranging from 34– 66 %) compared to CpGs located within\nand next to islands (island, shores and shelves, ranging\nfrom 24– 42 %), when the CpG distribution relative to CpG\nislands was analysed. The lowest number of CpGs was seen\nin shelves (ranging from 4 – 7 %), whereas higher number\nof CpG sites was located within CpG islands (ranging from\n5– 11 %) and the highest number of CpG sites was located\nin shores (ranging from 9 – 28 %). When the distribution of\nCpGs in relation to genes was examined, it was evident\nthat large proportions of CpGs were located in intergenic\nregions and gene bodies (ranging from 38 – 75 %) and only\na minority of CpGs (ranging from 8 – 25 %) were in gene\npromoter areas. However, when enrichment analysis of\nDMRs based on their location (promoter/gene body) was\ncarried out, no GO terms or KEGG pathways characteristic\nto specific menstrual cycle phase were found.\nTable 1 General characteristics of the study participants\nMicroarray study Patients with endometriosis ( n = 31) Disease-free women ( n = 24)\nEstonian patients Oxford patients Estonian controls Oxford controls\n(n = 24) ( n =7 ) ( n = 17) ( n =7 )\nAge (years ± SD) 31.0 ± 4.0 36.0 ± 5.0 30.1 ± 3.2 34.2 ± 6.2\nBMI (mean, kg/m 2 ± SD) 21.8 ± 3.1 23.6 ± 2.0 23.6 ± 4.2 26.0 ± 4.3\nSmoking (n)0 2 0 0\nStage I–II ( n) 16 3 NA NA\nStage III–IV (n)8 4 N A N A\nOnly endometrioma ( n)0 4 N A N A\nOnly peritoneal lesions ( n) 14 3 NA NA\nPeritoneal lesions together with endometrioma ( n) 10 0 NA NA\nMenstrual cycle characteristics\nMenstrual phase (days 1 –5), (n)0 4 0 1\nProliferative phase (days 6 –14), (n)02 0 3\nEarly-secretory phase (days 15 –20),(n)7 0 0 1\nMid-secretory phase (days 21 –23), (n)8 1 1 7 0\nLate-secretory phase (days 24 –28), (n)9 0 0 2\nValidation study Patients with\nendometriosis\n(n = 15)\nDisease-free\nwomen (n = 14)\nAge (years ± SD) 31.0 ± 3.39 32.0 ± 2.7\nBMI (mean, kg/m 2 ± SD) 20.0 ± 3.92 23.1 ± 5.73\nSmoking (n)1 2\nMid-secretory phase (days 21 –23), (n)7 7\nLate-secretory phase (days 24 –28), (n)8 7\nStage I–II ( n)1 3 N A\nStage III–IV (n)2 N A\nNA not applicable\nSaare et al. Clinical Epigenetics  (2016) 8:2 Page 3 of 10\n\nFig. 1 Schematic representation of the study design and main steps of the data analysis\nFig. 2 Hierarchical clustering analysis of all endometrial samples included into the study. Sample codes starting with E indicate patients with\nendometriosis and H indicates healthy individuals. Samples with the same index number are duplicates\nSaare et al. Clinical Epigenetics  (2016) 8:2 Page 4 of 10\n\nNext, the complete lists of differentially methylated genes\nfrom each cycle phase comparisons were subjected to Venn\nanalysis to reveal genes characteristic to specific menstrual\ncycle phases (Additional file 6). The results showed 5 hypo-\nand 5 hypermethylated genes for M phase and 127 hypo-\nand 113 hypermethylated genes for LS phase (central\nintersection in the Venn diagram, gene lists are given in\nAdditional file 7) but no enrichment of specific GO terms\nor KEGG pathways was found.\nDifferentially methylated genes between patients and\ncontrols—effect of menstrual cycle phases\nAs menstrual cycle phase comparisons revealed several\ndifferences in the methylation pattern, we compared the\nlists of endometriosis-specific differentially methylated\ng e n e sa n dr e g i o n st ot h em e n s t r u a lc y c l e - s p e c i f i ca l t e r -\nations. Results showed that ei ght out of 16 differentially\nmethylated genes found in pa tients with endometriosis\noverlapped with the menstrual cycle-related genes: seven\ngenes (PI3, SLC43A3, MGAT5B, MUC4, HIVEP3, FGG and\nCANT1) from comparison between MS and LS phases and\no n eg e n e(LTK) from M to P comparison. The remaining\neight differentially methylated genes — AHRR, AKR1B1,\nAPEH, CST11, ELOVL4, CLCF1, HBE1 and NEGR1w e r e\nnot related to the menstrual cycle changes. From DMRs\nthat were not related to any genes, five were also found in\nthe lists of menstrual cycle-specific genes. However, the\ntop-ranking DMR near the HOXA gene cluster was not\nfound to be associated with any specific menstrual cycle\nphase. To eliminate all potential confounders that may\ncome from menstrual cycle phase differences, we also com-\npared patients and controls only from MS phase group\nbecause this was the group with the largest number of indi-\nviduals (8 patients vs. 17 controls) in our dataset. Interest-\ningly, the MS phase group analysis revealed no DMRs.\nValidation of methylation data by direct bisulfite\nsequencing\nTo confirm the results of microarray analysis, four CpG\nsites located in the promoter regions (two CpGs from\nTable 2 The number of differentially hyper- and hypomethylated DMRs and genes between menstrual cycle phases\nDMR and genes\n(hyper-/hypomethylated)\nM( n =5 ) P ( n =5 ) E S ( n =8 ) M S( n = 26) LS ( n = 11)\nM( n = 5) DMR 1009/1775 130/189 363/855 288/368\nGenes 632/1066 92/116 254/512 208/222\nP( n = 5) DMR 1009/1775 0/0 1/2 3045/1650\nGenes 632/1066 0/0 1/0 1768/936\nES (n = 8) DMR 130/189 0/0 0/5 2806/1208\nGenes 92/116 0/0 0/3 1015/635\nMS (n = 26) DMR 363/855 1/2 0/5 2806/1208\nGenes 254/512 1/0 0/3 1616/704\nLS (n = 11) DMR 288/368 3045/1650 1727/1050 2806/1208\nGenes 208/222 1768/936 1015/635 1616/704\nDMR differentially methylated regions, M menstrual phase, P proliferative phase, ES early-secretory phase, MS mid-secretory phase, LS late-secretory phase\nFig. 3 Pie charts of DMRs between different menstrual cycle phases in relation to CpG island and relative to gene. CpG content together with\nneighbourhood context was defined as (i) open sea; (ii) island— DNA sequence at least 200 bp and GC content greater than 50 %, island shores— 2k b\nregions upstream and downstream of the CpG islands and shelves— 2 kb regions upstream and downstream of the CpG island shores and (iii) others\n(DMRs with several annotations). Gene context was defined as promoter region (TSS1500— 201 to 1500 bp upstream of transcription start site,\nTSS200— 200 bp to transcription start site and 5′UTR), the 1st exon of transcript; the gene body; 3′UTR and NA— non-island and others (DMRs with\nseveral annotations)\nSaare et al. Clinical Epigenetics  (2016) 8:2 Page 5 of 10\n\nCST11 gene, one from PI3 gene and one from SLC43A3\ngene) with differential methylation between patients with\nendometriosis and healthy women were selected for\nvalidation analysis by conventional bisulphite Sanger se-\nquencing in an extended group of patients and controls\nfrom LS ( n = 15) and MS phase ( n = 14). The correlation\nanalysis between microarray and bisulphite sequencing\ndata showed strong correlation (Pearson ’ s correlation\ncoefficient, PCC > 0.85, P < 0.001) for SLC43A3 and\nCST11 and moderate correlation (PCC = 0.58, P = 0.07)\nfor PI3. From four analysed CpG sites, only the CpG\nfrom SLC43A3 gene showed statistically significant dif-\nferential methylation between MS patients and controls\n(P = 0.03).\nDiscussion\nTo the best of our knowledge, this is the first study\nassessing the methylome of endometria of endometriosis\npatients and controls using Infinium HumanMethylation\n450K BeadChip array and taking into account DNA\nmethylation changes during the menstrual cycle. The re-\nsults of this study suggest that overall endometrial DNA\nmethylation signature is highly similar between patients\nwith endometriosis and healthy women but largely influ-\nenced by the menstrual cycle phases. Additionally, our\nstudy describes normal endometrial methylome through-\nout the menstrual cycle and shows that the largest\nchanges in epigenetic signature occur in late-secretory\nand menstrual phases.\nThe usability of epigenetic biomarkers in clinical setting\nhas been accepted and new and simple methodologies\nallowing straightforward DNA methylation biomarker\ndetection in routine diagnostics have already been devel-\noped [23]. Previous endometriosis studies have provided\nevidence that the epigenetic changes not only occur in ec-\ntopic endometriotic lesions but are already present in the\neutopic endometrium of endometriosis patients [15].\nTherefore, the combination of eutopic endometrium that\nis easily obtainable by the semi-invasive sampling proced-\nure and assessment of DNA methylation markers could\noffer an excellent source for epigenetic biomarker discov-\nery. So far, four microarray-based studies in eutopic endo-\nmetria [15], eutopic/ectopic endometria [12] and primary\nstromal cell cultures of eutopic and/or ectopic endometria\n[13, 14] have been performed focusing rather on disease\npathogenesis than on clinical usability. Only one study\nconcentrated on eutopic endometria [15] in the perspec-\ntive of using epigenetic markers as potential targets for\ntherapeutic agents. Despite finding a large number of\ndifferentially methylated genes, the authors concluded that\nmethylation and demethylation are both common events in\nendometrium, making the broad use of therapeutics\naffecting the methylation level impractical [15]. In our\nstudy, we tried to find endometrial epigenetic markers\nuseful for diagnostic purposes. However, the results of our\nstudy indicated that endometrialt i s s u ee p i g e n e t i cs i g n a t u r e\nin patients and controls is highly similar and only a few\nDMRs were found, indicating that alterations in endomet-\nrial methylation pattern are not common in endometriosis.\nThe lack of substantial differences in endometrial epigen-\netic signature in endometriosis was proposed also by\nanother study [13], where cultured primary stromal cells\nfrom eutopic and ectopic endometria of endometriosis\npatients and healthy women were used. Therefore, we\nsuggest that endometrial DNA methylation differences do\nnot provide good biomarkers with acceptable sensitivity\nand specificity for discrimination of patients with endo-\nmetriosis and healthy women.\nThe further validation of selected CpGs in extended\nsubsets of patients and controls from MS and LS phase\nconfirmed differential methylation of CpG in SLC43A3\npromoter, however, only between the MS patients and\ncontrols. The SLC43A3 hypermethylation was noticed in\nMS phase in our menstrual phase study and also while all\npatients were compared to controls, indicating influence\nboth from disease and menstrual cycle phase. Our results\nare supported by the study conducted by T amaresis et al.\n[24] who found that several genes showing differential\nmethylation in our study (such as AHRR, APEH, ELOV4,\nPI3,S LC43A3, MUC4, CANT and CLCF1) revealed also\ndifferential expression between patients and controls from\ncertain menstrual cycle phases. Interestingly, SLC43A3\nwas found to be differentially expressed only between MS\nphase patients and controls suggesting that small-scale\nmethylation alterations can probably affect the expression\nof this gene. There is only some data about the function of\nSLC43A3, but very recently, it was proposed that SLC43A3\nis a purine-selective nucleobase transporter [25]. SLC43A3\nis expressed during embryogenesis [26] but the possible\nrole in endometrium or endometriosis development re-\nmains to be elucidated.\nThere is an evidence both on transcriptome [27, 28]\nand epigenome levels [20] that endometrial molecular\nsignature is largely influenced by the menstrual cycle.\nThe significant impact of menstrual cycle phases to\noverall endometrial methylome was confirmed also in\nour menstrual cycle phase-specific analysis where we\nsaw that major epigenetic changes occurred while MS\nphase turned to LS phase, by which point, the endomet-\nrial tissue has reached its maximal thickness and\nsecretory capacity, predecidual changes begin in the\nstroma and the endometrium is ready for embryo im-\nplantation. However, if no implantation takes place, the\ndegradation processes are initiated. Also, significant\nchanges were found between LS and M phases, when the\ndesquamation of the tissue is followed by endometrial shed-\nding and menstruation, and between M and P phases, when\nactive repair and regeneration processes in endometrial\nSaare et al. Clinical Epigenetics  (2016) 8:2 Page 6 of 10\n\ntissue are taking place. In light of these results, it is evident\nthat normal endometrial methylation level fluctuations dur-\ning the menstrual cycle should be taken into account while\nsearching endometrial biomarkers.\nHowever, we believe that the relevance of epigenetic\nmarkers in the context of disease pathogenesis or menstrual\ncycle biology cannot be underestimated. For instance,\none of the most statistically significantly hypomethy-\nlated DMRs in patients was an intergenic CpG island\nabout 13 kb upstream from the HOXA gene cluster.\nWhether small but statistically significant differences in\nmethylation levels could affect gene expression levels is\ncurrently unknown, but previous studies have shown\nthat the members of HOXA cluster, HOXA10 and\nHOXA11 were differentially methylated in stromal cells\nobtained from endometriomas [14] and in eutopic en-\ndometria from patients [6, 7, 9, 29, 30] compared to\nhealthy controls, and hypermethylation of the HOXA\ngenes was accompanied by lower transcript and protein\nlevels in endometrium of endometriosis patients [6, 9].\nIn a recent review, Kobayashi et al. [31] assessed aberrantly\nexpressed genes in endometriosis during the process\nof decidualization and normal window of implantation.\nAuthors suggested that impair ed decidualization and dys-\nfunctional expression of genes related to Müllerian embryo-\ngenesis (like the downstream targets of HOXA10) could be\ncritical to the development of endometriosis. Also, it was\nproposed that DNA methylation of specific genes could\npartly explain the link between early exposure to a detri-\nmental fetal environment and an increased risk of develop-\ning endometriosis later in life [31]. Furthermore, it has been\nproposed that in utero exposure to endocrine disruptor\nbisphenol could be one potential cause triggering the\nabnormal fetal endometrial cell migration into ectopic\nlocation, as mice exposed in utero to bisphenol exhibited\nendometriosis-like phenotype [32]. One of the differentially\nmethylated genes in our study was AHRR,w h i c hs h o w s\nincreased gene expression in fetal tissues exposed to en-\nvironmental or even lower levels of bisphenol [33]. It has\nbeen proposed that developm ental exposure to environ-\nmental toxins may induce irreg ular methylation patterns\nand thereby permanently alter the expression of AHRR\n[34]. The relevance of AHRR methylation to theory of\nendometrial origin of endometriosis is intriguing and worth\nfurther examination.\nSome limitations of our study should be highlighted.\nAlthough analysed samples c overed the whole menstrual\ncycle, the size of some study groups (e.g. M and P phases)\nwas rather small. Moreover, the limited number of samples\nfrom particular menstrual cycle phases restricted the possi-\nbility to compare patients and controls from each phase\nseparately. Furthermore, histological endometrial dating\nwas available only for healthy volunteers from MS group\nand therefore, as the self-reported day of menstrual cycle is\nless accurate for phase dating, it could have some negative\nimpacts on menstrual cycle phase-specific analysis.\nConclusions\nThe results of this study demonstrated that endometrial\nDNA methylation profile of women with and without\nendometriosis was highly similar and thus, epigenetic\nmodifications in endometria are probably not the pri-\nmary source contributing to endometriosis development.\nAlthough some DMRs between patients with endometri-\nosis and controls were found, the magnitude of the\nmethylation differences was too small to enable discrim-\nination between patients and controls. The findings of\nthis study provide new knowledge about the normal epi-\ngenetic changes occurring across the menstrual cycle\nphases and accentuate the importance of considering\nnormal cyclic epigenetic changes when looking for dis-\nease specific endometrial DNA methylation changes.\nMethods\nEthics statement\nThe study was approved by the Research Ethics Committee\nof the University of Tartu (219/M-15) and written informed\nconsent was obtained from all p articipants. Tissues from\ncases and controls from Oxford originated from the\nENDOX study, which was approved by the NRES Commit-\ntee South Central-Oxford Research Ethics Committee (09/\nH0604/58).\nStudy subjects and tissue processing\nAltogether, 31 patients and 24 disease-free women were\nrecruited into the microarray study (T able 1). General\ncharacteristics, such as age and BMI were similar between\npatients and all controls (Student’ s t test, P >0 . 0 5 ) .\nEndometrial tissue samples were collected from 31\npatients undergoing laparoscopy at the T artu University\nHospital Women ’ s Clinic, Elite Clinic (T artu, Estonia,\nn = 24) and John Radcliffe Hospital (Oxford, UK, n =7 ) . I n\nall cases, the diagnosis was histologically confirmed and\ndisease severity was determined according to the American\nSociety for Reproductive Medi cine revised classification\nsystem [35]. All patients were in reproductive age, having\nreceived no hormonal medication during the previous\n3 months before laparoscopic surgery and had a regular\nmenstrual cycle (28 ± 5 days). Self-reported menstrual cycle\nday was used to estimate cycle phase.\nControl group consisted of 24 disease-free women from\nwhom 17 were self-reported healthy volunteers (Elite\nClinic, Tartu and Nova Vita Clinic, T allinn, Estonia) and\nseven were undergoing laparoscopy for pelvic pain, subfer-\ntility or tubal sterilisation and confirmed to be endometri-\nosis free (Oxford control group). Healthy volunteers were\nall in reproductive age, had not used hormonal medication\nat least 3 months before the recruitment, had regular\nSaare et al. Clinical Epigenetics  (2016) 8:2 Page 7 of 10\n\nmenstrual cycle (28 ± 5 days), had normal serum levels of\nprogesterone, prolactin and t estosterone, normal vaginal\nultrasound, negative screening results for sexually transmit-\nted diseases and no presence of endometriosis or polycystic\novary syndrome. Endometrial biopsies for the Estonian con-\ntrols were collected under local anaesthesia, and menstrual\ncycle dating was confirmed by combining menstrual cycle\nhistory, luteinizing hormone (LH) peak (estimated by the\nBabyTime® hLH urine cassette, Pharmanova), vaginal ultra-\nsound and by the histological evaluation of biopsy accord-\ning to the Noyes ’ criteria [36]. The menstrual cycle phases\nfor Oxford controls were estimated according to their\nself-reported menstrual cycle day.\nFor validation study, an extended group of patients\nand controls from LS ( n = 15) and MS phases ( n = 14)\nwas used, and in addition to endometrial samples from\nmicroarray study ( n = 11), further endometrial samples\nfrom patients with endometriosis from LS phase ( n =4 )\nand MS phase ( n = 3) and healthy controls ( n = 8) from\nLS phase and MS ( n = 3) phase were collected (Table 1).\nEndometrial biopsy samples from patients and con-\ntrols were collected using an endometrial suction cath-\neter (Pipelle, Laboratoire CCD).\nPre-processing and normalisation of the methylation\nmicroarray data\nDNA bisulfite treatment using EZ DNA Methylation kits\n(Zymo Research) and DNA hybridization to Infinium\nHumanMethylation 450K BeadChip were performed at\nUSC Epigenome Center (Los Angeles, CA) according to\nthe manufacturer ’ s specifications.\nMicroarray data from Estonian and Oxford datasets\nwere combined for the data analysis. The raw intensity\nfiles were imported into the R statistical computing\nenvironment using Bioconductor package minfi [37].\nThe methylation value (beta value) for each probe was\nthen computed into beta value using Illumina ’ s formula\nM/(M + U + 100), where M and U represent methylated\nand unmethylated signal intensities, respectively [38].\nThe delta beta ( Δβ) value was calculated as difference in\nβ-values between the two groups. Methylation values\nranged from 0, fully unmethylated, to 1, fully methylated\ncytosine. Multiple quality control measures were then\napplied to filter out unwanted probes. Probes containing\nSNP sites (n = 65), probes with the detection P value >0.01\nin more than one sample ( n = 11055) and probes with the\nbeadcount <3 in at least 5 % of the samples ( n = 2074)\nwere removed. The remaining 461,286 probes were\nnormalised for adjusting type1 and type2 probes using\nBeta-Mixture Quantile (BMIQ) normalisation method\n[39]. Finally, the batch effect was corrected using ComBat\nnormalisation method [40]. The preprocessing, quality\ncontrol and batch effect analyses were performed using\nthe Bioconductor ChAMP package [41]. Two Estonian\nsamples and all Oxford samples were run as duplicates\n(technical replicates). The Pearson correlation coefficient\nwas >0.99 for all replicates, confirming a good level of\ntechnical reproducibility. The duplicate beta values were\naveraged and used for further data analysis. PCA and\nunsupervised hierarchical clustering were performed as\na part of quality control and to provide a visual over-\nview of methylation differences between the samples.\nAll analyses were performed using statistical computing\nenvironment R.\nIdentification of DMRs\nDMRs were identified using ‘ seqlm’ package (https://\ngithub.com/raivokolde/seqlm) in the R environment,\nutilising MDL-based approach described earlier [18]. The\nBenjamini– Hochberg FDR was calculated for each probe,\nwith an FDR corrected P value <0.05 used to define DMRs.\nThe DMR analyses were performed to assess the differ-\nences between (i) endometria of healthy and endometriosis\npatients and (ii) menstrual cycle phases. In order to get\noptimal DMRs, we limited our search in regions where\ndistance between at least three consecutive probes was\n≤500 bp. Venn analysis, to determine overlaps between\nDMR genes, was performed using the web-based program\nVENNY 2.0 (http://bioinfogp.cnb.csic.es/tools/venny/).\nValidation of methylation array data by direct bisulfite\nsequencing\nFour CpGs with differential methylation, two from\nCST11 gene (cg06197930, cg12480562), one from PI3\ngene (cg19931348) and one from SLC43A3 gene\n(cg13046608) were selected for validation analysis. Bi-\nsulfite modification of the endometrial DNA samples\n(500 ng each) was carried out with the EZ DNA\nMethylation-Gold ™ kit (Zymo Research) according to\nthe manufacturer ’ ss p e c i f i c a t i o n s .P C Rp r i m e r sf o rt h e\nbisulfite-treated DNA were designed using MethPrimer\n[42]. PCR conditions and list of primers are provided in\nAdditional file 8. The sequencing results were analysed\nas described in [43] and using Mutation Surveyor soft-\nware (Softgenetics, State College, PA, USA).\nFunctional enrichment analysis\nA web-based tool g: Profiler was utilised to query genes\nfrom DMRs for GO category and KEGG (Kyoto\nEncyclopaedia of Genes and Genomes) pathway enrich-\nment (http://biit.cs.ut.ee/gprofiler/) [44]. The FDR P\nvalue <0.05 was considered statistically significant.\nAvailability of supporting data\nThe datasets supporting the results of this article have\nbeen deposited at NCBI Gene Expression Omnibus data\nrepository with accession number GSE73950.\nSaare et al. Clinical Epigenetics  (2016) 8:2 Page 8 of 10\n\nAdditional files\nAdditional file 1: Principal component analysis describing DNA\nmethylation data across all studied endometrial samples. The large\ndots and triangles mark overlapping samples. (TIF 852 kb)\nAdditional file 2: Differentially methylated regions between women\nwith endometriosis and healthy women. The methylation data of all\npatients was compared to controls. Menstrual cycle day has not been\ntaken into account. (XLSX 15 kb)\nAdditional file 3: Differentially methylated regions between\ndifferent menstrual cycle phases. The methylation data of each\nmenstrual cycle phase was compared to other phases. (XLSX 2676 kb)\nAdditional file 4: Functional annotation clustering of hypo- and\nhypermethylated genes in endometrial tissue using g:profiler\nbioinformatics tool. The complete lists of DMRs between different\nmenstrual cycle phases was used to create the lists of Gene Ontology\nterms. (XLSX 27 kb)\nAdditional file 5: Pathway analysis of hypo- and hypermethylated\ngenes in endometrial tissue using g:profiler bioinformatics tool. The\ncomplete lists of DMRs between different menstrual cycle phases was\nused to create the list of Kyoto Encyclopedia of Genes and Genomes\n(KEGG) pathways. (XLSX 12 kb)\nAdditional file 6: Venn diagrams of differentially methylated genes.\nDiagrams show the total number of hypo- and hypermethylated genes\nidentified in each comparison. (TIF 3142 kb)\nAdditional file 7: Menstrual cycle phase specific genes. The list of\nlate-secretory and menstrual phase specific hyper- and hypomethylated\ngenes. (XLSX 28 kb)\nAdditional file 8: PCR primers used in the methylation validation\nanalysis. PCR primers for the bisulfite-treated DNA were designed using\nMethPrimer[42]. (XLSX 9 kb)\nCompeting interests\nThe authors declare that they have no competing interests.\nAuthors’ contributions\nMSa was involved in the conception and design of the study, performed\nexperiments, interpreted the results and drafted the manuscript; VM performed\ndata analysis, generated figures and edited the manuscript; MSu and KR\nperformed the experiments and edited the manuscript; BR was involved in data\nanalysis; DSõ, PS and ASõ recruited study participants, collected and interpreted\nclinical data; HK, CML, CMB, KTZ and AS were involved in the conception and\ndesign of the study and revised the manuscript; NR performed experiments,\ncollected patient’ s medical data, and edited the manuscript; AD performed data\nanalysis and edited the manuscript; MP was involved in the conception and\ndesign of the study, interpreted the results and helped to draft the manuscript.\nAll authors have read and approved the final manuscript.\nAcknowledgements\nWe are grateful to the staff of Tartu University Hospital’ sW o m e n’ sC l i n i c ,N o v a\nVita Clinic and Elite Clinic for recruiting the patients and collecting the samples\nand to the women who participated in the study. We are also grateful to the\nstaff of Endometriosis CaRe Centre, Oxford, and to all women participating in\nthe ENDOX study.\nFunding\nThis research was funded by grants IUT34-16 and IUT34-4 from the Estonian\nMinistry of Education and Research, by European Regional Development\nFund through the Estonian Centre of Excellence in Genomics, by Enterprise\nEstonia, grant no EU30020 and EU48695, by the EU FP7-PEOPLE-2012-IAPP\ngrant SARM (grant no. 324509), by EU-FP7 Eurostars Program (grant NOTED,\nEU41564) and by ERDF through CoE EXCS and BioMedIT projects.\nAuthor details\n1Competence Centre on Health Technologies Tartu, Tartu, Estonia. 2Tartu\nUniversity Women ’ s Clinic, Tartu, Estonia. 3Institute of Bio- and Translational\nMedicine, University of Tartu, Tartu, Estonia. 4Institute of Computer Science,\nUniversity of Tartu, Tartu, Estonia. 5Elite Clinic, Tartu, Estonia. 6Women’ s Clinic,\nTartu University Hospital, Tartu, Estonia. 7Wellcome Trust Centre for Human\nGenetics, University of Oxford, Oxford, UK. 8Endometriosis CaRe Centre,\nNuffield Department of Obstetrics & Gynaecology, University of Oxford,\nOxford, UK.\nReceived: 14 October 2015 Accepted: 30 December 2015\nReferences\n1. Slieker RC, Bos SD, Goeman JJ, Bovee JV, Talens RP, van der Breggen R, et al.\nIdentification and systematic annotat ion of tissue-specific differentially\nmethylated regions using the Illumina 450k array. Epigenetics\nChromatin. 2013;6(1):26. doi:10.1186/1756-8935-6-26.\n2. Muangsub T, Samsuwan J, Tongyoo P, Kitkumthorn N, Mutirangura A.\nAnalysis of methylation microarray for tissue specific detection. Gene. 2014;\n553(1):31–41. doi:10.1016/j.gene.2014.09.060.\n3. Xue Q, Xu Y, Yang H, Zhang L, Shang J, Zeng C, et al. Methylation of a\nnovel CpG island of intron 1 is associated with steroidogenic factor 1\nexpression in endometriotic stromal cells. Reprod Sci. 2014;21(3):395 –400.\ndoi:10.1177/1933719113497283.\n4. Wu Y, Strawn E, Basir Z, Halverson G, Guo SW. Promoter hypermethylation\nof progesterone receptor isoform B (PR-B) in endometriosis. Epigenetics.\n2006;1(2):106–11.\n5. Xue Q, Lin Z, Cheng YH, Huang CC, Marsh E, Yin P, et al. Promoter methylation\nregulates estrogen receptor 2 in human endometrium and endometriosis. Biol\nReprod. 2007;77(4):681–7. doi:10.1095/biolreprod.107.061804.\n6. Lu H, Yang X, Zhang Y, Lu R, Wang X. Epigenetic disorder may cause\ndownregulation of HOXA10 in the eutopic endometrium of fertile women\nwith endometriosis. Reprod Sci. 2013;20(1):78 –84. doi:10.1177/\n1933719112451146.\n7. Fambrini M, Sorbi F, Bussani C, Cioni R, Sisti G, Andersson KL.\nHypermethylation of HOXA10 gene in mid-luteal endometrium from\nwomen with ovarian endometriomas. Acta Obstet Gynecol Scand. 2013;\n92(11):1331–4. doi:10.1111/aogs.12236.\n8. Andersson KL, Bussani C, Famb r i n iM ,P o l v e r i n iV ,T a d d e iG L ,\nGemzell-Danielsson K et al. DNA methylation of HOXA10 in eutopic\nand ectopic endometrium. Hum Reprod. 2014. doi:10.1093/humrep/\ndeu161.\n9. Szczepanska M, Wirstlein P, Skrzypczak J, Jagodzinski PP. Expression of\nHOXA11 in the mid-luteal endometrium from women with\nendometriosis-associated infertility. Reprod Biol Endocrinol. 2012;10:1.\ndoi:10.1186/1477-7827-10-1.\n10. Wang D, Chen Q, Zhang C, Ren F, Li T. DNA hypomethylation of the COX-2\ngene promoter is associated with up-regulation of its mRNA expression in\neutopic endometrium of endometriosis. Eur J Med Res. 2012;17:12.\ndoi:10.1186/2047-783X-17-12.\n11. Izawa M, Taniguchi F, Uegaki T, Takai E, Iwabe T, Terakawa N, et al.\nDemethylation of a nonpromoter cytosine-phosphate-guanine island in the\naromatase gene may cause the aberrant up-regulation in endometriotic\ntissues. Fertil Steril. 2011;95(1):33–9. doi:10.1016/j.fertnstert.2010.06.024.\n12. Borghese B, Barbaux S, Mondon F, Santulli P, Pierre G, Vinci G, et al.\nResearch resource: genome-wide profiling of methylated promoters in\nendometriosis reveals a subtelomeric location of hypermethylation. Mol\nEndocrinol. 2010;24(9):1872–85. doi:10.1210/me.2010-0160.\n13. Yamagata Y, Nishino K, Takaki E, Sato S, Maekawa R, Nakai A, et al.\nGenome-wide DNA methylation profiling in cultured eutopic and ectopic\nendometrial stromal cells. PLoS One. 2014;9(1):e83612. doi:10.1371/journal.\npone.0083612.\n14. Dyson MT, Roqueiro D, Monsivais D, Ercan CM, Pavone ME, Brooks DC, et al.\nGenome-wide DNA methylation analysis predicts an epigenetic switch for\nGATA factor expression in endometriosis. PLoS Genet. 2014;10(3):e1004158.\ndoi:10.1371/journal.pgen.1004158.\n15. Naqvi H, Ilagan Y, Krikun G, Taylor HS. Altered genome-wide methylation in\nendometriosis. Reprod Sci. 2014. doi:10.1177/1933719114532841.\n16. Bouquet de Joliniere J, Ayoubi JM, Lesec G, Validire P, Goguin A, Gianaroli L,\net al. Identification of displaced endometrial glands and embryonic duct\nremnants in female fetal reproductive tract: possible pathogenetic role in\nendometriotic and pelvic neoplastic processes. Front Physiol. 2012;3:444.\ndoi:10.3389/fphys.2012.00444.\n17. Signorile PG, Baldi A. Endometriosis: new concepts in the pathogenesis.\nInt J Biochem Cell Biol. 2010;42(6):778 –80. doi:10.1016/j.biocel.2010.03.008.\nSaare et al. Clinical Epigenetics  (2016) 8:2 Page 9 of 10\n\n18. Lokk K, Modhukur V, Rajashekar B, Martens K, Magi R, Kolde R, et al.\nDNA methylome profiling of human tissues identifies global and\ntissue-specific methylation patterns. Genome Biol. 2014;15(4):r54.\ndoi:10.1186/gb-2014-15-4-r54.\n19. Zhang B, Zhou Y, Lin N, Lowdon RF, Hong C, Nagarajan RP, et al. Functional\nDNA methylation differences between tissues, cell types, and across\nindividuals discovered using the M&M algorithm. Genome Res. 2013;23(9):\n1522–40. doi:10.1101/gr.156539.113.\n20. Houshdaran S, Zelenko Z, Irwin JC, Giudice LC. Human endometrial DNA\nmethylome is cycle-dependent and is associated with gene expression\nregulation. Mol Endocrinol. 2014;28(7):1118 –35. doi:10.1210/me.2013-1340.\n21. Rogers PA, D ’ Hooghe TM, Fazleabas A, Giudice LC, Montgomery GW,\nPetraglia F, et al. Defining future directions for endometriosis research:\nworkshop report from the 2011 World Congress of Endometriosis in\nMontpellier. France Reprod Sci. 2013;20(5):483 –99. doi:10.1177/\n1933719113477495.\n22. Edgar R, Tan PP, Portales-Casamar E, Pavlidis P. Meta-analysis of human\nmethylomes reveals stably methylated sequences surrounding CpG islands\nassociated with high gene expression. Epigenetics Chromatin. 2014;7(1):28.\ndoi:10.1186/1756-8935-7-28.\n23. Wee EJ, Ha Ngo T, Trau M. A simple bridging flocculation assay for rapid,\nsensitive and stringent detection of gene specific DNA methylation. Sci Rep.\n2015;5:15028. doi:10.1038/srep15028.\n24. Tamaresis JS, Irwin JC, Goldfien GA, Rabban JT, Burney RO, Nezhat C et al.\nMolecular classification of endometriosis and disease stage using high-\ndimensional genomic data. Endocrinology. 2014:en20141490. doi:10.1210/\nen.2014-1490.\n25. Furukawa J, Inoue K, Maeda J, Yasujima T, Ohta K, Kanai Y, et al. Functional\nidentification of SLC43A3 as an equilibrative nucleobase transporter\ninvolved in purine salvage in mammals. Sci Rep. 2015;5:15057. doi:10.1038/\nsrep15057.\n26. Stuart RO, Pavlova A, Beier D, Li Z, Krijanovski Y, Nigam SK. EEG1, a putative\ntransporter expressed during epithelial organogenesis: comparison with\nembryonic transporter expression during nephrogenesis. Am J Physiol Renal\nPhysiol. 2001;281(6):F1148 –56.\n27. Talbi S, Hamilton AE, Vo KC, Tulac S, Overgaard MT, Dosiou C, et al.\nMolecular phenotyping of human endometrium distinguishes menstrual\ncycle phases and underlying biological processes in normo-ovulatory\nwomen. Endocrinology. 2006;147(3):1097 –121. doi:10.1210/en.2005-1076.\n28. Ponnampalam AP, Weston GC, Trajstman AC, Susil B, Rogers PA. Molecular\nclassification of human endometrial cycle stages by transcriptional profiling.\nMol Hum Reprod. 2004;10(12):879 –93. doi:10.1093/molehr/gah121.\n29. Celik O, Unlu C, Otlu B, Celik N, Caliskan E. Laparoscopic endometrioma resection\nincreases peri-implantation endometrial HOXA-10 and HOXA-11 mRNA\nexpression. Fertil Steril. 2015;104(2):356–65\n30. Wu Y, Halverson G, Basir Z, Strawn E, Yan P, Guo SW. Aberrant methylation\nat HOXA10 may be responsible for its aberrant expression in the\nendometrium of patients with endometriosis. Am J Obstet Gynecol. 2005;\n193(2):371–80. doi:10.1016/j.ajog.2005.01.034.\n31. Kobayashi H, Iwai K, Niiro E, Morioka S, Yamada Y. Fetal programming\ntheory: implication for the understanding of endometriosis. Hum Immunol.\n2014;75(3):208–17. doi:10.1016/j.humimm.2013.12.012.\n32. Signorile PG, Spugnini EP, Mita L, Mellone P, D ’ Avino A, Bianco M, et al.\nPre-natal exposure of mice to bisphenol A elicits an endometriosis-like\nphenotype in female offspring. Gen Comp Endocrinol. 2010;168(3):318 –25.\ndoi:10.1016/j.ygcen.2010.03.030.\n33. Nishizawa H, Imanishi S, Manabe N. Effects of exposure in utero to\nbisphenol a on the expression of aryl hydrocarbon receptor, related factors,\nand xenobiotic metabolizing enzymes in murine embryos. J Reprod Dev.\n2005;51(5):593–605.\n34. Aragon AC, Kopf PG, Campen MJ, Huwe JK, Walker MK. In utero and\nlactational 2,3,7,8-tetrachlorodibenzo-p-dioxin exposure: effects on fetal and\nadult cardiac gene expression and adult cardiac and renal morphology.\nToxicol Sci. 2008;101(2):321 –30. doi:10.1093/toxsci/kfm272.\n35. Revised American Society for Reproductive Medicine classification of\nendometriosis: 1996. Fertility and sterility. 1997;67(5):817 –21.\n36. Noyes RW, Hertig AT, Rock J. Dating the endometrial biopsy. Am J Obstet\nGynecol. 1975;122(2):262 –3.\n37. Aryee MJ, Jaffe AE, Corrada-Bravo H, Ladd-Acosta C, Feinberg AP, Hansen\nKD, et al. Minfi: a flexible and comprehensive Bioconductor package for the\nanalysis of Infinium DNA methylation microarrays. Bioinformatics. 2014;\n30(10):1363–9. doi:10.1093/bioinformatics/btu049.\n38. Bibikova M, Barnes B, Tsan C, Ho V, Klotzle B, Le JM, et al. High density DNA\nmethylation array with single CpG site resolution. Genomics.\n2011;98(4):288–95. doi:10.1016/j.ygeno.2011.07.007.\n39. Teschendorff AE, Marabita F, Lechner M, Bartlett T, Tegner J, Gomez-Cabrero\nD, et al. A beta-mixture quantile normalization method for correcting probe\ndesign bias in Illumina Infinium 450k DNA methylation data. Bioinformatics.\n2013;29(2):189–96. doi:10.1093/bioinformatics/bts680.\n40. Johnson WE, Li C, Rabinovic A. Adjusting batch effects in microarray\nexpression data using empirical Bayes methods. Biostatistics.\n2007;8(1):118–27. doi:10.1093/biostatistics/kxj037.\n41. Morris AP, Voight BF, Teslovich TM, Ferreira T, Segre AV, Steinthorsdottir V,\net al. Large-scale association analysis provides insights into the genetic\narchitecture and pathophysiology of type 2 diabetes. Nat Genet.\n2012;44(9):981–90. doi:10.1038/ng.2383.\n42. Li LC, Dahiya R. MethPrimer: designing primers for methylation PCRs.\nBioinformatics. 2002;18(11):1427 –31.\n43. Parrish RR, Day JJ, Lubin FD. Direct bisulfite sequencing for examination of\nDNA methylation with gene and nucleotide resolution from brain tissues. Curr\nProtoc Neurosci. 2012;Chapter 7:Unit 7 24. doi:10.1002/0471142301.ns0724s60.\n44. Reimand J, Kull M, Peterson H, Hansen J, Vilo J. g:Profiler — a web-based toolset\nfor functional profiling of gene lists from large-scale experiments. Nucleic Acids\nRes. 2007;35(Web Server issue):W193–200. doi:10.1093/nar/gkm226.\n•  We accept pre-submission inquiries \n  Our selector tool helps you to find the most relevant journal\n  We provide round the clock customer support \n  Convenient online submission\n  Thorough peer review\n  Inclusion in PubMed and all major indexing services \n  Maximum visibility for your research\nSubmit your manuscript at\nwww.biomedcentral.com/submit\nSubmit your next manuscript to BioMed Central \nand we will help you at every step:\nSaare et al. Clinical Epigenetics  (2016) 8:2 Page 10 of 10","source_license":"CC0","license_restricted":false}